Audio Critic the 19 r schematic

Issue No. 19

Retail price: $7.50

The pros love it. The tweaks tried

to kill it. (See the loudspeaker reviews.)

In this issue:

We review nine loudspeaker systems from $829
to $8400 the pair, mostly very good and all very
different, plus an amazing low-priced subwoofer.

The promised guest article on very high-efficiency
speaker systems makes its belated appearance.

David Rich does his professorial number on still
more preamps and several new D/A processors.

The David Ranada interviews with major thinkers
in audio bring new insights in Part II of the series.

Plus delectable exposés of audio bull,
tons of short CD reviews, letters to the
Editor, and an entirely new column.

pdf 1

For subscription information and rates, see inside back cover.

Contents of this issue copyright © 1993 by Critic Publications, Inc. All rights
reserved under international and Pan-American copyright conventions. Repro­duction in whole or in part is prohibited without the prior written permission of
the Publisher. Paraphrasing of product reviews for advertising or commercial
purposes is also prohibited without prior written permission. The Audio Critic
will use all available means to prevent or prosecute any such unauthorized use
of its material or its name.

The Audio Critic is an advisory service and technical review for

consumers of sophisticated audio equipment. Any conclusion, rating,
recommendation, criticism, or caveat published by The Audio Critic
represents the personal findings and judgments of the Editor and the

Staff, based only on the equipment available to their scrutiny and on

their knowledge of the subject, and is therefore not offered to the reader

as an infallible truth nor as an irreversible opinion applying to all extant
and forthcoming samples of a particular product. Address all editorial
correspondence to The Editor, The Audio Critic, P.O. Box 978,
Quakertown, PA 18951-0978.

The Audio Critic® (ISSN 0146-4701) is published quarterly for $24

per year by Critic Publications, Inc., 1380 Masi Road, Quakertown, PA

18951-5221. Second-class postage paid at Quakertown, PA. Postmaster:
Send address changes to The Audio Critic, P.O. Box 978, Quakertown,
PA 18951-0978.

Editor and Publisher Peter Aczel
Contributing Technical Editor David Rich
Contributing Editor at Large David Ranada
Technical Consultant Steven Norsworthy
Columnist Tom Nousaine
Cartoonist and Illustrator Tom Aczel
Business Manager Bodil Aczel

Issue No. 19 Spring 1993

pdf 2

Contents

11

Divergent Design Philosophies:
Nine Speaker Systems and a Subwoofer

By Peter Aczel, Editor and Publisher

12 ACI (Audio Concepts, Inc.) G3
12 ACI Sapphire IIti (Reviewed by David Rich)
13 MACH 1 Acoustics DM-10
15 Monitor Audio Studio 10 (Reviewed by David Rich)
16 Tannoy 615
17 Thiel CS2.2
18 Waveform Mach 7

19 Westlake Audio BBSM-8VF
20 Win SM-10 (follow-up)
21 Subwoofer: Hsu Research HRSW10

24

35

Nostalgia and Loudspeakers

My response to a hundred requests for the lost sound of yesteryear.
By Drew Daniels

Miscellaneous Electronics for Audio and Video

By Peter Aczel, Editor and Publisher
& David A. Rich, Ph.D., Contributing Technical Editor

35 Full-Function Preamplifier: Bryston 11B
36 Cable Enhancer: Duo-Tech Model CE-1000
37 52 " Rear-Projection TV with Surround Sound: Magnavox RM8564A
38 Line-Level Preamplifier: Monarchy Audio Model 10
39 Full-Function Preamplifier: Rotel RC-980BX

42

Another Look at Outboard D/A Converters

By David A. Rich, Ph.D., Contributing Technical Editor (with an interruption by Peter Aczel)

42 Audio Alchemy DTI/XDP/PS2
44 EAD DSP-7000 (follow-up by the original reviewer, Peter Aczel)
49 Monarchy Audio Model 22A
50 UltrAmp D/A Converter

52

From Tweak to Geek or Satanic Realism

By Tom Nousaine

55

Interviewing the Best Interviewees in Audio: Part II

By David Ranada, Contributing Editor at Large

55 5. Interview with Mark F. Davis, Audio Designer
62 6. Interview with Bob Carver, Manufacturer and Audio Designer

66

Hip Boots Wading through the Mire of Misinformation in the Audio Press

66 Gerard Rejskind in The Absolute Sound
67 Robert Harley in Stereophile

68

Recorded Music Catching Up on the New and Not-So-New CDs

By Peter Aczel, Editor and Publisher

3 Box 978: Letters to the Editor

ISSUE NO. 19 • SPRING 1993 1

pdf 3

From the Editor/Publisher:

Whatever Happened to Fall 1992 and Winter 1992-93?

No, your mail didn't go astray. No, we haven't skipped two issues. This is Issue
No. 19. The last one you received was No. 18. Only the dates are a little strange.
Here's what happened:

Issue No. 18 was dated Spring/Summer 1992; however, it was delivered to our
subscribers at the end of summer, in mid-September, very close to the beginning of

fall. It would have been unrealistic to date the next issue Fall 1992 and the one after

that Winter 1992-93 because even in the ideal case (meaning: without our usual
delays) we would have been stuck in an almost-next-season pattern forever. We were
therefore going to call No. 19 the Fall/Winter 1992-93 issue and try to publish it at the
beginning of winter, still not much more than a 3-month interval after No. 18, but that
didn't quite work out, either.

Further delays were caused by a number of desirable but time-consuming new
developments. We had to organize our distribution at newsdealers, bookstores, audio
stores, and other retail outlets, this being our first issue with coast-to-coast retail
distribution. On the advice of our distributors we made some changes in the appearance
of our cover pages for greater recognizability and sales appeal on magazine shelves
with overlapping rows of publications. In consequence of our larger print run, we had
to make small changes in our page dimensions and other specifications in order to

permit printing on a web press. Finally, we had to get ready for second-class mailing

for the first time in our history; on top of it, our authorization to mail at second-class

postage rates was delayed way past all deadlines.

Doing things differently for the first time always takes longer than expected;

add to that our old—should I say traditional?—lateness problems due to the lack of a

full-time staff, and here we are again—late winter as I write these lines. So, for the

reasons already given, Fall/Winter 1992-93 is out and Spring 1993 is the only realistic
designation. That at least shifts us to the beginning of the season and gives us a fighting
chance to have Summer 1993, Fall 1993, etc., issues whose publication actually
corresponds to those dates. As I told you last time in this same space, I have given up
making promises; all I'm willing to say is that I don't expect the nonrecurrent problems
discussed above to cause new delays.

2

THE AUDIO CRITIC

pdf 4

Box 978

Letters to the Editor

We get quite a few intelligent, well-informed, and well-written letters that should never have been
sent and will never be published in this column. Why not? Because they ask questions and bring up
arguments that have already been answered in our pages, usually in the very article or review that
elicited the letter. It's amazing how many people would rather write than read. Halfway through the
article they get excited and rush to their typewriter or word processor. Please read what we have to
say, every word of it, before attempting reciprocal punditry. Letters printed here may or may not be
excerpted at the discretion of the Editor. Ellipsis (...) indicates omission. Address all editorial
correspondence to the Editor, The Audio Critic, P.O. Box 978, Quakertown, PA 18951.

The Audio Critic:

We were amused by the letter on

page 8 of Issue No. 18, in which the ridic-

ulous suggestion was made that neither

magnetic fields nor vibration have any ef­fect on the performance of an audio sys­tem. [That is indeed ridiculous but very

far from what the letter actually said. See

my reply below.—Ed.] This type of mis-

understanding is to be expected from a
nontechnical reader, but we found it sur­prising that the Editor agreed with him.
The letter must have been a last-minute
insertion. [A sarcastic reference to my

lame excuse for letting a tweako ad slip
through.—Ed.]

We at MSB Technology have de­signed all of our products on the basis of
solid engineering principles, including
our EMA Isolation Plate. At the risk of
boring the more technically astute read­ers, we will review the basics of motion,
current, and magnetic flux. This review
clearly shows why magnetic fields and vi­bration affect sonics, and why such a
plethora of products has been developed,
either by chance or design, in this field.

We will start with the simple equa­tion

Φ = F/R

where Φ is magnetic flux, F is magneto­motive force (remember also that mag­netomotive force = 0.4πNI, where N =
number of turns and I = current), and R is
reluctance (magnetic resistance).

One cannot help but see that changes
in magnetic flux will change magneto­motive force, and on the other hand
changes in current will change magneto­motive force and thus magnetic flux. The
principle is the basis for power genera­tion, motors, and of course all speakers.
In ribbon speakers, for example, as the
current changes in a film conductor with­in a magnetic field, the conductor moves,
creating sound. So current in a conductor
in a magnetic field can cause the con­ductor to move.

The same principle can also be re­versed. When a conductor is moved in a
magnetic field, a current is induced in the
conductor. This is how generators work,
as well as early microphones and record
players. As the needle vibrates in the
record groove, a magnet is vibrated, caus­ing the magnetic field to move relative to
the conductor, generating current in the
conductor—the audio signal. In any audio
product, as a conductor carrying an audio
signal is externally moved in a magnetic

field, extraneous currents are induced
within that conductor! These currents add
to or subtract from the audio signal.

At each end of your most basic au­dio system are clear examples of the role
of vibration and magnetic fields in play­ing back and generating sound. Within an
audio component, such as a CD player,
are many interesting sources of magnetic
fields, including transformers, drive mo­tors, and servos. Many of theses sources
change dramatically during playback,
while others are more predictable. Ampli­fiers have very large transformers and
fields. Fields are even created around in­dividual wires and components. If anyone
is interested in a detailed characterization
of these sources, an excellent source book
for further study is Volume 8 of the
Handbook Series on Electromagnetic In­terference and Compatibility. Many
sources of vibration also exist, including
direct feedback coupled through the air or
floor from the speakers, drive motor and
servo vibration, and magnetically induced
power supply vibration (ever felt a large
power transformer hum?).

At MSB Technology Corporation,
we understood these basic principles and
have created the EMA Isolation Plate to

ISSUE NO. 19 • SPRING 1993

3

pdf 5

best deal with both the vibrational prob­lems and the magnetic field problems.

Let's start with magnetic fields and a
simple illustration. Magnetic field lines ra­diate out into space from a horseshoe mag­net. (Remember the iron-filings-on-a-sheet­of-paper experiment in grade school?) [I

know when I'm being patronized, and so
do our readers.—Ed.] When an iron bar

is brought into proximity of the magnet,
the field finds less resistance in the iron,
and the field is coupled through the iron.
It no longer radiates into space. In the
same way an EMA Isolation Plate, on one
side of a large magnetic field, can couple
that field. A simple experiment with a
magnet will illustrate this principle.

As you can see, a solid iron plate
would be totally effective in coupling
magnetic fields. This is a problem, how-

ever. The magnetic sources are electrical,

and losses will occur in a solid plate.
Excessive losses could mean reduced per-

formance of the component. (We ob-

served this with CD player design). Eddy

current losses and hysteresis losses are
the principal source of loss in magnetic
coupling. Hysteresis losses are primarily
material-dependent, with the Steinmetz
coefficient providing an indication of the
quality of the material. (An 8 to 1 differ­ence in performance is seen just among
different types of steel). Eddy current
losses are given by

Pe = (πtfB)2/6pl0

16

where t = thickness, f = frequency, B =
flux density, p = resistivity. Notice that
by reducing sheet thickness, losses are re­duced. This is why transformer cores are
made of laminated sheets—to reduce
eddy current losses. This is also why the
EMA Isolation Plate is made up of thin
layers of low-Steinmetz-coefficient ma­terial. The plate is designed to couple
magnetic fields with a minimum of loss,
or external load on the component.

The second principle is vibration
control. Simplifying a discussion of vibra­tion is difficult. A more rigorous dis­cussion on the subject can be found in the
Theory of Vibration with Applications by
William T. Thomson. As vibration in an
object is damped, the amplitude of each
cycle of vibration is reduced. The ratio of
the amplitude of each cycle to the next
cycle is called the amplitude ratio and is

x1/x2 =

e

δ

where δ is the logarithmic decrement

δ=2πζ/ (l - ζ2)

where ζ is the damping factor

ζ=c/(2mwn)

where c = a constant of proportionality, m

4

= mass, and wn = the natural frequency of
the system.

At the risk of becoming too complex,
it is important to understand that primari­ly two elements determine the vibrational
damping of a system. First is the system
mass. Mass determines the natural fre­quency wn. Second is the damping factor
ζ of the isolation material. With these
simple tools in hand one can evaluate the
most outrageous "vibration control" prod-

uct, from clamps to gooey stuff, and eval­uate its contribution to mass and damping
factor. The EMA Isolation Plate weighs
about 50 lb. and is damped with Isodamp,
a superb material produced by E-A-R
Corporation, with a damping factor of

0.6.
Finally, we have demonstrated the

theoretical importance of vibration and
magnetic fields, and have shown how the
EMA Isolation Plate directly controls
both. The last test is subjective evalua­tion. We have found the EMA Isolation
Plate is effective in improving sound
quality under a wide variety of products.
Its effectiveness depends on the degree of
magnetic shielding in the component, the
component's mass and isolation material,
and its proximity to other components.
Most electrical engineers have little un­derstanding of shielding and vibration,
unless experienced in RF and microwave
applications. Most audio products are in­adequately shielded and damped.

At MSB Technology Corporation we

have utilized the EMA Isolation Plate in

all our products. We have used it as a
base to build our CD player and transport
on for many years. Recently, we de­veloped the MSB Processor, which is
built entirely within an EMA Isolation
Plate. We hollowed out three sections
within the core of the plate, one for the
power transformers, one for digital, and
one for analog. Our circuits are potted
within the core for complete isolation and
vibration control. This unique product has
outperformed every other D/A it has been
compared with.

We hope this simple tutorial will
help you and the reader better understand
the important role vibration and mag­netism play in high-end audio. I can be
reached at [phone] if anyone
wishes to discuss this topic in more detail.

Larry S. Gullman,

B.S.M.E., M.S., P.E.
General Manager
MSB Technology Corporation

It seems fairly obvious to me that

you're just playing games here. Your
opening game is to restate and alter what
the letter writer (Graham Ross of San
Mateo, CA) was actually saying, in order
to make it sound absurd. Read his letter
again. He never said that magnetic fields
or vibration will have no effect on an
audio system, period. He said, in essence,
that your particular product is unlikely to
have an effect. Quite a difference.

After that straw-man game, you play
the we-engineers/you-laymen snob game.
Well, we don't know whether or not Mr.
Ross is an engineer—he could be one,
couldn't he ?—but I have at least as many
engineers in my corner as you do, so I'm
not impressed.

Then you proceed to your big steam­roller game. You cut-and-paste standard

formulas from a reference book and offer

that as scientific proof of the effectiveness
of your EMA Isolation Plate. It's a classic
non sequitur, about as logical and con-

vincing as saying, "E = mc2, therefore
cold fusion works." (Come to think of it,
those Utah cold-fusion fantasists at least
reported some measurements.) We all
agree about the laws of electromagnetism,
vibration, etc. Citing them doesn't prove
a specific product claim. Instead of trying
to intimidate your opposition with Greek
letters, you should have shown what kind
of signal comes out of the back end of a
CD player, first without and then with
your Isolation Plate. If you can show a
significant difference in a typical audio-

phile environment when all other condi-

tions are equal, then I'll begin to believe.
So far you haven't proved anything. I still
think the EMA Isolation Plate is a high­end marketing gimmick, not a solution to
a real-world problem. The fact that
there's a highly qualified technical team
behind it doesn't change my perception.

As for your "simple tutorial"—it

simply isn't one. A simple tutorial would
(I) define all terms and (2) leave the con­stants out of the basic formulas. You were

deliberately trying to be complicated

rather than simple, to be difficult rather

than easy to understand, for purposes of
professional intimidation. I may not be a
professional engineer, but I can tell a
stratagem of adversarial dialectic when I

see one.

—Ed.

The Audio Critic:

Many thanks to you and Dr. David
Rich for including the Sumo Athena II in
your exhaustive preamplifier survey in Is­sue No. 18. In reading the review, it was

THE AUDIO CRITIC

pdf 6

clear The Audio Critic has a unique un­derstanding of the technical details of pre­amplifier design, so we were very excited
the Athena II was one of models rec­ommended to your readers. We would
like to make the following comments:

(1) The Athena II chassis, subpanel,
and top cover are made of 20-gauge steel.
The knobs, rack mounts, and front panel
are solid aluminum. The front panel is not
made of plastic.

(2) Some running production chang­es were made to the Athena II [around
February 1992]. The PC board is now
double-sided with plated-through holes.
Also, capacitors in the primary signal
path have been eliminated, and the line
gain stage uses a DC servo. By the way,
the front panel is not made of plastic.

(3) Sumo has never manufactured,
considered, thought of, looked at,
dreamed of, or even been in close proxim­ity to a front panel made of plastic. The
Athena II front panel is a high-grade cus­tom aluminum extrusion that is machined,
brushed, and black-anodized.

(4) There is a remarkable amount of
substance to Dr. Rich's presentation of
preamplifier design. Once we recover
from the ignominy of having our pride
and joy described as having a plastic front
panel, we look forward to making further
technical comment.

Okay, I guess we have to admit
that's all the self-righteous indignation
(our tongue held firmly in cheek) we can
squeeze out, since we try not to take our­selves too seriously. After all, this busi­ness is supposed to be about having fun
listening to music.

Sincerely,
Michael Custer
President
SUMO
Agoura Hills, CA

David Rich never, never listens to
anything but classical music. Don't ex-

pect him to know about heavy metal—not

even in front panels.

—Ed.

The Audio Critic:

To begin with, I'd like to express my

enjoyment of The Audio Critic in general

and of the latest (No. 18) issue especial-

ly....

I do have a problem with a couple of
areas in Dr. Rich's otherwise excellent re­view of preamps. The first is his refusal to
test tube preamps. The "buggy whip"
analogy is flawed. Car and Driver, Road

ISSUE NO. 19 • SPRING 1993

& Track and other "high-end" auto mag-

azines have reviewed the Dodge Viper,
the Corvette, the "new" Shelby Cobra,
and similar autos which are based on
technology as "new" as tubes are. I would
argue that the high ratings given to these
cars are for the same reasons that tube
electronics are still highly rated in audio
circles. To simply "write off affordable
preamps from Conrad-Johnson, Van Al­stine, Joe Curcio, and Sonic Frontiers
tends to show more about the reviewer's
prejudices than whether such preamps are
competitive with those reviewed. This de­tracts from an otherwise fine article. Two,
I think it is important that when a re­viewer implies the use of "subjective" lis­tening sessions in a review that the entire
setup be listed. The cartridge/arm/
turntable combination is especially crit­ical when evaluating how well the phono
section of a preamp works. The amplifiers
and speakers are also a useful tool in eval­uating how a particular component will
sound with other components. Finally,
does Dr. Rich have a problem with Ad­com? He implies that because the Adcom
GFP-565 uses linear switches it is flawed,
or that this contributes to the overall dis­satisfaction with the unit. Yet he implies
that in the Sumo Athena II the same
switches are fine. In fact, the same Alps
rotational-to-linear control/switch as­sembly is used in both preamps.

Overall the review was very in­formative and enjoyable. Dr. Rich has a
very clear writing style and his article
was never a chore to read. I look forward

to more....

Once again let me thank you for

such an interesting publication. You are
able to strike a balance against the "far
left" of audiophiles without resorting to
name-calling or paranoia. I look forward
to the next issue.

Sincerely,
Michael Baker
Falls Church, VA

Much as I appreciate your compli-

ments, I have a sneaking suspicion that
you still have one foot (or at least a few
residual toes) in that left-leaning camp of
audiophiles, otherwise you wouldn't be
saying some of the things you put in your
letter.

Whether or not the buggy-whip anal-

ogy is apt is rather beside the point. (As
an erstwhile car buff, I could also ques­tion your choice of analogous auto­mobiles, but this is an audio magazine.)
The point is this: the vacuum tube is an

outmoded device, at least for audio ap­plications. Any audio circuit that can be
done with tubes can be done better, or at

the very least just as well, with transis­tors. There is simply no credible technical

reason to go the tube route. I'm willing to
concede that a faultlessly engineered tube
preamplifier is essentially as good and

useful as a similarly well-engineered sol-

id-state preamplifier, but the tube preamp
will lose performance as the tubes age,
and even when new its distortion plus
noise will never be as low as 0.002%.

Why do it, then? The fact is that vacuum­tube audio circuit design has little or no
support in the professional engineering
community; all, or nearly all, the tube
amplifier companies are owned and run
by tweaky audiophiles without engi­neering degrees, and the rave reviews
come from their groupies at the tweaky
magazines. What drives the tube amplifier
market is a belief system, not a superior
technology.

As for critical listening setups, mine
changes (of necessity) all the time, and
David's also changes fairly often. Of
course, we would never use any com­ponents we have found fault with, but we
don't get as dogmatic about reference
systems as the tweako reviewers and their
readers because, unlike them, we perform
subjective listening tests for confirmation
and verification, not for revelation. When
the specifics of the listening setup become
critical for some reason, we do get spe­cific.

Lastly, your skeptical comment on
the Adcom and Sumo reviews is based on
totally false information. The switch as­sembly in the Sumo Athena II sample we
reviewed was made by Nobel, not Alps as
you incorrectly believe. As pointed out in
the review, the Nobel rotational-to-linear
converter is quite robust, whereas the

Alps unit used in the Adcom GFP-565 is
much less so. Perhaps more important is

the fact that the rotational-to-linear con-

verter is connected directly to the switch

in the Sumo but indirectly through a long
unsealed band of metal in the Adcom.

Rest assured that The Audio Critic and

its staff don't "have a problem " with any

particular brand—unlike some publica-

tions I could name, we genuinely don't
care who ends up looking good in our
tests and who doesn't.

—Ed.

The Audio Critic:

Enclosed please find an IAR Hotline!
article (61-62, Aug. 1991) touting the the-

5

pdf 7

oretical and "observed" superiority of
multibit D/A converters over delta-sigma
D/A converters. Since I do not have the
technical sophistication to see what is
wrong with Mr. Moncrieff's argument, I
would be delighted if you or Dr. Rich
would explain what is wrong in a future
TAC article...or published letter. Please
...enlighten me with a short explanation
or direct me to an explanation.

Thank you.

Peter Bandurian
Louisville, CO

The article you enclosed is too long
to be reprinted here, even if there were no
copyright problems, and too inconse­quential to be refuted in a full-length ar­ticle of our own. The question it raises,

however—is delta-sigma a comedown

from multibit?—keeps cropping up in our

mailbox and deserves an answer.

As usual, Peter Moncrieff starts out
with sufficient technical understanding to
give him an entry-level grasp of a com-

plex subject—for which he keeps patting
himself on the back—and then boldly
sails his ship into tweako waters because
his understanding is incomplete, alas. I
asked our Technical Consultant, Steven
Norsworthy, who is a recognized IEEE
authority on digital signal processing in
general and delta-sigma converters in
particular, to comment briefly on Mr.
Moncrieff's errors and on the true nature
of delta-sigma conversion. His somewhat
technical exegesis follows; if you find it a
bit abstruse, just cut straight to his con­clusions. Delta-sigma DACs have their
potential stumbling blocks, as has been
pointed out in these pages before, but
lack of 16-bit resolution definitely isn't
one of them.

* * *

With regard to Mr. Moncrieff's ar-

ticle on ΔΣ (delta-sigma) converters, there
are some extremely obvious flaws. In an
attempt to make ΔΣ converter theory ac­cessible to the layman, the popular audio
press has presented simplistic models
which do not even begin to describe what
is actually going on inside such a convert­er. This is tantamount to misinformation.
So, nonengineers like Mr. Moncrieff try
to make sense out of this nonsense, which
results in jumping to false conclusions.
Unfortunately, it seems that Mr. Moncrieff
has not consulted with professionally de­greed and experienced engineers who are
knowledgeable on the subject.

The ΔΣ D/A converter (DAC) pro-

cess begins in a manner similar to that of

6

an ordinary multibit DAC. The sampling
rate of the 16-bit 44.1 kHz PCM-encoded
signal is increased, typically by a factor
of 8, up to 352.8 kHz. This causes the fre­quencies between 0 Hz and 22.05 kHz to
repeat at multiples of 44.1 kHz all the
way up to 176.4 kHz, which is one half of
the new sampling frequency. These re­peated frequencies are commonly known

as spectral images, and they are removed
by a digital lowpass filter. This whole
process is known as interpolation. Typ­ically, the data in the digital lowpass filter
is processed with greater than 16-bit ac­curacy so that the net accuracy of the re­sult is not less than 16 bits. Usually the
output words are 18 bits, sometimes 20
bits.

Now, the next steps that follow are
different for the multibit DAC versus the
ΔΣ DAC. The multibit DAC takes these
18-bit words at 352.8 kHz and converts
them directly to a corresponding analog
voltage level. This means that the DAC
must be capable of generating 218, that is
262,144 levels. Following this, an analog
lowpass reconstruction filter removes
spectral images, which repeat at multiples
of 352.8 kHz.

In the case of the ΔΣ DAC, the 18­bit 352.8 kHz output of the digital low­pass filter has its sampling rate increased
further, by as much as 32 times that rate,
to 11.3 MHz, so that now the word rate is
as high as 256 times the original 44.1 kHz

rate. This signal is fed into what is known
as a digital ΔΣ modulator. Within the
modulator, the noise introduced by a 1-bit
quantizer is digitally highpass filtered so
as to suppress noise in the signal band
from 0 Hz to 22.05 kHz, while the noise
above 22.05 kHz is greatly increased.
This suppression of inband noise is what
enables the converter to be capable of
achieving its ultimate resolution of 16 or
more bits. An analog lowpass filter then
converts the 1-bit output of the modulator
and removes the high-frequency noise. In
order to assure the preservation of the
original 0 Hz to 22.05 kHz signal, one
only needs to measure the signal-to-noise
ratio in this frequency band from a power
spectrum, which is relatively straightfor­ward to do with proper laboratory equip­ment on an actual DAC system.

It is interesting to note that the actual

information content is much greater for a

1-bit signal at 256 times the original sam-

pling frequency than for the original 16­bit signal. In simple terms, there are 256
bits of information for every original 16­bit sample! (This is completely contrary

to Mr. Moncrieff's claim. He says there

are 256 possible levels, which are repre­sented by only 8 bits, i.e., 28 combina-

tions. There are actually 2

256

possible
combinations of 256 bits!) Of course,
much of this information is the high­frequency noise introduced by the ΔΣ
modulator and is somewhat unrelated to
the original 16-bit input samples.

In summary, there are no technical
reasons to doubt that a properly designed
ΔΣ DAC system is any less capable of

16-bit (or greater) resolution than a com­parable multibit system. This is not to say
that all commercially available ΔΣ DACs
are flawlessly designed. Indeed, some
DACs that are advertised as having 16-bit
quality have subtle flaws inherent in either
the basic architecture, or the actual prac­tical implementation, which can cause
loss of resolution. I plan to address this
subject in a forthcoming issue of The Au-

dio Critic.

—Steven R. Norsworthy

* * *

As for those funny waveform traces
shown in the article, they illustrate some­thing altogether different from what Mr.
Moncrieff thinks; they are the necessary
consequence of the different analog filter
configurations used in the different pieces
of equipment, with different rise times,

different ringing characteristics, etc. They
have nothing to do with the DACs.

I must add that I, personally, am not

necessarily more "sophisticated" in such
purely technical matters than Mr. Mon­crieff, but it seems that I associate with
more sophisticated practitioners than he
does. I must further add that none of us
here would ever dream of referring to our

publication as TAC; that kind of clubby
alphabet soup is in the style of the light­weight tweako journals.

—Ed.

The Audio Critic:

...I have often wondered how many

impending improvements to equipment
that manufacturers disclose to editors re­sult from a wish to deflect criticism grace­fully and how many were actually con­templated before the product's deficien­cies were pointed out by journalists.

[On another subject], on page 13 of

his most interesting and valuable article
[Issue No. 18, "Reasonably Priced Pream­plifiers..."], David Rich refers to an LP
record that is "no longer in print." You
are implored not to allow this inappro­priate expression to creep into your fine

journal. Since LPs and CDs are not print-

THE AUDIO CRITIC

pdf 8

ed in the first place, when no longer avail-

able are properly referred to as deleted,

which of course describes their status in
their respective catalogues. If this seems
too arcane, we are willing to give condi­tional approval to "discontinued" or "un­available."

I only point this out because the pen­alties for such faulty terminology can, in
certain jurisdictions, be quite severe...

József Izsák
Toronto, Ont., Canada

/ often tell David Rich, who always

wants to redesign the electronic equip­ment he is reviewing, that our job is to
evaluate what exists, not to find better so­lutions for the manufacturers. It is true,
however, that some manufacturers look to
the audio journals as a source of design
guidelines, sometimes regardless of the
reviewer's credentials. Each case is dif-

ferent, so I can't make wholesale general-

izations here, but a negative review by,
say, Bob Harley shouldn't make a gradu­ate E.E. change his carefully designed
circuit just to get a better review next
time. That would be sad.

The lapse in David Rich's terminolo­gy is my editorial oversight; his copy
comes to me replete with the (ahem) sty­listic impurities (ahem) that an American
engineering education tends to leave
unremoved. Your admonition is well­taken and will be heeded. The funny part

of it, Jóska, is that it seems to take at least

two Hungarians to keep The Audio Crit-

ic's English out of trouble.

—Ed.

The Audio Critic:

In recent editorials in Stereophile,
Robert Harley has suggested that techni­cal performance is overemphasized when
reviewing audio components. Instead, we
should concentrate on the equipment's
ability to convey the "emotional content"
of the music. Imagine if we applied that

logic to other types of reviews:

"I was moved to tears by the hard­bound edition of Gone with the Wind, but
the paperback left me cold."

Or:

"I laughed hysterically watching The

Simpsons on the 27" Sony, but the 26"
Toshiba rendered the same episode dull

and lifeless."

The quality of a device designed to

convey information must be assessed us-

ing parameters that measure the accuracy

of the information conveyed. The emo­tional content of the subject has little rela-

ISSUE NO. 19 • SPRING 1993

tion to the design of the typeface or the
quality of the picture—it's inherent in the

subject itself! And as long as it's accu­rately presented (measured using parame­ters such as legibility, color contrast,
etc.), the inherent emotional content re­mains unchanged. Just as television sets
are measured on their ability to accurately
present a picture, audio components must
be evaluated based on their ability to ac­curately present the sound. Nothing more,
nothing less. Robert Harley would do
well to leave the emotion to the music.

Mark L. Swierczek
Great Mills, MD

/ not only agree with you but have

editorialized in the same vein more than
once. (See, for example, Issue No. 17,

page 45, first indented paragraph.)

Of course, Robert Harley can't be
objectively tested on what he feels when
he listens, the way he could be tested on
what he actually hears (if he didn't refuse
to be so tested). Thus, for all we know, he
is moved to ecstasy by certain logos and

front panels, and is left cold by others.

That's entirely possible. We do know that
he must look at the brand name on the

front panel before he is able to form an

opinion of the sound.

—Ed.

The Audio Critic:

I have received every single issue of
your publication, and I very much want
you to continue publishing. You are the
only magazine that I trust to give me the
necessary facts to make an intelligent de­cision about what audio equipment to
buy. I am totally unable to spend the kind
of money other high-end publications rec­ommend, and I have a natural distrust in
someone who thinks spending $10,000
for a preamp will make them happy.

Many of my components have been
purchased after reading about them in
your magazine. For instance, I still enjoy
listening to LPs through an Advent Model
300 receiver. It is not my main preamp,
but it still works well.

I also receive almost every other au­dio publication around, as I am not only a
musician but a music lover, and I like to
keep up with the latest in equipment even
though the newest component I purchased
is over two years old. Yours is the only
publication whose reviews discuss the
practical reality of design features (and
flaws). I really get tired of reading about
how the cymbal crash at measure 149 in
Mahler's Third sounded with this speak-

er, or that amplifier, or how the hushed
murmur of the strings moved this re­viewer or that one. I want to know what
those unbuffered tape outputs mean if I

buy that preamp, and you deliver when it

counts. Keep up the good work.

Charles Hardgrave
Goldthwaite, TX

I gratefully acknowledge and genu-

inely appreciate your kind words, but let's
not go overboard here. A competent audio
component review is necessarily focused
on the engineering of the unit, but the
sound of that cymbal crash (or of those
murmuring strings) can be an important
clue to an engineering fault or advantage,
especially in the case of loudspeakers.
Self-indulgent subjective reviewing is
worthless, but technical analysis without
listening is also of rather limited value.

We at The Audio Critic measure every­thing, listen to everything, and try to get
the best professional opinion on the engi­neering strengths and weaknesses of
everything—because only such a compre­hensive approach meets our standards of
accountability.

—Ed.

The Audio Critic:

Congratulations on the addition of
David Ranada to your staff. His inter­views were well focused, and I am looking
forward to more of them. I particularly
appreciate Kevin Voecks's statement that
loudspeaker models of the same type
should not vary in quality.

The letters column in Issue No. 18
discussed your modified views on the im­portance of absolute phase. Another point
made in an early issue of The Audio Crit-
ic is that electronic components need to
warm up for at least an hour before they
sound their best. Do you still believe this?
I once did but no longer do.

The only CD player that you have
found to be audibly superior to its com­petition was a Sony unit modified by Pre­cision Audio. I am curious why you have
not used this unit (or another Precision
Audio modified model) as reference in
your more recent tests of CD players and
D/A converters.

Keep up the great work!

Sincerely yours,
David Altman
Maywood, NJ

/ must confess that we haven't so far

investigated the warm-up question scien­tifically. The right way to do it would be

7

pdf 9

to go through our complete measurement
protocol on a dead-cold unit immediately
after turn-on, then again after a one-hour
warm-up, and once again after 24 hours.
It could very well be that the differences
are trivially small or nonexistent, but just
in case I'm wrong I always warm up the
equipment for at least 30 minutes before
any serious listening. (I also knock on
wood occasionally.) One of these days I'll
do the tests, at least on the components
that are semipermanent in my system.
(No, I don't trust my earlier subjective

perceptions on this anymore.)

As for the Sony CD player modified
by Precision Audio (New York), they took
it back from me shortly after my review.
(Incidentally, I corrected their name in

your letter; you wrote Audio Precision,

which is a much larger company in Ore­gon, making state-of-the-art audio mea­surement gear.) Many of the more recent
CD players and DACs reflect the same
circuit design philosophy as the Precision

Audio mod, so I don't think I lost an ir-

replaceable reference.

—Ed.

The Audio Critic:

...I have struggled over the years,
desperately trying to hear the "astound­ing" differences attributed to various sys­tem tweaks by the high-end community. I
always wanted to hear these differences
so that I could justify spending large
sums on beautiful toys. I have usually
failed to hear these "differences," so I
made purchasing decisions on the basis of
friends' and trusted dealer recommenda­tions.

I now find myself with a decent sys-

tem... [The letter lists the very respect-

able components in the system, but they
are of no relevance to the issue raised be­low.—Ed.]

I genuinely enjoy the sound of my
system but have been anxious to have re­mote control, at least of volume, from the
listening chair. This idea has seemed to
be anathema to the high-end community.
I cannot understand why. I was excited to
hear about the Forte 44, but then heard
from my local dealer that since the re-

organization at Threshold he no longer
plans to carry the line because he was dis­appointed in the way the company has put
together the product.

In your most recent preamplifier sur­vey, none of the units had remote control.
I would love to hear some discussion in
the magazine about why only Jeff Row­land and Krell seem to be able to offer
"properly done" remote control, and also
the opinion of you and your editors about
the possible availability of such units in
the near future.

Thanks.

Yours truly,
Eric Brody
Lake Oswego, OR

Remote control has never been im-

portant or even mildly interesting to me

because I am a pacer rather than a couch

potato. I jump up, pace around, go to the

front panel, adjust the volume or select

another source, pace around again, and
sit down only later. The finer points of re­mote control are therefore beyond my
ken, so I asked Dr. Rich (who is even

more hyper than I am but at least knows

the subject cold) to answer you.

* * *

Two distinct problems must be
solved in a remote-control preamp. The
first is the switching of signals. One meth-

od is to use a CMOS solid-state switch.
CMOS switches restrict the maximum in­put swing of the switched signal to some
fraction of the power-supply voltage of
the switch. This is typically ±5 V, al­though some expensive switches allow
more that that. If the input signal exceeds
the power supply of the switch, the switch
can be driven into a potentially de­structive latchup mode. Another problem
with a CMOS switch is that it has a rel­atively high on resistance. This, in con-

junction with the nonlinear junction ca-

pacitance of the switch, can lead to signal
distortion. The on resistance of the switch
is itself nonlinear, and it is critical not to
have the CMOS switch terminated into a
resistive load. Bipolar devices are not
usually used for switches because the
voltage drop,

V

CE(sat)

,

across a bipolar de-

vice cannot be brought to zero. Another
method of performing switching is to use
a relay. There can be reliability problems

caused by the mechanism of the relay or
by corrosion of the electrical contacts due
to an insufficient wetting current. Sealed
relays with switch contacts made of rare
earth materials can minimize these prob­lems, but these devices are expensive.

The second problem involves the
design of the remote-controlled volume
control. One method is to use a tapped re­sistor string, with the desired tap selected
by a CMOS switch or relay. The afore­mentioned problems associated with these
components will then arise. A second
method is to use a voltage-controlled
amplifier (VCA). The VCA is an elec­tronic circuit in which a DC control volt­age sets the voltage gain of the circuit.
These circuits often have limitations in
dynamic range and have relatively high
distortion. Those interested in a detailed
analysis of VCAs are referred to Ben
Duncan's series in Studio Sound magazine.
The third method is to drive a standard
potentiometer with a servo-controlled mo­tor. This solution extracts no performance
penalties.

—David Rich

* * *

I'm pretty sure those cute little ser-

vo-controlled motors ordered in limited

quantities aren't cost-effective for small
audio manufacturers. It seems remarkable,
therefore, that Forte puts the industry's

first motorized input selector in the 44 at

the $1095 price point, whereas Krell still
proudly advertises relays for the switch­ing functions in the KRC at $6000 and
motorized control only for the volume. I
believe that Jeff Rowland also uses relays

for the switching functions. Apparently

there's always more than one "properly
done " engineering solution. (Incidentally,
David Rich was unaware of Forté's mo­torized input switching when he wrote the
above but later said it sounds like a good
solution to him.) As for your local dealer,
how do you know whether he dropped
Forte or the new Threshold organization
dropped him ?

—Ed.

Erratum

Our review of the Acurus L10 preamp, printed as an advertisement

apparently contained an error. According to Mondial Designs,

William Snyder is not "now a consultant to Mondial." He did, however, design the

basic circuit of the Acurus L10, whereas the actual hardware implementation

was the work of Mike Kusiak. Our rule is that advertising reprints must be exactly

what was published, warts and all, hence the form of this correction.

THE AUDIO CRITIC

8

pdf 10

Divergent Design

Philosophies: Nine Speaker

Systems and a Subwoofer.

By Peter Aczel

Editor and Publisher

While the electronic components in the audio chain are gradually
converging toward a few basic design architectures, audiophile­quality loudspeakers are still trying to make highly individual

statements. Is that good? Read the reviews and decide.

Loudspeakers have been my number one topic of

interest since the earliest days of The Audio Critic; I have
written on the subjects of speaker design and speaker test­ing at great length; longtime subscribers know exactly
where I stand, and newcomers are finding out rapidly—so
I have no intention to repeat myself here just because a
new crop of speakers came in for review. Readers are re­ferred to Issues No. 10, 11, 14, 16, and 17 for background
information about my approach and predilections.

The only new thing I want to note here is that I
have started to make use of the MLS (Maximum Length
Sequence) testing capability of the Audio Precision "Sys­tem One Dual Domain" in order to obtain the equivalent
of anechoic response measurements above approximately
300 Hz. (The more widely used and much less costly
MLSSA add-on system for the PC is based on the same
principle.) I can't say that it has been a revelation; my
older, cruder methods also showed me what I needed to
know, albeit less conveniently and a little less accurately.

At any rate, since we don't really have a mathemat­ical model for the "perfect" loudspeaker operating within
room boundaries—as we do, e.g., for the "perfect" ampli­fier—all loudspeaker testing remains a bit tentative and
open to argument on certain points, although we can
readily (and incontrovertibly) measure most of the im­portant performance characteristics. The measurements
invariably show nontrivial differences between input and
output—i.e., distortion—in all speakers, and that means
(1) that we don't know what a totally nondistorting speak­er would sound like and (2) that the preferred choice be­tween one kind of distortion and another is necessarily
subjective, at least to some degree. The soon forthcoming
DSP-corrected loudspeaker systems will perhaps begin to

ISSUE NO. 19 • SPRING 1993

change that situation, but don't count on amplifier-perfect
speakers just yet.

It is interesting to note that, even though loudspeak­ers are obviously the one remaining component category
open to insufferably self-indulgent subjective reviewing,
most of the insufferable subjective reviewers freeze up
when it comes to describing the sound of a speaker and
become amazingly conservative—the speaker has, for ex­ample, a "recessive" midrange, whereas the midrange of
the amplifier just reviewed is "chocolaty" or "dark-hued,"
right? It would appear that reality is more inhibitive to
colorful description than fantasy.

I also want to reiterate here my great reluctance to
publish the printouts of my measurements, although I
keep getting requests to do so. I have a great fear of mis­interpretation by the technically semiliterate, which could
then be straightened out only by publishing even more
printouts, resulting in page after page of little or no value
to the average reader. I believe that a basic truth is better
expressed in a few well-chosen words than a picture, ex­cept in certain rare instances when, of course, I'll make
an exception and publish the picture. I further believe that
the plethora of "scientific" graphs in certain high-end

journals is a cosmetic cover-up for the lack of science in

their basic approach to equipment evaluation.

Who wrote which review?

Two of the reviews that follow are by David Rich,
who loves small speakers, whereas I find them a source of
frustration in my large listening room. His byline appears
at the head of those two reviews; the other seven are my
humble handiwork. All of the measurements discussed by
either one of us were taken in my laboratory.

11

pdf 11

ACI (Audio Concepts, Inc.) G3

Audio Concepts, Inc., 901 South 4th Street, La Crosse, WI

54601. G3 floor-standing 3-way loudspeaker system, with base,
$829.00 the pair (direct from ACI, fully assembled, including
shipping charges). Full kit (without base), $709.00 the pair
(including shipping charges). Tested samples on loan from man­ufacturer.

ACI is how Audio Concepts, Inc., would now like
to be known. As longtime readers know, I look with dis­favor on all forms of alphabet soup, of which we have far
too much in audio (as we do of companies having names
starting with audio, for that matter). Hey, what's wrong
with Mike Dzurko, Inc., named after the owner? Isn't that
more distinctive? Mike has a great deal to be proud of,
but perhaps he is too modest to tout his own name. Or
maybe he thinks it sounds too Slavic—but did that hurt
Stravinsky in the music business? Anyway, ACI it is.

The G3 is one of the things Mike can be proud of
because for the price of so many other mediocre, not­quite-full-range speakers this model offers you a 40 Hz to
20 kHz system without significant faults, good enough in

just about every important respect to satisfy the dis-

criminating audiophile. That's an achievement not to be
sneezed at. If you opt for the full kit version, the value per
dollar becomes even more remarkable, since relatively lit­tle work is involved.

The speaker consists of a 10" woofer in an "aperi-

odic" enclosure vented rearward through five 9/16" holes,

a 5" midrange driver in an isolated subenclosure with a
single 1" hole in the back, and a 1" ferrofiuid-damped alu­minum-dome tweeter firing through a flat doughnut of

felt. The whole structure stands about 40" high, which in­cludes the 4" high screw-on base. The width and depth
are about a foot each. The cabinet is solidly constructed
of ¾" stock and finished in wood veneer (several choices
available) on the sides and top only. The black knit grille
cloth is stretched on a beveled frame, not as ingeniously
antidiffractive as that of the Thiel CS2.2, for example, but
showing some attention to the diffraction problem none­theless.

The impedance characteristics of the speaker pre­sent an easy load to the amplifier. Above 100 Hz or so,
past the box-tuning convolutions, the magnitude stays be­tween 4.2 and 10 ohms, the phase within ±22.5°. Can't
ask for anything simpler.

The bass response of the G3 goes down to 40 Hz,
the approximate -3 dB point, and declines at the rate of

12 dB per octave below that. Don't confuse such a profile
with that of a vented box tuned to 40 Hz; the latter would
have considerably less output in the 20 to 30 Hz region.
The G3 delivers genuine low bass without boom; the ape­riodic tuning assures good damping. (Yes, it's a "fast"
woofer in untutored tweako terminology.) The woofer is
at the bottom of the enclosure, near floor level—to avoid
the "floor bounce" identified and demonized by Roy Alli-

12

son—leaving about 13 inches ( meter) of land between
woofer and midrange, which made it a bit difficult for me
to measure the 1-meter frequency response of the system
between 200 and 1200 Hz. The crossover to the midrange
is in that region (350 Hz, 12 dB per octave); above 1200
Hz my 1-meter MLS measurements with the microphone

aimed at the midrange/tweeter boundary should be com­pletely trustworthy, reading ±2.5 dB on axis, all the way
up to 20 kHz. The response is actually ±0.75 dB between

10 and 20 kHz! That obviously good tweeter (made by
Vifa) is crossed over at only 6 dB per octave; fortunately
it comes in quite high, above 4 kHz, so that the lack of
bottom-end filtering creates no major power-handling
problem. Off axis I measured some significant dips in the
vicinity of that 4-kHz-ish crossover, of the order of 5 dB
at 30° off horizontally and 15 dB at 30° off vertically.
That's not at all surprising with a first-order crossover.

In the time domain I observed no anomalies. A pos­itive-going pulse makes all three drivers push forward;
square pulses of any width cannot be coherently repro­duced, but then I've never seen it done in a 3-way system
regardless of the crossover; and neither pulses nor tone
bursts revealed any storage patterns. It's a clean machine
(to quote Paul McCartney).

In general I discern no grand design behind the

driver placement, enclosure geometry, and crossover con­figuration of the G3, just good, pragmatic decisions based
on results vs. cost. The drivers are good, the overall con­struction is good, the crossover is good (at least on
axis)—so the sound is good. Indeed, the sound is more
than just good; it's quite remarkable. I never listen to
speakers in this price range for more than just a few
hours, long enough to form an opinion, but this one I left
in my system for five days and remained basically happy,
although I'm accustomed to the sound of much costlier
speakers. The sound of the G3 is open, smooth, finely de­tailed, and quite precise in imaging. I prefer it to the
sound of the Sapphire IIti (see below), which in its cur­rent incarnation has a much less pleasant top end and, of
course, needs a subwoofer. If I had to find fault with the
G3—as I'm not really inclined to—I'd say it sounds just a
little too smooth and rounded in the treble, lacking suf­ficient bite (as distinct from edginess), probably as a re­sult of those off-axis suckouts in the most sensitive range

of the ear, where we naturally expect to be overstimulated
rather than spared. It's not a serious objection; at least the
sopranos never scream at you. For the money, and then
some, this is an outstanding speaker.

ACI Sapphire IIti

(Reviewed by David Rich)

Audio Concepts, Inc., 901 South 4th Street, La Crosse, WI

54601. Sapphire IIti compact 2-way loudspeaker system,
$964.00 the pair (direct from ACI, fully assembled, including

THE AUDIO CRITIC

pdf 12

shipping charges). Full kit, $884.00 the pair (including shipping
charges). Tested samples owned by the reviewer.

This speaker is a modification of the Audio Con­cepts Sapphire II minimonitor reviewed in Issue No. 16.
The modification involves the replacement of the
modified Focal T120KT Kevlar inverted-dome tweeter
with a modified Focal T120Ti, which has an inverted
dome made of titanium. An additional modification to the
speaker is that its price has been increased to $964.00 the
pair. It is also available as a kit for $80.00 less. Should
you decide to return the assembled speaker—remember,
the speaker is only sold direct—after the at-home audition
period (which has been reduced from 30 to 15 days), ACI
will no longer refund the assembly charge.

The tweeter modification gives us a chance to re­assess the speaker, this time using the Audio Precision
MLS test setup to measure frequency response. The new
tweeter appears to have rearranged the large resonant
peak of the original tweeter at 17 kHz; in one of my sam­ples there is now a 4 to 5 dB peak at 14 kHz followed by
a comparable dip at 17 kHz; in the other sample the peak
is fairly well suppressed but the dip is not. Tone burst test­ing showed significant energy-storage effects above 12
kHz in both samples. This is hardly the performance one

expects from a state-of-the-art tweeter design; indeed,
many soft-dome tweeters perform better. The woofer also
showed significant energy-storage effects on tone burst
tests above 2 kHz.

The titanium tweeter may be more efficient than the

old Kevlar unit, since the on-axis MLS frequency re­sponse curves of the IIti indicate that the average tweeter
level is 1 or 2 dB above the woofer level when the micro­phone is aimed at the apex of the woofer. A 5 to 8 dB dip
(depending on the sample and the microphone position)
in the amplitude response between 2 and 3 kHz is the
only other significant error in the frequency response
curve. Frequency response runs at 30° off axis show less
difference in level between woofer and tweeter but an ad­ditional dip of 5 dB between 4 and 5 kHz appears in the
curve. Whether the grille is on or off the speaker does not
significantly affect the amplitude response. The 30° off­axis measurements also show a phase variation of less
than 45° from 300 Hz to 15 kHz (phase variation is great­er on axis), justifying the claim that this is close to a mini­mum-phase design. But a large price is paid for minimum
phase—the drivers exhibit very significant interference
effects when the speaker is used above or below its op­timum horizontal plane. (See Issue No. 16 for more de­tails on the cause of this effect.) Aiming the microphone
at the apex of the tweeter resulted in an amplitude dip be­tween 4 and 7 kHz with a minimum value of 16 dB.
Clearly, great care must be taken when these speakers are
set up to insure the woofer is at ear level. The two sam­ples were typically matched within 1 dB, but variations of
more than 2 dB were observed in some frequency ranges.

Before the tweeter update the Sapphire II was the

ISSUE NO. 19 • SPRING 1993

principal speaker in my reference system. The unmodified
speaker, while lacking the transparency of high-priced
state-of-the-art speakers, gave an excellent account of it­self over a wide range of program material. I used the
equalization function of the Cello Palette (more on this
preamp-equalizer will follow if a loan can arranged from
the manufacturer) to verify that the transparency loss is
partially due to the amplitude dip in the 2 to 3 kHz region.
I had my Sapphire II speakers updated to the IIti by ACI,
so a direct comparison between the new and old tweeters
was not possible. From my memory of the unmodified
Sapphire II, it appears the new tweeter has brightened the
sound of the speaker and the treble range is more colored.
The speaker is now less forgiving of problems in program
material, but the reproduction of audiophile-quality re­cordings has also been degraded.

Given the speaker's price increase and performance
decrease, the Sapphire II is no longer the remarkable bar­gain it once was. The speaker would still be good enough
to warrant a risk-free audition, but the audition is not risk­free under ACI's new policy. The cost of shipping and
nonrefundable fees add up to over $115.00. The speaker
is, in my opinion, not good enough to justify taking this
chance. I can thus only recommend audition of the speak­er in its kit form. Assembly of the kit requires significant
care to prevent damage to the fragile drivers. The speaker
costs $80.00 less as a kit, which reduces the risk of return
to the $35.00 shipping charge and the time required to as­semble the speaker.

[Having not only measured but also listened to Da­vid Rich's modified speakers in my laboratory, I agree
with his opinion about the decline of this product. The
original Sapphire was an awkwardly packaged, somewhat
impractical but marvelous speaker that gave the Quad

ESL-63 a hard time in side-by-side comparison. The Sap­phire II was almost as good and a much more sensible
package. The Sapphire IIti sounds slightly wiry and sib-

ilant—a very untypical quality in an ACI speaker—and

lacks the sonic refinement and class of the original de-

sign. Maybe Mike Dzurko listened to some not-so-good

advice instead of his own good ears.—Ed.]

MACH 1 Acoustics DM-10

MACH 1 Acoustics, RR 2, Box 334A, Wilton, NH 03086. DM-10
floor-standing 3-way loudspeaker system, with accessory granite

base and grille, $7995.00 the pair. Tested samples on loan from

manufacturer.

Let me say it before we get involved in the details:
this is one of the finest loudspeakers known to me, re­gardless of price. Like everything else in this world, it has
its limitations, but those limitations are intrinsic to the ba­sic concept and intended purpose of the speaker; they
aren't design faults. The speaker is intended for extremely
refined, high-resolution playback at not excessively high

13

pdf 13

levels in not excessively large spaces, and it accomplishes
that faultlessly.

The key to the design is the choice of drivers. Marc
McCalmont, the designer of the DM-10 (he is a Marine
flier turned Pan American pilot turned audio entre-

preneur), chose the Accuton 1" inverted-dome tweeter
and 3½" inverted-dome midrange, and a 9½" Dynaudio
woofer. The Accutons have ceramic diaphragms made by
vapor deposition and are billed to the manufacturer at ap­proximately $160 and $200, respectively. They are quite
fragile and need to be crossed over just so to keep them
out of trouble. The Dynaudio is also ridiculously ex-

pensive, so that Marc pays over $1000 up front for drivers

before he has even started to put other parts into a pair of

speakers. Welcome to the world of High End. I must say,
however, that these are better drivers than you get in, say,
a Wilson WATT.

The cabinet of the DM-10 has 1" walls, except the
front bafflle, which is made of 1¾" damped laminate.
This guy doesn't fool around. The dimensions of the box
are 44" high by 11" wide by 14½" deep; the front edges
are rounded; the finish is in your choice of veneers; the
grille is optional, the basic design having been conceived
with fully exposed drivers. The woofer is located only a
few inches above floor level to avoid "floor bounce" (see
the ACI G3 review above); the midrange and tweeter sit
high and are offset inboard, resulting in a mirror-image
pair. The woofer is in a sealed enclosure; the crossover
slopes are fourth-order (24 dB per octave); the network is
made with air-core inductors (except in the woofer cir­cuit) and polypropylene capacitors; the crossover fre­quencies are approximately 250 Hz and 3 kHz. The gen­eral design philosophy is to be textbook correct and never
mind the cost. No tricks, no surprises, no compromises.

I found only two basic design characteristics that I
—putting myself in the place of a purchaser—would have
wished to see improved at this exalted price level. One is
the bass, which is very clean and well-controlled but
could go deeper in a box of this size. (An off-the-shelf
woofer, no matter how costly and how magnificently
made, hardly ever has the exact Thiele-Small specs for
the particular system optimization one needs.) The near­field response I measured was flat down to an f3 of 44 Hz
and declined 12 dB per octave below that—a classsic
sealed-box profile. The impedance curve indicates that
the box is tuned to 35 Hz, so the system must be slightly
overdamped. As the nearfield response at 30 Hz is only 10
dB down, the "room gain" in smallish rooms should bring
it up a tad, but in my big room I would have preferred
stronger bass. The other small weakness of the design is
that the little Accuton tweeter is somewhat deficient in
power handling, so that you have to watch the level in op­era recordings, for example, because the soprano's for­tissimo high notes tend to sound a bit strained if you turn
up the volume. This is a medium-signal, rather than a
large-signal, transducer.

That said, I must then immediately add that at nor-

14

mal to moderately high levels the sound of the DM-10 is
exquisitely beautiful and transparent, absolutely world­class. Both texture and structure—to use the John Eargle
terminology which is so superior to the high-end tweako
vocabulary—are as accurately reproduced as anyone
could wish for. Furthermore, the crossover design and
driver mounting/positioning are such that the speaker
isn't the least bit temperamental when it comes to place­ment—the soundstage doesn't collapse and the balance
doesn't go to hell when you move the cabinets eight inch­es this way or that way. (Marc McCalmont has written an
entire manual on room acoustics and speaker placement,
by the way.)

In measuring the 1-meter response of the speaker
with the MLS technique, I didn't run into the same prob­lem as I did with the ACI G3, although the vertical dis-

tance between woofer and midrange is even greater in the
case of the DM-10. The much steeper crossover slopes
are probably the reason. On the tweeter axis, the response
was ±3 dB from 300 Hz to 20 kHz, which is even better
than you'd think because in the crucial three octaves from

1 kHz to 8 kHz the deviation from absolute flatness was

only ±1.25 dB. And that's not all. There's hardly any
change in the response up to 10 kHz at 30° off axis; only
the 10 to 20 kHz response starts to slope downward a bit.
No wonder the speaker sounds great.

In the time domain, I observed nothing that could

change my high opinion of the DM-10. Pulse coherence

was of course nil; it's a spread-out 3-way system with
high-order crossovers to begin with, and a positive-going
pulse pushes the tweeter diaphragm inward, whereas the
midrange and woofer diaphragm move outward. It's aca­demic; the proof of the pudding is in the superior fre­quency response on and off axis. I did see just a tiny bit
of garbage between tone-burst envelopes but not enough
to attribute any importance to.

The impedance characteristics of the speaker in­dicate the need for an amplifier of good but not excep­tional current capability; above the impedance swings due
the box, the magnitude stays between 3 and 8 ohms and
the phase within ±25°. Efficiency is of the order of 87 dB,
which is about average for speakers in this format.

Where do I rank the MACH 1 Acoustics DM-10?
If sheer transparency, refinement, and naturalness of sound
are the top priorities, it ranks very close to the top. I
haven't tested everything, of course, but its only competi­tion known to me in that super-finesse category is the Win
SM-10. If, on the other hand, the big sound, life-size dy­namics, deep bass, and generally awesome impact are the
desired traits, then it lags behind the Waveform Mach 7,
the Carver "Amazing Loudspeaker" Platinum Mark IV,
and others of that ilk not yet reviewed, which are slightly
cruder in sonic texture, at least in my opinion. In any
event, although the DM-10 is fairly priced considering the
manufacturer's cost of parts and labor and the dealer's
normal markup, it's still a classic case of "if you have to
ask the price you can't afford it."

THE AUDIO CRITIC

pdf 14

Monitor Audio Studio 10

(Reviewed by David Rich)

Monitor Audio Loudspeakers, Kevro International, Inc., P.O.
Box 1355, Buffalo, NY 14205. Studio 10 compact 2-way loud­speaker system, 2549.00 the pair (without stands). Tested sam­ples on loan from distributor.

Editor's Note: This review had already been written and
was moving slowly through our editorial pipeline when
Monitor Audio announced their new Studio 6, replacing
the Studio 10. We already have the new model on hand as
this issue goes to press, and the differences appear to be
minor, certainly not significant enough to require the
withdrawal of this review. The Studio 6 introduces some
improvements, which will be the subject of a briefer re­view in Issue No. 20, using this review as background.

* * *

Monitor Audio is a small British speaker manufac­turer. A distinct advantage of Monitor Audio over small
American manufacturers is that the company custom de­signs its own transducers. The Studio 10 has custom­designed drivers exclusive to this unit. Monitor Audio
was one of the first manufacturers to use metal-cone driv­er technology. In the Studio 10 both the woofer and
tweeter diaphragms are made of metal. The use of an alu­minum/magnesium alloy in the tweeter dome and a pro­prietary anodizing process are claimed to reduce tweeter
resonance. Ceramic coatings and a three-stage metal
drawing process for stress release are claimed to reduce
cone resonance in the woofer. The large R&D expense
involved in developing the drivers and the associated
manufacturing process is reflected in the price of the Stu­dio 10. What is unacceptable to me at the speaker's price
point is the fact that the five-way binding posts at the in­put of the speaker are not 0.75" spaced, so you cannot use
double banana plugs. [Britannia no longer rules the

waves in high-end audio but she still waives the rules. Ac­tually, David, old boy, the measure used to determine the
distance between those ruddy binding posts, don't you
know, is King Henry VIII's thumb.—Ed.]

The cabinet is made of ¾" thick Medite medium­density fiberboard. The speaker's woodwork is of the
quality found in high-end furniture. As an example, the
cabinets are veneered inside the box as well as outside.
This prevents the wood from warping by balancing the
stress on the wood exerted by the veneer. For $500.00
more a rosewood piano finish is available. I would sug­gest the availability of a lower-cost vinyl walnut version
(that way the speakers would match the rest of the fur­niture in my apartment), but the master wood craftsman at
Monitor Audio would almost certainly be insulted by my
even suggesting such an idea. The structural rigidity of
the cabinet is more important to a speaker's performance
than the cabinet's looks. The cabinet lacks the extensive
crossbracing of the ACI Sapphire II and appears (using
the high-tech knuckle-rap test) to be more resonant. D. B.

ISSUE NO. 19 • SPRING 1993

Keele Jr. in his review of the speaker in the July 1991 is­sue of Audio found a slight cabinet resonance at 445 Hz,
using a more sophisticated test metrology. Perhaps to
minimize the effect of the resonance, a pair of optional
lead-filled metal speaker stands, weighing over 65 pounds
each, is available. Any small improvement in the sound
quality of the speaker is outweighed by the speaker
stands' $850/pair price. In addition, the use of these
stands cancels out one of the chief advantages of a mini­monitor—the ability to move the speaker easily. Another
problem with these stands is that the height of the speak­ers above the floor is not adjustable. When a loudspeaker
uses low-order crossovers (second order in the case of the
Studio 10) and is auditioned at relatively close distances
(as would be the case in the small rooms this speaker is
designed to be used in), large dips can occur in the re­sponse if the height is not adjusted optimally.

Except for a broad peak between 3 and 7 kHz, the
anechoic frequency response of this speaker fits in a ±2
dB window from 300 Hz (the low-frequency limit of the
test) to 20 kHz. This measurement was made with the mi­crophone aimed between woofer and tweeter. The peak,
which is due to the resonant behavior of the large "phase

plug" of the woofer, has a maximum value of 5 dB at 5.5

kHz. The peak was almost identical for both samples of
the speaker. Tone burst testing showed energy storage
problems in the woofer's phase plug near the 5.5 kHz
peak, but the woofer's performance was exemplary below
the peak's frequency. The tweeter showed no energy stor­age problems at all below 20kHz, an excellent result. The
tweeter's resonance occurs at 26 kHz. This is almost an
octave above the resonance measured in the Focal tweeter
used in the ACI Sapphire IIti. Matching between the sam-

ples was superb, as it was held to within 1 dB.

Changes in amplitude response near the crossover

region are significant but much less severe than with the
ACI Sapphire IIti in vertical off-axis tests, and the chang­es are symmetrical. This is indicative of the drivers' being

inphase at the crossover point. An even order (2nd, 4th,

6th...) crossover will drive the speakers inphase at the
crossover, but Monitor Audio claims that the electrical
crossover is 6 dB per octave, which would result in the

drivers' being 90° out of phase at the crossover. The dis­crepancy, as explained by Monitor Audio design engi­neers, occurs because the mechanical rolloffs of the driv­ers are designed to exhibit a 6 dB per octave slope in the
crossover region. The sum of the mechanical and electrical
rolloffs yields a 2nd-order crossover response. Custom
designing the drivers' rolloffs is possible only because
Monitor Audio does its driver design in-house.

Our nearfield measurements of the woofer and the
port showed the bass response of the speaker to be flat to
60 Hz (the -3 dB point); then it falls at 24 dB per octave.
Given the small size of the woofer and the cabinet, dis­tortion in the bass region becomes high as drive level is
increased. Sonically this translates into the complete loss
of the bottom octave of the orchestra, and the octave

15

pdf 15

above that is reproduced with less definition than by larg­er speakers costing far less than the Studio 10. Sub woofer

users be warned: it is very difficult to mate a 4th-order

vented box to a subwoofer without incurring significant
amplitude variations near the crossover region.

With the speakers placed on the optimal vertical
axis they are slightly forward-sounding but much less so
than would be expected given the 5.5 kHz peak measured
in the speakers' amplitude response. The explanation may
be found in the room response measurements made by D.
B. Keele in his review of the speaker. The amplitude of
the peak appears to be reduced in the room response
curve. These speakers produced a remarkably transparent
sound. In comparison with the ACI Sapphire IIti, in­strumental timbres were reproduced with much greater

clarity and inner voices were more easily distinguished.
Dynamic contrasts were enhanced by the Studio 10. The
audible differences were significant enough to justify the
price difference between the two speakers.

At first I thought the transparency of the Studio 10s
would make less than optimally recorded CDs unlisten­able, but just the opposite occurred. With less distortion
coming from the loudspeaker, many of these recordings
proved to be more listenable than before. The ability of
this speaker to bring out the best in old CBS, Philips, and
RCA remasters was its most remarkable quality. Of
course, nothing could be done to make truly bad record­ings (such as most Deutsche Grammophon digital CDs)
listenable. On modern source material these speakers pro-

duced the best sound I have experienced in my listening
room. Thanks to the speakers' small size and careful driv­er matching, imaging was excellent. In your Editor's lis­tening room the speakers were bested by the MACH 1
and Win loudspeakers, but both of these speakers cost
more than twice as much as the Studio 10s. The principal
difference was the elimination of the forward-sounding
character of the Studio 10s (the sophisticated diaphragm
materials used by MACH 1 and Win do not show the res­onance effects of metal cones) and a more extended and
less distorted bottom end. To my ears the Studio 10s were
more open and transparent than the Waveform speakers,
but the Waveforms can play much louder and have sub­woofer-quality bass.

A favorite trick of high-end dealers is to claim that
a speaker of the quality of the Studio 10 can only be driv­en with electronics costing five figures. Do not believe
this. The Studio 10s are easy to drive. The impedance
never goes below 5.5 ohms and the phase angle varies
less than ±30 degrees. I temporarily replaced my current
electronics with my 15-year old Audire Legato/Crescendo
electronics and heard virtually no degradation in the Stu­dio 10s' performance. The combination of the Studio 10s
and a good-quality mass-market integrated amplifier and
CD player could cost less than $3200. The difference in
sound quality between this system and a similarly priced
high-end dealer system would be almost comical. If you
have a small room, listen at sound levels that will not

16

break a lease, and do not require extended bass response,

the Monitor Audio Studio 10s are highly recommended.

[In my large listening room, connected to my Boul-

der electronics, the Studio 10 didn't delight me nearly as

much as it did David, although I do not disagree with his

point-by-point analysis of the speaker's characteristics.

The 5.5 kHz peak and ringing were obviously more of-

fensive to my ears than his, but I happen to be very touchy

in the 3 to 6 kHz octave, more so than most people. In any
event, the woofer of the new Studio 6 has no comparable

phase plug, so that may turn out to be a dead issue.—Ed.]

Tannoy 615

Tannoy Ltd, c/o TGI North America, Inc., 300 Gage Avenue,

Unit 1, Kitchener, Ont., Canada N2M 2C8. Model 615 floor­standing 3-way loudspeaker system, $1599.00 the pair. Model
6s1 base, $99.00 the pair. Tested samples on loan from dis-

tributor.

Tannoy is one of the oldest English brands in loud­speakers, and Model 615 is the top of the Tannoy "Sixes"
line. All models in the line are six-sided columns of vari­ous heights—the cross section is a kind of chopped-off
hexagon—and the costlier models feature drivers in Tan­noy's classic Dual Concentric format. The 615 has an 8"
Dual Concentric main woofer/tweeter, an additional 8"
woofer, and an 8" passive radiator, each with a separate
snap-on, floating-type grille. In the main driver, a com-

pression-type high-frequency transducer is horn-loaded
by the specially contoured woofer magnet and woofer
cone. All three 8" cones are made of a very inert high­tech plastic material.

The remarkable thing about the 615 is that, while its
footprint is slightly smaller in area than this page and its
height is only a couple of inches over three feet, it sounds
like a big speaker! It has some colorations, to be sure, but
these are of the brassy, pro-sound variety, not the closed­down, nasal, unmusical kind exhibited by unsuccessful
audiophile-oriented designs. I like this speaker, despite its
small shortcomings, and the younger generation in my
house also preferred it in some ways to the more puristic,
neutral-sounding Thiel CS2.2, for example, because of its
dynamics and impact. It's definitely not a wimpy speaker.

One reason is that the 615 is very efficient; its 1­watt/1-meter SPL rating was a little hard for me to nail
down because of its wide swings in impedance, but the
figure is well above 90 dB and could be close to 92 dB.
About that impedance—once above the range of the tuned
box, the magnitude rises from 3 ohms to 28 ohms over
five octaves, then falls back to 5 ohms over two octaves.
The phase also swings widely within ±45°. Not exactly a
resistorlike load for the amplifier, but the good ones can
handle it with aplomb. The box with its passive radiator is
tuned to 29 Hz; maximum passive radiator output is at 25
Hz; the nearfield summed response is -3 dB at 35 Hz and
within ±2 dB up to 200 Hz. That's remarkable bass re-

THE AUDIO CRITIC

pdf 16

sponse for a small, efficient speaker system in which the
woofer is fairly well level-matched to the midrange. The
low bass doesn't hold up too well at very high levels, but
what did you expect—this is a very compact floor­standing speaker. Power handling above the bottom two
octaves is, on the other hand, truly excellent.

The frequency response from 300 Hz up is a little
humpy and notchy but still fits into a ±3 dB strip up to
about 16 kHz; I've seen better and I've seen worse. At
least nothing sticks out terribly. Off axis the picture is
very nice; up to almost 10 kHz you have to go more than
45° off axis to see a serious rolloff, and the pattern is of
course symmetrical up and down and sideways because
of the coaxial wave launch. It should also be noted that
the first-order electrical crossover at approximately 2 kHz
between the two coaxial elements doesn't come with the
usual penalties (lobes, tweeter distortion, etc.), thanks to
the coaxial geometry and the power handling of the com­pression driver. The crossover network also rolls off the
separate woofer above 400 Hz; the two woofers are in
parallel, but only that of the coaxial unit is active up to
the tweeter crossover. (The passive radiator isn't con­nected to anything; it acts as a vent.)

My time domain measurements indicated that the
two woofer cones move outward in response to a pos­itive-going pulse, but the tweeter diaphragm moves in­ward. The tweeter is wired with reverse polarity, perhaps
to compensate for the total second-order profile resulting
from the acoustical rolloff added to the first-order electri­cal slope. In any event, the resulting pulse coherence is
quite good, though far from perfect; tone bursts reveal no
evidence of storage but plenty of interference between the
coaxial drivers, which of course are not coplanar. That
may be one reason for the slight colorations heard, in ad­dition to the not quite smooth frequency response. Im­aging, however, is very good, probably because of the
symmetry of the radiation pattern.

One practical feature of the 615 is its fake-marble
top, perfect for your drink while you change CDs; an im-

practical feature is—here we go again, God save the
Queen and pass the crumpets—the Britannic spacing of
the binding posts at the input, making it impossible to use
double banana plugs. I would gladly trade the 615's two
pairs of inputs—a genuflection to the tweako doctrine of
biwiring—for one pair with ¾" spacing (believe me,
blokes). Even so, this is a special speaker with a special
flavor, and after getting my reservations off my chest I'm
quite willing to endorse it. After all, how many waist-high
speakers with a small footprint have large-signal capabil­ity, decent bass, and outstanding dispersion?

Thiel CS2.2

Thiel, 1026 Nandino Boulevard, Lexington, KY 40511. CS2.2

floor-standing 3-way loudspeaker system, $2250.00 the pair.

Tested samples on loan from manufacturer.

ISSUE NO. 19 • SPRING 1993

According to a PR release received in mid-December,
the Thiel CS2.2 will henceforth be known as the CS2 2,
prounounced "two-two." Huh? What's that? Yeah, a space
in place of the decimal point. You see, Bose Corporation
owns the trademark to the number 2.2; they currently sell
a speaker called a 2.2 II. The press release didn't say that
Bose demanded the change, but to me it sounds typical of
their corporate mentality. Well, I'm not selling anything
called 2.2; I'm an editor in a country that guarantees free­dom of the press and I can call the Thiel speaker Irving,
Attila, or even Amar if I want to; so for the purposes of
this review I'll just stick to the old CS2.2 designation.
(Whew, that was a close one, Amar, wasn't it?)

I also want to say, before anything else, that I have

the greatest respect for the Thiel company as an audio

manufacturer, for Jim Thiel as a speaker designer, and for

Thiel speakers as products of absolute integrity—but also
that I disagree with the Thiel philosophy of loudspeaker
design on a very fundamental level. I admire the product,
but it isn't what I want for myself as a music lover and
audiophile. I fully understand, however, those who are of
the opposite opinion.

Now then, the Thiel CS2.2 is a 3½-foot high
speaker with a footprint of just over a square foot and a
slanting front. Every detail of its design reflects a single­minded pursuit of quality, as if the designer had never
heard of money-saving solutions, not even in this low­end-of-the-high-end price range. The front baffle is 2"
thick, the cabinet walls 1" thick, the bracing truly massive.
The back is finished just as beautifully as the sides, and
the specially contoured baffle fits deep into hollow of the
specially beveled grille to form the neatest, most inge­nious nondiffractive front of any speaker I've seen—I
would have died for such a design in my speaker man­ufacturer days. Every inch of the speaker shows serious
thought and loving care. The driver complement consists
of an 8" woofer with a novel two-layer diaphragm, a 3"
midrange driver, and a 1" metal-dome tweeter. The woofer
is in a tuned enclosure using a 6" by 9" elliptical passive
radiator. The crossover network is the heart of the design,
consisting of 26 circuit elements and built with a total of
35 components, including the highest-quality air-core
coils, polypropylene and polystyrene capacitors—nothing
but the best. Its purpose is to synthesize perfect first­order acoustical crossovers and achieve perfect phase and
time coherence. That is Jim Thiel's obsession, for which
he is willing to trade off many other important perfor­mance characteristics, and that is where I part company
with his design philosophy.

On the axis of the midrange driver, the CS2.2 has
the most perfect frequency response of any speaker I have
tested so far. The 1-meter MLS measurements show the
amplitude response to be within ±1 dB from 1.5 kHz to

19 kHz and ±2.5 dB from 300 Hz to 20 kHz; the phase re­sponse stays within +15° from 300 Hz to 20 kHz. I sus­pect that even better results are obtainable at 2 or 3 meters
because the center-to-center distance between the mid-

17

pdf 17

range unit (which is crossed over at 800 Hz and 3 kHz)
and the woofer (which takes over below 800 Hz) is about

meter, i.e., a significant part of 1 meter. The off-axis
response at a 30° horizontal angle from the midrange axis
is also spectacularly good, almost as flat as the on-axis re­sponse, just 4 or 5 dB lower in level. But—a very big
but!—move the microphone up or down vertically, away
from the "sweet spot," and all bets are off. The response
goes to hell. Welcome to the world of first-order cross­overs. (Siegfried Linkwitz didn't spend all that time and
effort perfecting fourth-order crossovers for nothing.) In
time domain tests, square pulses of various durations

confirm the claims of outstanding coherence at the sweet

spot—and only there.

And that's not all. At fairly moderate levels, sine

waves in the 300 to 400 Hz range, a full three octaves be-

low the tweeter's nominal 3 kHz crossover, make the
tweeter buzz! Why? Because the first-order highpass filter
is only 18 dB down in that range, and that's apparently
not enough to protect the tweeter from out-of-band over­load. Musical program material containing a fair amount
of 300 to 400 Hz energy will do the same thing; luckily it
doesn't happen very often. (No, neither tweeter was de­fective; they sounded very clean in their working range.
It's possible that Vifa, their manufacturer, doesn't con­sider buzzing to be an issue because the higher-order
crossovers used in most speaker sytems provide adequate
protection from low-frequency excitation.)

Regardless of the tweeter buzz, the overall power
handling of the CS2.2 is unimpressive. The bottom octave
isn't reproduced cleanly at high signal levels, although
the small-signal bass response is very good—I measured

-3 dB at 32 Hz and essentially flat response above that.
When the music gets really loud over a wide frequency
range, as in a Mahlerian climax, the little 3" midrange

driver begins to sound mildly distressed, its nominal two­octave passband being in reality more like six octaves

with those first-order rolloffs at either end. It's a case of
cruelty to small creatures.

Part of the power-handling problem is the low
efficiency, necessitating more than the usual amount of
drive for realistic sound levels. The standard input of 2.83
V produces an SPL of 86 dB at 1 meter, but when you
look a little closer that's actually 2 watts because the

nominal impedance of the speaker is 4 ohms. So the 1­watt/1-meter efficiency rating is in reality more like 83
dB. On the other hand, the load presented by the CS2.2 is
a piece of cake for almost any amplifier to drive. The

magnitude of the impedance is almost flat, hugging the 4­ohm line and staying between 3.3 and 5 ohms at all fre­quencies above 100 Hz. The impedance peaks of the
tuned box are only 7 ohms. As for the phase of the im­pedance, it stays within ±10° at all but the lowest fre­quencies. Jim Thiel loves flat response curves.

Having said all that, I must not fail to emphasize
that the sound of the CS2.2 at moderate levels—and with
the listener's ears within a certain window, not too high

18

and not too low—is outstandingly good. Nothing sticks
out, everything is wonderfully smooth and neutral, and
the imaging is nothing short of superb, as good as I have
heard in my setup. I know that Jim Thiel attributes that
subjective perception to the Miracle of Saint Coherence,
but I don't agree. I suggest the following experiment,
which is beyond the purview of an audio journal but not
of a speaker designer:

Build three versions of the CS2.2, or of any other

comparable Thiel model for that matter. No. 1 would be

the standard factory version. No. 2 would be the standard

version with a well-designed fourth-order Linkwitz-Riley
crossover substituted for the first-order Thiel crossover.
No. 3 would be the standard version with the first-order
crossover but in a square-cornered cabinet with sharp
edges, no bevels, no contoured front baffle—just a regular
"monkey coffin." I'm willing to bet the ranch that No. 2
would image just as well as No. 1 (and, incidentally, have
better power handling and vertical polar response),
whereas No. 3 would image much less well and generally
sound a little rougher and less focused. In other words, I
contend that Thiel's deservedly prized imaging is due to
the superior management of diffraction, reflective sur­faces, second arrivals, etc., not to the first-order crossover.
I also know that Jim is most unlikely to change his mind
on this subject.

In conclusion, I want to mention that long before I
got around to measuring and listening to the CS2.2 in my
laboratory, I had it hooked up to a more modest but fairly
high-powered system in another room for my two sons
(both in their twenties) to use as their regular stereo. Their
verdict: "It sounds very accurate and uncolored but it

doesn't kick ass."

Waveform Mach 7

Ötvös Industries, RR #4, Brighton, Ont., Canada K0K 1H0.

Waveform Mach 7 floor-standing 4-way loudspeaker system,

$8400.00 the pair, complete with dedicated electronic crossover.

Tested samples on loan from manufacturer.

This is not a new speaker system but it has changed

sufficiently since my exhaustive and highly favorable re­view of the original version in Issue No. 14 to warrant a
second look. The new Mach 7 designation after the brand
name (couldn't John Ötvös have found a more original
tag?) indicates the following new features:

The dedicated electronic crossover is now made by

Bryston and uses the same basic topology as the superb
Bryston 10B general-purpose crossover. The "boost of a
little over 3 dB at 16 to 17 kHz to equalize the super­tweeter where it begins to roll off (my words in No. 14)
has been eliminated. The 1" textile-dome lower tweeter
has been replaced by an MB QUART 1" titanium-dome
unit. That, of course, necessitated a new passive crossover
network for the upper section of the speaker. There is also
a new grille incorporating a special felt ring that fits over

THE AUDIO CRITIC

pdf 18

the dome tweeter. In addition to the incredible virtuoso
woodcrafter's custom cabinet (what I called the Bernini
version) and the black piano-finished cabinet, there is
now a crackle-finished professional-style cabinet—the
one I received. The update to Mach 7 is available to own­ers of the older model for the amazingly small sum of
$600.00 (Canadian). John Ötvös is obviously not of the
William Z. Johnson school (which holds that mods should
not only be offered frequently but also end up being just
as profitable as the original sale).

Not that there are many candidates for the update.
The speaker, despite its outstanding qualities, just didn't
sell—Stereophile made sure it wouldn't. (Yes, one more
example of a superior audio product railroaded by the
Atkinson, Archibald, and Santa Fe. See the review by
Larry Archibald and John Atkinson in their November

1989 issue and my comments on the politics of the situa­tion in my original review.) What is truly remarkable is
that the endorsement of the Waveform by the professional
audio community didn't help. Jack Renner and Michael
Bishop at Telarc are using it to monitor their latest and
greatest recordings (see the equipment credits in the cover
brochures of Telarc CDs). Craig Dory at Dorian has also
started to use it. Jack Renner has gone so far as to com­mission for his own and Mrs. Renner's personal use, in
their house, a five-channel surround-sound system using
Waveform speakers—in the "Bernini" version! Nev­ertheless, only 30 pair are currently in the field—such is
the influence of the Santa Fe railroaders in the high-end
audio market. It didn't help, either, that Harry Pearson
had also taken a cheap shot at the speaker in a typically
smug, half-assed editorial footnote—not a review—in
The Absolute Sound. (The review never materialized.) I
ask you, friends, who knows more about good sound, and
who stands to lose more by being wrong about it—Jack
Renner, Michael Bishop, Craig Dory, and their pro­fessional crews, or Larry Archibald, John Atkinson, and
Harry Pearson? Tough choice, isn't it?

The improvements in the Mach 7 version are quite
significant. The frequency response is flatter and smoother.
The highs are better balanced with the rest of the spec­trum. The difference between on-axis and off-axis re­sponse is smaller. Also, the Bryston-made electronic
crossover is a more advanced and reliable design, less
likely to cause ground loops and more likely to have an
unlimited life span. Since the review in Issue No. 14 cov­ered every detail of the design, which is still exactly the
same except for the changes mentioned, I have no reason
to go over the same ground again. The measurements
above the bass range, however, have improved. The on­axis response is now within ±2 dB all the way up to 18
kHz, and the off-axis response up to 45° hews amazingly
close to that line, deviating by only 1 to 4 dB under 10
kHz and only 4 to 7 dB in the octave above that, up to 20
kHz. (Yes, I said 45°, not 30°.) That begins to approach
the constant-directivity model, and the total radiated pow­er measurements made in the acoustical laboratories of

ISSUE NO. 19 • SPRING 1993

the National Research Council in Ottawa (better than my
lab!) paint a corroborative picture. Distortion above the
bass range is minuscule, typically -48 dB relative to an
SPL of 95 dB (which is unbearably loud as an average),
but the bass is slightly more distorted than it would be
with a motional-feedback woofer like the Velodyne. That
could be the next design update, but the bass as it stands
is still pretty awesome, as I originally wrote.

What is the overall impact of the Waveform Mach 7
in terms of subjective listening? I would say that I haven't
heard another forward-firing monolithic speaker that
equals it as a total package. Its distortionless handling of
the most dynamic and complex program material, its deep
and unshakably tight bass, its lack of resonant coloration
in any frequency band, its extended but neutral top end,
its natural tonal balance, its high efficiency, its absolutely
clean delineation of whatever music is fed into it—all
these virtues together give it the highest composite score
in my book. On various individual counts—3-D imaging,
see-through inner detail, finesse of texture in the treble,

forgivingness in placement—I can name speakers that

will beat it, some in this very article. Also, if you prefer

planar and/or line-source speakers, the Waveform is un-

likely to convert you because it just doesn't launch sound
waves the same way. As a sonic decathlon champion,

however, I think it's very hard to beat. Paul Barton, the
Canadian engineer who designed it for John Ötvös, cer­tainly did his homework.

One small warning. Because of its flat response and
wide dispersion in the treble range, the Waveform can
sound too bright if highly reflective walls and other bare
surfaces are near it. In a very large room, where the walls
are far away on all sides, it's not an issue (the delayed
second arrivals actually sound nice); in a smaller room
wall treatment and careful aiming may be necessary.

I should also add that if your curiosity is aroused
now and you would like to check out the Waveform in a

dealer's showroom, there are no dealers, alas. To the best
of my knowledge, all sales so far have been manufacturer

to end user. The bright side of that situation is that the

official "retail price" of $8400 the pair is at this point
rather theoretical and probably negotiable to some degree

in a direct sale. That's just a hint, not a promise.

Westlake Audio BBSM-8VF

Westlake Audio, Manufacturing Group, 2696 Lavery Court, Unit
18, Newbury Park, CA 91320. Model BBSM-8VF floor-standing
3-way "reference monitor" loudspeaker system, $4050 the pair,
including pedestals. Tested samples on loan from manufacturer.

Here we have what in my opinion is a highly spe­cialized speaker. For certain users with highly specific
requirements it provides the best possible answer—or at
least one very good answer. For the general audiophile
with four big ones to spend on a pair of speakers there are

19

pdf 19

many other options to explore before taking this route.

What are those specific requirements that make the

BBSM-8VF (whew, what a mouthful of alphabet soup!)

an attractive choice? Tremendous efficiency, bulletproof
power handling, wide dynamic range, low distortion,
powerful and well-controlled though not extremely deep
bass—and all that, most importantly, in a fairly compact
package. If that's what you want, rather than ultraprecise
imaging, structural detail, finesse in the treble, gorgeous
instrumental textures—the more delicate audiophile prior­ities—then you'll be a "happy camper" listening to a pair
of these Westlake speakers, which have a distinct studio
pedigree. Westlake Audio is basically a professional
brand; their typical customer is a professional who earns
his/her living with audio; their interest in the audiophile
market is relatively recent (though currently quite keen).
Their audiophile speakers reflect this heritage; for one
thing, they're built like the proverbial brick rest room­solid, thick-walled, functional in appearance (oiled wal­nut, brown grilles, no frills).

To hear the BBSM-8VF at its absolute best, play

through it one of the more dynamic recordings by Tom

Jung on his dmp label. These close-miked, not very rever-

berant, typical studio recordings of various contemporary
bands are generally dominated by hard left, hard center,

and hard right information, and have stupendous dynamic
peaks. The Westlake speaker eats up that kind of program

material and projects it into the room with surgical clean­liness and precision. Complex symphonic material re-

corded in a large space does less well on the BBSM-8VF;
the subtle spatial and timbral cues tend to get lost in a
kind of homogenized soundstage, although the climaxes
are life-size and undistorted.

The dimensions of this powerhouse are, as I said,
quite modest: 31" high, 19¼" wide, 12" deep—something
like a large bookshelf speaker doubled. For proper listen­ing height the speaker should be mounted on the matching
pedestal supplied, which raises it about a foot off the
floor. (My pedestals came with the wrong mounting hard­ware, by the way. No big deal, just annoying.) The driver
complement consists of two 8" woofers mounted side by
side in a vented enclosure, a 3½" midrange driver in a
separate sealed subenclosure, and a 1" dome tweeter. The
crossover network is a somewhat peculiar affair that I had
trouble figuring out strictly from my measurements, with­out a circuit diagram (or taking the speaker apart); it is
specified to have crossover frequencies of 600 Hz and 5
kHz, and slopes of 24 dB per octave "minimum," but I'm
not so sure after my microphone probings. There seem to
be all kinds of slopes. All I know is that the network
creates unusually large time displacements; in my time
domain measurements a square pulse of 0.5 ms duration
was stretched out to 2 ms and one of 1 ms duration to 4
ms. These measurements also indicated that all four driv­ers move outward in response to a positive-going pulse,
but that by itself isn't sufficient for coherence, which in

this case is minimal (as in all steeply crossed-over 3-way
systems).

The impedance curve of the BBSM-8VF clearly
puts it into the specialized category; the magnitude is 4
ohms at 1 kHz but dips well below 2 ohms at 150 Hz and

7.5 kHz, and doesn't exceed 8 ohms even at the vented­box peaks. The phase, above the box range, fluctuates
within ±45°. Obviously, the speaker sucks current like an
electronic vampire and needs an amplifier with high cur­rent capability to feed that habit. Indeed, the whole concept
of the design is to draw the maximum power possible out
of a good amplifier that doesn't necessarily put out a lot
of volts. Westlake even supplies optional speaker cables
with plus and minus leads as thick as cocktail wieners and

of virtually zero resistance (though highish inductance)

just to draw even a tiny bit more current. The speaker has

no standard red and black binding posts with banana jacks

at its input but comes with a block of screw terminals in­stead, in order to accommodate the spade-lug termina­tions of these tweako cables. Terminals are provided for
biwiring as well as passive biamplification. (None of that
tweaking can hurt, of course, but I have little patience
with it, as all but our newest readers know, and I refuse to
spend time and energy on it until I see scientific proof of
its worth.) The right amplifier may be easier to find than
you think, however, because the speaker is very efficient,
of the order of 90 dB (1-watt/1-meter SPL reading).

The frequency response measurements gave out­standing results. At the best summing junction of the
woofers and the vent, the nearfield small-signal bass re­sponse was flat down to a -3 dB corner at 25 Hz, closely
tracking the box tuning frequency and the maximum­output frequency of the vent. That's quite a bit better than
what the spec sheet says. The overall frequency response
above the bass range was within ±2 dB up to 12 kHz,
with the microphone aimed at the midpoint between mid-

range and tweeter. A tiny elevation at 15.5 kHz (an extra

1.5 dB or so) prevented that reading from being applica-

ble all the way up to 20 kHz. At 30° off axis horizontally,
there was hardly any change up to 10 kHz except a little
notch at 3 kHz; between 10 and 20 kHz there was about 5
dB attenuation. These excellent figures explain the basi­cally neutral tonal balance of the speaker.

I think the imaging of the BBSM-8VF would be im­proved if there were less bare "real estate" on the front
baffle and if the edges were rounded or beveled. Westlake
will soon be coming out with a new line of speakers
called the BBSM VNF series, which will be taller and
much narrower, without side-by-side woofers, resulting in
a wave launch that should be more to the liking of audio­philes. There will be an almost exact equivalent of this
particular speaker in the series. Meanwhile the important
thing to remember is that Westlake Audio has its own
highly individual and identifiable approach to loudspeak­er design, that it's a valid approach, and that you either
need exactly that kind of speaker or you don't. I don't.

20

THE AUDIO CRITIC

pdf 20

Win SM-10

(follow-up)

Win Research Group, Inc., 7320 Hollister Avenue, Goleta, CA

93117. SM-10 Broadcast Monitor (2-way coaxial loudspeaker
system), $6250 the pair, including stands. Tested samples on
loan from manufacturer.

I retested this unique speaker from a new production
run that corrected a minor component-mounting error on
the circuit board of the original crossover module. This
error was claimed to have been the cause of the minor
glitch in the crossover region that I had reported. I also
wanted to repeat some of my measurements using our
newly phased-in MLS technique.

The reader is referred to the original review in Issue

No. 17 for a detailed discussion of the design. Here I must,
first of all, remedy the failure of that review to specify the
impedance of the speaker. The magnitude fluctuates be­tween 3 ohms and 9 ohms; the average is about 6 ohms.
The phase fluctuates between +30° and -40°. All in all, a
load of fairly limited severity—but not quite a piece of
cake—for the amplifier.

The frequency response as originally reported needs
no revision in the bass range but has to be qualified fur­ther up in the spectrum. All is well on axis up to about 2.5
kHz; I stand by the original +2 dB. Between 4 and 17 kHz
the ±2 dB tolerance is again still valid. In that narrow 2.5
to 4 kHz band, however, there is a definite discontinuity,
centering on a peak at approximately 3.3 kHz. If that
wrinkle is included, the overall frequency response on
axis is no better than +3.5 dB from 45 Hz to 20 kHz—a
very respectable specification but not amazing. Off axis
(measured at 30°) the peak moves down to 2.4 kHz—
strange but true!—and the average level of the tweeter
drops by 3 to 4 dB, with steeper rolloff above 15 kHz. I
am beginning to think that a properly massaged fourth­order Linkwitz-Riley crossover would be better for the
Win SM-10 than the variations-on-a-theme-by-Spica net­work currently used. I also have a feeling that Sao Win is
beginning to think so, too. Then again, the whole question
may be rendered moot by the dedicated active crossover
and amplifier system for the SM-10 he is planning to
come out with. All you'll need then will be money.

For the moment, here is the bottom line. Even with
its less than perfect passive crossover, the Win SM-10 is
the best-sounding loudspeaker known to me in terms of
transparency, definition of inner detail, spatial cues, tonal
balance, lack of coloration, nonfatiguing top end—in oth­er words, all the finesse criteria. If it could play at very
high SPLs and had flat bass down to 20 Hz, it would be
the ultimate speaker system, bar none. I think the two
main reasons for its superiority are the inherently dead
materials used in the construction of the diaphragms and
the perfect symmetry of the wave launch. Those appear to
be even more important determinants of sound quality
than a dB here or a dB there in the amplitude response.

ISSUE NO. 19 • SPRING 1993

(Can't give away too many dB, though.) Anyway, that's

just a hypothesis, not a proven fact.

What is a fact is that the overwhelming majority of
audiophiles who have heard the Win SM-10 would like to
own a pair.

Subwoofer

Hsu Research HRSW10

Hsu Research, 20013 Rainbow Way, Cerritos, CA 90701.
HRSW10 vented-box 10" subwoofer, $750.00 the pair (with wal­nut top and standard 12 dB/octave passive crossover). Tested
samples on loan from manufacturer.

This very impressive product started its life as the
Definitive Research SW10; subsequently discovered and
debated name conflicts were resolved by changing both
the company name and the model prefix. Dr. Poh Ser Hsu,
the scholarly Singaporean technologist responsible for the
design, should have had his name on his creations from
the beginning; the days when you needed a Waspy name
like Lansing or a techie acronym like Altec for your
speaker brand are gone forever. We live in the age of

Hyundai and Häagen-Dazs. So, in this case, Hsu is the

name and deep bass is the game. Very deep bass.

It so happens that I know Poh Ser from his Boston
Audio Society days (his tracks stretch from Singapore to
the Massachusetts Institute of Technology to the Boston
audio mafia to Southern California—a recent move). One
of my more frequently quoted bons mots is about the Bos­ton Audio Society—in 1977 I wrote that their members
"would like to discover an audio Nirvana for forty-nine
dollars and ninety-five cents." Poh Ser is very much part
of that tradition of penny-pinching audio romanticism,
and in his subwoofer design he has brilliantly vindicated
it. The HRSW10 is low-frequency Nirvana for $375 per
side—one of the two best subwoofers known to me, the
other being the Velodyne, comparable models of which
range from $1095 to $2750.

How did he do it? The enclosure is a cheap but ex­tremely strong and acoustically inert paper tube with an
inside diameter of 14", a wall thickness of ¼", and end
pieces made of ¾" fiberboard. This structure is 27¼ " tall
and stands on four ordinary ¼" thick hardware-store
bolts that raise it 2½" off the floor, for a total height of
not quite 30". An inexpensive "sock" made of black knit
fabric covers the entire tube except for the top end piece,
which has a walnut finish. The downward-facing bottom
end piece holds the 10" driver and the 3½" port, which is
ducted by a 23" long paper tube. Basically it's a Thiele­Small vented box.

That 10" driver is quite special, however. Since it
doesn't have to reproduce any frequencies over 100 Hz or
so, it can be totally optimized for low bass. It has a big,
heavy 2" voice coil wound in four layers, with sufficient
overhang to permit unusually large linear excursion. The
cone is made of heavy paper, which is both cheaper and

21

pdf 21

better for the purpose than fancy polypropylene and such.
A driver like that, with a properly designed magnet struc­ture, in a ducted enclosure having an internal volume of

about 2 cubic feet, can do some fancy woofing. What's
more, the relatively thin-walled but rigid tube can't bulge
or flex under pressure, even though it weighs practically
nothing and isn't braced, because its cross section is a cir­cle and a circle already has the largest possible area for a
given circumference, so there's no place for the bulge to
go. There are more ways to build a solid woofer cabinet
than throwing money at it.

I did my usual nearfield measurements with the
driver facing up, a procedure that yields a close approx­imation to the free-field anechoic response. In the normal
position of the subwoofer, the proximity of the floor plus
the room gain will of course influence the response, but I
wanted to explore the raw input/output capability of the
system. At the best summing junction of the driver and
the vent, the small-signal response was ±0.5 dB from 20
to 100 Hz! At 10 Hz I measured -6 dB, but that included
some contribution to the rolloff by the lab-bench amplifier,
which isn't flat to DC. As I kept cranking up the input,
the output above 26 Hz remained just as flat; just before I
ran out of amplifier power the 20 Hz response was down

-2.5 dB, but the 10 to 15 Hz response actually went up.
Never, never have I seen anything like it. The distortion
may have been slightly higher than with the Velodyne
ULD-15 Series II at equal SPLs (I didn't retest the latter
side by side), but the Hsu HRSW10 appears to be capable
of somewhat higher absolute levels—and who cares about
a tiny difference in harmless second-harmonic distortion
at 20 Hz? Boston Cheapie is at the very least the silver
medalist! Unbelievable.

The vented box is tuned to approximately 15 Hz,
where there's a deep null in the output from the driver,
and the maximum output of the vent stretches from about
17 to 27 Hz. That's not exactly classic fourth-order But­terworth tuning—I don't know what it is precisely—but
the results speak for themselves. The impedance curve in

the 10 to 100 Hz range is of course a roller coaster, both
in magnitude and in phase (the biggest swings are roughly

110 ohms and ±60°), but the resistive component is 8

ohms.

The subwoofer needs to be biamplified and should
be used in pairs (no singles sold and no L + R matrixing
recommended). I used a Bryston 4B power amplifier to
drive a pair of HRSW10's and a Bryston l0B-sub elec­tronic crossover to match them to various main speakers
at different frequencies and with different slopes—but
that's traveling first-class and not absolutely necessary
(nor in the frugal spirit of the product, for that matter).
Just about any old amplifier that can drive 8 ohms will do,
and Hsu Research includes with each pair of subwoofers
a simple passive network (a second-order RC lowpass
filter or third-order if you wish) that goes between the
output of the main amp and the input of the sub amp. It's
a bit on the Mickey Mouse side in my opinion (I'm not of
the Boston school) but it works, and there's really nothing
wrong with the concept as long as you're content to drive

your main speakers without highpass filtering. For finicky
audiophiles there's also a special Hsu Research electronic
crossover at a Boston price: $350.00. I haven't tested it.
There's some flexibility as to crossover frequency in both
the passive and active Hsu crossovers, but you have to
specify your needs before you buy.

Readers who have been waiting all this time for a
quasi-pornographic subjective description of the bottom
end obtainable with the HRSW10—how big, how firm,
how rumbling, etc.—will have to be disappointed, as is
usually the case in this publication. Flat, correctly damped,
undistorted bass down to well below 20 Hz is just that; it
can only sound one way. Once you've heard it you know
it—and you never again want to be without it. It brings
you the real world of music, not a preshrunk facsimile.
And now, for the first time, it costs relatively little. If you
have a listening room of reasonable size, nothing can im­prove your stereo system as dramatically for $750 as the
Hsu Research HRSW10. •

"Qui ne gueule pas la vérité, quand il sait la vérité, se fait le complice des menteurs et des faussaires."

He who does not bellow the truth when he knows the truth makes
himself the accomplice of liars and forgers.

—CHARLES PÉGUY (1873-1914)

22

THE AUDIO CRITIC

pdf 22

ISSUE NO. 19 • SPRING 1993

23

In Your Ear

pdf 23

Nostalgia and

Loudspeakers

By Drew Daniels

My response to a hundred requests for the lost sound of yesteryear.

Editor's Note: Drew Daniels was until recently Principal
Electroacoustic Engineer at Walt Disney Imagineering.
Before then he was the Applications Engineer for Pro
Sound products at JBL Professional. He is one of the
leaders of the Los Angeles section of the Audio Engi-

neering Society; he is the author of various AES technical
papers; in other words, he is a genuine pro, not a retired
dentist who likes to build speakers in his hobby shop.

* * *

Memory.

Nostalgia isn't what it used to be. Our recollection
is proactive. My 1970 Webster's doesn't include that
word, but the newer one says it's a psychological term
coined in 1933. It's an adjective meaning "relating to,
caused by, or being interference between previous learn­ing and the recall or performance of later learning ([as in]
proactive inhibition of memory)."

Memory is not a constant but, rather, a fuzzy collec­tion of impressions swimming in the chemical reactions
of the brain. There are strong, precise recollection pro­cesses like recalling facts while watching Jeopardy, and
then there are indistinct tenuous "recollections" of things
like odors, touch sensations, colors and sounds. But, to be
specific, even the fuzzier recollections of sound can be

divided into strong and weak recollections. For example,
the sounds of familiar voices on the phone form strong
enough recollections to allow us to recognize people
quickly, while the timbre of individual voices is virtually
impossible to recall except by a small number of people

who are "tuned-in" to timbre, such as impressionists, who

use it in their work .

A test for the presence of B.S.

If you know any pompous "experts" who claim to
possess a calibrated ear, there is an easy way to humble
these golden ears. Send five bucks to Riverbank Labs in
Geneva, Illinois, and get yourself one of their fine milled
aircraft-aluminum tuning forks. I have found that the A-

880 works well for this test. Strike the fork and place it an
inch from one ear, then quickly move it to the other ear
and back and forth. The object is this: People are living
organisms (most people), not test equipment. Most of the
time a person's two ears don't hear the same way. At
some times, the left ear will hear a pitch as flat or sharp

24

compared to the right ear at that same time. At other
times, this pitch perception is reversed, or the two ears
hear the same pitch, depending on brain activity and the
blood pressure in the individual ears, which can vary de­pending on mood, head orientation, or even facial expres­sion. If you fail to obtain a different pitch perception with
this little test, wait a minute or two and then repeat the
test. Most likely, the result will be different perceived

pitches.

Anyone who claims to be "calibrated" or have ma­chine-like abilities of hearing is either lying or mentally
incompetent. We are not machines, nor can we be.

The request.

During the five years I was JBL's Applications En­gineer, many audio "old-timers" called looking for in­formation about old speaker systems. After just a few
months on the job, I began to realize that their interest in
these old systems was often a considered and educated

personal listening choice, established with good knowl­edge of more modern "advances" in loudspeaker tech-

nology. So strong and adamant was their insistence that

the old systems were simply better and more musical than
any of today's systems—including the modern $40,000+
supersystems—that I began to experiment with some of
the components, and with some of the older systems
when I came across them.

As a result of five years of such experiments, com-

puter modeling, culling and matching of components and,
of course, listening, I'm convinced that the old systems
and their technology contained the germ of the correct ap­proach to music reproduction using loudspeakers. Modern
devices are available to add the pieces needed to complete
what they started—provided you are not trendy, have
plenty of cash lying around, and have an understanding
family and neighbors.

Some explanatory background.

Audio power amplifiers these days can be very

economical indeed. If the purchaser isn't too fussy about
designer labels, a power amp can be had for as low as a
dollar per watt—even less in the larger economy sizes—
and this represents, at least in my mind, one of the best
values one can obtain in the audio marketplace. As a mat-

THE AUDIO CRITIC

pdf 24

Basic configuration and dimensions

of the main speaker (without subwoofer)

discussed in the article.

25

ISSUE NO. 19 • SPRING 1993

pdf 25

ter of engineering policy, I generally recommend that the
largest amp available, affordable or practical, be used in a
given system. This is because the largest amplifier design
within a manufacturer's line is usually the flagship model
and, more importantly, usually has more output devices
(transistors) and greater capability to deliver large
amounts of current into low-impedance loads.

One of the amplifier designer's most difficult chal­lenges is second-guessing what kind of load his product
might face once it ends up in the hands of the end user.
This is a fact that would seem to argue for speakers with

purpose-built amplifiers attached, but since most audio

consumers tend to think they can get better sound by
selecting their own components, the idea has met with
marketing failure.

In general, an amplifier selected for low output re­sistance, high damping, low distortion, and so on, will be
fine for most speaker loads it may encounter. This am­plifier type will produce the least variation in system re­sponse due to the slings and arrows that amplifiers are
subject to.

When I was a kid and my ears were at their best and
my musical sophistication near nil (ca. 1955-57), I was
easily impressed by sounds that were simply unusual, that
is, the sound of a system with "crisp" highs, since this
was something one didn't find very often in the main­stream of audio systems. Low bass from loudspeakers
was utterly unknown—subwoofers were not yet even a
concept. About the best one could ever expect to hear

were systems way beyond the economic reach of con-

sumers, namely theater systems with enormous en­closures, large recording studio monitors, and the like.
There were a few "home" models made from the avail­able pro-style components in furniture-quality cabinetry
that made them semitolerable to understanding, or def­erential, '50s wives.

Of course, in the late '50s amplifiers available to
the nonprofessional public were small in size, mostly for
price reasons, but even "professional" amplifiers designed
for industrial installations only provided between 5 and
30 watts, with the exception of the new wave of "super
amplifiers" boasting a gigantic 60 watts. Speaker man­ufacturers were making voice coil bobbins out of ordinary
kraft paper, and the adhesives used to hold speaker mov­ing assemblies together were unsophisticated, not yet hav­ing benefited from the great advances science and engi­neering were to enjoy in the coming "Space Age."

This period, in loudspeaker history, is very im­portant for a number of reasons. First, decisions were
made that would guide the form designs would take and
affect all subsequent progress right up to the present.
These decisions took speaker design in the obvious direc­tion of taking advantage of the falling prices of amplifier
power, thanks to the introduction and implementation of
the transistor. With plentiful available power, designers
could virtually ignore speaker efficiency. They could add
mass to speaker moving assemblies to drive down res-

ISSUE NO. 19 • SPRING 1993

onant frequencies and reduce driver size, allowing them
to stuff pretty good bandwidth into small boxes. A case in
point might be the early AR series bookshelf systems
from Acoustic Research, which offered efficiency well
below 1%—as opposed to the larger counterparts that
could offer up to 20% efficiency figures. Further, as adhe-

sives improved, loudspeaker drivers became more rugged,
inviting amplifier manufacturers to keep increasing pow­er, and this technological spiral continues still.

Second, the "High Fidelity" movement/industry
really began in earnest, as compared to the esoteric hob­byist status it had occupied through its early history.

Third, the technologies I referred to earlier made
possible things that would not have been considered be­fore, opening up the loudspeaker design field to methods
that would eventually provide inexpensive products of
astounding performance-per-dollar value.

Today, we are basically stuck with such high-value
speakers. I say we are "stuck" with this type of speaker,
but let's understand why this is so. It has become easy
and inexpensive (read that as high profit margin) to make
one-inch dome tweeters that perform very nicely, thank
you. It has become easy to mill speaker enclosures auto­matically from pallet-loads of lumber—allowing the box­es to be shaped to take advantage of the shape of the lum­ber supply to get maximum yield and minimum waste. It
has become easy to stuff into these mass-produced boxes
loudspeaker components specifically designed to operate
near optimally in the box volumes provided, thanks to
computer optimization for the needs of the model, its size,
sensitivity, bandwidth and price. The prices are held low
by reducing or shortening worker operations, as you
would expect. I have seen some small, moderately priced
speakers being made in a total of about 200 seconds!

Looking fondly at the past.

I used to be an "audiophile." Thankfully, I've out­grown it. When I realized I wasn't enjoying music any­more—just playing with hardware—I mostly abandoned
the pursuit of system perfection, put together a passable
system, and began listening again. My ears, however, still
search for a sense of realism, for which I really have only
one criterion: the output (from the speakers) should sound
like the input (the live acoustic instruments). The system
should be transparent to the source material and par­ticularly to the performance.

I was educated in music. I studied opera in the bel
canto voice style, performed roles in college, light opera,

played in big bands and folk groups in the '60s, and rock

and jazz groups through the present. If there's one thing I
can say from a musician's point of view, it's that music
from loudspeakers is not right! More to the point, music
from loudspeakers can never be right. The sound that we
hear is the result of all the limitations speaker designers
feel constrained to put up with. Despite all the progress in
materials and techniques over the years that people have
devoted to making speakers, the results are all trade-offs.

29

pdf 26

They have to be, because musical instruments don't have
dispersion characteristics like speakers. Pop music re­cordings are mostly monaural anyway. Everything is
panned up the middle with some stereo reverb, so it really
doesn't much matter what a mass-market speaker does as
far as "imaging" is concerned, since there's nothing to
image in most recorded material. In the case of pop music,
it's worth remembering that the sound placed on the re­cording is the art form itself. The sound of the individual
instruments is often obtained by shoving microphones
into the instrument to get signal-to-noise ratio improve­ment, decrease mike-to-mike leakage, or to produce a mi­crophone-proximity-effect bass boost, or to reduce ambi­ent sound from the room. All this forms a part of the pop
recordist's art but ends up preventing listeners from hear-

ing anything like an image of the instruments that pro­duced the recorded sounds. As far as actual "stereo" is
concerned, there is only volume level panning if you're
lucky, and only some artificial electronic reverberation if
the recordings are typical pop. (See my article on this
subject in the November/December 1991 issue of db

Magazine.)

Having played in an orchestra and a big band, I can
tell you firsthand that what comes out of speaker-repro­duced recordings of these ensembles bears little resem­blance to what happened originally.

The fidelity of reproduction to the original sound
source has always been the heart and soul and the core
reason engineers pursue audio fidelity, notwithstanding
Sony MiniDiscs, Philips Digital Compact Cassettes and
other forms of squashed audio—but that's another tirade
for another article. [We'll see how "squashed" they are
when we test them.—Ed.]

And now, the nitty-gritty.

During the course of progress toward producing
high-quality, low-cost loudspeaker systems for con­sumers, designers have ignored what I consider to be the
worst and most important form of distortion, dynamic dis-
tortion. Dynamic distortion in loudspeakers is caused by

the instantaneous local heating of the voice coil wire.
This is not the global motor-structure heating cited as the
cause of power compression in large, professional drivers,

but the very local heating that takes place in the smallest
volume of wire, that volume which is comparable to the
size of the wire cross section. If you were able to obtain
an infrared detector sufficiently sensitive at long IR wave­lengths and point it at a voice coil, you would find that

the IR (heat) output of the wire would be somewhat pro­portional to the audio power input to the wire. I say some-

what, because there is thermal inertia, and the transfer

function is not perfect. This imperfect transfer function is
also nonlinear, and gets worse as higher power levels are

applied to the wire and global heating takes place. The es­sence of this phenomenon is that power pumped into a
loudspeaker is lost in nonuniform proportions to the input

power. [This explanation is new to me and my associates.

30

It may be perfectly valid, but I would have preferred to
see it confirmed here by an authoritative reference or
two.—Ed.]

How to improve this situation? Well, one way is to
use less power. Make less heat. But if we use less power,
we will need more efficient transducers.

Now, in the world of touring sound companies, the­aters, discos, stadiums and other big sound installations,
designers follow the seemingly human principle that if
you pay a lot for a speaker, it should be able to deliver

endlessly higher volume without burning up or breaking.
We know this is a flawed principle, but the attitude per­sists. And so, pro loudspeaker manufacturers keep trying
to supply drivers that won't break under standard oper­ating conditions, namely abuse. I suppose, as an academic
or engineering exercise, it's valuable to make drivers of
this sort; after all, if it can handle a sufficient amount of
power, then it can support added mass and lower res­onance, and be used as the low end of a two-way system,
making it cost-effective and relatively small—both rea­sonable and marketable attributes.

Without constraints.

If we now decide that the idea of a highly efficient

loudspeaker system is interesting and that the sound could

prove satisfying, we must look at the trade-offs (re-

member, there are always trade-offs).

The first trade-off to deal with is driver bandwidth.
Higher efficiency, less bandwidth. It's like putting a tur­bocharger on your car's engine; you get more torque, but
over a narrower range of engine revolution speed. If we
choose high-efficiency drivers, they will individually cov­er narrow frequency bands, requiring perhaps 4-way or 5­way systems, instead of 2-way or 3-way.

The second trade-off is that to achieve high effi­ciency, driver motor structures need the highest possible
magnetic flux, which means enormous and/or expensive
magnetic assemblies; also, driver motors must not waste
any resource—no extra mass such as overhanging voice
coil windings. This means that coils and air gaps will be
close to the same size, allowing for little or no linear coil
excursion. On the positive side, excursion in this situation
causes even-order distortion, which sounds musical, adds
brightness, and in speech systems that are driven hard,
produces an increase in speech intelligibility. The third
and fourth trade-offs, as alluded to before, are expense
and size.

High-efficiency systems will need larger drivers
and will press air volume into service. Remember, there is
a good reason why bass fiddles and tubas are bigger than
violins and trumpets.

There is also a rather nontrivial drawback to build­ing a high-efficiency speaker system that is only fair to
tell you about. There is no such thing as a high-efficiency
subwoofer driver. You obtain high efficiency in the low
bass range by coupling lots of cones together and moving
lots of air. You will want a subwoofer to match the rest of

THE AUDIO CRITIC

pdf 27

the system, and I'm warning you right now, before you
get ready to allocate space and funds, that you're prob­ably looking at four 18" woofers in about 50 cubic feet of
box, and a 1200-watt amplifier to drive them, in order to
keep up with the system I'm about to describe. You may
never play it loud enough to justify four drivers, but I in­clude them as a recommendation, because my goal in
building this system was to keep distortion below 1% at
any musically realistic playback level, at any frequency in
the human hearing range.

I have chosen my system drivers to be close to the
spirit of the old-technology drivers. JBL has kept this
spirit alive in some of their models, the 2123, 2202, 2220,
the "E" series "MI" speakers, and of course their com-

pression drivers. The drivers are:

(4) 2245H subwoofers (2.1% conversion efficiency

each or 8.4% in tandem)

(4) 2220J high-efficiency 15" woofers (one pair

each side, about 17% efficient each pair)

(2) 2123H high-efficiency 10" midranges (3.5%
efficient each)

(2) 2382A horns with 2450J compression drivers
(30% efficient each)

By efficiency, I mean for example, that if you put
100 electrical watts in, you will get 30 acoustical watts

out, as in the case of the horn.

If you're wondering, can the poor 10" mid with a
mere 3.5% efficiency keep up with the horn, rest assured,

I needed 10 dB of attenuation on the mid to get flat fre-

quency response.

For amplification, I used two BGW SPA-3 tri-

amplifiers built for me by BGW to provide highpass
filtering for the two fifteens, lowpass and horn EQ
filtering, switched attenuation, and built-in delay to

acoustically align the cones and compression driver. It
turned out to be an elegant and simple alternative to a
large rack of gear. Although the amplifiers each take up
only 5¼ inches of rack space, they each produce up to a
total output of 1000 watts, providing 600 watts for each

pair of fifteens, 200 watts for the mid and 100 watts for

the horn (because the compression driver is 16 ohms).

This represents an average of over 30 dB of headroom

above normal living-room listening levels, which gener­ally range in milliwatts for these speakers.

To give you a better idea of the difference in
efficiency I'm describing here—between low-efficiency
audiophile-type speakers and these high-efficiency speak­ers—for example, if you played, say, Magnepans at 100
mW, you would have a hard time hearing and under­standing speech from your listening seat, particularly if
your refrigerator is running in the kitchen or if your kids
are on the phone. At a one-watt average input, the Magne­pans will be producing what you would most likely re­gard as background music. At the same one watt into the
high-efficiency speakers, you will find some musical pas­sages to be uncomfortably loud, and your spouse will be
moved to exclaim, "lower that fi!"

ISSUE NO. 19 • SPRING 1993

Theory of operation.

I will labor each point here because there are likely
to be some statements I make that will require jus­tification in the mind of those readers who have never
been exposed to the type of technology or components
I'm describing.

Horns.

First, let me tell you why and how I believe horns
have acquired their unjustly poor reputation. The first rea­son is their cost; the second, their size; the third, their tra­ditionally intended application; and the fourth, their mis­application. It has been some years now since speaker
system manufacturers have offered horns in systems in­tended for sale to home users. Real horns, not schlock­shack plastic piezo types, are designed to impart direc­tional characteristics to the output of compression-driver
transducers. Such transducers are acoustic transformers,
and generally represent science and experimentation of
the highest order when they are designed and executed
correctly. It has taken over 40 years, aerospace-style dy­namic finite-element analysis, the highest-tech materials,
ten-year development cycles (I'm talking careers here),
and endless tweaking and testing to get the near­theoretical performance the largest "professional" com-

pression drivers provide. Proper horns couple the low

acoustic impedance output of the compression driver to
the air we sit and listen in. If the designs are good, the
coupling provides output which is the same as the input,
that is, the frequency response in the so-called acoustic
free-field is the same as the frequency response of the
compression driver when it is mounted on a terminated
tube (totally absorbing transmission line). In addition, and
in concert with the goal of achieving this technically non­trivial ideal of operation, horns can only control the dis-

persion pattern of sound when the horn dimensions are

larger than the sound wavelengths. As the mouth or the
depth of the horn gets smaller with respect to the in-

creasing wavelength of progressively lower frequencies,

the dispersion begins to "bloom out" and tend toward om­nidirectional. A rule of thumb I use for a lot of my "off
the top of the head" engineering is that the dispersion will
shrink to about 90 degrees (at the -6 dB off-axis angles)
when the wavelength is the same size as the source trans­ducer it's coming from. You can see right away that if, for
example, you want to control the horn's dispersion down
to 1000 Hz (wavelength 1.13 feet), the horn should be
over a foot in mouth size. This makes for big, often unac­ceptably industrial-looking devices occupying the wall
space on either side of the TV. This is a perception not
ignored by most speaker manufacturers' marketing de­partments.

In my opinion, many if not most speaker systems
are marketed on the concept of good sound from small­sized packages. This necessarily causes the design of such
products to rely on drivers which are small and have
small voice coils, which, as I mentioned earlier, heat up

31

pdf 28

too much and produce that worst form of distortion,
dynamic distortion.

To use a horn properly in an extremely high-fidelity
speaker system, it must be transparent to the sounds it
will be asked to reproduce. This seems an obvious state­ment, but it begs explanation and dispelling of the rep­utation for "horn sound" or "honkiness" so common
among all but the cognoscenti. The explanation is that old
cliche, "trust me" (I try to avoid cliches like the plague),
but a thorough explanation could easily fill a succinctly
written 2200-page college text book and is still the sub-

ject of scholarly investigation by engineers and physicists

doing their doctoral thesis.

The essence of the explanation is this: Horns that
have a directional control over the sound they are pro­ducing exhibit what engineers call a "Q factor," which is
a number used to describe how the output angle differs
from omnidirectional. [Don't confuse this with the Q of a
resonant circuit!—Ed.] For example, if we place an omni­directional loudspeaker in free space, at the top of a 10­story flagpole, it would radiate everywhere with the same

radiation intensity (Q=l). If we place the same omnidirec­tional loudspeaker on a hard reflecting surface, it's radia­tion will be forced to reflect into half the space it did on
top of the pole (Q=2). To an observer or a measurement
microphone, the loudspeaker will behave as though it has
twice the acoustic power output, or 3 dB more than it did
on the pole.

Horns force their output radiation into fairly narrow
angles, such as 90 degrees horizontally and 40 degrees
vertically, and can thus achieve acoustic gain in trade for
more omnidirectional sound dispersion. If the Q factor of
a horn is 10, then it will produce about 10 dB more acous­tic output on axis (where it's aimed) than would an omni­directional device with the same original acoustic power
output. This allows us to get extremely high sound levels
with very little power relative to, say, a dome tweeter.

There are some tricks that are essential for elim­inating horn "honk." The first is to use the cone driver
placed just below the horn all the way up to a frequency
where it begins to "beam" in accordance with the relation­ship of sound wavelength and cone diameter. At a fre­quency where the resulting Q factor of the cone matches
that of the horn, the transition from cone to horn will be
smooth, and not abrupt—as it can be in systems where the
cone is too large and the horn is too small. If this condi­tion is met, and the frequency response of the cone is
good well beyond the frequency up to which it is used (a
well-behaved upper-end rolloff), then the horn will enjoy
a seamless transition from the cone and will not honk, as­suming its frequency response is good and uniform over
its output angle. This latter condition is referred to as be­ing "power-flat" and is very important to the transparent
operation of the speaker system in rooms, with their con­comitant acoustic implications. If the speaker system is

power-flat, the sound in the room will be as good as that
particular room will allow it to be.

32

The midrange.

The midrange driver must be a cone, unless you
live in a theater and don't mind a 5-foot high horn (I
crossed my mid at 300 Hz into the woofers). As it turns
out, a mid cone supplying 300 Hz to 1200 Hz gives the
proper effortlessness with very little power, and thus has
extremely small cone excursions and low distortion. As I
mentioned earlier, I had to trim the 2123H mid cone back

10 dB on the amp's gain control to get the response

through the band flat. The power absorbed by the mid

cone driver amounts to milliwatts most of the time, which

helps to hold harmonic distortion to very low levels.

The driver is mounted on the baffle as close to the
horn as I could get it with my inexpensive mid-chamber
geometry. You could do better if you are willing to cut

out the lower lip of the horn and snug the mid frame up
into the cutout, and figure out a mid-chamber arrange­ment that would clear the horn and driver behind the
baffle, but this is not measurably better than just a touch­ing fit. The enclosure for the mid cone consists of a 10­inch diameter concrete casting tube made of plasticized
paper. The tube is mounted to the baffle by gluing into a
counter-bored shoulder cutout, routed in the back of the
baffle around the mounting hole. The tube is about 12
inches long (deep); it is filled completely but loosely with
a "jelly roll" of fiberglass cut from a roll about 4 feet
long. The back end of the tube is sealed with a disk of 1­inch medium-density fiberboard—the same material used
to build the rest of the box.

I experimented with a dozen midrange drivers before
I was confident that the 2123H with its high efficiency and
limited excursion linearity would produce sufficiently low
distortion. It is a wonderfully transparent driver and a
large part of the reason why this speaker system sounds
like listening to live music rather than loudspeakers.

The woofers.

The 2220 fifteen-inch cone driver should be thought
of as a low-midrange, not really a woofer. Yes, it's a big
driver with a big voice coil. In fact, I use two of them in
my bass guitar rig, but its QTS and moving mass are so
low that when you put it through Thiele-Small calcula­tions and plot curves on a computer, as I did a hundred
times, the device ends up looking more like a midrange it­self. To be accurate, the Keele exponent-corrected pro­gram I use (because it tends to give me systems that mea­sure the same as the model predictions) calls for about 1.5
cubic feet per driver, tuned to about 85 Hz—not exactly
organ pedals. I ended up opting for a slight overdamping
of two units in a 3-cubic-foot volume tuned to 80 Hz.
Even so, the unassisted output of the box is flat to 65 Hz
and droops only slightly at 40 Hz, in the middle of a

40,000-cubic-foot room. [I didn't know you lived in

Windsor Castle, Drew.—Ed.]

You must make sure the highpass filter on the am-

plifier is set to 80 Hz and rolls off at a rate of at least 18

dB per octave. The 2220J drivers are high-efficiency, lim-

THE AUDIO CRITIC

pdf 29

ited linear excursion devices (in fact, they are one of the
highest efficiency cone drivers made anywhere). The
crossover frequency of 80 Hz is the design target to limit
cone excursion and produce a good transition to the sub-

woofers.

The enclosure is built to be as rigid and non­resonant as possible and then lined with fiberglass over
the entire interior surface area, except around the ports,

where air turbulence might spray fiberglass around. The
woofer portion of the enclosure is the only real structure.
The midrange tube is extremely rigid and exceptionally
nonresonant. My goal in designing the woofer section
was to minimize spurious panel vibration and acoustic
output—within reason. There is more panel output, in
fact, from the back cover of the compression driver and
horn walls.

I used four two-by-fours for bracing inside the
woofer compartment. These were counterdrilled for wood
screws and glued on-edge to the compartment panel inter­ior surfaces. I tried to space the braces at random—so that
no two unbraced panel areas were the same size. I also
glued the two cutout disks from the woofer holes to the
back panels to make the total panel thickness 2 inches,
plus braces!

Setting up the system.

This can be tricky. I used a TEF analyzer and a

1/24-octave Brüel & Kjær real-time analyzer. First I did
energy-time measurements to set the delays in the am­plifier. This proved to be difficult, since I had to take dis­tance-ranging measurements of each driver separately to
make sure I was looking at arrival times from the in­tended measurement object. After I got the delays set, I
checked frequency response the best I could in the space
available, then resorted to the real-time analyzer. Be
aware that the frequency responses you get with these two
methods are very different because TEF windows its mea­surements to try to exclude reflections and examine only
direct sound from the source, while real-time analysis in­cludes all returning room energy information along with
that from the speaker. I like a balance of both mea­surement methods, because one lets you fine-tune the
energy output of the speaker, and the other lets you adjust
large trends like the general "too-bright" high end you
will likely notice if you try to obtain flat output to 20 kHz
from the direct-sound readings of a truly power-flat or
"constant-directivity" horn. Such horns can produce more
high-frequency energy in the first place, and then they can
deliver it to even fairly large rooms. Most people are not
used to listening to a power-flat top end and will find it
too brassy. Only minimal-miked big-band recordings will
sound right with the system adjusted that way.

For your playback system, I recommend that you
use an extremely low-noise preamp with simple Baxandall­type "tone controls." The bass and treble turnover fre­quencies are a matter of taste, but 100 Hz and 10 kHz
seem to work well to adjust this system to music on re-

ISSUE NO. 19 • SPRING 1993

cordings. To my ear, parametric equalization is less pleas­ing and is certainly prone to putting phase aberrations in
more audible frequency ranges than are simple tone con­trols. One-third-octave "graphic" equalizers are complete­ly useless for high-fidelity use and one-octave units are
even worse. [How about a Cello Audio Palette? If you
can afford the speaker, you can afford the Cello.—Ed.]

You should start adjusting with the amp gains set
low and turn the preamp all the way up. Slowly advance
the amp gains until you can achieve proper balance and
slightly louder than necessary output. This will ensure
that you will have the least possible amplifier hiss from
the speakers. Amplifier hiss is a phenomenon that rarely
troubles owners of low-efficiency speakers, but these
monsters are efficient enough to make the transistor junc­tion noise of poorly designed amplifiers quite audible.

If any of you reading this decide to build this sys­tem, it will cost around $10,000, including amps and all.

If you get that serious, if you are rich and adventurous

and don't care what stereo salesmen think, you are the
type of person who would really enjoy this system.

[Rich? I know poor audiophiles who have accumulated

five-figure stereo systems. But maybe that's why they're

poor.—Ed]

The subwoofers should generally occupy the space

between your main speaker systems. Because of the ear's
forgiveness, you'll find there's a "window" of space for
physical placement that allows a good deal of flexibility

in fitting the speakers into your listening space.

What you have when you're done.

These loudspeaker systems could easily be used be-

hind a perforated movie screen, providing sound to an au­dience of hundreds of people in a small movie theater.
They are also equally capable of causing you, in a home
listening-room setting, permanent hearing loss, and doing
so quickly.

I have two tips for you, and you would do well to

pay heed:

First, play music at no more than realistic levels. I

assume that if you choose to build these things, you've
done so because you're interested in the fidelity of the re­produced sound to the originally recorded sound. You
will get the best representation of the original sound if
you play the reproduction at the original sound level.
Playing too loud is as detrimental to fidelity as playing
too softly.

If you play predominantly rock music, you need to

keep a sound level meter handy. You can get a perfectly

adequate one at your local Radio Shack store for around

thirty bucks, and for your ears' sake, don't ignore this ad-

vice. These speakers make so little distortion that you will
be tempted to believe that the 120 dB sound you are lis­tening to is only playing at 90 dB. This is not good. You

will lose your hearing. Don't let this happen!

If you find that the clean sound causes your favorite

(continued on page 53)

33

pdf 30

Miscellaneous Electronics for

Audio and Video

By Peter Aczel

Editor and Publisher

&

David A. Rich, Ph.D.

Contributing Technical Editor

Not really an article but a catchall assortment of individual reviews,
by two editors, of unrelated or loosely related audio components,
grouped together here for mere convenience.

As our regular readers surely understand by now,
and our new readers soon will, we at The Audio Critic
expect no audible differences in our tests between high­quality preamplifiers, amplifiers, and other linear analog
electronics, and report such differences, if and when dis-

covered, with some degree of surprise. That's one of the

main differences between us and the golden-eared high­end journals, where they just know that the $6000 unit
will sound better than the $1500 one and, of course,
confirm that expectation in their tests. We believe, on the
other hand, that an audible difference always has a cause
and that in the absence of an explicable cause any report
of an audible difference is highly suspect—unless verified
by a series of double-blind comparisons at matched levels
(and who does that except us and archvillains like David
Clark and Ken Pohlmann?).

Thus the emphasis in the applicable reviews that
follow is on measurable accuracy and quality of construc­tion. Accuracy—i.e., transparency to the signal—so we
won't even have to think about the electronic component
when evaluating the sound of the total system that in­cludes it; quality of construction, so that the accuracy can
be expected to continue for a long, long time. We talk
specifically about the sound of the electronic component
only if there's something new or different to be said. All
pretty obvious.

—Ed.

Full-Function Preamplifier

Bryston 11B

(Reviewed by Peter Aczel)

Bryston Ltd., 57 Westmore Drive, Rexdale, Ont., Canada M9V
3Y6. Model 11B preamplifier, $1450.00 with optional balanced
outputs. Tested sample on loan from manufacturer.

The difference between this model and the lower-

ISSUE NO. 19 • SPRING 1993

priced Bryston .5B reviewed in Issue No. 18 is greater

flexibility and additional features; the basic engineering
concept, circuit design, and build quality are exactly the
same and require no further reviewing. In essence, this is
merely an editorial follow-up, without new insights by
David Rich.

Model 11B is actually the same preamp as Model

12B, which David Rich mentioned in passing as a pos­sible recommendation at the end of his series of reviews;
the difference is only that the 12B incorporates a moving­coil step-up transformer, whereas the 11B I tested does
not. Any 11B can be retrofitted, however, to become a

12B. Together these models represent Chris Russell's cur­rent thinking on top-of-the-line preamps—and in my
book he is the most levelheaded thinker in this crazy
mixed-up industry.

The 11B has more inputs and outputs than the .5B;
it can accommodate two tape recorders, although the tape
outputs are still unbuffered, so the same caveats apply re­garding a connected tape deck which is not powered up.
(Bryston has recently added specific instructions about
this.) In addition, the 11B I tested had both unbalanced
and balanced outputs. The substantial price difference you
must pay for all this is explained by the superb construc­tion of the expanded switching facilities and additional
connectors. (Remember, Bryston products are guaranteed
for 20 years.)

The basic Audio Precision measurements on the 11B
yielded results virtually identical to those reported in the

.5B review and need not be repeated here. Two excep­tions: (1) Channel separation in the line stage was greatly
improved, though still not very impressive; I measured -44
dB cross talk at 20 kHz in the less good channel, drop­ping at 6 dB per octave to -102 dB at 20 Hz. That's a 12
dB improvement across the board, the result of recent
work on the printed-circuit board. (2) The balanced out­put adds a tiny bit of distortion because of the additional

35

pdf 31

active stage; the minimum THD plus noise of the line

amplifier was about 6 dB higher through the balanced
output than through the unbalanced (main) output. The
message is clear: use the balanced output only when the
power amplifier is far away and the interconnect is so
long that the signal needs extra protection—or when the

power amp has only a balanced input. (Sorry, tweaks, I
couldn't hear the difference. Can you hear the difference
between -88 dB and -94 dB distortion? Yeah, right.) It
should be added that the balanced output clips at 30 V, as
against 15 V for the unbalanced output.

In actual use this preamp is truly a joy—no hum, no
hiss, no pops, no funny noises of any kind, just smooth
operation and flawless sound. That total endorsement ap­plies only to the current version; an earlier production
sample gave me some minor problems that are no longer
relevant. I think even the most finicky audiophile would
be satisfied with the 11B, except perhaps for its price,
which is quite high but not high enough to impress some
insecure high-end gurus. My regular standby, the Boulder
MS, has perhaps even better measurements by just a hair
(in some but not all categories) and a few nonessential ex­tra features (polarity inversion, mono/stereo switching,
channel reversal, and such) plus modular construction—
but it costs so much more! If I didn't already have a
world-class preamp, I know I could live happily with the

Bryston 11B or 12B in my main system.

Cable Enhancer (it says here)

Duo-Tech Model CE-1000

(Reviewed by Peter Aczel)

Duo-Tech Corp., 37396 Ruben Lane, Building F, Sandy, OR

97055. Model CE-1000 Cable Enhancer, $179.00. Tested sam-

ple on loan from manufacturer.

Tweako cult items in audio fall into four broad cat­egories: (1) ridiculously priced but gets the job done—
e.g., silver cable; (2) ridiculously priced and screws up
the job—e.g., UltrAmp D/A Converter; (3) ridiculously
priced and accomplishes nothing—e.g., Tice clock; and
(4) fairly priced and accomplishes nothing—e.g., this so­called cable enhancer. When you take it apart and look in­side it, you see electronic circuitry and parts that should
sell for approximately as much as they ask for the prod­uct. That's good. When you examine the manufacturer's
claims and rationale for it and run some simple tests, you
realize that it's pure B.S. That's bad. I don't think there's
fraudulent intent behind the Duo-Tech, just utter silliness
and self-deception.

I'm sure that many of our readers have already
heard of this magic device. The tweako testimonials and
anecdotal raves have been out there for a while. (Hey,

there's never a shortage of wishful thinkers in the audio

world.) In case the whole thing is new to you, here's the
gist of it: You get this cute little chassis, about as big as a
ten-pack of 3.5-inch computer diskettes. It has a pair of

36

phono jacks at each end and a front panel with two
switches and two LEDs. You also get two plug-in adapter
modules that convert the phono jacks into speaker-cable
terminals, plus an AC power adaptor that supplies the 12
volts DC used by the unit. To "enhance" a pair of cables,
you connect them between the jacks/terminals on the left
and right, energize the system, and let the cables "burn
in" for at least 48 hours, or more for "quality cables" (it
says in the manual). The "signal flow direction" is terribly
important; if you have plebeian cables without directional
arrows on them, the treatment permanently locks in their
directionality (from left jacks/terminals to right), and you

must label them with little arrow stickers that come in the
package. (How's that for anal, Sigmund?) The following
before/after differences in the sound are claimed (are you
ready?): Better focus, improved coherence and dynamics,
reduced glare and ring, more realistic soundstage, better
defined and tighter bass, more extended highs. What
about "burning in" with music instead? Not nearly as
good!

I was of course curious what kind of signal has the
extraordinary capability of producing such results, so the
first thing I did was to connect an oscilloscope across the
signal path of the device. Well, it's a pulse train consisting
of squarish pulses of completely random duration but
equal amplitude, about 14 V from peak to peak, the tops
and bottoms steeply tilted as if by highpass filtering. I
could see no difference when I switched from the Inter­connect mode to the Loudspeaker mode. I must add that I
lost very little sleep trying to pry further into the secrets
of the design; it's some kind of simple digital circuit.

Then I went out and bought two brand-new pairs of
identical 3-foot interconnects. Nothing fancy but of decent
quality, with good shielding and nice gold-plated plugs,
the kind a noncultist like me is happy with. I took one

pair and burned it in (if you'll pardon the expression once

again) for 96 hours on the Duo-Tech. I left the other pair
untouched. Then I did my listening test. I plugged both

pairs into my trusty Boulder MS preamplifier, so that I
could switch from one to the other with a single click of
the source selector switch. I plugged the other end of one
pair into the line output of the Sony CDP-X779ES CD
player (the best in my tests) and left the remaining plugs
of the other pair dangling free near the line output, ready
to be swapped instantly. Thus I was able to switch back
and forth between the burned-in and the virgin pair of
cables in about two seconds while a CD was playing—not

a blind test, to be sure, but I was ready to set up a formal

ABX comparison in case I thought I heard the slightest
difference. But—you guessed it!—I heard no difference
whatsoever on any kind of music, even though I knew
which cable was which, so how could I possibly have
heard a difference under double-blind conditions? And
how could anyone, under any condition, have heard a dif­ference that had no reason within the laws of physics to
exist? Afterwards, I felt a little sheepish for having spent
time and energy on such a silly exercise—but somebody

THE AUDIO CRITIC

pdf 32

had to do it, right?

I flatly refuse to dignify the muddleheaded techno­babble offered by Duo-Tech as the scientific rationale for
the Cable Enhancer by printing it and then refuting it
here. Instead, I'll make the same offer as I did to George
Tice regarding his magic clock. If Duo-Tech can produce
three electronics experts with university graduate degrees
in engineering or physics who are not commercially
linked to the company or its products and who will certify
in writing that Duo-Tech's claims for the Cable Enhancer
are scientifically valid, then I shall devote a special issue
of The Audio Critic exclusively to the explanation and
celebration of the Cable Enhancer technology and mail it
as a free bonus to all subscribers. Any bets on the out­come of my offer?

52" Rear-Projection TV with Surround Sound

Magnavox RM8564A

(Reviewed by Peter Aczel)

Philips Consumer Electronics Company, One Philips Drive,
P.O. Box 14810, Knoxville, TN 37914-1810. Home Video The­atre 1992: Magnavox RM8564A, with JBL speaker system and
Dolby Pro Logic Surround Sound, $3400.00. Tested sample on
loan from manufacturer.

Of the three available large-size TV formats—
direct view, rear projection, front projection—I strongly
lean toward rear projection at this time. Even the 35" di­rect-view CRTs are too small for maximum viewer in­volvement in, say, football or opera (to name only two of
the many kinds of panoramic TV fare), whereas the un­questionable impact of the big front-projection format is
canceled out by the immense inconvenience it creates in
any room that isn't exclusively dedicated to home theater
viewing. A 52" rear-projection set is big enough to satisfy
my craving for lifelike dimensions in audiovisual en­tertainment, yet it still fits into a multipurpose family
room in which kids, dogs, friends, neighbors, etc., come
and go. Admittedly, the resolution of detail isn't quite as
high as on, say, a 19" Trinitron, but the overall pre­sentation imitates life more convincingly because of the
scale. (See Issue No. 12 on the subject of small versus
large video.)

This particular Magnavox unit is not only an ex­cellent rear-projection TV but also unusually sophisticat­ed in terms of audio, interesting enough to be reviewed in
an audio journal. Unfortunately, by the time this is in
print, it will be on its way out of the stores, to be replaced
by a 1993 Philips model (Philips and Magnavox being
virtually interchangeable brands, like Dodge and Ply­mouth at Chrysler). Unlike high-end audio components,
TVs in all price brackets undergo a yearly model change,
leaving a slowpoke reviewer like me panting as I bring up
the rear. No matter; it's the current Philips approach to
home theater packaging that's under review here, not just
a 1992 model.

ISSUE NO. 19 • SPRING 1993

Rear projection, not long ago rather low in video
fidelity, has been getting a great deal better lately, and the
Magnavox RM8564A illustrates just how much better.
The picture isn't significantly less good than on a 32" or
35" direct-view set, and that's quite something. (No, I
didn't say it was as good.) The trade-off is between the
highest possible definition and the greatest possible re­alism in dimensional perception. As I said, I like the
trade-off. Let me, however, discuss the audio system first,
as it is the long suit of the set. It would actually pass as an
upper-mid-fi home music system by itself, without the
video, and that's more than I can say for any other TV
known to me.

On each side of the 52" screen, there is—get this!—
a JBL pure-titanium 1" dome tweeter, the same I enthused
about at some length in Issue No. 14. I never expected to
find it in a TV, I must say. Below the screen, on each
side, a JBL long-throw 8" woofer in a vented enclosure is

crossed over passively to the tweeter to form a very cred-

itable full-range speaker system, with response from

about 40 Hz to well beyond 20 kHz. A 25-watt amplifier
channel drives each side; in addition, there are two more
25-watt amplifier channels available for two surround­sound speakers in the rear.

There is a wide range of stereo and surround-sound
modes to choose from. The simplest is plain stereo out of
the two built-in speaker systems. The most elaborate is
full Dolby Pro Logic, which requires an additional stereo
amplifier feeding two speakers flanking the TV set plus
two rear speakers fed from the built-in extra channels.
The set's own speakers and their amplifiers then become
the Dolby center channel. Other combinations between
these two extremes can be easily set up, depending on the
available program material and ancillary equipment. In
each case, the audio quality is just as good as you would
expect it to be with a component system using similar
amplifier power and the same surround speakers. Home
theater can—and should—sound just as good as com­ponent audio of comparable sophistication.

The microprocessor-controlled audio and video set­tings that can be selected from the screen menus of this
set are as varied, elaborate, and versatile as any I have
seen and then some—not really surprising when you con­sider Philips's deep involvement in microprocessor tech­nology. Remote control jockeys will surely experience an
unprecedented sense of power as they issue all those
push-button commands. One small complaint: to switch
between program sources (antenna, cable, VCR, laser
disc, etc.) requires going to a submenu of a submenu of
the main menu. That's a bit ridiculous; program source
selection is one of the most frequent user operations, and
most remote controls have separate buttons for it. Some
totally unrealistic computer nerd must have thought up
that one—and I don't even think it has been fixed in the

1993 successor model. Once you get used to it, though, it

matters very little.

To get back to video quality—that's why you buy

37

pdf 33

an expensive TV, after all—the first thing you notice
about this rear-projection system is that you can't view it
too far off to the side or standing up. The set's really ex­cellent picture quality is evident only when you sit in
front of it or at a small angle to it. I would say that four

front-row seats are the maximum for proper viewing. Of
course, you can always set up a second row. Again, no
big deal, but viewers accustomed exclusively to direct-

view CRTs are initially bewildered.

As usual, I put the set through its video paces with
the various excellent tests on the Reference Recordings
laser videodisc A Video Standard. Black level retention
(i.e., the ability to hold black at black, regardless of the
picture content) was quite good for consumer equipment
but not studio-perfect—and I didn't expect it to be. Con­trast is not the long suit of rear-projection TVs; the peak
linear capability of just about every set will be exceeded
before the contrast is as high as one would ideally like it
to be, and such was the case here. The default setting by
the factory was a very acceptable compromise, however.
Color performance via the S-video input was very good,
and the factory settings of Color and Tint were pretty
much on the money. Geometry was also quite precise,
with no significant distortion on checkerboard patterns,
circles, etc. Convergence as adjusted at the factory need­ed no trimming. On a subjective basis, the picture with a
good program source was truly beautiful in color and had
excellent definition in my opinion—always keeping in
mind that it's rear projection on a large screen.

The advertised horizontal resolution of the set is
600+ lines, but that's a nebulous area of specifications
where laissez-faire reigns—with my best laser disc player
and the RR test disc I can resolve only about 400 lines,
and the Magnavox easily passes that test. (The theoretical

best for NTSC broadcasts is 336 lines.) More interesting
is the set's PIP (picture-in-picture) capability: you can,
for example, watch your videocassette of Terminator 2
(or do you prefer Wings of Desire?) and at the same time
tune in the news on CNN within a cutout in the main pic­ture. Unfortunately, you can't simultaneously view two
programs from set's own tuner; you need an outside
source for the second program.

The 1993 successor model to the Maganavox
RM8564A is the Philips 52NP51FS. The price is the
same; the amplifier wattage has been increased and the
audio functions expanded (among other things, you no
longer need an outboard amplifier for full Dolby Pro Log­ic, although you can add one for even more power); the
video circuitry and optics are also claimed to be improved
(800+ lines of horizontal resolution is the new spec!?). It
looks like an even better deal to me, except that the new

1" tweeter and 8" woofer are by Philips, not JBL, so I

can't predict the speaker performance.

To sum up, if you like large-screen rear-projection
TV as much as I do—and that's a purely personal prefer­ence—then give serious consideration to the rear­projection home-theater packages by Philips (under its

38

various brand names) because they offer the rare com­bination of excellent video and audio performance, with
options for integration into even more elaborate home
audio systems.

Line-Level Preamplifier

Monarchy Audio Model 10

(Reviewed by David Rich)

Monarchy International, Inc., 380 Swift Avenue, Unit 21, South
San Francisco, CA 94080. Model 10 Buffered Control Center,
$980.00. Tested sample on loan from manufacturer.

This preamp is an example of the best of science
and the worst of tweako loony tunes. The most significant
difference between this preamp and pure tweako products
is its relatively moderate price ($980).

Mechanical construction is excellent. The front pan­el is massive, the sheet metal thick, and the metal volume
control knob is 4 cm in diameter and 2 cm deep. The
high-quality RCA jacks are directly mounted to the rear

panel. The PC boards are double-sided with plated-

through holes. The RCA jacks are wired directly to selec­tor switches, which are directly wired to the volume con­trol. The amount of hand wiring in this unit significantly
adds to the construction cost. The wire is claimed to be

pure silver. I consider silver wire an extravagant waste of

money, but it could not have added very much to the cost

of this preamp given the unit's reasonable price. The vol­ume control is a huge (4 cm in diameter and 9 cm deep)
stepped attenuator manufactured by Shallco. The control,

which would be more at home in a spacecraft than a

preamp, has 30 steps, each 2 dB. This is the most expen­sive control I have encountered in a commercial preamp,
including those priced in the high four figures. A balance
control is not included on this preamplifier. The switches
are of the finest quality I have seen on a preamp, being
both sealed and gold-plated.

The selector switch arrangement is a dim-bulb
tweako cultist idea. A pair of inputs is directly connected
to a three-way toggle switch. The up position puts one of
the two input signals on the bus, the down position the
other; the middle position disconnects the switch from the
bus. A complement of three toggle switches is used for
the six inputs of this preamp. It is possible in this arrange­ment to connect more than one input signal to the bus.
Once, when we engaged the preamp in this fault condi­tion, the line amplifier went into a latch state, rendering
the preamp inoperative until the power supply plug was
pulled (the unit has no power switch), but we were subse­quently unable to recreate this. The preamp will consis­tently go into a latch mode if the power supply is inter­rupted briefly. This is the result of the power supplies
collapsing asymmetrically. The problem will be fixed in
production units according to the manufacturer (we tested
a very early sample). Believe it or not, this $980 preamp
has no tape monitor function—tweak-tweak!

THE AUDIO CRITIC

pdf 34

The line stage uses an Analog Devices AD744
BiFET IC for voltage gain, in conjunction with an 8­transistor discrete unity-gain output stage. The AD744 is
a high-performance device with a slewing threshold of

0.9 V and a 13 MHz gain-bandwidth product. The op­amp is used in a novel topology developed by Walt Jung,
which bypasses the AD744's output stage. A DC servo is
used to reduce the preamp's output offset. The PC board
on our preproduction test sample had a significant amount
of component rework on it. The manufacturer claims this
will be corrected in production units. Parts quality on the
board is typical for a unit in this price class. In true tweako
cultist fashion, a muting relay to prevent power-up pulses
from passing to the power amplifier is not included.

The power supply is dual mono, including the trans­formers. An AC line filter is included in this unit, so you
do not have to spend megabucks on an audiophile­approved external one. A separate button quad rectifier is
used for the positive and negative supplies of each chan­nel. The unregulated filter caps are 4700 µF, and the regu­lated filter caps are 8200 µF. LM317 and LM337 IC regu­lators are used. The regulators have an output voltage of
+24 V. 78L18 and 79L18 regulators are used to subregu­late the power supplies to the ICs down to ±18V. A pair
of subregulators is used for each channel.

Performance of the preamp was for the most part

exemplary. Channel separation changed at 6 dB per oc-

tave from 110 dB at 20 Hz to 50 dB at 20 kHz. Clipping
occurred at 12 V rms. Distortion reached a minimum of

-94 dB at 6 V rms across the entire audible band. Switch­ing to a 600-ohm load did not change these results. The
state-of-the-art performance of the line amp should drive
adherents of the tweako camp away in droves, since

tweaks equate good numbers with bad sound.

Clearly this preamp cannot be recommended un­conditionally for the reasons outlined above, despite the
fact that it represents exceptional value. If a power switch
and muting relay were added, and the loony-tune input se­lector switches were replaced with a rotary function
switch and rotary record selector switch, we would have a

product that could be unconditionally recommended.

Full-Function Preamplifier

Rotel RC-980BX

(Reviewed by David Rich)

Rotel of America, P.O. Box 653, Buffalo, NY 14240. RC-980BX
preamplifier, $499.90. Tested sample on loan from distributor.

This preamp is at a lower price point than the

preamps we looked at in the last issue. It is made in Japan

but it is very different from most Japanese designs, which
typically have a more complex control panel than a 747.
The Rotel has just four plastic controls on the front pan­el—power switch, volume control, source selector (la­beled Listening), and record selector (labeled Recording).
A headphone jack is also on the front panel. The head-

ISSUE NO. 19 • SPRING 1993

phone amp is a JRC4556 op-amp. There is no balance
control because at this price point only a low-cost unit
could have been selected. The volume control is a 2-cm
wide unsealed Alps unit similar to the control used in the
much more expensive Adcom GFP-565. The left- and
right-channel levels can be adjusted individually. A fric­tion plate on the volume control allows both channels to
be adjusted together. The record selector has a novel Off
position which disconnects Tape 1 Out and Tape 2 Out
from the main bus. This is a poor man's replacement for a
tape monitor buffer circuit. Tape 1 Out is disconnected
when Tape 1 is selected on the record selector, preventing
system damaging oscillations. For an unknown reason
there is no Tape 2 position on the record selector, and it is
thus impossible to record on tape deck 1 from tape deck

2. The function and record selectors are unsealed linear
switches. The front panel has a rotational-to-linear con­verter which drives the switches through a long unsealed
band of metal. This arrangement could be a long-term re­liability problem if the band of metal starts to bind or
stick in its unsealed channel. The switches are located
near the input connectors to minimize jumpers. This is re­quired because the PC board is single-sided. The PC
board does not have plated-through holes. The sheet met-

alwork is slightly thicker than is typical of mass-produced
consumer electronics. The unit is held together by sheet
metal screws.

The power supply is massive for this design. A 8.5­cm diameter, 4-cm high toroid transformer drives a high­current composite bridge rectifier. The unregulated rails
are filtered with a 4700 µF capacitor. Power supplies are
regulated by series-connected pass transistors which have
no global feedback. The bases of the pass transistors are
connected to a zener diode reference. Each channel has a
separate regulator pass transistor, but both channels share
the same diode reference. Heat sinks are not used on the
regulator pass transistors. The regulated power supply rail
voltage of ±17.4 V is too close to the absolute maximum
voltage rating (18 V) for the ICs used in this preamp to
insure maximum reliability. Separate 78M15 and 79M15
regulators are used for the headphone amplifier.

The line stage uses an Analog Devices AD711 Bi­FET op-amp (4 MHz unity gain bandwidth and 0.8 V
slewing threshold). The input (C1) and output (C3) are 10
µF electrolytic coupling capacitors. The C2 capacitor is
not present in the feedback loop. Time-delayed relays
connected to the outputs prevent turn-on transients from
reaching the power amp. In our Audio Precision tests, the
line stage clipped at 12 V rms. Low-frequency distortion
plus noise of the line stage (we always report the "worse"
channel) reached -95 dB before clipping, an excellent re­sult. The 20 kHz distortion of the line stage reached a
minimum of -83 dB at 2 V rms, rising to -80 dB at clip­ping. Adding a 600-ohm load to the line stage reduced the
clipping level to 5 V rms. The 20 kHz distortion reached
a minimum of -73 dB at 0.3 V rms and remained at that
level until clipping. The fact that the line stage is based

39

pdf 35

on the AD711 IC and not a more advanced IC or a dis­crete transistor circuit can be seen from the last three
measurements. Channel separation of the line stage is
greater than 98 dB below 200 Hz. It then rises 6 dB per
octave to 66 dB at 20 kHz. The simplicity of the signal
path and the elimination of a balance control are partially
responsible for this excellent result.

The phono equalizer is a two-stage design. The first
stage is a linear-gain amplifier with gain selection by a
switch on the rear panel. The 2120 Hz pole is imple­mented passively between the first and second stage. The
second stage implements the 50 Hz pole and 500 Hz zero
of the RIAA equalization curve actively. The first stage
uses a discrete differential pair in conjunction with a Tex­as Instruments TL071CP. The required noise performance
of the phono stage could not be achieved with an IC op­amp alone. The TL071 is a low-cost, general-purpose,

JFET-input op-amp. The AD711, used in the line stage, is

an enhanced version of the same. The TL071's low (3

MHz) unity gain bandwidth restricts its use in high­performance gain stages (it is widely used in mass-market
equipment), but it is occasionally used, with acceptable
results, as a low-cost unity-gain buffer in high-end de­signs. As will be seen below, the TL071 performs in an
exemplary manner in the function it is assigned in this
preamp. A Signetics NE5534AN is used in the second

stage. The Signetics part has better noise performance
than the AD711 used in the line stage, but it has a lower
slewing threshold, making it less desirable to some de­signers for use in a line stage.

The performance of the phono stage proved to be

state-of-the-art. Phono equalization is held to better than

±0.05 dB in both the MM and MC modes. The channels

were balanced within 0.05 dB. These are the best results
we have ever measured and should embarrass the de­signers of more costly preamps. The phono stage clipped
at 12 V rms, just like the line stage. Low-frequency dis-

tortion plus noise was -74 dB and -90 dB for the MC and
MM modes, respectively. These numbers are the result of
noise, not distortion, since the lowest measured numbers
occurred just before the onset of clipping. At 20 kHz the
phono stage distortion reached a minimum of -86 dB at

5.5 V rms.

The Rotel RC-980BX lacks the look, feel, and build
quality of the best preamps reviewed in Issue No. 18 but
it costs significantly less. It is a clear choice over the
more expensive Adcom GFP-565 preamp reviewed in
that issue. The phono-stage performance is, as I said,
right up to the state of the art. The preamp is therefore
highly recommended. The only change I would make
would be to use a less overbuilt power supply and apply
the cost savings to a better-quality IC in the line stage. •

40

THE AUDIO CRITIC

pdf 36

Another Look at

Outboard D/A Converters

By David A. Rich, Ph.D.

Contributing Technical Editor

(with an interruption by Peter Aczel, Ed.)

Everybody is doing something different in this product category,

but a faultless design at a realistic price still remains to be seen.

This is a continuing chapter in the search for a CD

decoder box that can truly produce analog signals at 16-

bit resolution. As will be seen below, it is not the final
chapter in that search. Readers who find my analysis of
these devices a little too technical are directed to my tu­torial paper, which appeared in the now out-of-print Issue
No. 15. An updated reprint of the article should be avail­able soon.

Please note that the measurements we report below
do not check for limit cycle oscillations, which can be a
problem with the delta-sigma (1-bit) DACs. The prom­ised examination of limit cycle oscillations in delta-sigma
DACs has been postponed to the next issue.

Recently Stereophile has begun measuring the jitter
of the internal word clock driving the DAC. We have not
done this here for three reasons: (1) Jitter on the word
clock will cause discrete tones or an increased noise floor
at the output of the decoder. Both of these effects will be
seen in the THD + N measurements. (2) Power supply
noise and digital input signal feedthrough at the DAC
may contribute additional jitter from the DACs internal
logic, which will be completely missed by this test. (3)
The probe placed on the word clock pin will present an
additional load to the word clock driver. Since the word
clock driver is not designed to drive the additional load, a
false reading can result.

Audio Alchemy DTI/XDP/PS2

Audio Alchemy, 30879 Thousand Oaks Boulevard, Suite 222,

Westlake Village, CA 91362. Digital Transmission Interface
(DTI), $349.00. Extended Digital Processor (XDP), $300.00.
Power Station Two (PS2), $120.00. Tested samples on loan from
manufacturer.

These three separate components add up to one D/A
decoder box. As will be seen below, Audio Alchemy uses
this à la carte approach to increase the potential number
of users of its products. All the extra boxes and cables re­quired add to the cost of producing the Audio Alchemy

42

D/A system. One annoying problem is that the feet of the
XDP are not high enough for the XDP's front panel to sit
flush with the DTFs front panel.

The Power Station Two consists of separate analog
and digital power supplies. Two power transformers, two
DIP-sized full-wave rectifiers, and four 4700 µF ca­pacitors create the unregulated supply rails. The digital
rails are then regulated with 7808 and 7908 regulators and
further filtered with 4700 µF capacitors. Heat sinks are
used on these regulators and all regulators in the DTI and
XDP. Analog regulation is identical, except that 7818 and
7918 regulators are used. The DTI uses the digital supply
only. The XDP uses both supplies. The Power Station
Two can also be used as an upgraded power supply for
some other Audio Alchemy products .

The DTI (Digital Transmission Interface) in­corporates an S/PDIF decoder, an S/PDIF encoder, a
switch and circuitry to invert the phase of the digital sig­nal, and coaxial as well as optical inputs. The S/PDIF de-

coder is the Crystal Semiconductor CS8412-CP. This is
the same chip as EAD uses in their DSP-7000 processor.
While this chip provides good jitter performance, Crystal

Semiconductor data sheets show that it can be improved
with the addition of a VCXO (voltage-controlled crystal

oscillator) based PLL (phase-locked loop) placed after
this chip. Audio Alchemy does not include this second
PLL. A very early press release for this unit indicated it
would use a GaAs (gallium arsenide) based VCO running
at a very high clock rate. The VCO clock was then to
have been divided down to the system clock frequency.
The division process reduces jitter from the VCO. This
intriguing approach is, unfortunately, not used in the pro­duction version of the DTI.

The DTI has two outputs: (1) a modified I2S bus

output and (2) an S/PDIF output. The I2S bus is one of a

series of data transmission bus standards used to connect

chips in a CD player or D/A decoder. I2S has no ad-

vantage over other bus standards. The I2S bus does not
transmit subcode information; thus Audio Alchemy adds
an additional signal line to the I2S bus to transmit the de-

THE AUDIO CRITIC

pdf 37

emphasis code. The S/PDIF output is generated from the
I2S bus data by a Crystal Semiconductor CS8402 S/PDIF
transmission chip. Separate 7805 subregulators are used
for the main digital circuitry, the PLL, and the data re­ceiver and transmitter.

The process of receiving the S/PDIF data and then
retransmitting it is claimed to reduce jitter levels when the
signal is ultimately decoded in another D/A decoder box.
While this approach has the potential to reduce the jitter
levels of the auxiliary decoder box, it cannot eliminate jit-

ter completely. To understand what the DTI can and can­not do we need to introduce some basic concepts of dig­ital data transmission. Given the limited space I have to
do this here, and to prevent the MEGO (My Eyes Glaze
Over) effect, we will use very simplistic models.

For a digital data stream to be meaningful, it must

be accompanied by a clock signal which partitions each
data bit cell. Thus a minimum of two wires is required,
one which contains the clock and one which contains the
data. To represent the clock and the data in a single wire,
an encoding method must be used. The method used in
the S/PDIF format is called biphase mark. In this en­coding scheme, a clock transition always occurs at the be­ginning of a bit cell. If a 7 is to be transmitted, a clock
transition occurs in the middle of the bit cell as well. Note
that even if a long string of 0s or 7s is transmitted, at least
one clock transition will occur during each system clock
period. These added clock transitions allow the S/PDIF
decoder to recover the system clock from the single wire.
One method to decode the reference clock is to use a
PLL. A PLL can be thought of as a flywheel. The en­coded digital signal sets the speed of the flywheel. The

speed of the flywheel's rotation represents the PLL's out­put—the recovered clock. One rotation represents one
clock cycle. The flywheel will continue to operate at the

clock frequency even if the clock transitions are not al-

ways present. Now we have three sources of error: (1)

Short-term changes in the speed of the flywheel can occur

when the clock transitions are absent. Recall that the

number of clock transitions per cell is data-dependent. (2)

A jittery incoming data stream can cause short-term
changes in the flywheel's speed. (3) The flywheel's speed

may change randomly as a result of deviations from ideal

operation. (Flutter would be an example in our mechan-

ical analogy; a noisy VCO in the actual PLL.) Now, error

sources (1) and (3) occur in the PLL of the S/PDIF de-

coder independently of the jitter in the incoming data

stream. Only error source (2) can be reduced by reducing

the jitter in that data stream. Error sources (1) and (3) can
only be reduced by improving the performance of the S/

PDIF decoder.

Please note that the encoded S/PDIF signal is not a
pure digital signal. The performance of the S/PDIF de­coder can be degraded if the signal is distorted by the data
transmitter, through the S/PDIF cable, or at the data re­ceiver. It is thus possible to see performance degradation
in some S/PDIF decoder designs if the communication

ISSUE NO. 19 • SPRING 1993

channel is not optimum. Inexpensive solutions are avail­able to make sure that the S/PDIF signal is not distorted.
Very expensive glass fiber or coaxial cables are not re­quired for optimal performance. Better S/PDIF decoders
are less sensitive to waveform distortions in the received
S/PDIF signal. As I have explained previously, the best
solution is to eliminate the problems associated with the

single-wire encoded signal and use a two-wire system.
This approach is used in the newly introduced Denon
DA-X D/A converter. Audio Alchemy had the option to
do this too, since they also make a CD transport, but they
chose not to do so.

The XDP (Extended Digital Processor) converts the
I2S bus data into an analog signal. An earlier Audio Al­chemy product, the DDE, also generates the I2S bus data
required by the XDP. First the I2S data is passed to a
Burr-Brown DF1700P digital filter. (This is a remarked
version of the NPC SM5813. See Issue No. 15 for more
details on the NPC part. NPC has just introduced a sec­ond-generation filter chip, the SM5842AP. More details
on this chip will appear in the next issue.) The output of
the digital filter goes to the Philips SAA7350 delta mod­ulator D/A chip (again, see Issue No. 15 for more details
on this chip). This chip generates a one-bit data stream,
which is filtered by the Philips TDA1547 chip (see Issue
No. 16, page 45, for more details on this chip). The output
of the TDA1547 is balanced. It is converted to a single­ended signal by a discrete six-transistor operational amp­lifier stage. Finally, the signal passes through a second­order Sallen-Key filter which uses a two-transistor unity­gain buffer. The buffer is an npn emitter follower biased
by a current stage formed with a pnp transistor. Separate
supply subregulators are used for the digital chips (a
78M05) and the TDA1547 (two 79M05s, one for the an­alog and one for the digital section of the chip, and a
78M05.), and separate regulators for the left and right
channels of the discrete output stage (78M15 and
79M15). This brings the total number of regulators to 15!
All this circuitry just fits on the double-sided PC board.
The board is clearly not the product of a garage operation,
since the parts are autoinserted and the board has plated­through holes (the DTI board has similar construction).
The de-emphasis is performed by a relay. Power-up tran­sients are passed directly to the output, since no muting
circuit is included. Small noises would also occur at the
output of the unit under some conditions when the
S/PDIF signal was interrupted. LEDs on the front panel
indicate lock and de-emphasis. It is also possible to put
the decoder in the de-emphasis mode with a front panel
switch. The DDE device does not have the extra de­emphasis code at its I2S bus output; thus the de-emphasis
must be engaged manually when the DDE is used.

Measured performance of the DTI/XDP combo was
a for the most part very good, but a major problem was
identified in the distortion tests. The frequency response
was down by 0.3 dB at 20 kHz. For a disc with de­emphasis, the frequency response is down 0.5 dB at 20

43

pdf 38

kHz. Channel separation was 120 dB or better below 200
Hz. It then decreased at a 6 dB per octave rate to 80 dB at
20 kHz. With a bigger chassis a better result might have
been possible at 20 kHz. Noise spectrum analysis showed
no hum components down to -125 dB. This shows that
the external power supplies and the extensive regulation
are achieving the desired result. Gain linearity remained
within ±0.2 dB down to -100 dB. Full-scale THD + N

proved to be disappointing; from 20 Hz to 1 kHz it was

-87 dB and it then rose to a maximum of -77 dB at 10
kHz before dropping again as the harmonics began to fall
into the stopband of the reconstruction filter. Reducing
the signal level by 20 dB resulted in a reduction of the
THD + N to -95 dB relative to a full-scale signal. (Refer­enced to the fundamental signal level this figure differs,
of course, by 20 dB.) At the reduced signal level the an­alog output stage is no longer contributing distortion com­ponents to the output signal. Noise from the analog elec­tronics and nonideal performance of the DAC itself
account for the remaining 3 dB deviation from the theo­retical minimum of -98.08 dB. The poor full-scale per-

formance is very likely the fault of the two-transistor buf­fer circuit. Perhaps a more complex buffer could not be
fitted into the small enclosure.

The distortion of the buffer circuit prevents an un­conditional recommendation of these Audio Alchemy
units. The performance was more than adequate to insure
that the unit has no audible colorations, and ABX test re­sults confirmed this. If all the Audio Alchemy units were
combined on one 19-inch chassis, then the total cost of
the unit could be reduced and the buffer stage could be
improved. Such a unit would receive a high recommenda­tion by this journal.

EAD DSP-7000

(follow-up)

(By the original reviewer, Peter Aczel)

Enlightened Audio Designs Corp., 607 West Broadway, Fair-

field, IA 52556. DSP-7000 outboard D/A converter, $1399.00.

Tested sample on loan from manufacturer.

The main reason for this follow-up is the muddle­headed, irresponsible review of the EAD DSP-7000 by
Robert Harley in the September 1992 issue of Stereophile
(Vol. 15, No. 9). Believe it or not, the man reports 20 dB
more noise across the audio spectrum with a digital zero
input than with a dithered 1 kHz input at -90.31 dB.
That's like saying that the rainfall was 2 inches more un­der the roof of the carport than in the driveway. On top of
it, he actually notes that this is peculiar and then—
standing there with his bare face hanging out—he com­ments, "Whether this correlates with the disappointing
sound quality remains to be seen." Reading that rubbish I
had little doubt that what it correlates with is Harley's
questionable qualifications as an audio equipment re­viewer, but just to prove my point I thought, what the

44

hell, I'll ask the nice people at EAD to send me another
sample of the DSP-7000. This time I got a factory-sealed
box out of production stock, containing a unit with the
more restrained black front panel I had wished for.

Well, what do you know, when I measured the new
unit on the Audio Precision "System One Dual Domain,"
the noise spectrum from 30 Hz to 200 kHz with digital
zero input ranged, in the less good channel, from -140 dB
on the bottom end to -117 dB at 20 kHz and -93 dB near
the limit on top. That's 34 dB better on the bottom, 27 dB
better at 20 kHz, and at least 13 dB better in the ultrasonic
region than Harley's less good channel. Let's face it, Bob,
old buddy, you messed up bigtime. There's nothing
wrong with the noise floor of the DSP-7000.

While I was at it, I investigated the THD + N versus
frequency performance of this unit in somewhat greater
depth than in my original evaluation (see Issue No. 17). It
turns out that the digital circuitry is capable of literally
perfect 16-bit resolution but the analog circuitry intro­duces a teensy bit of nonlinearity at maximum output.
The evidence is that with an input of -20 dB, the THD +
N as normalized to 0 dB (full scale) reads almost exactly
the theoretical minimum of -98.08 dB across the entire
audio spectrum, up to where the analog filter begins to
roll off the output. With a 0 dB input, however, the read­ing ranges from -95 dB at 20 Hz to -90 dB between 3
kHz and 11 kHz, and that difference has to be the con­tribution of the analog circuitry. These results are slightly
better than what I obtained with the first sample; the new­er production unit appears to be free from the 60 Hz prob­lem I reported at the time. (Stereophile, by the way, never
reports THD + N versus frequency in digital playback

equipment. No test is more important, but the ultrahigh­end stuff doesn't always pass it with flying colors, so it

becomes unmentionable.)

Gain linearity in my new sample was off by +0.5
dB at -80 dB and +1.5 dB at -90 dB; however, as David
Rich explained long ago, gain linearity specs must not be
confused with the more important and stringent integral

and differential linearity criteria. Channel separation
ranged from 130 dB at 20 Hz to 82 dB at 20 kHz, chang­ing at 6 dB per octave. (Again, Harley's measurements
were considerably worse.) A minor flaw I overlooked in
my original review was the not quite dead-accurate de­emphasis, with -0.2 dB error at 10 kHz and -0.4 dB at 16
kHz. That, however, includes the inherent top-end rolloff
of the unit, which is almost certainly due to the analog
filter and amounts to -0.07 dB at 10 kHz and -0.2 dB at

16 kHz. I could not hear any "softening" of the highs as a
result; it takes a larger fraction of one dB up there to be
reliably audible.

Thus the EAD DSP-7000 can be declared to be
close to perfect in the digital domain, only ever so slightly
imperfect in the analog domain, and unusually full­featured as well as nice to look at in the user domain. It
matters very little to me at this point exactly how much of
a "breakthrough" the so-called AccuLinear I-to-V con-

THE AUDIO CRITIC

pdf 39

verter is. The radiated RFI problem, on the other hand, is
still as bad as ever; my FM tuner, sitting four feet from
the DSP-7000, chirped like a bird when the latter was ei­ther on standby or fully turned on, but especially when on
standby. The remedy was to pull the plug.

The next version of the EAD DSP-7000 will use the

20-bit Burr-Brown PCM63P-K DAC instead of the 20-bit
Analog Devices chip in the original design. An upgrade
kit is also being made available. I expect the low-level

gain linearity measurement to be essentially perfect with

the new chip, among other benefits.

One other interesting little sidelight. EAD's re-

search director, John Hagelin, the Harvard Ph.D. who
originally conceived the design of this D/A processor,
was the presidential candidate of the Natural Law Party in
the recent national election. You may have seen the TV
commercials. The Natural Law Party advocates, among a
number of other things, transcendental meditation as a so­lution to the problems of this messed-up world. And that's
not all. Dr. Hagelin currently teaches at The Maharishi
International University in Iowa, where transcendental
meditation is the unifying discipline under the leadership
of the Maharishi Mahesh Yogi. Alastair Roxburgh, the
Englishman who is engineering director of EAD, is also
associated with the university. Now, this is an audio mag­azine and not a forum for the discussion of political or
spiritual matters, but isn't this fascinating? Here is a far­out group with a pretty extreme belief system—and I say
that merely as an outside observer, without any judg­mental implications whatsoever—and yet they are doing
solid engineering work based on mainstream science
without tweako distractions. How strange, then, that there
are Episcopalian Republicans out there who believe in
CD rings and silver cable!

Monarchy Audio Model 22A

Monarchy International, Inc., 380 Swift Avenue, Unit 21, South
San Francisco, CA 94080. Model 22A Dual 20-Bit D/A Convert­er, $1200.00. Tested sample on loan from manufacturer.

This unit represents one of the first affordable D/A
processors to use the Burr-Brown PCM63P-K multibit
DAC. Of all available monolithic DACs, this chip has the
lowest specified distortion levels as published in its data
sheet. A novel topology (see Issue No. 16, page 49) al­lows the DAC to have excellent linearity around digital
zero. It is also used in the $4000 Theta DS Pro Genera­tion III. Monarchy hired a team of highly qualified con­sultants to aid in the design of this unit. Different con­sultants were involved in the design of the S/PDIF
decoder, the DACs' peripheral circuitry, and the analog
section. The PC board is double-sided and the board has
plated-through holes. The components on the board are
autoinserted, and modern surface mounting is used for the
S/PDIF decoder. Extensive attempts to limit RFI radia-

ISSUE NO. 19 • SPRING 1993

tion, including an AC line filter, are part of the design of
the 22A. Sheet metal work is of a high quality. A phase
inversion switch is on the front panel, but there is no lock
or de-emphasis indicator. In addition, the unit has no
power switch. The unit accepts both optical and coaxial
inputs, switchable from the front panel.

The power supply consists of separate analog and
digital power supplies. The digital transformer has two
secondaries. Separate button-sized full-wave rectifiers
and 15,000 µF capacitors are used to for the positive and
negative digital supplies. A 78M05 regulator is used to
supply the digital circuits. Separate 78M05 and 7905 reg­ulators drive the PCM63P. The analog supply consists of
another button regulator and two 8200 µF capacitors on
the unregulated supply rails. An RC1515 dual-tracking
regulator is used for the ±15 V supplies. I am not familiar
with this part but it appears similar to the Raytheon
RC4195. A separate 78M05 regulator connected to the
analog transformer is used to power the analog supply of
the S/PDIF decoder. All regulated rails have either 330
µF or 100 µF capcacitors connected to ground.

The S/PDIF decoder is a new second-generation
chip from Yamaha, the YM3436C. The first-generation
Yamaha chip, the YM3623B, is often maligned by Robert
Harley in Stereophile. Contrary to the assertions by Mr.
Harley, the YM3623B can provide good performance if
properly applied (the Yamaha CX-1000 preamplifier is a
good example). The preliminary data sheet for the
YM3436 does not give specifications for its recovered
clock jitter. The output of the YM3436C is passed to a
Burr-Brown DF1700P digital filter. The current-to­voltage converter is the new Analog Devices AD811
high-speed transimpedance amplifier (2500 V/µs slew
rate, 60 dB PSRR at 100 kHz, 65 ns settling time to

0.01%, and 140 MHz bandwidth). To get this speed, the
chip burns a half watt, and Monarchy Audio uses a heat
sink on the chip.Two PMI OP275 op-amps (9 MHz
bandwidth, 7 nV/ Hz voltage noise at 30 Hz, and a slew­ing threshold of 0.4 V) are used in the analog filter stage.
As explained in Issue No. 18 (page 37), this op-amp uses
a composite bipolar/JFET input stage, which attempts to
combine the low noise of a nondegenerated bipolar input
stage with the high slewing threshold of a JFET op-amp.
While the OP275 is a good op-amp, I would like to see a
state-of-the-art op-amp such as the AD797 used instead in
the 22A, given its high price point. The OP275 opera­tional amplifiers are used to form a GlC-based (gener­alized impedance converter) reconstruction filter. A Burr­Brown application note has shown that this filter topology
has the potential for slightly lower noise levels in com­parison with the Salien-Key topology, but it is nearly
twice as complex. The gain of the filter stage can be
changed by 6 dB by changing shorting links on the PC
board. An open-loop discrete buffer follows the filter. I

prefer the closed-loop arrangement used in the same com-

pany's Model 10 preamp to this open-loop approach. A
DC servo prevents direct current from appearing at the

49

pdf 40

output of this unit. Another OP275 is used to drive the

balanced outputs. A transistor switch connects the passive
de-emphasis network. No muting relay is used at the out­put, a significant omission in my opinion. Power supply

transients were small, however, and surprisingly no tran-

sients occurred when the S/PDIF signal was connected or
removed in a variety of nefarious ways.

The measured performance of the Model 22A was a
mixed bag. The frequency response was tipped up by 0.1
dB at 20 kHz. For a disc requiring de-emphasis, the maxi­mum frequency response error was +0.2 dB. Channel sep­aration was 110 dB or better below 3 kHz, then decreased
to 98 dB at 20 kHz. Noise spectrum analysis showed sig­nificant power-supply components at multiples of 60 Hz;
the 300 Hz component was particularly strong at 96 dB
below full scale.. In addition, the overall noise floor ap­peared higher than required to achieve 16-bit signal-to­noise ratios. Gain linearity remained within ±0.4 dB
down to -100 dB. Full-scale THD + N proved to be dis­appointing. From 20 Hz to 20 kHz, it remained flat at -85

dB. Lowering the input signal level by 20 dB or even 40
dB does not change this result. This indicates that noise,

not the distortion, is causing the degradation in signal-to­noise ratio. All analog components used in the 22A are
specified to have noise levels significantly lower than
what we measured; thus I cannot explain the origin of the
noise.

The disappointingly high noise level prevents a
strong recommendation of this unit. Monarchy is in­vestigating the cause of the noise and will hopefully be
able to lower it to acceptable levels, since the rest of the
design appears to be excellent. A follow-up review will
be forthcoming if a revised unit is made available for test.

UltrAmp D/A Converter

Mobile Fidelity Sound Lab, 105 Morris Street, Sebastopol, CA

95472. UltrAmp D/A Converter, $1295.00. Tested sample on
loan from manufacturer.

This is one of the three units in the Michael Yee de­signer collection (his signature is on the back of each
unit). This processor accepts only coaxial inputs. It has a
phase inversion switch on the front panel. For some un­known reason, the power switch is located on the rear of
the unit but at least this function is included. The PC
board is a high-quality double-sided board with plated­through holes. Two inexpensive transformers are used for
the analog and digital supplies. Three DIP-sized full­wave rectifiers are on the board. A single 7805 regulator
services all the digital circuitry. A separate 79M05 gener­ates the DAC's -5 V supply. All analog circuitry shares
the same master-slave topology regulator; 7818 and 7815
devices are used in the positive regulator; 7918 and 7915
devices are used in the negative regulator; 2200 µF filter

capacitors are used on the regulated and unregulated rails.
A muting relay circuit is included in this design but it
does not work correctly. Power-down transients are still
passed directly to the output. This is the same problem we
also had on Mr. Yee's preamp.

The S/PDIF decoder is the new Philips SAA7274P.
Unlike other S/PDIF decoders, the SAA7274P is de­signed with much of the analog circuitry for it off the
chip. This gives the potential for the generation of a very
low-jitter word clock. The phase detector and loop filter
are fully balanced, increasing the PLL's ability to ignore
power supply noise. Some of the advantage is lost in Mr.
Yee's design, since a single regulator is used to service all
circuits operating at 5 V. The high point of the Philips de­sign is the crystal-based voltage-controlled oscillator
(VCXO) that can be implemented with the SAA7274P.

The high effective Q of the crystal insures that the VCO

will contribute a very small amount of jitter. Mr. Yee

does not use the crystal, even though a space was created
for it on his PC board. An inductor replaces the crystal in

a last-minute board change. The digital filter is the Philips

SAA7220P/B, and the DAC is the Philips TDA1541A S1
"Golden Crown." These old parts were last used in state­of-the-art CD players at the end of the Reagan administra­tion. Mr. Yee says they "sound better" than modern chips.
The I/V converter and analog filter stages appear to be
similar in topology to the poorly performing circuitry
used in the UltrAmp preamp. We never did get schemat­ics from UltrAmp.

The measured performance of the UltrAmp pro­cessor was worse than what we would have expected to
measure in a $200 mass-market CD player. The fre­quency response starts to decline at 2 kHz and is down by

2.5 dB—yes, 2.5 dB!—at 20 kHz. This is of course quite
audible. Channel separation was 100 dB below 1 kHz.
Above that frequency it decreased at a rate of 6 dB per
octave to 70 dB at 20 kHz. Noise spectrum analysis
showed significant hum components. The 60 Hz com­ponent was at -88 dB relative to full scale and the 180 Hz
component was at -93 dB. Gain linearity error was -1.5
dB at -80 dB and -4 dB at -90dB. This is an out-of-spec
result for a "Golden Crown" TDA1541A S1 DAC. Ap­parently no incoming QC is done by UltrAmp. Full-scale
THD + N was -80 dB from 20 Hz to 10 kHz and then
rose to a maximum of -67 dB at 20 kHz. This is equiv­alent to the distortion at 20 kHz that would result if the
data were encoded to just 11 bits. Would you want your
name to be on the back of this unit if you designed it and
it performed this badly?

In the last issue we promised a listening evaluation
of the complete three-component UltrAmp hookup, but
this was abandoned when the frequency response errors
in this unit were discovered. The UltrAmp D/A converter
is a very expensive treble control. Clearly, for all of the
foregoing reasons, it cannot be recommended. •

50

THE AUDIO CRITIC

pdf 41

Editor's Note: Tom Nousaine,
strange visitor from the planet of the
Southeast Michigan Woofer and

Tweeter Marching Society, came to

the audio world with objectivistic

powers and abilities far beyond those

of mortal audiophiles. Disguised dur­ing working hours as a mild-man­nered specialist for a great metropoli­tan telephone company, he fights in
his spare time a never ending battle

for truth, justice and the scientific

way of looking at audio. The plan is
to make this column one of our regu-

lar features. Our readers have prob­ably seen Tom's name in a number of
small audio journals; as a matter of

fact, he has told the "Julian story" in

print before, but this is the complete,

unabridged, philosophically append­ed version. It's a story every audio-

phile should be exposed to, no matter

where he sees it first.

* * *

I finally realized at the last CES
that I no longer want to be called an
audiophile, tweak or purist. Those
neat terms, once an endearment to
me, have been usurped by people
who want us to spend needless thou­sands on unproven spikes, wires, pil­lows, clamps, clocks, tubes and other
weirdness. The same people violently

react when asked to prove their
claims.

Let me expand. I will do prac­tically anything to improve my sound
system. A quick look at my eight­channel (10 channels of ampli­fication) Lexicon CP-3 system with
23-driver dipole line arrays, 22­cubic-foot tubular 18" subwoofer,
and 5-way rear surround speakers
will attest to this. But, I still have to
take abuse from friends, the under­ground press and people hawking ri­diculous products and ideas, about
"refusing to just listen for myself
and being "insensitive" or "uncaring
about music." I beg your pardon?

52

When the Evil Gunslinger said
"I'm faster than you" to the Sundance
Kid, the Kid said "Draw, sucker!"
When Bob Falfa said "I'm aimin' to
blow his [John Milner's] ass off the
road," Milner dropped the clutch.
Why then do audio salesmen and
manufacturers get away with not hav-

ing to verify their claims?

Oh, people have "just listened"
and heard the news you say? Yeah,
well why does the magic disappear
when the blindfolds appear? Why
can't some audiophiles prove their
claims when listener bias controls are
employed? When that happens I say
it's because the claims aren't real.

Now, many illusions that that go
away when the lights come on can be
considered to be real. Stereo for ex­ample is a perfect example of a "real"
illusion. The effect can be duplicated,
remains perceptible under controlled
conditions and can therefore be
verified to others. Anything that can't
isn't real enough for me to waste
time, money or energy chasing.

And I am getting impatient with
those who don't require verification
from the proponent or who want to
vilify me for asking for it. It just ain't

my job to prove the Golden Ear
claims. (Even though I have often
tried!) If they can't or won't, that's
their problem. Would you let a sports
car salesman sell you the "fastest car
in the valley" if he refused a race? Or
claimed that a stopwatch made him
too nervous to prove it?

Speaking of proving it, here's a

neat little story about a guy named
Julian [not Hirsch—Ed.] who had the
balls to put his beliefs on the line. In
the fall of 1991 he told me about how
great his new Sumo Andromeda II
sounded compared to the Adcom
GFA-555 it had replaced. I told him
the Sumo was probably a fine piece
but I doubted that it sounded much
different from the Adcom or any oth­er quality amplifier. Push came to
shove and Julian accepted my offer
of an ABX double-blind comparison
at a Prairie State Audio Construction
Society (PSACS) meeting in Novem­ber 1991.

The Sumo was matched against
my $200 (used) Parasound HCA­800II 80-watt-per-channel solid-state
amplifier in a "short" system of just a
Sony D-15 CD player, the amps, an
ABX switchbox and a pair of Dahl-

quist DQ-lOs. Julian felt the audio
character of the Sumo was apparent

under open conditions prior to start­ing. He picked the recordings and
controlled the switchpad (which al­lows an unlimited number of com­parisons and unlimited time for each
decision or trial) for the first three
trials.

After nine decisions he quit the
test. Other audiophiles completed the
remaining three of the 12 total trials.
Results: Julian correctly identified
amplifiers 2 out of 9 times (not even
once out of the three trials where he
picked the records). Best score was 7
out of 12, and the overall correct rate
was 48% of about 100 trials. I con­cluded that when levels were care­fully matched in each channel, the
Sumo and Parasound sounded exactly
alike.

Julian was not impressed. After
further discussion he felt the test con­ditions must have somehow inter­fered with his ability to tell amplifiers
apart. He felt this way even though

he had been able to hear the character
of the Sumo in the test setup under
open conditions and was allowed to
tell the other listeners what they
should have been hearing. Anyway,
Julian got another chance.

On April 18, 1992, I took the

Parasound HCA-800II and the ABX
switchbox and relay module to Ju­lian's place. I installed the ABX ma­chine and the Parasound in his system,
which included a NYAL hybrid pre­amplifer, JSE Infinite Slope speakers,
$700 worth of interconnects and Tara
Labs speaker cabling, in addition to
the Sumo.

Julian then had as long as he
needed to (1) determine how these
amplifiers sounded different from
each other, (2) determine if the ABX

equipment interfered with the revela­tion of those differences and (3) se­lect music programs which high­lighted the differences. By May 16 he
had concluded that the amplifiers did
sound different from each other and
that the ABX equipment did not ob­scure those differences, and he had
selected CDs and LPs that high­lighted those differences.

On Saturday, May 16, 1992, I

adjusted the level controls on the
Parasound to match the Sumo within

0.1 dB at 1 kHz in each channel. I
also verified that each of the ampli-

THE AUDIO CRITIC

pdf 42

fiers was flat within 0.2 dB from 20
Hz to 20 kHz. No other controls were
changed and nothing on the primary
system was touched.

We agreed that if he could cor-

rectly identify amplifiers 12 out of 16

times using any music he wanted,

controlling all the switching and the

duration of trials, I would agree that
he was right, i.e., the amplifiers were

indeed sonically different. We also
agreed that the results, no matter
what the outcome, would be sub­mitted for publication.

The test began at approximately 1
p.m. The equipment had been turned
on the day before and allowed to
warm up overnight. I suggested Ju­lian start with the music selectio­where the differences were most ap­parent. He decided to use the phrase
repeat on his CD player for the first
selection, so he could use the same
music passage for identification. That
first trial took about 15 minutes.

Twelve minutes into trial number
two, Julian looked up and said, "I
have to admit I just can't tell them
apart." I asked him for at least 10
trials just to give us enough data to
analyze and to assure himself that he
didn't leave out any music material
where differences indeed could be
heard.

Another hour and ten minutes
later we had 10 trials. There was no
pressure to hurry. Julian had com­plete control over program selection
(including LPs), switching intervals,
switching direction and trial duration.
And he'd had four weeks of practice
with the switchpad.

He correctly identified amplifiers
5 out of 10 times or at the rate ex-

pected if he were just guessing. And

he was just guessing by his own ad­mission. He didn't get either of the

first two right, by the way. Thus, a
confirmed longtime audiophile, choos­ing his own music, controlling every

variable in his personal reference sys­tem with a month-long warm-up, was
unable to correctly identify his $1599
Sumo Andromeda II compared with a
$200 (used) transistor-piece-of-crap
Parasound HCA-800II when the face­plates were, in effect, removed.

Now there were two small but in­teresting technical differences between
these amplifiers. The Sumo had a 2
dB channel imbalance. Further, the
voltage sensitivities of the two amp-

lifiers were markedly different (0.7
volts for the Parasound and 1.3 volts
for the Sumo). But Julian didn't hear
volume differences or channel imbal­ances. He reported quality differences
in sound between the two.

The moral of this story is that a
couple of dozen blind amplifier com­parisons have pretty much proven
that a modern, reasonably well­designed amplifier, with reasonably
flat frequency response and operating
within its power capacity, will sound

just like any other well-designed

amplifier. This latest test shows that
amplifier differences apparent under
the best possible long-term compari­son conditions disappear when vol­ume differences and channel imbal­ances are dialed out.

Becoming a geek is really fun.
You don't have to chase those fragile
differences that disappear when you
close your eyes or the "coach" leaves
the room. Now you can devote time,
energy and money to system up­grades that can be demonstrated to
others even when you're not there.

Today that means better speakers,
better recordings and DSP surround

systems. Hey, man, it's freedom. Try

it out.

* * *

Postscript by David Rich:

The experiences of Tom Nou-

saine in ABX testing are very similar
to my own. I conducted my tests with
dealers and manufacturers. Profes-

sionals in this field should, in theory,
be able to make an accurate evalua­tion of the sound of a component. A
custom ABX box was used for these
tests. This box differs from the stan­dard David Clark ABX box in some
respects, one of which is self­contained level matching with a built­in set of passive attenuators. This one­off custom unit was built to tweako
standards, including megabuck jacks,
MIL-spec potentiometers, and Teflon
wire. The ABX box was given to the
test participants for use in the par­ticipants' own room with their own
equipment. The participant was al­lowed as much time as necessary to
become used to the ABX box and to

find the most revealing program ma-

terial.

At the start of the experiment all

participants found the ABX box to

have a minimal effect on the per-

formance of their system. All par-

ticipants felt confident that they could
identify the differences between the
equipment when the switch box was

set to A or B. After the participants

had worked with the ABX box for

about a week, they would report that

they were having more difficulty than
expected in the blind test, but they re-

mained confident they would even-

tually be able to pass the test. As time

passed, the participants had to admit

they could not achieve a statistically
significant result. No participant
agreed that this indicated that the
components sounded identical; in­stead they blamed the ABX box for
hiding the differences. They now
claimed that the ABX box was so col­ored that differences could only be
heard when the box was out of the
system. They were unwilling to ac­cept the evidence of their own ears.

They had to find a scapegoat, and the

ABX box was the only thing available

even though they originally had no
objection to it.

Nostalgia and Loudspeakers (continued from page 33)

rock artist to be emasculated, you can go out and get an
Aphex Aural Exciter to add distortion back in, so that it
sounds loud again. (Seriously.)

In summary.

If you are a true high-fidelity junkie (not an "audio-

phile"), if you have an insufferable doctor friend with

Mcintosh equipment who needs to learn a lesson about

ISSUE NO. 19 • SPRING 1993

the assumptions made by (mass-market) speaker manu­facturers, or if you're simply tired of commercial offerings
available to just anybody, you can have the peace of mind
that comes from knowing that you have built something
truly impressive, based largely on technology ignored for
thirty years because of economic considerations.

[Detailed construction diagrams are available from

the author. Write to him c/o The Audio Critic.—Ed.] •

53

pdf 43

Interviewing the

Best Interviewees in Audio

Part II

By David Ranada

Contributing Editor at Large

Here are two more of the acknowledged deep thinkers in the field,

sharing with you their highly original and enlightened insights into

the present state and future promise of audio technology.

Editor's Note: It so happens that The Audio Critic has a
history regarding both interviewees here. Bob Carver has
been an editorial stormy petrel in these pages for years
now, as most of our readers know. The color of his hat—is
it good-guy white or bad-guy black or possibly pearl
gray?—has been debated endlessly by audiophiles in gen­eral and our correspondents in particular. I think David
Ranada manages to draw out the essence of the man in
his interview, allowing Bob's tremendous enthusiasm and
overflowing creativity to come through loud and clear.

Mark Davis, on the other hand, is hardly ever men­tioned in our articles, reviews, or correspondence, although
he is certainly a brilliant practitioner in his own right.
Long, long ago, however, he occupied center stage in just
one letters-to-the-editor column, and the treatment he re­ceived there at my hands has been preying on my mind for
the last few years, threatening to become a major guilt
trip unless I expiate it right here and now. What happened
was that Mark wrote to The Audio Critic in 1977 that any

two competently designed preamplifiers with identical fre-

quency responses will be audibly indistinguishable from

each other at matched levels. He had extensive research

and experimental data under his belt to support that

claim, and he complained that "I'm beginning to think it

unfair that I should be the only one to have to carry the
burden of proof." And how did I answer him? I made
cruel fun of him, that's how. I just knew that the Mark Le­vinson had to sound better than the Dynaco, Yamaha, etc.

So let's get this straight once and for all, Mark. You

were right, and I was wrong. You were objective and unin-

fluenced by tweako belief systems; I was still thinking like

a typical audiophile on many (though not all) subjects. I
hope you have meanwhile forgiven me. More recent issues
of The Audio Critic explain in detail my current per­ception of reality, and today I refuse to defend some of the
views expressed in those earliest issues. I still don't like
cheap preamps but not because of the sound. (See also
David Rich's comments on that subject in Issue No. 18.)

5. Interview with
Mark F. Davis,
Audio Designer

RANADA: When I first met you, more
than 15 years ago, you were already deep­ly into audio and psychoacoustics. How
did you get that way?
DAVIS: I was into electronics as a kid.
RANADA: Just electronics or audio­related electronics specifically?
DAVIS: It started out as just electronics, I
thought. I took a long electronics course
in 10th, 11th, and 12th grade. But I guess
I started getting into audio when I was
five. At the age of five I can recall a little
girl down the block having her phono­graph break and asking me to fix it. And I
remember taking it apart, not knowing
what the hell I was doing. But it looked

ISSUE NO. 19 • SPRING 1993

pretty with all the tubes.
RANADA: Why did she ask you to fix it?
DAVIS: Because I was interested in
audio.
RANADA: Did you fix it?

DAVIS: No. At that point I didn't know
that it was really broken. I think we were
trying to figure out where the singers
came out of. There was this little book
called Basic Electronics that I took out of
the library at one point. A lot of it had to
do with audio because there just wasn't
that much RF yet. My father had this
enormous collection of old records, and

in high school I would play them by just

holding a phono cartridge in my hand and

had the wires from the phono cartridge

run to a crystal earphone—no amplifiers,

very pure reproduction.
RANADA: And no de-emphasis either...

DAVIS: They used pre-emphasis in 78-

rpm records?

RANADA: Yes, they did.
DAVIS: In this case, it was whatever you
got was what you got.
RANADA: The case of a human tonearm,
literally.
DAVIS: Exactly. Later on we had a Lafa­yette Radio in Syosset on Long Island. I
used to go up there by bicycle on week­ends and spend far too much money on
stuff. I was very fascinated by tape re­corders. I tried to build a tape recorder for
a science project in sixth grade. It didn't
work.

RANADA: Why was that?
DAVIS: Because I didn't realize that a
tape head had to have two pole pieces that
came around like a U magnet and came
close together. I was trying to use an iron
nail, which didn't work nearly as well. I
couldn't find much written about tape re­corders back then and I didn't know any­body who was particularly expert in

55

pdf 44

them.
RANADA: When was this?
DAVIS: Sixth or seventh grade, around

1958. It's too bad I didn't live out here in
Redwood City where Ampex was because
I could have sat around their offices and
asked questions. But this was all that Le­vittown had to offer.
RANADA: So you were destined for an
electronics career at an early age?
DAVIS: So it would seem. I did see to it
that I got into this electronics course in
high school. And then I went to MIT and
majored in electrical engineering and so
forth. I tried to learn how to design cir-

cuits—because MIT doesn't like you to

learn how to design circuits. They want

you to learn the principles behind the de­signing of circuits. I have now come
around to their point of view but at the
time I wanted to learn how do design cir­cuits.

RANADA: How did you get into psycho­acoustics? There isn't necessarily a con­nection between electronics and psycho­acoustics.
DAVIS: Absolutely. Through under­graduate years I had a strictly under­graduate curriculum, which had nothing
specifically to do with audio.
RANADA: The basic EE kind of stuff.
DAVIS: Yeah, pretty much. I think there

was this assistant associate professor,
Barry Blesser, who gave a course on prac­tical audio design. I think I took that.
RANADA: Did you ever take Bose's
course?

DAVIS: Yeah, somewhere along the way
I took Bose's loudspeaker design course.
It was a very good course. Audio was still

my hobby, and there was this local hi-fi
shop—Tech Hi-Fi—that I worked at part­time as a salesman and so on. But it was
strictly a hobby. I was meandering down
the hall one day with my master's thesis
supervisor-to-be, Campbell Searle. He
brought up some problems he was having
with his hi-fi system, and we got into a
long, involved conversation about his hi­fi. And I somewhere sort of brushed off
audio as this silly side-hobby that one
doesn't really talk about in the halls of
MIT. Campbell said, "No, no, no, it's a
perfectly serious and reasonable area of
study" and so on, and that I should take it
seriously. That was at lunchtime. I went
home and over lunch hour I typed up a
four-page critique of his hi-fi system and
brought it back to him. He had made some
Klipschorns at home. When Klipschorns
had first come out, they were very pop-

ular. There had been a little group of MIT

students—this was way before my time—
who had made Klipschorns, so I think he
still had a couple of them. And he was

biamping them with these little Heathkit

amplifiers, with 30 watts per channel or
something like that.

RANADA: That should be enough for a

Klipschorn.
DAVIS: And they were biamped, so they

could really put out a lot of sound. But I
think I had critiques about how he was

setting the relative levels of his tweeters
and his woofers, and how he was doing

the crossovers and things like that. That

56

sort of started a more-or-less serious asso­ciation between Campbell and me, ex­ploring audio.
RANADA: He was a psychoacoustician?
DAVIS: He had been a straight circuit­theory guy. He and Paul Gray, who for a
time was a president of MIT, had written
a book on circuit theory, a great big book
that was known to everyone as "Gray and

Searle." Having done that, he reached a
point where it was evident to him that cir­cuitry per se was going to become very
"cookbook," and something that only
chip designers worried about, where you
manipulate the stuff using computer cir­cuit-design programs. Since it [circuit de­sign] was clearly drying out as a major
area of research, he decided that he need­ed to shift his academic focus and pro­ceeded to go into psychoacoustics, where
the intent was to apply the sort of signals­and-systems theory that had evolved in
the electronic disciplines to human hear­ing—to treat [hearing] as a signal-pro-

cessing system and then try to character-

ize it as best as possible to try to establish
causative relationships between what
sound went in your ears and what the re­sulting perception was.

RANADA: This is a very different way of

looking at perception, as sort of an an-

alogue of an electrical circuit.

"Each person's pinna
characteristics were very

individual.... If we gave

somebody a new set of
pinnas, in some amount of
time they'd ' learn' [how to
use] them."

DAVIS: Exactly. You are not caring pri­marily about the biological functions or
anything like that. You want to know how
the whole thing acts as a system. And al­though tracing things down to basic phys­iology is a useful endeavor and a check to
make sure that your model of the system
is accurate, it's the models that evolve
that are really important. So for my Ph.D.
I got seriously into psychoacoustics, and
there was a very strong combination of
applying electrical engineering and psy­choacoustics. I was using computer-based
DSP to try to create signals that would
impart the illusion of a sound coming
from a predetermined direction [a phe­nomenon called "localization"]. I was
kind of knee-deep in the tying together of
EE, signal processing, and perception.
RANADA: Let's go into this a bit. I re­member being an experimental subject of
yours and you pouring gook into my ears
in an effort to make molds of my outer
ears. I assume that this was part of your
dissertation.

DAVIS: That's right. That actually pre­ceded the formal dissertation. That was
exploring to what extent the pinnas im­parted localization. And one of the things
that was done was that we made these
gooey models of all of our ears—first
negatives and then positives—and made

recordings with microphones in a dummy
head with these ears fitted to the outside

of the head. We found that you could
make recordings that did impart a greater
amount of spatial orientation than if you

just used a plain-vanilla KEMAR head [a

standard psychoacoustical manikin]. When
you tried to play one person's recordings
to another person [we found] that it didn't

work generally very well. Each person's
pinna characteristics were very individ­ual.

RANADA: Can you explain how the out­side of your ear affects what's going on at
the eardrum?
DAVIS: As far as we can tell, the little
curves on the ear cause little reflections
that affect the frequency range from about
5 or 6 kHz out to maybe 14 or 15 kHz.

The effect of these [reflections], depend-

ing on the direction of arrival of the
sound, tends to impart features like deep,

narrow notches at various frequencies in
the spectrum [as received at the eardrum].
And as the sound moves around, the

notches move around [change frequency],

and your brain learns the pattern of these
notches and how they correspond to di­rections.

RANADA: So that explains why one per­son's ears won't work on another person,
because the notches are in the wrong
place.
DAVIS: That's right. What notch goes
where is a function of how the sound
moves around. In time, you sort of learn
your own ears because you have visual
and other feedback to tell you [which di­rection] the notches correspond to, and so
on. You're constantly moving your head
around and often looking at sounds that
you care about and subconsciously—but
actively—tracking the notches. If we
gave somebody a new set of pinnas, in
some amount of time they'd "learn"
them.

RANADA: What would happen if you
gave somebody their own pinnas but
when they were five years old and much
smaller?
DAVIS: That's an interesting question.
RANADA: I would assume that the rela­tive positions of the notches would re­main where they were but that their fre­quencies would change.
DAVIS: Right—things would sort of

slowly lower [in frequency with age] but
basic shape of the variations with position
would remain the same.
RANADA: And you'd be able to tell
whether it's the relative positioning that
was important or whether it's the absolute
frequency that matters.
DAVIS: That's a very interesting sugges­tion. I don't ever remember anybody try­ing that. You have raised what I think is
still an outstanding unknown of research
about all this, namely what is the im­portant information. Is it the notches,
their depth, or their width, or their center
frequency or what?

RANADA: Along with all the psycho­acoustics, you had to learn the basic
mathematical and statistical principles for
evaluating how things are.
DAVIS: That was a major stumbling

THE AUDIO CRITIC

pdf 45

block. I knew signal theory going into the
thesis and I knew something about hear­ing but I didn't know a whole lot about
statistical analysis. And a great deal of
time had to be spent on just coming up to
speed on that.
RANADA: There are a lot of people who
complain about controlled listening test
for all sorts of reasons. But they always
complain about the statistics' being not
meaningful. What are your thoughts
about that?

DAVIS: There are admittedly some condi­tions that apply to just casual listening
that don't quite get faithfully reproduced
in listening tests. The usual complaint is
the long-term listener fatigue kind of
thing, where listening tests are generally
relatively short in duration. Listener fa­tigue is something that is very difficult to
measure under any conditions because it's
tied in not just with the sound but the lis­tener's physical condition at the time.
You've got a lot of variables that are
difficult to pin down. In general, I think
that controlled listening tests—done by
people who have some experience in con­ducting them, [as well as] an under­standing of what the underlying mech­anisms being tested are—can be pretty
valid and pretty revealing. If you can get
a multiplicity of testers to agree on a test­ing paradigm and to agree to administer a
test in similar fashion, you can get usually
pretty similar results across a pretty broad
spectrum of subjects [listeners]. We all do
have human ears and there are pretty
well-established norms as to what those
ears are and are not capable of doing.
RANADA: Your dissertation topic was
still some distance from the hi-fi world.
How did you get from MIT into audio?
DAVIS: The Boston Audio Society is the
major culprit in that. I started going to
their meetings and interacting with people
who are actually in the hi-fi biz. I would
go and listen to these people give pre­sentations on their hot new products. In
some cases it was very impressive, but in
other cases I'd find myself sitting there
thinking, "This is silly—I can do better
than that." So I actually started messing
with designing hi-fi circuits well during
the time I was in college. I got interested
in noise reduction and built noise-

reduction systems—wideband compand­ers and things like that.
RANADA: Just for your own amuse-

ment?
DAVIS: Just to see what they did. I

briefly considered going into business and

making them and decided that was not

what I wanted to do. I did phono

preamps, trying various and sundry FET

and bipolar transistor phono preamps, and

loudspeaker equalization and stuff. In col-

lege I got really interested in what really

made one loudspeaker sound different

from another or whether they sound dif-

ferent at all. The lab at MIT had this 31-

band graphic equalizer with very steep

filters. It cost several thousand bucks at

the time. I would drag that home with

B&K calibrated microphones and related

equipment. We'd set up a whole bunch of

loudspeakers and try to equalize each of

ISSUE NO. 19 • SPRING 1993

them very, very carefully to the same V3­octave response four feet in front of them.
We'd listen to see whether we could hear
the difference between loudspeakers or
not. We found when speakers had similar
radiation patterns and when they could
produce similar acoustical output [volume
levels], you could make them sound
astonishingly similar. The larger, more
expensive speakers could deliver more
acoustic output, so there were times when
I could make a very cheap speaker sound
like a more expensive speaker as long as I

didn't turn the volume too far up. Once

you were past a certain point, the little
one would start "flapping." We also found
that if the radiation patters were materi­ally different, there was no way you could
make one speaker sound like another un­der any conditions. But it wasn't that hard
to imitate the radiation pattern. At one
point I recall comparing an AR LST loud­speaker, which was a speaker with a

front-facing panel and two diagonally
outward-facing panels, with drivers on all
three. We attempted to equalize an AR-

7—one of their smallest [two-way] speak-

ers—to have the same response, and we

found that when we did they did not

sound the same. But when we took two
AR-7s, one on top of the other, and cant­ed them a little bit so that they were

"We found when speakers
had similar radiation
patterns and when they could
produce similar acoustical
output [levels], you could [by
equalizing them] make them
sound astonishingly similar."

pointed in slightly different directions to
imitate the slightly different directions of
the AR LST drivers, then we found we
could produce an extremely good match
between those two. That started me think­ing about what would be good loudspeak­er radiation patterns and so on.
RANADA: What was your first product
to hit the unsuspecting consumer?
DAVIS: That was the Allison Electronic
Subwoofer. But I think that was preceded
by a little phono preamp—the Davis­Brinton phono preamp—though it was
not generally offered for sale. Jim Brinton

[former president of the Boston Audio So­ciety] and I did that. Jim was friends with
Roy Allison and sort of got us the con­tract to do the design for the electronic
subwoofer. That [product] was indeed
sold, albeit in small numbers. Out of the
noise-reduction work that I did came a lit­tle compander built around a Signetics
compander chip. Joel Cohen used that cir­cuit to noise-reduce his bucket-brigade
delay line for the Sound Concepts ST-550

[an early ambience enhancer]. Toward the
end of my college career, I got hooked up
with Francis Daniel in New York, who
had started Benchmark Acoustics, and

largely on the basis of Francis' well-tuned
ear we put together an ambience box that

was called "The Other Half." It was

vaguely like the Sound Concepts in that
you would take the signal and pass it
through this thing, and then it would either
feed front speakers and/or side speakers.
It was a much more subtle sort of effect
than previous ambience-type units. Fran­cis was trying to get rid of the harshness
that tended to accompany ordinary stereo
reproduction, particularly on direct-firing
speakers.

RANADA: I have always wondered why
some stereo systems sound harsh when
played loudly.
DAVIS: Because of the fact that the upper
midrange and high-frequency sounds are
being beamed very narrowly at the listen­er and not around the room, whereas the
rest of the sounds—due to the radiation
pattern of the speaker—are getting
bounced around the room. The middle
and high frequencies are like a pair of
headlights shining in your eyes, whereas
the rest of the spectrum is more like dif­fuse background lighting.
RANADA: Why does this sound worse
when it gets louder?

DAVIS: I'm not entirely sure. Partially
probably because the effect falls below
thresholds. In any case, even at lower lev­els it can add an unnatural amount of
false brightness. What Francis' observa­tions led to was that you had to try to
make, on the one hand, the radiation of
the energy a little bit more diffuse, but at

the same time you had to do it in such a

way that the imaging did not also become
more diffuse. So evolved "The Other
Half," which did a nice job of easing the
problem of making loud passages much

less harsh without either rolling off the
treble or compromising the imaging par­ticularly. I finished the design of that one
and Francis went about trying to get it
produced and promoted. In the meantime
I started on the design of a loudspeaker
intended for Benchmark Acoustics. But
before I got a little way into it, it became
evident that that project was going to be a
good deal larger, more expensive, and
more involved than poor little struggling
Benchmark could afford at that point. So
we arranged that dbx would buy the
rights for the speaker. I went over to dbx
to work on this speaker and to work on
noise-reduction systems.
RANADA: You do seem to have had the
perfect background to work there.
DAVIS: It did seem to kind of all come
together with all the systems theory, the
electronics, the psychoacoustics, and the
real-world experience designing stuff.
RANADA: Could go into the psycho­acoustic background of your speaker?
DAVIS: There were a couple of funda­mental considerations that went into the

[first] dbx speaker. One was the notion of
using cross-firing systems to try to widen
the imaging space [the listening area
where you can get a good stereo image].
That, in and of itself, was not unique to
dbx. [The idea] had been floating around
in hi-fi circles for a while—you could im­prove the imaging area by cross-firing

whatever speakers you had slightly [that
is, aiming them so that their front axes
cross in front of the listening position, not

57

pdf 46

at it]. But what happened was, we started
looking at how well that worked and how
to make it work as well as it could. We
found, not very surprisingly, that a con­stant radiation pattern was needed. As you
moved around the speakers, the balance
between the left and right speakers had to
vary according to a specific characteristic.
We could measure this characteristic readi­ly enough by moving people around, hav­ing them stand still, and adjusting left/
right levels [so that the image still seemed
come from between the speakers].
RANADA: So this experiment took in

both level and timing differences.

DAVIS: It took them into account in that

the timing differences were present, and

the listeners would turn the left/right bal­ance so that the image was most central

and the timing [differences] went into
their judgment. At that point, not only
were we interested in measuring what the
settings were, we were also interested in
how good that image could be. We found
that when using conventional speakers to
do this measurement, as long as we kept
the speakers pointed at the listener so that
the frequency response did not change,
we could obtain astonishingly good im­aging at distances way off axis. From this
comes the notion of a constant radiation
pattern.

RANADA: When you say constant radia­tion pattern, you mean over frequency?
DAVIS: That is right, it's not over angle.
[What you want is a frequency response
that does not change with direction of
radiation.] That also came about from the
earlier experiments showing that non­constant radiation patterns led to harsh
high-frequency response. It was also my
observation that typically, on one of these
direct-firing loudspeakers, the sound ap­pears to come from somewhere near the
speaker, except for the high frequencies,
which seem to come directly from the
tweeter. I felt that by doing a constant ra­diation pattern you would get rid of this
clue that you were listening to a tweeter,
and that the high frequencies would be
better integrated with the entire rest of the
image. The point, and the difficultly, was
that if you did these experiments, what
you came up with was a radiation pattern

that, as you walked around the speaker,
had to get louder at some points and soft­er at others. But it had to do that at all fre­quencies by about the same amount, at

least above 200 to 300 hertz, below which

point there's nothing really localizable.

Since real-world drivers simply do not

have constant radiation patterns, we set

about trying to make a system with a

combination of drivers individually con­toured so that their net acoustic output

was this constant radiation pattern. [The
frequency response in any direction was

the same, but the overall loudness of

sound radiated in that direction had to

change by a specified amount.]
RANADA: This was similar in principle
to phased-array radars, then.
DAVIS: That's right. What resulted was
this enormously complicated crossover
network that got put into the base of the

speaker, and fourteen drivers (four woofs,

58

four mids, and six tweets), and also an
outboard equalizer because the aggregate
frequency response of the system, par­ticularly with all this crossover stuff, was
far from flat in and of itself. In sub­sequent models of the dbx speaker we got

better at doing it with fewer drivers and

we concentrated on front-firing [radia­tion], whereas the original one was more
omnidirectional. Eventually I did an

[acoustic] lens-based system that got by

with just a couple of drivers, but we never
got that into production.
RANADA: Your acoustical lens was
made out of foam? Plastic?
DAVIS: It was sort of built out of hunks
of wood, a little plaster of pans, and a
few socks stuck into the grille holes at ap­propriate places.

RANADA: You don't recall the brand of
sock, do you?
DAVIS: No, I don't.
RANADA: This would make a great con­struction project for a magazine.
DAVIS: Yeah, it actually would. I kind of
would like to go back and build one of
these systems myself because the imaging
was really excellent; it was superb. It

used just a pair of coaxially mounted

drivers with an individual acoustic lens
for each one, and it produced very, very
good imaging. One of these days...

"...as long as we kept the

speakers pointed at the

listener so that the frequency

response did not change,
we could obtain astonishingly
good imaging at distances
way off axis."

RANADA: That was one of your prongs
of research at dbx. The other was noise
reduction. You have the unique honor of
being the only name on the patent for the

[United States'] stereo TV noise reduc-

tion system.
DAVIS: At that time nobody knew if any-

body really wanted stereo sound for TV.
They still didn't have widespread stereo
for AM radio; nobody seemed to want it.
So I puttered away on that [dbx's] system.
I got to bring to bear a lot of the ex­perience and a few of the prejudices that I
had developed in college in playing with
noise reduction.

RANADA: What were some of those
prejudices?
DAVIS: At the time, there were two class­es of noise reduction around: the dbx and
the Dolby. The dbx was a wideband sys­tem that could do enormous amounts of
noise reduction under the proper condi­tions, but there were times when it would
be imperfect and you could hear the
noise—and the fact that it [the noise] was

[audible] only at times was very ob-

trusive. The Dolby was a variable-filter
kind of noise reduction system; it was
much more conservative. It worked pri­marily at high frequencies, which is
where the primary noise components
were. Although it did not totally eliminate

the noise, it was much less obtrusive and
seemed to reduce the noise by a constant
amount. So one of the prejudices was—I
felt that combining those two systems and
using the best points of them was a viable
way to go. It was also becoming practical
because the cost of high-performance
VCAs [voltage-controlled amplifiers] was
coming down, dbx now had theirs in a
chip form instead of a circuit module.

There were three key tests I did at dbx

that really specified the system. The first
thing I did was measure the TV-audio
channel as a function of frequency for its
overall [frequency-response] character­istics and its noise floor. This gives you
two curves you can plot on the same
graph. One was the maximum signal you
could put in as a function of frequency,
which was like a flat line that rolled off at
higher frequencies. The other was the
noise floor that started maybe 50 dB be­low. [This curve] was more or less a flat

line, but it came up at higher frequencies.
You had less overload [margin] and more
noise at higher frequencies. At midband
you may have 50 dB of signal-to-noise ra­tio; at high frequencies you may only
have 30 dB. I seem to have been the first
one to analyze the problem in this way,

[in terms of] dynamic range as a function
of frequency. Prior noise reduction sys­tems pretty much dealt with the channel
as having a single overload point (in­dependent of frequency) and a single ag­gregate amount of noise (independent of
frequency). Masking is the operative psy­choacoustic principle that you are ex­ploiting in noise reduction. But even
though it was acknowledged that masking
is frequency-dependent—[leading you to]
a frequency-dependent system like Dolby
A—it was a semiarbitrary decision as to
how much noise reduction was applied.
Here I was saying, "Let's measure exactly
how much noise reduction you need and
provide a system that provides exactly
that much and no more." If I have a chan­nel that has 50 dB of dynamic range at
midband and 30 dB at high frequencies,
and I needed 90 to 100 dB when it's in
noise-reduced mode, then I derive from
that that I need 2:1 compression from the
midfrequencies on down, and at high fre­quencies I need about 3:1. Well 3:1 is a
lot of compression—even 2:1 is—but that
was the first time that somebody said
you've got to bite the bullet if you really
want to have a chance at getting this sig­nal through the channel. So that was one
measurement: the overload point and the
noise floor. The second measurement was
a characterization of the class of signals

that would be going through each chan-

nel, which consisted of me watching a

real-time 31-band spectrum analyzer (an

Eventide unit that was plugged into an
Apple II computer) for a wide variety of
material—speech, music and so on.
RANADA: You didn't take any data; you

just looked at it?

DAVIS: For long periods of time. I tried
to figure out how I would try to get this
moving bar graph of sound through this
rather narrow channel. What came out of
that was a characterization of typical au-

THE AUDIO CRITIC

pdf 47

dio signals as consisting of relatively
strong mid-to-low-frequency funda­mentals and a rolling off series of har­monics above 1 kHz or so. What varied,
in a first-order approximation, was the
overall height of this funny mountain and
the slope of the high frequencies. If you
could control the up-and-down motion of
the whole thing and you could control the

[high-frequency] slope, you could fit the
resulting spectrum through the channel
most of the time. What came out of this
was the notion of combining a wideband
compressor with a variable high-frequency
filter, something like putting a dbx and a
Dolby B together. But the characteristics
of that high-frequency filter were def­initely not the characteristics of a Dolby
B filter, which is a sliding shelf. The char­acterization that came out of watching the
signals on the analyzer was that you need­ed something that would more or less
hinge around 1 kHz and have more effect
at higher frequencies because you had
progressively narrow dynamic range. The
third thing was, what should happen in
the absence of high frequencies in the
program material? How sharp a filter do
you need? Here's the case where the pre­emphasis [treble boost] on the encoding
side is at its maximum and the de­emphasis [treble cut] is maximized on the
playback side. How sharp a filter do you
need to get rid of the noise so you won't
hear it? I did a simple test: I put a 400 Hz
tone into the electronic simulator of a TV­audio channel and adjusted a filter until I
couldn't hear the noise anymore. It turned
out that the filter had to be a second-order
(12 dB per octave) filter in order to get rid
of the noise. This was important because
that is the sort of thing you need to get a
piano to come through without "breath­ing." [If] somebody hits a piano note,
which is mostly midrange, you don't
want the high-frequency filter to "open
up" [and let through all the noise]. My
funny little variable filter that I was going
to tack on to the wideband compressor
had to do a 12-dB-per-octave boost or
cut. But it also had to go down to no
boost or cut when the signal had a lot of
high frequencies. It needed this range
from no boost to a 12-dB-per-octave
boost, which was unusual for a noise­reduction filter. The noise-reduction
filters before, and most of them since, are
pretty much 6-dB-per-octave animals. So
this was a sharper filter than had been
used before. Anyway, those three things
all put together, along with a limiter that
was put inside the noise-reduction loop so
that you wouldn't hear its effect, became
dbx's entry against CBS Labs' and Dolby
Labs' entries for the selection of the TV
NR system.

RANADA: One of the accomplishments
of your system is that stereo TV is actual-

ly quieter than mono TV—is that correct?

DAVIS: No, it's no noisier. The stereo TV

system uses basically the same kind of

signal flow as in FM stereo. You have a

main audio channel, which is driven from

the sum of the two stereo channels. On a

second [transmitted] channel, which is de-

coded by the stereo decoder, you send the

ISSUE NO. 19 • SPRING 1993

[stereo] difference information, the left­minus-right signal. Then by alternately
adding and subtracting those signals in
playback you can recover left alone and
right alone. In both the TV and the FM
stereo systems, the L-R channel is in fact
a good deal noisier than the L+R channel.
So that if you just listen to the L+R, it's
generally pretty quiet. But as in FM,
when you put in the L-R and try to listen
in stereo, it will generally get at least 15
dB noisier. However, by using this fairly
aggressive noise-reduction system on the
TV L-R signal, the resulting noise is ac­tually lower, in most conditions of use,
than the noise on the main channel, which
is not noise-reduced. (As the main chan­nel was already in use, we couldn't add
noise reduction to it, since then it would
not have been backward compatible.) So
we left the main channel alone and we ap­plied the noise reduction only to the dif­ference channel. The difference channel

[when] noise-reduced is quieter by a good
amount than the main channel, so in
switching from mono to stereo things
don't get any noisier. It's too bad that we
didn't have things like that around when
we did FM stereo.

RANADA: Soon after your TV stereo NR
system was indeed selected as the best
submission, you were hired by Dolby and

"It's not that you are
necessarily throwing relevant
information away; it's just
that you are only sending
information that the ear is
responding to and not extra,
redundant information."

have been working on digital approaches
to noise reduction and signal coding. I
wish you could explain how the new dig­ital-audio data-reduction schemes can ac­tually work, since you are throwing out
much of the data.
DAVIS: This gets into audio coding,
which is kind of the DSP equivalent of
analog noise reduction.
RANADA: But in noise reduction you
aren't throwing away any of the signal,
are you?

DAVIS: In a sense you are, in that when
the signal gets into the channel you have
all this noise that the channel adds that
wipes out any part of the signal that is
equal to or below it [the noise]. On a sim­plistic level, when you digitize something,
each additional bit of accuracy is equiv­alent to 6 dB of audio dynamic range. So
if I start with 16-bit PCM samples, multi­plying 6 dB per bit by 16 bits gives a total
theoretical dynamic range of somewhere
around 96 dB. In digital coding you are
throwing away bits, and if I try to throw
away ¾ of those bits, if I have only 4 bits
per sample, that's 24 dB [of dynamic
range]. [This 4:1 compression] is like I'm
trying to compress a signal with a 96-dB
dynamic range through a channel with a

24-dB dynamic range. Many of the same
psychoacoustic principles apply and many

of the same or similar techniques apply.
But where an analog noise reduction sys­tem might have two or three or four bands
or something like that, a DSP system can
have 40 bands or so. This enables us to
effectively deal with [transmission] chan­nels that are much noisier than anything
we have dealt with in the past. This has
been a very exciting area and as the world
increasingly carries audio information

around in digital form, coding [as this
process is called] is becoming an in­creasingly important element.

RANADA: There are people who say,

"How can you possibly preserve anything

when you are getting rid of all of this
data?" Why wouldn't the ear be able to
detect that?

DAVIS: Basically what the coders do is,
they transform the signal into the domain
that the ear is using and then they send
only the actual information that the ear
will respond to. For example, if I put a 1­kHz sine wave through a simple 16-bit

PCM system with 44.1-kHz sampling (as
in the Compact Disc system), the fact that
I have a 1-kHz sine wave is sort of ir­relevant. The system will send 44,100

samples per second times 16 bits per sam­ple, some 700,000-odd bits per second

[705,600 to be precise]. That's how much
I'm sending, regardless of whether there's

a 1-kHz tone there or nothing or a sym­phony orchestra. But the ear is going to
hear the 1-kHz tone as an isolated tone.

So we do a Fourier transform [spectrum
analysis], and what comes out of that is

all these narrow frequency bands, and all
of them are zero except the one for 1 kHz.
It takes very little information to send to
the decoder the fact that all the bands are
zero except for the one at 1 kHz. It's not
that you are necessarily throwing relevant
information away; it's just that you are
only sending information that the ear is
responding to and not extra, redundant in­formation. Systems without this kind of
processing often get into the situation
where they are sending completely re­dundant information that's many, many
bits but very little audio. I guess the worst
case would be if I put nothing—utter si­lence—in. One of these plain-vanilla sys­tems will still send 44,100 16-bit samples
per second—all being zeros. A coder
wouldn't do that; a coder would send
"everything is zero"—one little message.
There wouldn't be any need to send any­thing more on the channel until some­thing actually happened. But you're not
losing anything in that case; there's no
implicit distortion of the audio. If what

you put in is 1 kHz, what comes out—if

it's properly coded—is that 1 kHz. For
the most aggressive coders we actually go
beyond that. We are actually throwing

away information. While it's not audible,

it's at least measurable. My first point was

that you don't even have to throw away

anything that's measurable; simply ex-

press the information succinctly, and you

can send fewer bits. But we do get into

cases where we are throwing away

enough bits that the differences are mea-

surable by sufficiently sensitive in-

struments. However, the differences are

59

pdf 48

meant not to be audible. What this in­volves is going back to psychoacoustics
and using somewhat more elaborate mod­els that predict what parts of the signal
are audible and what are not. For ex­ample, in the case of my 1-kHz tone, sup­pose I had an additional tone at 1050 Hz
that was 50 dB quieter. I could pretty
much skip transmitting anything about
the presence of that 1050-Hz tone be­cause you simply won't hear it next to the

1000-Hz tone. If the 1000-Hz tone goes
away, then you do have to send the 1050­Hz tone because it [the 1050-Hz tone] is
no longer being masked. So we do make
use of masking to throw away in-

formation. But the net result is that you
can still do an A/B listening test of your
original signal against what comes out,
and they should sound identical. [PASC,
the 4:1 digital coder used in] DCC is by
no means the only audio coder that will
become a widely held standard, although
it may become one too. One of the other
ones that seems to be bubbling around is
a standard for computer-based storage of
music and interactive multimedia play­back. There is a working group, with
about 170 corporations participating, that
is defining standards for coders that will
run on PCs and that will provide bit re­duction on the fly on PCs in an inter­changeable format across [computer] plat­forms. These coders, because they
actually run on the CPU and do not make
use of a formal DSP chip, are much less
elaborate than the sort of coders that are
being put into DCC. These standards are
being evolved now, and you'll see them
in the next year or so.
RANADA: These aren't going to claim
that you can get CD-quality sound, are
they?

DAVIS: Right. The question is how close
can you come—can you make the coder
unobtrusive?
RANADA: I want to get into the philo­sophical issue we covered once in another
conversation: whether you think perfect
high fidelity is possible at all, whether
simply trying to create in the listener's

brain the same perceptions of something

that has occurred somewhere else at an­other time violates any physical laws?
DAVIS: There's this theorem by Laplace

that says effectively—imagine yourself

sitting in a concert hall, and you surround

yourself with an imaginary array of an
infinite number of microphones located
on this big sphere surrounding you. Each
one is connected to a loudspeaker just on
the other side of this imaginary spherical
membrane. You convey the sound
through the membrane via a zillion little
microphones on one side and a zillion lit­tle loudspeakers on the other. If you had
an infinite number of them, in theory you
would have a perfect sound conveyance
system.

RANADA: Is this a mathematical theo­rem?
DAVIS: Yes, the basic theorem operates
within a region—in any wave-bearing
medium, it doesn't have to be sound
waves in air—in which there are no
sources. Assuming that you have wave

60

energy arriving from strictly outside this
sphere—there are no internal sources of
wave energy—then the pattern of waves

within the space is entirely specified by
just knowing the normal [perpendicular]
component of the impinging waves on the

surface of the space.

RANADA: Now this would seem to be
one of those mathematical theorems that
would require, for perfect reproduction,

an infinite number of everything.

DAVIS: Right, in and of itself, it would
require an infinite number of everything
and would also mean that you would end
up with an infinite bandwidth. Potentially,
if you fulfilled the strict requirements of
this theorem, you would need an infinite
bandwidth [for mikes, loudspeakers, and
any recording medium].
RANADA: You'd be able to put your dog
inside the sphere, and it would respond to
a dog whistle outside it?
DAVIS: Exactly. So we very quickly get
to the point of saying that you can get by
with fewer than an infinite number of
these things and still not hear any differ­ence—and still have a perceptually per­fect system. Then the question becomes
how far below infinity you can take it.
You can't hear anything above 20 kHz,
and the wavelength of 20 kHz in air is
something like ¾ inch or so. I think, in

"In theory, you should
be able to encode a
three-dimensional [coherent]
sound field sampled at

three or four dozen points
into a [reasonably low]
data rate."

all likelihood, if I have a microphone/
speaker pair only every ½ inch around
the sphere, that that would probably be
enough. If you work that through and you
still wind up with a god-awfully high
number of channels, then it's still not
practical at that point.
RANADA: This would be hundreds or
thousands of channels? This would, I take
it, depend on the size of the sphere.
DAVIS: This is true too. I suppose it
would not have to be any bigger than
your head. So that might not be too bad,
having ½-inch resolution around a
sphere with a one-foot diameter; I'd have
to work that out but it'll probably come to
the hundreds of channels.
RANADA: So you'd be putting on your
stereo helmet...

DAVIS: Of course, you don't want to
have to wear a stereo helmet. If you want

to move your playback transducers back

to the walls around your room, then that
becomes the theoretical boundary and

you might need a lot more channels in
that case. You know that in signal pro­cessing there is the digital sampling theo-

rem that says you don't have to sample at

a rate more than twice the highest fre­quency contained in your signal. There's

a kind of perceptual corollary to that re-

garding [this kind of] spatial sampling.

Each one of these channels is basically a
sample in space of the sound at that point.
The ear can only sense a certain mini­mum change in angle. It varies as a func­tion of position and direction, but for hor­izontal directions—say for a sound right
in front of you moving left to right hor­izontally—the just noticeable difference

is on the order of about one degree of arc.
If the source was directly in front of you
and moving up and down, instead of back

and forth, the minimum difference would
be about five degrees of arc. [The sound
source] would have to move that much
before you could clearly hear a change in
position. Just from these specifications, if
you chopped up the solid space around a
person into one-degree segments in the
horizontal direction and five-degree seg­ments in the vertical direction, you would
again get a few thousand channels. But
that's still quite a bit fewer than if you
had a speaker every ½ inch around your
walls. And so that is a first, simplistic cut
at applying what might be considered a
spatial sampling theorem to reduce the
number of channels.

RANADA: Now, the object of this ex-

ercise is to create around the listener's

head the exact same sound field that he

would have experienced in the original re­cording environment?
DAVIS: Yes, to the extent that the listener
can perceive it as being different or the
same. Now there's another thing you can
use which is certainly used in present au­dio systems, and that is phantom imaging.
In fact, I can have a pair of sound sources
farther apart than one degree and still
create the illusion of sounds coming from
the space between those speakers, by ad­justing the timing and amplitude of the
sounds coming from them. And that's ba­sically all we've got in two-channel ste­reo. It's pretty crude. But if you had a
speaker every 5 degrees of arc horizontal­ly and every 15 degrees vertically, that
would sound pretty close to perfect.
RANADA: Now these are just simple
systems, just microphones and loudspeak­ers—no other processing?
DAVIS: Right. Where the processing
comes in is that, even with only a few
dozen channels, the storage requirements
are enormous and beyond what existing
storage media can comfortably handle if
you are going to make a digital recording.
And [they are] also hopelessly beyond
what the ear can possibly absorb. Assume
that three dozen channels would some­how cut it. You've got 36 channels, each
running at 700 kilobits per second—
you're talking about oodles and oodles of
bits per second. You're talking 25 mega­bits per second. There's no way that the
human auditory system can possibly ab­sorb 25 or so megabits per second.
RANADA: We talked once about what
the maximum data output rate of the hu­man ear is; it's well below that figure.
DAVIS: In theory, you should be able to
encode a three-dimensional sound field
sampled at three or four dozen points into
a much, much lower data rate. [But with­out the employment of special coding
techniques] effectively what you are cod-

THE AUDIO CRITIC

pdf 49

ing [in this case] is three or four dozen
channels, each playing a totally different
program. There is no way a human listen­er could absorb all those programs si-

multaneously. All that is going to come
out is this horrible cacophony, and no­body will understand a thing. We don't

want or need a system that can play 36
different, simultaneous, discrete, in­dependent programs to a listener. What

we want is a system that will transmit

faithfully a three-dimensional, coherent

sound field that is one ongoing sonic
event. It is clear that we ought to be able

to apply some sort of relevant coding to

[compress] this entire sound field down to

some sort of reasonable number of bits,
commensurate with what the human ear
can actually absorb. This is a very active
area of investigation. You [should be able
to record] on a standard Compact Disc,
instead of just two discrete channels, a
complete three-dimensional sound-field
event—in coded form—that could then
be reproduced with arbitrary accuracy. If
you've only got a mono speaker, then
what can you do but play it through that
speaker. If you've got two speakers,
you'll get stereo. If you happen to have
taken the time and trouble to set up 36
speakers in your room, this thing will de­code the 36 channels and—in theory—
you should have, within perceptual limits,
a perfect recreation of the original re­cording.

RANADA: Thirty-six speakers is still a
lot of speakers. Is any research being di­rected to make such a system more prac­tical?
DAVIS: We're still not at 36 in terms of
any real-world systems but we are at six,
and that is the Dolby AC-3 coding tech­nology, among others, which is being
used in Dolby SR·D films. The AC-3 cod­er is my little invention; I'm the principal
inventor. The Dolby AC-3 coding tech­nology is derived from Dolby's AC-2
coding technology. AC-2 technology is
one-channel-at-a-time coding using a
filter bank based on a running transform.
The description of a 40-band system I
gave is based on AC-2, and that is the
starting point for AC-3, but AC-3 takes in
multiple channels at a time. (With the cur­rent implementation there are six; they
are intended to be left front, center front,

right front, left surround, right surround,
and a limited-bandwidth subwoofer chan­nel.) The AC-3 coder takes running trans-

forms of each of those channels and then
encodes the entire mess of transform
coefficients in such a way as to use a min­imum data rate—at least for the state of
the art at the moment—without com­promising fidelity or introducing coder ar­tifacts, while using fewer bits than would
be possible by using six channels of AC-2
coding. In fact, we save about half the
bits that way. By specially coding re­dundancies that exist across channels—by
knowing, going in, that because you've
got a coherent sound presentation there
will be significant redundancy—we can
acutally encode those redundancies in a

way that doesn't impair the sound but
does reduce the required bit rate. And, of

ISSUE NO. 19 • SPRING 1993

course, this system is being actively used
now to encode movie soundtracks. The
ingoing bit rate is six channels going at
48,000 16-bit samples per second
[4,608,000 bits per second total]. (The
system will shortly be using 18-bit con­verters which will increase the incoming
bit rate but not the coded bit rate.) The
coded bit rate for the audio is 325,000
bits per second for all six channels.
RANADA: That's less than a quarter of
the standard CD bit rate!
DAVIS: That is quite correct; it is a data­rate reduction by more than a factor of
12, from 16 bits per sample to about 1.25

bits per sample.
RANADA: PASC used in the DCC sys-

tem gives only a factor of 4 reduction in
bit rate compared to a CD. Using AC-3
technology at the normal CD data rate,
then, you could have 24 channels?
DAVIS: You can see that it is now prac­tical at this point to code it [onto CD]. To­morrow afternoon we could make a CD
that had 24 channels of [sonically related]
stuff on it if we wanted.
RANADA: Has anybody done experi­ments using standard audio recordings, as
opposed to movies, to see whether this
perfect-reconstruction theory is the right
direction to go?
DAVIS: We have obtained six-channel

"You can see that it is now

practical...to code it [24

channels on a CD]. Tomorrow
afternoon we could make
a CD that had 24 channels of

[sonically related] stuff on it

if we wanted."

mixes of various and sundry audio re­cordings, jazz combos, and what have
you. And they come through just fine,
thank you.
RANADA: The trick is now to combine
this with some sort of sound pickup tech­nique to start approximating the infinite

sphere of microphones.
DAVIS: There's two parts to this, I think.
One is just pickup techniques that can
take advantage of having lots of channels.
Since commercial recordings are often
made in layers and stuff—you put up four
microphones to mike this part of the or­chestra and four others to do this and so
on, and you array those in various chan­nels—I think some people might just

want to make 24- or 36-channel record­ings. I think eventually you will want to
have honest-to-god microphone "trees"

with an awful lot of microphones on
them. I don't think the microphone com­panies are going to mind selling groups of

24 microphones at a pop. I think synthe-

sizers will start to be multichannel in their

orientation and the control that they will

allow. You will have little mouse pads

and things like that to control the trajec-

tories of sounds and so on. I think it

would open up music composition tre-

mendously; I think it would be a real rev-

olution, from an artistic point of view, to

give this kind of control to the artists: to
enable them to put any sound they want
anywhere in space they want and have it
reliably come out at the end.
RANADA: So at least in the number of
channels available, we are approaching
the level of technology needed to transmit
to the home everything needed for
perceptually perfect "3-D" sound repro­duction. The question is: how can it be
made practical in the home, where getting
only two speakers installed is often a
problem?

DAVIS: I think if you really want to be
something like perfect, you're going to be
looking at a few dozen loudspeakers.
RANADA: There's no way to get around
that using very clever psychoacoustic
trickery?
DAVIS: Not yet; there's only a finite dis­tance that you can count on phantom im­ages to be pretty solid. You can't get too
far afield before the room begins to im-

pose its character. If you want to dom­inate the room characteristic, you're go­ing to need a lot of loudspeakers. But I
think that this is going to be very strongly
a matter of personal taste. I think that for
a lot of people by the time they have, say,
five speakers around them in a horizontal
arc and maybe one on the ceiling, that a
lot of them are going to feel quite happy.
Our experience to date in simply listening
to movies produced in six-channel dis­crete compared to the previously avail­able Dolby 4:2:4 matrix system is that a
significantly enhanced perception of sonic
reality is imparted by just going from the

four matrixed channels to six discrete
ones. That's already a big step. For a lot
of people the approximation to reality
will be sufficiently close by the time they
get a half dozen speakers. I do think that
there will be audio equipment that comes
out that will help support this [multi­channel effort]. For example, if you've
got 24 speakers, you're going to want 24
amplifiers, but none of them individually
has to be particularly powerful. So you
could see receivers with 24 amplifiers in
them, each of which would have to put
out maybe only five watts. It's not that
hard to do.

RANADA: All this sounds tantalizingly
close. But to be a commercial success,
there would probably have to be some
standardization to prevent market chaos.
Do you know of anybody working toward
such standardization of multichannel cod­ers?

DAVIS: Standardization is a very big is­sue. At least with our Dolby AC-3 tech­nology, we are working toward trying to
establish that as a standard. Certainly, if it
is used in motion pictures it would be to
the benefit of the motion picture industry
to adopt it as a worldwide standard. As it
is, 35-mm films are a worldwide stan­dard; you can take any 35-mm print and
play it anywhere. It is highly desired to
continue that. It's one of the reasons we
put the digital AC-3 sound track in be­tween the perf holes on the film[!], so that
it wouldn't disturb the analog sound track
or the picture, so that everything is com­patible. In addition, I think I can reveal

61

pdf 50

that there are already one or more chip
makers who are working on building a
cheap single-chip decoder for this system.
Starting from the movies, this system
could become a standard for broadcast
use, since broadcasts often show movies.
Why not just take the bit patterns that are
on the film and send them digitally? This
could be done tomorrow for NTSC!
There's enough room on an NTSC carrier.
We've demonstrated the ability to send a
couple of hundred kilobits per second on

an existing NTSC signal without dis­turbing any of the other components, in­cluding the MTS stereo [signal]. So you
could have six-channel discrete digital
sound on your TV! With a cheap chip to
do the decoding, and a low bit rate, it

[AC-3] can be piggybacked onto video­discs, it can be back-engineered into VHS
tapes, and things like that. So we are cer­tainly looking at the possibility. This is, at
the moment, a six-channel system. But
our conversation has indicated that for
any sort of "perfect" system you might
want a couple of dozen channels. The
techniques involved [with AC-3] are per­fectly extendible. The same coder, in rel­atively few additional bits, could code 24
channels if desired. One of the things that
digital coding offers that may not be im­mediately obvious to the casual observer
is that the bit stream can be logically sub­divided into any number of information
streams. This is something that you can't
do with analog. Up to now, all of the spa­tial audio systems that we have been
messing with—stereo, quad—all try pret­ty much to piggyback the spatial in­formation on the audio, such as matrixing
the audio channels together or something
like that. That of necessity compromises
both the spatial information and the au­dio. An advantage of a digital channel is
the ability to divide it very cleanly into
separate information streams, sending the
audio as part of one stream—and the spa­tial information, to the degree to which it
can be segregated, can be sent as a separ­ate data stream, where neither data stream
corrupts the other. This is a very valuable
and powerful solution. But to make use of
it requires a substantial amount of work
on optimum ways of coding spatial in­formation into multiple data streams. Our
AC-3 is our first effort along that line,
and I can see the possibility of further and
more powerful spatial coding as time
goes on.

RANADA: Much of what we've been
discussing revolves around the basic low­er limit of the ear's data rate. What is the
basic data rate of the ear? Even though
you can code what was originally gigabits
per second down to megabits per second
without any perceptual difference, there
must come a point at which you cannot
code below—a minimum data rate.
DAVIS: One very simple way to estimate
that, which will give a very conservative
answer, is to say that the ear has ap­proximately a 96-dB dynamic range (a
"16-bit range"), and it has kind of a 20­kHz bandwidth. But if you just assume a
channel can handle 16-bit data with 20­kHz bandwidth, that right there is

62

700,000 bits per second times two chan-

nels, giving 1.4 megabits per second—the
CD data rate. If you allow for audio
thresholds changing with frequency, you
can knock off from that probably another
factor of two and say 350 kilobits per sec­ond per each ear, total. But that's a fairly
simplistic analysis, and I think it's con­servative. [David Ranada's Note: The

point is that the entire infinite­microphone imaginary recording sphere
talked about earlier should be codable
down to this 350 kilobits/second data
rate.] Now I've recently heard claims—I

think it was from Anderson from MIT—
who claimed that in fact the internal data
rate that you can actually absorb is way
down at an unbelievable 100 bits per sec­ond! But someone pointed out that 100
bits means 100 flips of a coin every sec­ond. It's actually like you have 2-to-the-

l00th-power different possible outcomes,
and you are differentiating one out of 2­to-the-100th outcomes every second. And
you can't ask a person to do any more
than that—it's hard to ask a person to do
even that much. Two to the 100th power
is a huge number. And yet, to do any of
these coders, so far we still need many or­ders of magnitude more bits per second.
Any time we try to get anywhere near
even several thousand bits per second,

"I think [coding a very large
number of related channels]
is doable, and you'll see this
coming along within the next

five to ten years. And I'll be

one of the people trying to
make it happen."

much less several hundred bits per sec-

ond, the audio quality goes away quite

substantially, so far.

RANADA: At 100 bits per second, the

waveform entering your ear will not be
anything like the original.
DAVIS: That's a question about what you

do with those 100 bits. The thing is that to
get down to substantially lower bits per
second than we have now, we have to go
to what might be called event coders. If
someone strikes a piano note and it lasts

4.5 seconds and they let go of the key,
you send that information. Here's a piano
and it has this harmonic content and
here's this note that lasted 4.5 seconds.
And at the playback side you have this
very elaborate decoder which amounts to
a synthesizer that puts the signal back to­gether. I think that this will be the sort of
coders we will have.

RANADA: But this is less psycho­acoustics and more synthesis...
DAVIS: But in order to do it properly the
analysis of the original sound will have to
be in psychoacoustic terms. You won't
have the luxury of having an isolated pi­ano note you can do this processing to.
You're going to have arbitrary sounds,
and the coder will have to break the com­plex sounds into parts that can be char­acterized psychoacoustically in very con-

cise terms.
RANADA: And you think this is doable?
DAVIS: I think it is doable, and you'll see
this coming along within the next five to
ten years. And I'll be one of the people
trying to make it happen.

* * *

6. Interview with
Bob Carver,
Manufacturer and
Audio Designer

RANADA: How did you get interested in
audio?
CARVER: My mom was a musician and
my dad was an engineer, so here I am. It
sort of emerged naturally.

RANADA: Was there any revelatory ex-

perience that showed you the audio path­way?
CARVER: When I was about 13 or 14

years old, I heard a Heathkit hi-fi and it
just bowled me over. I could not believe
these incredible sounds that were coming
out of this big giant speaker that was
about the size of a refrigerator. I had nev­er heard a high-fidelity system before and
it was wonderful. I was hooked ever since
then. I started reading magazines and
stuff like that.

RANADA: But at that stage you were
probably too young to know what you
wanted to do with your life.
CARVER: No, actually, that's not true.
I've known ever since I was little—since
I was perhaps four—that my life was go­ing to be having something to do with
electricity. The way I expressed it then
was that I was going to be an electrician.
Electricians did interesting and fun things
with electricity. So when I was a Cub
Scout I got merit badges by building any­thing that had to do with electricity. As I
recall, that included making a crystal ra­dio set. It included making a capacitor out
of tin foil, and other scientific experi­ments. So I knew I was going to be an
"electrician."

RANADA: So you directed your higher
education in that direction?
CARVER: Right.
RANADA: You have a degree in what?
CARVER: Physics.
RANADA: You specialized in electronics
I assume?
CARVER: No, actually, the electronics
was always an avocation. I've trained as a
physicist, but circuit design is my first
love.
RANADA: So you have a scientific train­ing which gives you a slightly different
background than is the case with those
who have simply gone into engineering?
CARVER: I've found that to be absolute­ly true. Physicists learn a couple of
things. One is not to take themselves very
seriously. As soon as one tries to learn the
secrets of the universe... I'll tell you
something—studying the secrets of the
universe is an extremely humbling ex­perience. As we go along studying or

THE AUDIO CRITIC

pdf 51

learning the secrets of the universe, we
learn another very important thing—that
whoever designed this universe of ours
really had his, or her, act together. We
feel terribly insignificant just trying to un­derstand it... Physicists learn to take
themselves with a grain of salt and with a
little bit of humor, hence we name things
funny. Physicists always give things fun­ny names, like quarks, charm, spin...
RANADA: Magnetic field amplifier...
CARVER: Oh yeah, magnetic field am­plifier. That just goes to show that we like
to have fun.

RANADA: How would you say you look
at designing differently than a "normal"
engineer?
CARVER: Well, I think I approach it
from fundamental first principles. My de­sign approach starts with first principles
as opposed to perhaps a set of application
notes. As a physicist, I have such a ter­rible memory, I never remember any­thing. I can't even remember the table of
integrals, and that's terrible since it
means that I always have to start—when I
want to design something—from first
principles. Like, for a capacitor, let's see,
here's an electrical field between these
two plates, and then I'd have to derive the
equation on the spot because I can't re­member it. I'm exaggerating a little bit
but not much. So when I design some­thing, it really does start from absolutely
the ground up. And then as I go along and
as I'm traversing the design stages, that's
when I start looking up the work of others
to see what they've done. But it's sort of
halfway through the process.
RANADA: Let's go back to the very first
Phase Linear product. How did that
spring into your mind?
CARVER: As a hobby I had built ampli­fiers—tube amplifiers—and had even
built a giant tube amplifier. So I knew
about tube amplifier design. And when I
decided I wanted to make a living de­signing amplifiers, I decided to do a tran­sistor amplifier.

RANADA: What made you decide to
make a living out of designing amplifiers?
CARVER: All through school, especially
through graduate school, I found that de­signing amplifiers was so much fun. I
can't describe how much fun it is to sit
down at a workbench and design an am­plifier and have it work. It's hard to de­scribe. I have no idea why it is so much
fun; it just is. The Phase Linear 700 was
my first transistor amplifier. Again, start­ing from first principles, I didn't under­stand how difficult it would be to make it
work, to make it hold together at the very
high voltages that were required to pro­duce a powerful transistor amplifier. My
tube amp had 150 watts, but when I
played it into a pair of loudspeakers and
put a scope across the loudspeaker ter­minals, I found that the output voltage
went very high. When I decided to make
the transistor amplifier, I said, well, I'm
going to make the transistor amplifier
have as much voltage swing as my tube
one. I found out quickly that to make that

voltage swing it had to be 350 watts into

8 ohms. There were a couple of things

ISSUE NO. 19 • SPRING 1993

that made the Phase Linear possible: the

high-voltage transistors that Delco made
for automobile ignitions and a com­prehensive protection scheme that made
the thing "flameproof." And that pro­tection scheme has subsequently been
copied by almost every amplifier man­ufacturer there is, at least until recently. I
call it an energy limiter. It turns out that
the thermal time constant of a transistor
chip, in a TO-3 transistor package, is
about the same time as one quarter of a
cycle of 20 Hz. And if you play music, it
turns out that all you have to do is design
a protection circuit that will allow that
large volt-time integral to pass through
the transistor chip. If you make it be ap­proximately the same as the thermal time
constant, it'll work. You can play music
and the amplifier won't have to go into
limiting. Prior to that time, transistor
amplifiers either had to have a loophole in
their protection circuits whereby an un­wanted protection initiation would occur
and cause a snapping sound in the music,
or a loophole had to exist in the pro­tection scheme itself whereby, if you
shorted out the amplifier or hooked it up
to a faulty load, the amplifier could blow
up because it was inadequately protected.
In fact, the Crown DC-300 used to have a
switch on it labeled "hysteresis" and "nor-

"I can't describe how much
fun it is to sit down at a
workbench and design an
amplifier and have it work.
It's hard to describe.
I have no idea why it is so
much fun; it just is."

mal," and the instruction manual that
came with this beautiful Crown DC-300
amplifier said, when you're first hooking
up the amplifier throw the switch to "nor­mal." That gives you the maximum pro­tection. If you accidentally short it out
while you are hooking it up and first test­ing it, you won't blow it up. Once you are
secure in the installation, you are sup­posed to throw the switch over to hystere­sis so you can drive a load. But don't
tempt fate. The fact that the switch even
existed indicates that protecting an am­plifier back in those days was a real prob­lem. So that's why I came up with the
energy limiter. It's really an obvious con­cept, but my Phase Linear happened to be
the first one incorporating it.
RANADA: The next thing was a preamp
with many innovative features in it. How
did that come about?

CARVER: That harks back to my tube
amp days when I had my hi-fi set in my
small workshop. And I would listen to my
Beethoven symphonies with Bruno Wal­ter conducting, and there was a lot of hiss.
There was so much hiss it was un­believable. I thought, boy, somehow I've
got to figure out how to make this hiss go
away. One day I sat there with my hand
on the treble control and I noticed I could
turn the treble up when the music was

loud and I could turn the treble down
when the passages were low. As I listened
to the symphonies a lot I got to the point
where I would sit there with my hand on
the treble control and gain-ride the treble.
When the music played loud I'd turn the
treble up so that the highs were there.
And when the music was soft I turn it
down so the hiss would go away. I
thought, "Aha! This is really slick." From
that developed the Autocorrelator, which,
as you know, is a series of voltage­controlled bandpass-filter gain stages that
are sensitive to the musical spectra. Ac­tually it works kind of like the way the
new DCC works.

RANADA: I remember your getting a lit­tle flack for the name Autocorrelator. You
often get into trouble with the names you
pick for things.
CARVER: Physicists always pick good
names for things. Autocorrelator seemed
like a good name to me. Actually, there is
a basis in fact for the name because one
of the problems of controlling a series of
a bandpass gates is to dynamically raise
and lower the thresholds at which they
operate. And those thresholds are related,
from microsecond to microsecond, to our
hearing thresholds, which change with the
musical spectrum because of masking. I'll
use two extreme cases to illustrate this
point. If you play white noise, there's no
point in having a noise-reduction system.
But if the musical information you are lis­tening to is, say, a struck glass goblet
ringing with a nice pure tone, more or
less, it's going to be easy to hear any hiss
surrounding that tone. It turns out that
when a signal is highly correlated, like a
turning fork, it's necessary to have the
gate thresholds high, otherwise you'll
hear the hiss. But when your are listening
to music, the correlation coefficient of the
music is changing from moment to mo­ment. As an example, consider human
speech. During the sibilant portions of
speech, the correlation coefficient is ex-

tremely low—approaching zero—so the

gate thresholds under those conditions
can go down. Let's say you pronounce
the words speaking softly—during the s
sounds the thresholds can be very low be­cause the masking is very strong, but dur­ing the other portions of the phrase the
thresholds have to go up. The trick was to
control the gate thresholds dynamically
along with the information in the music

or voices. There's a circuit that de-

termines the correlation coefficient of the
incoming signal. That circuit in turn be­comes a control voltage that sets the gate
thresholds. If the sound is highly correlat­ed, the thresholds go one way and vice
versa. So that's why I call it an auto­correlator. The inventor gets to call it any­thing he wants.

RANADA: There's also the very famous
Sonic Holography circuit. How did you
decide to do something like that? There
were papers available on similar signal­processing techniques. Did you read them
or did you start with first principles
again?
CARVER: I went again from first prin­ciples. After Phase Linear, when I wanted

63

pdf 52

to make a comeback with Carver [Cor­poration], I figured I needed a couple of
new technologies to have people pay at­tention to me. At first I was thinking of

just one, but then I thought two would be

better for good measure. One was the
Mag amp and the other was Sonic Holog­raphy. Over the years I had become less
and less enchanted... I knew there was
something wrong with stereo. I under­stood early on that stereo presents us with
only a limited set of cues and clues that
prevent our ear/brain system from de­veloping a believable sense of a acoustic
space. That's because, when this stereo-

phonic system that we enjoy was con­ceived, very little thought, if any, was
given to preserving timing cues. All of
the thought was given to preserving am­plitude cues, that is, left/right separation
in the stereo system. Today, left/right sep-

aration is no big deal in any stereo sys­tem. But achieving a sense of spacious­ness and depth is a big deal and, in fact,
high-end audio in many respects has
grown up in the search for that difficult­to-find Holy Grail—that is, a sense of
depth and spaciousness. You can read in
high-end audio journals that "this ampli­fier gave a sense of layered depth; I could
tell which violin was in front of which vi­olin." So a sense of depth is obviously
searched for, and people die for it and so
on. It's hard to get and the reason is that
our stereo system was designed without
providing the proper set of spatial cues
and clues, the so-called timing cues. For
our brain to latch onto a set of timing
cues, it's necessary that each ear receive
the correct temporal information, which is
not possible with a simple pair of speak­ers. For example, in real life, when we
hear a sound it has the proper amplitude
cues and the proper temporal cues for
each sonic event. We hear sound arrivals,
and the temporal displacement of those
arrivals gives us the timing cues and al­lows us to locate the sound in three­dimensional space. But when that sound
is played back over a set of loudspeakers,
instead of a pair of sound arrivals (one for
each ear from a single sonic event), we
have four (two per speaker per ear). Four
arrivals into our ears instead of two is in-

correct and our brain doesn't quite know

what to do with it. Our brain has evolved

through evolutionary millennia to work

with two arrivals, not four. So what hap-

pens is our imagination has to work over-

time to make us believe that a stereo rep­resentation is real. A child can listen to it
and say, "Aha! That's not the real thing.
I'm not fooled!"—regardless of how
good the equipment is. So, Sonic Holog­raphy was a solution to the extra arrivals.
It simply acoustically canceled the un­wanted arrivals so that we're left with
two. And two arrivals are much more
believable than four arrivals. And the ex­pression Sonic Holography of course de­rives from a hologram, which builds a vir­tual image where none really exists. An
optical hologram does this by combining
two beams of light in an interference pat-

tern, a reference beam and another beam,

and by constructive and destructive inter-

64

ference an image is built. Sonic Ho-

lography works sort of the same way.
Two independent sounds—one from the
left speaker and one from the right speak­er—are combined in space around our
head. Constructive and destructive inter­ference cause the correct-side sound to be
accentuated and the incorrect-side sound
to be canceled. So the term Sonic Ho­lography refers to two things. First, peo­ple will hear the expression Sonic Ho­lography and go: "I know what that's
about; it's about images." And, of course,
sonic means that this is for sound. And it
does work in a similar fashion to a visual
hologram.

[Bob Carver's carefully constructed ra­tionale for Sonic Holography is some­what vulnerable to an argument I have
been throwing at him for years, namely
that the producer and engineer of a typ­ical stereo recording hang their micro-

phones and create their final mix to make

the sound as believable as possible
through the standard two loudspeakers

producing four arrivals. Removing two of

those anticipated arrivals alters the fine­tuned compromise and may under certain

circumstances be counterproductive. With
some recordings it works great.—Ed.]

RANADA: These innovative circuits
were all designed to solve very specific

"...when this stereophonic
system that we enjoy was
conceived, very little thought

... was given to preserving

timing cues. All of the thought
was given to preserving
amplitude cues..."

problems which existed ten years ago.
The audio scene is now very, very differ­ent. What do you see are the problems re­maining, or where can your inventiveness
be applied now?
CARVER: My big thing is psycho­acoustics. And psychoacoustics is an area
that audio designers have really not paid
much attention to, in my opinion. I be­lieve that is because it is a very difficult
arena to play in. We have to understand
how we hear things, why we hear things,
and know how to manipulate that without
fear of doing something wrong in the pro­cess.

RANADA: What is wrong, if what the
end result sounds like is what counts?
CARVER: That's my belief. There is
nothing wrong with that, of course. But
you have to be fearless to even have that
notion, it turns out, in today's audio
world.
RANADA: Why would you have to be
fearless if you're working scientifically?
CARVER: There's a notion that what ap­pears in the record groove or the CD bit
stream is somehow pristine and perfect.
That if only the amplifier and the pre­amplifier and the loudspeakers could pre­serve this pristineness, then we would
have believability in the sound field. That
somehow the sound coming off the disc is

so perfect and oh-so-delicate, and if you
mess with it you're going to screw it up
and ruin the believability. This is a belief
system that seems to have gained mo­mentum in the last decade.
RANADA: Is this due to the influence of
high-end magazines?

CARVER: I believe it is. I believe that
people just entering the fascinating field
of audio are hungry for information and
that the notion of a pristine signal is se­vere misinformation. It turns out that the
information coming off the disc is not all
that pristine. It's quite flawed to begin
with and, further, it's not delicate. It's not
going to be damaged by looking at it the
wrong way. You have to be fearless in
that you understand that the information
coming off the disc is flawed to begin
with, at least in format, and not be afraid
to work with it, to investigate it, to under­stand it, and to change it so that when it is
played back through a stereo system it'll
produce a more believable image of re­ality.

RANADA: So you are willing to forgo
what would normally be called accuracy

in order to obtain realism?
CARVER: I think accuracy and realism
go hand in hand. You can't have realism
without accuracy but you can certainly
have a lot of accuracy without any re­alism. So the comparison is an apples­and-oranges comparison. But back to the
original question. The audio scene has
changed a lot. We have super-power am­plifiers, we have loudspeakers that are
better than ever, everything's better than
ever. You might say, well, what's left?—
like the proposal to shut down the patent
office many years ago because everything
had already been invented. It turns out, I
think, that the things that should be tack­led today are the psychoacoustic events
that lead us into believing that a repro­duced soundstage is real. And there's a
lot of work to be done in that area. I be­lieve that it's going to involve somehow
the superposition of acoustic vectors in
space for the listener so that the sound
vectors will be essentially the same as
they were in a real-life recording venue.
But there's an important distinction here.
For something to be seen as if it were real
it doesn't have to be facsimile. It really
doesn't have to be identical. But the way
our brain reads it, it has to sense that it
could have been real. And that's the im­portant thing to remember. You know if
you go to the Holodeck—the computer­constructed illusion-recreation area of the
U.S.S. Enterprise in Star Trek, the Next
Generation—the Holodeck illusion seems
very real even though it never existed in
reality, and it's real to the touch, the
smell, the feel and all of that. And that's
what I'd like to do with the musical ex­perience in our listening room—basically,
have a musical Holodeck. It may not be a
facsimile reproduction of anything that
actually existed but it's so real to the
senses that it certainly could have been.
RANADA: In a lot of this kind of effort,
the problems are so enormous that it
would require a massive R&D effort,
mountains of DSP and computer power to

THE AUDIO CRITIC

pdf 53

do this kind of stuff. Are you in a position
to do this?
CARVER: Yeah. Powerful computers are
cheap now. And the beauty of audio de­sign is that it doesn't take an intensive
capital investment. It's not like building
an airplane where you have to have all
this machinery. You have a desk to work
on, a computer, somebody to help you as­semble circuits, a soldering iron, a scope,
and a few other odds and ends. It's ba­sically very cheap.

RANADA: So you don't feel intimidated
by the immense resources some com­panies can bring to a particular problem?
CARVER: No, I don't feel intimidated; I

feel jealous. Obviously, developing the
CD system is beyond my scope, but that's

also not my specialty. And I tell you the

wonderful work that Philips and Sony
have done on the DCC and MD is so re­freshing—to see somebody tackle it and
do some tremendously good work in that
area.

RANADA: You mentioned high-end
magazines a while back. Have high-end
magazines been good or bad for the in­dustry?

CARVER: I think that they've been both
good and bad. On balance, though, I think
they've been more good than bad.
They've been good in developing a lan­guage of high fidelity. For example, Har­ry Pearson of The Absolute Sound was the
founding father of the language that au­diophiles use to describe their experi­ences; in many ways he genuinely was.
And, even before him, Gordon Holt I re­member one time published a little dic­tionary of audio terms, and he used ex­pressions like woolly for the bass, and

fuzzy. And then Harry Pearson elevated

that and was a real class act in developing
a language for high-fidelity that was un­derstood throughout the world. As a sin­gle man he was very, very responsible for
giving us a language to talk about high
fidelity. And also, he and others like him
taught how to appreciate the dimensional
aspects of a soundstage. In fact, he wrote
an article on the stereo soundstage which
had influenced me when I was just getting
into appreciating this. So the high-end au­dio journals taught us how to listen;
they've taught us what to listen for;

they've been so entertaining as not to be
boring. So we've stuck with it and we've
learned. Where they've been an abom­inable failure is in the scientific dis­information and misinformation, and

that's been terrible. Some of these crazy
notions that are absolutely off-the-wall
with no scientific basis.
RANADA: Do you think these notions
are dangerous?

CARVER: I think that whenever fantasy
masquerades as truth it's dangerous. And
especially when it becomes a part of a be­lief system, because people are willing to
die for belief systems no matter how right
or how wrong. What happens is that peo­ple who are entering audio for the first

time really want to find out about hi-fi.
And they read anything and everything
they can get their hands on. When some­body says you should freeze a CD, or you
should plug this special magic clock into
the wall so your system will sound better,
they can't tell that that's not true. They
believe it. These are very smart people
and very bright people, and they want to
have an open mind. So they don't close
their minds to these notions. But it's real­ly an emperor's-new-clothes kind of
thing. I think there's some damage done.
But on balance I think it's good.
RANADA: Do you feel that you've been
done wrong by some high-end mag­azines? Have you been damaged by
them?

CARVER: Oh absolutely, tremendously
damaged. Because when a high-end mag­azine says that the Carver amplifier, be­cause it's so small, must not be any good,
and then listens to it and says this thing
sounds terrible—even though it has more
current, has more voltage, has more pow­er, can drive lower-impedance loads than
anything else around, for a fifth of the
price, and can even have tube output char­acteristics so it can sound like tube
amps—and make pronouncements like
that when they're patently incorrect,
that's harmful because many people be-

"I think that whenever

fantasy masquerades as truth

it's dangerous. And especially
when it becomes a part of a
belief system, because

people are willing to die for

belief systems..."

lieve it. It's like a restaurant reviewer
turning up his nose at an inexpensive dish
even though it might be a great, great
dish. It's not fair, but that's life.
RANADA: You learn to live with it?
CARVER: No, I haven't learned to live
with it. It hurts my feelings. But what can
I do? Actually, there's a lot I can do. I can
educate people. People want to under­stand how the world works, and if some­body out there tells them how the world
works they'll latch on to it, make sense
out of it. I think that there's a lot that can
be done and that I can do here.
RANADA: Where do you see the audio
industry going?

CARVER: I think what's happened is the
ratio of people interested in general sound
compared to component audio has gone
up. In other words, as a percentage of the
total population interested in audio, a
smaller percentage is interested in com­ponent audio. But the total population has
grown, so the interest in component audio
has grown. You can sort of verify that.
Take a look at the proliferation of high­end audio companies that are in the one­to four-million dollar range in sales.

There are lots of companies like that mak-

ing high-end audio components, and they
are enjoying success. Let's go back to the
mid-to-early '70s; hardly any of them ex­isted. There was a much, much smaller
number of them. So it's grown. It really
has. Just in my block, I sometimes during
the summer leave my doors open and

play my stereo. Like in the Pied Piper sto­ry, kids come in from the neighborhood
going "Oo, ah, wow!" They want to know

all about it; they get sucked in.
RANADA: Do you think audio will ever
reach a peak of development after which
further progress would not be possible?
CARVER: A deep, deep question, the an-

swer to which would be very presump­tuous, almost dangerous. But... if it's
possible to develop a stereo system in
which one could close one's eyes, sit
back and not be able to distinguish wheth­er you are actually in the presence of a
real-live musical performance or in the
presence of a sound system—if it's pos­sible to attain that, then you have to say,
"Well, that's it, you don't have to do any
more work." The problem is, imagine in
your mind's eye that the hardware for
such a system is successfully developed.
There's then the problem with the soft­ware; you'll bring home records and
some of them will sound real and some
not so real. There will always be more
work to be done. When it doesn't sound
so real, there'll be the temptation to tweak
one's system. So the fun will always be
there, and I don't foresee a solution in my
lifetime.

RANADA: Do you think, on an abstract
level, it is a solvable or unsolvable prob­lem?
CARVER: I think it's solvable.
RANADA: Some people say that humans
will always be better than the equipment
they are listening to. But one can take the
opposite viewpoint and say that the equip­ment is already better than what humans
can do, and the problem is what's being
fed through it and the exact way the cir­cuitry is employed.

CARVER: That's exactly right. The con­cept of limits has to be addressed. The no­tion that humans are always better than
the equipment ignores the limits that are
associated with human beings. And there
are limits. There are limits to all of our
senses, and those limits are very well un­derstood—or at least sort of well under­stood—and certainly well defined. So
we'll know when the equipment is better
than it has to be; we'll know when the
equipment is better than the limits that
our senses are able to deal with. Again
the notion of limits is an important one
because it is given zero—literally zero!—
credence by many high-end audio jour­nals.

RANADA: Do you think the equipment
will be good enough in your lifetime?
CARVER: Yes, I'm hoping. I want to be
part of making it good enough in my life­time; that's my plan. •

ISSUE NO. 19 • SPRING 1993

65

pdf 54

Hip Boots

Wading through the Mire of Misinformation in the Audio Press

Editor's Note: I turned my pet column here over to David Rich for this issue because he is so upset

about the monstrously ignorant "technical" articles currently appearing in various high-end audio

publications that he needs a printed outlet for his pent-up indignation on the subject. As he pointed

out to me when he requested this space, we aren't dealing here with the slower students in an
engineering class but with self-appointed "experts" who take your money for their stupidities. I

agree. I don't think all of our readers are fully aware of the utter contempt of degreed engineers and

responsible audio professionals for the untutored scribblings of these pitiful pundits. "How do they

get away with printing such garbage ? " those credentialled authorities keep asking me. How indeed.

Gerard Rejskind in The Absolute Sound.

The depths to which the underground journals will
sink to find someone to support the claim that linear PCM
coding is fundamentally flawed seem to be bottomless. In
choosing Gerard Rejskind, the editor of Ultra High Fidel-

ity, a Canadian underground journal, to address this topic,
TAS has sunk lower than ever before. [Rejskind, G. "The
Sound of Digital: The Present State-of-the-Art." The Ab­solute Sound 17.81 (July/August 1992): 28-36.] Mr. Rejs-

kind's qualifications as stated in the article are: "I have...
listened to countless [CD] players...and I have had the
privilege of talking with numerous designers." The result­ing TAS article is as flawed as any article I have encoun­tered that claims to be a scientific analysis. Not even Bob
Harley has so far written anything quite as absurd.

Mr. Rejskind states: "The usual claim of over 90 dB
of dynamic range is based on a common mathematical
blunder, coupled with what may be outright fraud." After
this statement he goes on to present the wrong formula
for the signal-to-noise ratio (dynamic range) of a PCM
system, namely 20 log (2b - 1), instead of the correct for­mula, which is 6.02b + 1.76. He then goes on to state the
formula is in error because the noise level is calculated as
a peak, not rms, value. Apparently he never looked at the
derivation of this formula in any standard text on com­munication systems or data conversion, since such a text
would have shown him that this statement is completely
false, in addition to showing him he was using the wrong
equation. Mr. Rejskind goes on to make the completely
false claim that the last bit of the 16-bit word in CD cod­ing is a parity check bit and only 15 bits of the word are

66

data. Mr. Rejskind blames the overstatement of signal-to­noise measurements (such as "over 90 dB") in CD players
to the use of a zero-code digital data stream to make the
measurements. The fact that sine wave tracks are used to
assess the signal-to-noise plus distortion characteristics of
a CD player is known to any reader of any audio mag­azine that performs electrical tests on CD players. Such a
reader also knows Mr. Rejskind's statement that "only the
best players can reproduce a sine wave at a level of -60
dB as anything but a caricature of the original" is also
completely false. But to quote Mr. Rejskind, "don't put
away your calculator just yet, because the fun is just be­ginning."

Gerard Rejskind uses the techniques of a patent
medicine salesman to claim that inband harmonics occur
when an unquantized sine wave is sampled above the
Nyquist rate. He offers proof of this by invoking concepts
for the analysis of amplitude modulation waveforms and
applying that approach to the points of a 12.5 kHz sine
wave sampled at a 44.1 kHz rate. He then claims to
identify a sideband at 2.321 kHz. It all looks very correct
to those unaware of the sleight of hand Mr. Rejskind is
using. The fact is that amplitude modulation theory sim­ply does not apply in this case. Interested readers looking
for a correct explanation, at a layman's level, of the sam­pling theorem (Mr. Rejskind spells it theorum) are di­rected to the Ken Pohlmann text, Principles of Digital

Audio. Another example of this sleight of hand can be

seen·in the following excerpt from Mr. Rejskind's article:
"Sophisticated designers...know that no [digital error]
correction system can be trusted to fix gross errors. Sen-

THE AUDIO CRITIC

pdf 55

sible amplifier designers are aware that they must build
circuits that work well even without feedback..." Mr.

Rejskind hopes the reader is unaware of the fact that ana-

log feedback theory and digital error correction are totally
unrelated fields.

Figure 1 of the article shows that Mr. Rejskind's
ego has no bounds. The first part of the figure shows the
analog and digital sections of a CD player connected
through zero-impedance ground connections. The caption
of the figure is, "How designers think they ground their
circuits." The second part of the figure adds resistors to
the ground leads. The caption reads, "How they actually
ground them." Come on, Mr. Rejskind, any C-minus stu­dent in the second year of E.E. school knows that ground
connections are not necessarily zero-impedance and may
cause problems. According to Mr. Rejskind you can hear
the effect of the ground interaction by slapping the CD
player. You can then, according to him, hear jitter caused
by this interaction. The fact that the player's sound chang­es because the error interpolation circuitry is activated
and the disc synchronization signals are momentarily lost
when the CD player mechanism is jarred apparently never
occurs to him.

From the above it should be clear that it is Mr. Rejs­kind who is in effect committing "what may be outright
fraud."

—David Rich

Robert Harley in Stereophile.

You may ask why we keep coming back to Bob
Harley in this column. The answer is that he has become
the most widely read and followed journalist in the field
of digital audio. According to manufacturers, a bad re­view by Harley, no matter how inaccurate, will send the
sales of the product to almost zero. That makes it impos­sible to ignore the dramatic errors in his copy that occur
because of his lack of training in electrical engineering
and his desire to believe in every claim for something that
makes an audible difference.

As an example, he virtually paraphrased the press
release for Sony's Super Bit Mapping (SBM) noise shap­ing system. ["Industry Update, US: Robert Harley." Ste-
reophile 15.8 (August 1992): 53-57.] He even included
misleading figures supplied by Sony. He described the
system as a Second Coming, with dramatic improvement
in sound quality. Any competent audio engineer reading
the Sony press release would have realized that the con­cept of noise shaping was not new and that the significance
of noise shaping was being seriously misrepresented, wit­tingly or unwittingly, by the author of the press release.

So inaccurate was Harley's original article that he was
forced to retract the claims he had made there after Dr.

Stanley Lipshitz of the University of Waterloo and Robert
Adams of Analog Devices had explained to him his errors.

["Industry Update, US: Robert Harley." Stereophile 16.1
(January 1993): 51-55.] Apparently Harley felt no need to
talk to these experts before running the original article be­cause he thought he heard a dramatic improvement in the
sound quality, and this alone validated any claims by
Sony. My question is, now that Bob Harley knows that
the SBM system is much less than it first appeared to be,
does he still hear such dramatic differences in the sound?

In another article, Harley attempts to explain the de-

sign innovations made by Robert Gendron in the new

Museatex Audio digital processor. ["Industry Update,

Canada: Robert Harley." Stereophile 15.9 (September

1992): 47-51.] Here it not a case of Harley parroting a
press release; instead, he fails to explain the operation of
the system because it is clear that he has not understood
what was being explained to him. In this article Harley
shows that he has no idea what the difference is between
an finite impulse response (FIR) digital filter and an
infinite impulse response (IIR) filter. He then goes on to
claim that the Museatex oversampling filter switches be­tween both FIR and IIR filters, when in fact it uses only
FIR filters in a novel algorithmic adaption scheme and
does not use an IIR filter at all. His explanation of Mr.
Gendron's new S/PDIF decoder is so garbled as to render
the idea virtually unrecognizable.

I obtained an explanation from Museatex of how
this novel S/PDIF decoder works, but since some of the
technical material that was discussed with me may be

proprietary I will not comment on the decoder here. I will

give more details on these design innovations if Museatex
decides to send us a sample of the new D/A converter.
The company has already supplied one to Bob Harley, ap­parently because his power to make or break a company
outweighs his manifest inability to explain what the com-

pany is doing. That is really a shame because the ideas

embodied in this product appear to be truly innovative
and important, and deserve to be presented lucidly.

—David Rich

[I wonder how Larry Archibald, Stereophile's own-
er and President, is able to look himself in the mirror in
the morning when he is shaving and tell himself that he is
running a credible and responsible publication. He is
being told over and over again—not just by us but by
strictly neutral and disinterested parties with impeccable
technical credentials—that Bob Harley is a loose cannon
in his organization and yet he does nothing about it. Not
that Harley is the only one at Stereophile writing techno­babble, but in his case the management appears to regard
that as part of his official job description.

—Ed.]

ISSUE NO. 19 • SPRING 1993

67

pdf 56

Recorded Music

The plan for this column was to have David Ranada and others take it over completely, but David
got busy with higher-priority assignments (although he keeps threatening to do a compleat Stravinsky
Le Sacre du Printemps disco graphy for us here), and those others—where are you? So, it's your

overburdened Editor "once more unto the breach."

Catching Up on the

New and

Not-So-New CDs

By Peter Aczel

Editor and Publisher

The condensed tabular review format I experi­mented with in Issue No. 17 attracted favorable comment
from all those who commented at all; it seems that broad
horizontal coverage is more in tune with the needs of our
readers than in-depth reviews of fewer releases. After a
large accumulation of new CDs on my shelves, I decided
to go back to capsule reviews; however, I had found the
tabular format with its precisely defined fields (in the
data-base sense) to be too rigid as well as needlessly
repetitious, so a capsulized version of my earlier columns
seemed to be worth trying. I went back to the label-by­label approach (as against composer by composer) because
the comments on producers, recording engineers, record­ing techniques, etc., are generally applicable to all recent
releases under the same label and need to be made only
once; it's a more efficient and concise way of organizing
the reviews. Besides, audiophiles are very label-conscious.

I no longer list the SPARS code for each CD because
it's nearly always DDD; only the exceptions need to be
noted. The year in parentheses after each listing always
refers to the recording session or sessions, not the release
or copyright date.

68

The 20-bit bandwagon.

The hot button in the CD world right now is the
conversion of 20-bit digital master tapes, now coming
into wide use as the professional studio standard, to the
required 16-bit standard used in CDs. It looks as if every
company were getting into the act and making sweeping
claims about its special proprietary technology for doing
this, whereas in reality all of them are obviously using
very similar though perhaps not identical noise-shaping
techniques. (See also the "Hip Boots" column in this is­sue.) So far I have heard the good news from Sony (for­merly CBS), Reference Recordings, Telarc, and Dorian,
but I know that there are others. I don't think it's as big a
deal as they want the world to think but it's certainly a de­sirable development because it makes the achievement of
genuine 16-bit resolution in the final product easier and
therefore more likely. I seriously doubt, however, that one
can hear the difference between perfect 16-bit and perfect
20-bit encoding/decoding of recorded music, but maybe
one can between imperfect 16-bit and careful 20-to-16­bit. We shall see; so far there has been only a very thin
trickle of the new CDs and nothing of major importance.

THE AUDIO CRITIC

pdf 57

Bainbridge

The following two au­diophile spectaculars are
actually the work of Man­tovani Production Asso­ciates, recorded in London
with different production
and engineering teams, and
distributed by Bainbridge.

"The Age of Swing," Vol-

ume 1. The BBC Big Band.

BCD 6291 (1991).

An authentic and styl­ish re-creation of the big-

band sound of the '30s and

'40s by the almost too vir-

tuosic BBC players. The

sound is close to perfection

—up-front, very clean, and

very dynamic.

•

"Golden Cinema Classics,"
Volume 1: The Adventure
Film. The BBC Concert Or­chestra. BCD 2521 (1992).

It opens with a "James
Bond Suite" and ends with

the music from Ben Hur. If
you like that sort of thing,
you'll love this CD. The
orchestral playing is first­rate, and the sound is again
super, in an appropriately
unsubtle way—big sound­stage, high definition, in­your-face brass.

Delos

John Eargle is still the
chief recording engineer at
Delos, and in my book he
is still numero uno in the
classical music field. Oth­ers may on rare occasions
equal him or even surpass
him, but they're not nearly
as consistent. The com­pany, under the leadership
of Amelia Haygood, is fur­ther distinguished by its
unique efforts on behalf of
mid-20th-century American
music, long neglected.

•

Paul Creston: Symphony
No. 3, Op. 48; Partita for
Flute, Violin & Strings,
Op. 12; Out of the Cradle;
Invocation and Dance, Op.

58. Seattle Symphony, Ger­ard Schwarz, conductor.
DE 3114 (1991-92).

Paul Creston used to be
performed almost as fre­quently as Aaron Copland;
then his conservative but
robust, colorful, and beau­tifully crafted music—
admittedly not as exciting
as Copland's—fell out of
favor. These are meticu­lous performances, and the
sound is as panoramic,
open, and finely detailed as
you'd expect from John
Eargle.

•

Howard Hanson: Sympho­ny No. 7; Symphony No. 5,

Op. 43; Piano Concerto in
G Major, Op. 36; Mosaics

[1958]. Seattle Symphony

& Chorale, Gerard Schwarz,

conductor; Carol Rosen-

berger, piano. DE 3130
(1992).

This is mostly late, or
late middle, Hanson and
not quite as unequivocally
convincing as some of his
earlier work. The Mosaics
variations are quite pow­erful, though, and Carol
Rosenberger in the con­certo is sensitive and sure­handed, as always. The
sound quality throughout is
strictly late Eargle, mean­ing high-end demo quality.

•

Walter Piston: Symphony
No. 4; Capriccio for Harp
and String Orchestra; Ser­enatafor Orchestra; Three
New England Sketches.
Seattle Symphony, New
York Chamber Symphony
(Serenata), Gerard Schwarz.,
conductor. DE 3106 (1990-

91).

Some of the earlier re­leases in this Piston cycle
on Delos presented more
exciting music, but com­pleteness is the goal here,
and Schwarz/Seattle are as
good in this sort of thing as
anyone could ask for. The
concluding maestoso move­ment of the New England

piece is alone worth the
price of admission, and
John Eargle's recording is
once again glorious.

•

William Schuman: Varia­tions on "America" (Ives,
1891; orch. by Schuman,

1963); New England Trip­tych; Symphony No. 5; Ju­dith (Choreographic Poem

for Orchestra). Seattle

Symphony, Gerard Schwarz,
conductor. DE3115 (1990-

92).

Stiff structures and su-

perbly crafted sound, with­out much feeling—that's
my image of Schuman (the
one with only one n), but
what a vehicle for a John
Eargle audiophile spec­tacular! The orchestrated
Charles Ives organ piece is
loads of polytonal fun,
though (hey, guys, the tune
is "God Save the King,"
not "America"—how pro­vincial can you get?), and
the orchestral climaxes
throughout the CD are
breathtaking.

•

Richard Strauss: Macbeth,
Op. 23; Serenade in E-flat
Major, Op. 7; Ein Helden­leben, Op. 40. Seattle Sym­phony, Gerard Schwarz,
conductor. DE 3094 (1990).
Richard Strauss: Der Ro­senkavalier, Op. 59, 1st &
2nd Sequence of Waltzes;
Burleske for Piano and Or­chestra; Die Frau ohne
Schatten, Op. 65, Sym­phonic Fantasy. Seattle
Symphony, Gerard Schwarz,
conductor; Carol Rosen­berger, piano. DE 3109

(1991).

Gerard Schwarz's fran­chise in the standard Aus­tro-German repertoire is
not as strong—at least in
the opinion of many dis­criminating music lovers—

as in contemporary Amer­ican music, but here he and
the Seattle forces provide a
pleasant surprise. Helden-
leben, for example, is
played with precision and
carefully gauged expres­siveness but without ex­aggeration—a fine effort,
and so are the other pieces.
In terms of sound, you
won't find better Richard

Strauss CDs than these
two, although the compos­er's massed violins are

more difficult to record

with accuracy and beauty
than the sparer textures of

Piston or Schuman—but

John Eargle knows how.

•

Igor Stravinsky: Le Sacre
du Printemps (The Rite of
Spring); Pulcinella (com­plete ballet). Seattle Sym­phony, Gerard Schwarz,
conductor, with soloists.
DE3100 (1990).

The performance of the
Sacre is a bit too careful,
though precise and finely
balanced (maybe only the
absolute best orchestras
can play it savagely and
precisely). The Pulcinella
is not only complete but
very faithful to the 18th­century spirit and in­spiration of the work. John
Eargle's recording of the
Sacre is unusually spa­cious and at the same time
has explosive "bite," not
an easy thing to achieve;
the Pulcinella was done
without him by his as­sociates and very nicely,
too.

"American Diva." Arias
by Verdi, Puccini, Cilea,
Charpentier, Wagner, Cat­alani. Alessandra Marc,
soprano; New Zealand Sym­phony Orchestra, Heinz Wall­berg, conductor. DE 3108
(1991).

What a surprise! What
a voice! Alessandra Marc
may not be a truly great
artist—not yet, anyway—
but in terms of voice pro­duction, power, range, and
just sheer beauty of tone
she is the equal of anyone,
anywhere. Her top notes
are sensational. The or­chestral accompaniments
are a bit on the plodding
side, and the sound—done
without John Eargle, in
Wellington, New Zea­land—is okay but not
great, better for the voice
(fortunately) than for the
orchestra. Regardless of
that, if you're an opera lov­er, get this CD!

"L.A. Guitar Quartet: Danc­es from Renaissance to
Nutcracker." The Los An­geles Guitar Quartet (John
Dearman, William Kanen­giser, Scott Tennant, An­drew York). DE 3132
(1992).

Frankly, I didn't expect
this to be as delicious as it
is. Tchaikovsky, Praetorius,
Gabrieli on four acoustic
guitars? Well, it's a blast!
These four musicians play
with such fantastic rhythm,
phrasing, and color that the
experience is not to be
missed. Don't be an up­tight purist; give it a listen.
Absolutely superb sound,
too, by John Eargle.

Denon

This is the label of Nip­pon Columbia, but unlike
other major international
record companies they
don't have a "corporate"
sound dedicated to market­able mediocrity. Some of
their best recordings sound
as good as the proudest
products of the small au­diophile labels—and, of
course, the artists are gen­erally of greater stature.
Denon's recording team
appears to vary from ses­sion to session, but the
name most frequently list­ed in the credits is that of
Yoshiharu Kawaguchi as
producer. Brüel & Kjær
microphones are standard
in Denon recordings, and
normalization to a single
microphone location by
digital delay is still the
technique used, to the best
of my knowledge. Denon
has never given up high­frequency pre-emphasis on

their CDs, and I'm all for it
(as long as the de-emphasis
in the playback is ac­curate).

•

Béla Bartók: Music for

Strings, Percussion & Ce­lesta, Sz. 106; Concerto for
Orchestra, Sz. 116. Orches­tre de la Suisse Romande,
Eliahu Inbal, conductor.
81757 9044 2 (1989 &

1991).

Two of the towering

masterpieces of twentieth­century music, both in
slower performances than
the composer specified.
The MSP&C, however, is
so finely detailed in texture
and so lovingly phrased
that I can't resist it, espe­cially since the recording is
wonderfully transparent and
silky, with a very strong
bass line. The Concerto is
the earlier taping of two
and could use greater vir­tuosity in the orchestral
playing; it is also less im­pressively recorded. Even
so, a recommended CD.

Joseph Haydn: Symphony
No. 6 in D Major ("Le Ma-

tin"); Symphony No. 7 in C
Major ("Le Midi"); Sym­phony No. 8 in G Major

("Le Soir"). Orchestre de

Chambre de Lausanne,
Jesús Lopez-Cobos, con­ductor. 81757 9612 2
(1991).

Early Haydn is differ­ent from the late sym­phonies everyone knows—
more concerto-grosso-like,
more coloristic, with lots
of virtuoso solo passages.
These are very well-played
performances, thoroughly
musical, not at all hard­driven, authentic in style
and sonority albeit with
modern instruments, and
very pleasant in sound.

•

W. A. Mozart: Symphony

No. 23 in D Major, K. 181;
Symphony No. 28 in C Ma­jor, K. 200; Symphony No.
35 in D Major, K. 385
("Haffner"). Sinfonia Var­sovia, Emmanuel Krivine,
conductor. 81757 9884 2
(1990).

More of the same as
the Krivine/Warsaw CD
reviewed in Issue No. 17,
but who can have enough
of this when it's so nicely
done? I find Krivine's ap­proach to Mozart vigorous
and at the same time sen­sitive. The "Haffner" is the
highlight here. Again, ex­cellent sound—suave and
transparent.

•

Arnold Schönberg: Gurre­lieder. Frankfurt Radio
Symphony Orchestra with
solo singers and 3 chorus­es, Eliahu Inbal, conduc­tor. 2 CDs: 81757 9066 2
(1990).

This supercolossal, ora­torio-like, late-romantic, qua­si-decadent work from 1913
(I'll see your Wagner and
raise you a Strauss and a
Mahler) needs a very firm
hand and an illuminative
musical mind on the po­dium. Inbal has both. If his
singers were better, this
would be a great perfor­mance; even as it is, I found
it very rewarding. The
sound is wonderful in the
quieter passages, but the
big climaxes are just a lit­tle bit strained.

•

Dmitri Shostakovich: Sym­phony No. 5, Op. 47. Frank-

furt Radio Symphony Or-

chestra, Eliahu Inbal,
conductor. 81757 4175 2
(1988).

For some reason I
didn't review this extreme­ly lucid, lovingly shaped,
perhaps too careful per­formance of one of my fa­vorite Shostakovich works
when I received it two

ISSUE NO. 19 • SPRING 1993

69

pdf 58

years ago. The sound is

very similar to that of In­bal's outstanding Berlioz
Symphonie Fantastique of
the previous year, as good
as a Denon can get, but the
Berlioz is more tautly,
more excitingly played.
The music itself is Stalin's
only beneficial legacy—his
brutal, ignorant meddling
was the indirect cause of
its composition.

Karol Szymanowski: String
Quartet No. 1 in C Major,
Op. 37; String Quartet No.
2, Op. 56. Carmina Quar­tet: Matthias Enderle and
Susanne Frank, violins;
Wendy Champney, viola;

Stephan Goerner, cello.

81757 9462 2(1991).

Modernized yet still

overripe romanticism com-

bined with an occasional
touch of almost Bartókian
acerbity but without Bar­tók's fascinating rhythms,
surprises, and bigness of
concept—this is not my fa­vorite music but I can see
why some like it. An early
Webern quartet movement
is also included as a filler.
The playing is extremely
polished, and the string
sound is wonderfully gutty
and vivid.

•

Peter Ilyich Tchaikovsky:

Violin Concerto in D Ma-

jor, Op. 35. Igor Stravin-

sky: Violin Concerto in D.
Jean-Jacques Kantorow,

violin; London Philhar­monic Orchestra, Bryden
Thomson, conductor. 81757
3325 2 (1987).

Another worthwhile re­lease I neglected when it
came out. Kantorow is a

brilliant French violinist
with a leaning toward fast
tempi and sensitive, ele-

gant playing rather than the

big line. The neoclassical,

superbly witty Stravinsky

concerto from the compos­er's middle period suits
that approach perfectly; the
Tchaikovsky could be
more heart-on-sleeve, but
then it's schmaltzy enough
when played with restraint.

(And my taste is vulgar

enough so that I love it, no

matter what!) The orches-

tral framework by Thom-

son and the LPO is highly
satisfactory, and the sound

is vintage Kawaguchi.

Peter Ilyich Tchaikovsky:

Symphony No. 5 in E Mi­nor, Op. 64. Boris Black-
er: Orchestervariationen,

Op. 26. Frankfurt Radio

Symphony Orchestra, Eli­ahu Inbal, conductor.

81757 6364 2(1989).

Peter Ilyich Tchaikovsky:

Symphony No. 6 in B Mi­nor, Op. 74 ("Pathétique ").
Richard Wagner: Vorspiel

und Liebestod, "Tristan

und Isolde." Franfurt Ra­dio Symphony Orchestra,
Eliahu Inbal, conductor.
817579715 2(1991).

The two Tchaikovsky
performances, recorded two
years apart, show a slight
shift by Inbal over that pe­riod toward a livelier, more
dramatic approach. Both
interpretations are in his
deliberate, carefully de­tailed, often illuminative,
and in the final analysis
un-Russian style, but the
Pathétique has greater im­pact. It also happens to be
a much better recording,
one of Denon's best, con­siderably smoother in the
climaxes than the older
CD. The umpteenth set of
variations on the famous
Paganini Caprice by Boris

Blacher (1903-75) is not
the best (nor the second
best) set; as for the Wag­ner, it's lovingly shaped
and tonally beautiful, but
the seething intensity is
missing. (Listen to Tos­canini's 1952 recording
and you'll know what I
mean.)

dmp

Tom Jung is the un­disputed master of New
York studio-style record­ing; his CDs of small and
medium-sized groups on
his dmp (Digital Music
Products) label fairly leap

out of the speakers, with

immense realism, very
high definition, tremendous
dynamics, yet no obvious

gimmickry. You couldn't
ask for better audiophile
demo material. Not surpris­ingly, Tom has been using

19-bit and 20-bit A/D con-

version for years now to

make dmp's master tapes.

"Different Strokes." The

Robert Hohner Percussion
Ensemble. CD-485 (1991).

I think the idea here

was to make the "ultimate"
percussion recording, and
in my opinion the effort
was successful. If you have
big speakers and lots of
watts, you must get this
CD—it will clean out your
ears better than Q-tips. Not
that it's a lowbrow audio­goon special, far from it.
How could it be with the
music of Darius Milhaud

(on five marimbas, yet!),

John Cage, Christopher
Rouse, and other good
20th-century composers?

Some of the sonorities are

delicate, some are over­powering, and the playing
by the 17 percussionists is
as musical as you could
ask for. The fidelity on a
scale of 1 to 10 is an 11.
Recommended to music
lovers, not just drum freaks.

Dorian

Some would argue that
this is the label for audio
quality; my perception is
that their best work is ab­solutely unbeatable but that
their sound is just a little
idiosyncratic in a few re­leases. Craig Dory, one of
the founders and chief re­cording engineer of Dorian,
is a relentless perfectionist
with a very keen and crit­ical ear; he is also a curi­ous combination of highly
credentialed technologist
and tweaky audio fantast.
He obviously doesn't like a

close-up, dry sound (I do
for certain kinds of music),
preferring a considerable
amount of ambience in his
recordings. In the Troy
Savings Bank Music Hall
he sometimes goes over­board in his celebration of
the great hall sound, but in
the marvelous Eugene Mc­Dermott Hall of the Mor­ton H. Meyerson Sym­phony Center in Dallas he
achieves the current pin­nacle of digitally recorded
orchestral sound. The pro­ducers are not always the
same in Dorian recordings,
and lately Craig Dory has
been delegating the en­gineering of lighter ses­sions to his associates.
Coming Dorian releases
will originate from 20-bit
A/D coded masters.

J. S. Bach: "The Organ

Works of Johann Sebastian

Bach," Volumes 2 through

5. Jean Guillou on the
Great Kleuker-Steinmeyer
organ of the Tonhalle,

Zürich. DOR-90149/50/51/

52 (1990).

After the initial Vol-

ume 1 recorded in 1987 at

the Église Notre-Dame des

Neiges in Alpe d'Huez,
France (see my rave re­view in Issue No. 13), this
monumental project moved

to Switzerland, where the

nearly five hours of music
on these four CDs were all
recorded in a marathon ses­sion in August 1990, then
released piecemeal. The
producer, Randall Fostvedt,

still speaks in awed tones
about Jean Guillou's non-

stop feat of musical endur­ance. The Zürich Tonhalle
organ is a much larger and

more imposing instrument
than the one at the French

ski resort, but both were
designed by Guillou and

share some of the same ir­resistible tonal qualities.
The big toccatas, preludes,

fugues, and other major
pieces in the Bach organ

canon fare particularly well

on the larger instrument,
but the smaller-scale works

also sound wonderful be-

cause the polyphony re-

mains crystal-clear regard­less of the nature of the
music. As for the per­formances, all organists I
am familiar with sound
boring next to Guillou,
who may not be most liter­ally accurate and respectful
Bach interpreter, just the
most musical and cap­tivating. He proves that
these icons of polyphonic
composition can be fresh
and exciting. The recorded
sound is simply the best
there is in organ music; the
same crew in the same
place was responsible two
years earlier for the Mus­sorgsky-Guillou "Pictures"
that the hippest audiophiles
still use as a test record.

•

"An American Panora-

ma." Leonard Bernstein:
On the Waterfront. Roy

Harris: Symphony No. 3.
Aaron Copland: Billy the

Kid. Dallas Symphony Or­chestra, Eduardo Mata, con­ductor. DOR-90170(1992).

I compared this version
of Billy the Kid with the
Schwarz/Seattle recording
on Delos, and it's a tough
choice. Mata has the better
hall, by far, and the better
orchestra, by a narrower
margin. Schwarz has a bet­ter understanding of the
Copland idiom—a much
snappier, jazzier perfor­mance. Mata is too old­world and careful (as he is
in the other two American
works). Both recordings
are superb, the Dorian a lit­tle smoother overall and
cleaner in the climaxes, the
Delos more close-up and
explosively dynamic. It's a
nice dilemma. As for Bern­stein and Harris—superior

movie music and some­what arid "serious" music,
respectively—Mata doesn't
quite have his heart in the
performances.

•

Johannes Brahms: Piano
Trio No. 1 in B Major, Op.

8. Antonín Dvorák: Piano
Trio in E Minor, Op. 90
("Dumky"). The Rembrandt
Trio: Valerie Tryon, piano;
Gerard Kantarjian, violin;
Coenraad Bloemendal, cel­lo. DOR-90160 (1991).

My personal prejudice:
most of Brahms's chamber
music is boring; most of
Dvorák's chamber music is
delightful. (So sue me.) I
find this pairing no excep­tion, but the playing by the
Canadian trio is first-rate
throughout. The recording
is quite lovely but would
be even better without the
far too reverberant Troy
hall signature (and Craig
Dory wasn't even in on
this one).

•

Antonín Dvorák: "Com-

plete Music for Violin and

Piano." Ivan Zenaty, violin;

Antonin Kubalek, piano.

DOR-90171 (1992).

This is one of the most
natural-sounding recordings
of the violin I know of,
again a Craig-Dory-less re-

cent production in the Troy

hall. The piano is a little

remote but the liveness is
just right. The program is
all of the 75 minutes of
music Dvorák composed

specifically for violin and

piano; none of it is great
but all of it is lovely. The
two Czechs play it con
amore; it's pure joy from
beginning to end.

Joseph Haydn: Symphony
No. 82 in C Major ("The
Bear"); Symphony No. 38
in C Major; Symphony No.

104 in D Major ("Lon-

don "). Los Angeles Cham­ber Orchestra, Christof
Perick, conductor. DOR­90168 (1992).

The immortal "London"

receives a somewhat less
than immortal performance
here; it's merely well
played—very carefully and
lucidly—as are the earlier
symphonies. The record­ing, however, proves that
Craig Dory tends to hit a
home run when he comes
to a new venue (in this
case Royce Hall on the
UCLA Campus, Los An­geles) and only later starts
to tweak around with the
sound. Here the sound is
all creamy goodness, trans­parency, and just-right am­bience. Please do it exactly
the same way next time,
Craig!

•

Joseph Haydn: Piano Tri-

os. Trio in E Minor, Hob.
XV: 12; Trio in C Major,
Hob. XV:27; Trio in E Ma­jor, Hob. XV:28; Trio in G
Major, Hob. XV:25. Ka-

lichstein-Laredo-Robinson

Trio: Joseph Kalichstein,
piano; Jaime Laredo, vi-

olin; Sharon, Robinson,

cello. DOR-90164 (1991).

Wonderful music, won­derful playing by a world­class trio, wonderful re­cording by a Dorian team
minus Dory. I could listen
to this all day long; each
trio, each movement is bet­ter than the last, and much
of it seldom performed.
These three superb musi­cians play as one, and the
Troy hall is being used cor­rectly here, as a frame­work, not as a feature. You
have to be tone-deaf not to
love this CD.

•

Franz Schubert: Winter­reise, D. 911. Victor Braun,
baritone; Antonin Kubalek,

piano. DOR-90145 (1990).

If Schubert was the

70

THE AUDIO CRITIC

pdf 59

greatest composer of lieder,
as he almost surely was,
and if Winterreise is his
masterpiece, as it is widely
considered to be, then if
you have only one lieder
cycle in your collection,
this should be it. The ques­tion is, should it be Victor
Braun's performance? Well,
he has a beautiful voice,
very strong and free on
top, and he sings most ex­pressively, with excellent
musicianship and without
any mannerisms. Fischer­Dieskau, for one, has more
"personality," but that isn't
necessarily the definitive

way of singing this. Kuba-

lek's sensitive accompani­ments are another point in

favor of this recording,
which has my highest rec­ommendation. It would
have made me even happi­er if Craig Dory had

moved in a little closer on
this intimate music and
kept the swimmy Troy hall

ambience out of it, but

c 'est la vie.

Dmitri Shostakovich: Sym­phony No. 7 in C Major,

Op. 60 ("Leningrad").
Dallas Symphony Orches­tra, Eduardo Mata, con­ductor. DOR-90161 (1991).

I think this music is of
quite limited artistic value,
possibly the composer's
weakest major work, al­though it served its purpose
as patriotic propaganda in

1942. Half a century later,
with even the name Lenin­grad gone, all that is left is
that insufferably long and
banal first movement and
then more in the same vein.
This performance is as

good as I have ever heard,

vigorous, precise, with
great playing by the brass­es, and the recording is
stunning, confirming the
status of the Craig Dory
sound in the Dallas hall as
one of today's major au­diophile experiences.

Erato

This distinguished and
at one time independent
European label originated
in France but is now, to the
best of my knowledge, part
of the Time Warner em­pire, tied in with Teldec.
The producers varied in the
Chicago recordings dis­cussed below, but the re­cording engineer in each
case was Larry Rock (great
name for a classical re­cordist). It should be men­tioned that Daniel Baren­boim took command of the

great Chicago orchestra
only a few months before
these recordings were made.

Gustav Mahler: Das Lied
von der Erde. Waltraud

Meier, mezzo-soprano; Sieg-

fried Jerusalem, tenor;

Chicago Symphony Or­chestra, Daniel Barenboim,
conductor. 2292-45624-2
(1991).

With better singing this
might have been a very sat­isfying performance of
Mahler's crowning master­work because Barenboim
understands the idiom and
the orchestra plays mag­nificently. Jerusalem, how­ever, screams instead of
singing when the going
gets tough, and Meier is

just mediocre. The record-

ing is very good though a
bit harsh in the climaxes.

•

Maurice Ravel: Daphnis

et Chloé (Suite No. 2);
Rapsodie espagnole; Pavane

pour une infante défunte;
Alborada del gracioso; Bo-

léro. Chicago Symphony
Orchestra, Daniel Baren­boim, conductor. 2292­45766-2 (1991).

With an orchestrator
like Ravel and an orchestra
like the Chicago, you can't
miss if you just play all his
notes, and Barenboim cer­tainly does that, with a
great deal of control and
beauty of sound. That extra
magic, however (the Pierre
Monteux kind), just isn't
there. The recording is
great, the cleanest and
most spacious of the Chi­cago releases on this label
so far, with a particularly
nice bass drum.

•

Richard Wagner: Der Ring
des Nibelungen (excerpts).
Deborah Polaski, soprano;

Chicago Symphony Or­chestra, Daniel Baren-

boim, conductor. 2292-

45786-2 (1991).

I grew up on Toscani­ni's and Szell's Wagner,
and currently I favor James
Levine. This just isn't as

exciting. The five excerpts

(Walkürenritt, Waldweben,

Rheinfahrt, Siegfrieds Tod,

and the final immolation
scene) are played with
great clarity and gorgeous
orchestral sound, but the
grandeur is missing and
Polaski is a highly forget­table Brünnhilde. The re­cording is excellent overall
but again a little strained
on brassy climaxes.

Harmonia Mundi

This is a French label,

but their scope is inter-

national and their USA
affiliate distributes many
more small international
labels. The audio quality of
their recordings has been
variable; those made by
Peter McGrath are gener­ally superb.

George Frideric Handel:

Messiah. Six soloists; U.C.
Berkeley Chamber Chorus;
Philharmonia Baroque Or­chestra, Nicholas McGegan,
conductor. HMU 907050­52 (1991).

Handel changed the
score of Messiah many
times to suit available sing­ers, and this is the first and
only recording that offers
every note of all the differ­ent versions, with pro­grammable indexing. The
performance is of course
superauthentic à la McGe­gan; the orchestral playing
is wonderful; the chorus is
weak; the soloists are ac­ceptable to pretty good; the
Peter McGrath recording is
marvelously lucid.

•

George Frideric Handel:

Giulio Cesare. Jennifer

Larmore, Barbara Schlick,
et al., Concerto Köln, René
Jacobs, conductor. HMC
901385-87 (1991).

If you get to know only

one Handel opera, this
should be the one, his best,
with many truly brilliant

moments. This is a gen-

uinely enjoyable but very
far from great period­practice performance; the
same can be said of the re­cording, which is clean and
transparent but often thin
and too bright. One big
point for Handelians: no
cuts.

London

Of all the big inter­national labels using more
or less standard multi­miking techniques, London
(i.e., English Decca) comes
closest to what could be
called an audiophile sound.
The recording engineer re­sponsible for that is usually
John Dunkerley, Colin
Moorfoot, or John Pellowe.

Hector Berlioz: Symphonie

fantastique; L'Invitation à

la valse (Weber, arr. Ber­lioz). The Cleveland Or­chestra, Christoph von Doh­nányi, conductor. 430 201­2 (1989).

Flabbergastingly bril­liant playing by the Cleve­landers that just bowls me
over—what a band!—but
Dohnányi is a bit too stiff
and controlled in this mu­sic, perhaps because he
wants to set some kind of
new record in precision.
The sound, recorded by
John Pellowe in the Ma­sonic Auditorium, is won­derfully detailed down to
the lowest notes, as good
as a Decca can get.

•

Robert Schumann: Sym­phony No. 1 in B-flat Ma-

jor, Op. 38 ("Spring");

Symphony No. 4 in D Mi­nor, Op. 120. Royal Con-

certgebouw Orchestra, Ric­cardo Chailly, conductor.
425 608-2 (1988).

No. 4 is the better work
in my opinion, with some
beautiful passages, but nei­ther one is a truly great
symphony. I prefer Schu­mann as a composer for
the piano. Chailly has no
special feeling for these
works, either; he conducts
them with his usual vigor
and that's that. The Am­sterdammers play with the
utmost virtuosity and beau­ty of tone, as always; the
recorded sound is round
and clean but could be
more transparent (probably
Schumann's fault, not the
Decca engineers').

Marco Polo

This European "label
of discovery" (as they call
themselves) is distributed
in this country by Har­monia Mundi USA; the
CDs are manufactured in
Germany.

•

Havergal Brian: Sympho-

ny No. 1 ("The Gothic").
Four soloists; seven cho­ruses; CSR Symphony (Bra­tislava) and Slovak Phil­harmonic, Ondrej Lenard,
conductor. 8.223280-281
(1989).

Havergal Brian, an

Englishman, died in 1972
at the age of 96 and left a
legacy of 32 symphonies,
hardly ever performed.
This, his first, is something

of a cult item because of
the gigantic forces needed
to perform it; the orchestra
alone requires almost 200
players, and then there are
the huge double choruses,

soloists, etc. (You can look

it up in The Guinness Book
of Records. No kidding.)
This recording evokes

Samuel Johnson's "like a
dog's walking on his hind­er legs—it is not done
well, but you are surprised
to find it done at all." Not
that it's done badly, far
from it, but I think only a

Solti/Chicago type of per­formance of a bloated and
somewhat incoherent, epi­sodic work like this could
give it genuine shape and
thrust, and this is not quite
on that level, although the
playing is highly pro­fessional. As an audiophil­ic curiosity, however (the

Strauss/Mahler/Schönberg
sound cubed), I can rec-

ommend this recording be-

cause the sound is amaz-

ingly good—transparent and

panoramic, with a wide dy-

namic range, clean highs,

strong bass, and only a few

moments of strain.

MusicMasters Classics

An affiliate of the BMG

group (which now also in­cludes RCA Victor), this
label has an international
stable of top artists and
works with world-class pro­ducers and recordists like
Max Wilcox.

•

J. S. Bach: Goldberg Vari­ations. Vladimir Feltsman,

piano. 01612 - 67093 - 2

(1991).

Recorded live in Mos­cow, this performance is
just a few seconds short of

80 minutes because every
repeat is observed. To
avoid tediousness, Feltsman
jazzes up the repeats by
playing them in a different
register, switching the
voices by crossing hands,
etc. Some will be offended
by this; I am not, but I still
like Glenn Gould's Bach a
lot better. Feltsman has the
chops but doesn't have ma-

jor insights to offer. The

sound is basically good but
a little dry and thin, with
some audible coughs in the
hall.

•

Igor Stravinsky: Pulcinella
Suite; Symphony in C; Rus­sian Peasant Choruses;
Russian Sacred Choruses;
Les Noces. The Orchestra
of St. Luke's & The Gregg
Smith Singers, Robert
Craft, conductor. 01612­67086-2 (1992).

All of this is first-chop
middle-period Stravinsky,
pungent and wonderfully
listenable. It is Volume II
of the complete Stravinsky
canon being recorded by
Robert Craft, who as the
composer's closest asso-

ciate is entitled to whatever
performance style he sees
fit. For my musical palate

he is a little sec—in the

Pulcinella he sounds like a
Nicholas McGegan on ste­roids demonstrating pe­riod-practice Frescobaldi—
but the orchestra is very
good, the singers are very
good, Craft's grasp of the
music is unquestionably of
a high order, and the whole
thing works as a musical
experience. The recording
is extremely lucid and
clean, just like the playing.

RCA Victor Red Seal

The granddaddy of all
labels is now part of the
BMG Music empire—

nothing is sacred anymore.
Their sound quality has
been good lately but ba-

sically middle-of-the-road.

•

Antonín Dvorák: Piano

Quintet in A Major, Op.
81; Piano Quintet in A Ma-

jor, Op. 5. Rudolf Firkus-

ny, piano; Ridge String
Quartet. 09026-60436-2
(1990).

Antonín Dvorák: Piano

ISSUE NO. 19 • SPRING 1993

71

pdf 60

Concerto in G Minor, Op.

33. Leos Janácek: Con­certino; Capriccio. Rudolf
Firkusny, piano; Czech

Philharmonic Orchestra,

Václav Neumann, con­ductor. 09026-60781-2
(1990-91).

Dvorák and Firkusny
go together like ham and
eggs, like Shakespeare and
Olivier. He was 78 and 79
when he recorded these
works, but the lovely Op.
81 quintet and the almost
as lovely piano concerto
still sing with incompa­rable lyricism under those
magic fingers. As for the
quirky but fascinating Ja­nácek pieces, he is the au­thority, and of course the
Czech Philharmonic can't
hurt in Czech music. The
recorded sound is un­spectacularly excellent in
each case. Recommended!

Reference Recordings

This is the small inde­pendent label founded and
run by Tarn Henderson; its
audio engineering wizard
is the legendary Keith
("Prof.") Johnson, whose
best recordings are as good
sonically as anything in the
world and whose worst are
still very respectable.

•

Malcolm Arnold: Over-

tures. The London Philhar­monic Orchestra, Malcolm

Arnold, conductor (David

Nolan, leader). RR-48CD
(1991).

This is rather light­weight but colorful or­chestral fare by England's
most famous trumpeter, al­most as famous film-music
composer, and less famous
symphonist. The recording,
however, is simply stag­gering, possibly Keith
Johnson's masterpiece and
a contender for the title of
Best Demo CD. Both the
hall acoustics and the bril­liant instrumental textures
are sensational—and this is
from before RR's switch to

the 20-bit HDCD system. I
won't dwell on it further;

just go out and get it!

•

"Fennell Favorites!" Fred-

erick Fennell live with the
Dallas Wind Symphony.
RR-43CD (1991).

"Fiesta!" Dallas Wind Sym-

phony, Howard Dunn, con-

ductor. RR-38CD (1990).

The Meyerson/Dallas
acoustic raises these terrif­ic, in-your-face Keith John­son band recordings to
demo level. Excellent band;
splendid Hispanic sonor­ities in "Fiesta!" (the bass
drum is awesome); solid
classics (Bach, Brahms,
Prokofiev, etc.) transcribed
for band on the Fennell CD.

Telarc

Michael Bishop has
taken over some of Jack

Renner's responsibilities as
Telarc's recording engi­neer; the Levi/Atlanta re­cordings, for example, are
entirely his currently. The
Telarc sound I have praised
is a house cuvée, however,
and remains constant.

J. S. Bach: "Switched-On
Bach 2000." Wendy Car-

los. CD-80323 (1992).

As a 25th-anniversary se­quel to the original Carlos
electronic Bach album, this
sounds perhaps a little less
fresh and startling musical­ly but a lot more sophisti­cated electronically. Check
out the concluding Toccata
and Fugue in D Minor for
the ultimate in synthesized
timbres for organ music.
Soundwise, your system is
the only limit here.

•

Sir Edward Elgar: Sym­phony No. 1 in A-flat Ma-

jor, Op. 55; Pomp and Cir-

cumstance Marches, Op. 39.
Baltimore Symphony Or­chestra, David Tinman, con­ductor. CD-80310 (1991).

Every audiophile needs
a blockbuster version of
Pomp and Circumstance
March No. 1, and this does
the job in spades. The aes­thetic fallout of such a
crass pursuit is a beauti­fully shaped, fine-sounding
performance of the much
more important Symphony
No. 1. This is a good or-

chestra, no doubt about it.

•

George Frideric Handel:

Messiah. Four soloists; Bos­ton Baroque Orchestra &
Chorus, Martin Pearlman,
music dir. & harpsichord­ist. 2CD-80322 (1992).

When it rains it pours;
here's another fine period­instrument, period-practice
performance of Messiah,
perhaps even better than
the one on Harmonia Mun­di but offering only the
standard version without

the extra goodies. The
chorus is clearly better; the
singers are somewhat better;
the conducting isn't better;
and the recording is more
conventionally creamy-rich
in texture.

Bedrich Smetana: Má
Vlast (My Country). Mil­waukee Symphony Orches­tra, Zdenek Macal, con­ductor. CD-80265 (1991).

That this is the son-

ically best-ever recording
of this beautiful and col­orful cycle of tone poems
is almost indisputable. Jack
Renner outdid himself; the
sound is rich, suave, and
powerful beyond belief.
The performance is argu­able. Some would surely
prefer a more lyrical, more
heart-on-sleeve, more typ­ically Czechoschmaltzian
approach. Macal's inter­pretation is warm yet

strong and firmly struc­tured, not at at all senti­mental. I buy it; you may
not. But his name is Zdenek
and yours is probably Bob.

•

Igor Stravinsky: Le Sacre
du printemps (The Rite of
Spring); Pulcinella Suite.
Atlanta Symphony Orches­tra, Yoel Levi, conductor.
CD-80266(1991).

Another title contender
for Best Demo CD. The
dynamic range and the
accuracy of instrumental
delineation are just amaz­ing. The various drum ex­plosions in the Sacre are
what audiophiles are will­ing to die for. The Atlanta
orchestra's playing under
Levi is extremely discip­lined, controlled, precise;
the pacing is good in the
Sacre; and the Pulcinella is
equally well played.

Teldec

The star label of the
Time Warner empire.

Ralph Vaughan Williams:

Symphony No. 6 in E Mi­nor; Fantasia on a Theme
by Thomas Tallis; The
Lark Ascending. BBC Sym-

phony Orchestra, Andrew

Davis, conductor. 9031­73127-2 (1990).

Truly great, profound
music, very knowledgeably
performed and well record­ed (for a big company).

72

THE AUDIO CRITIC

Coming:

In-depth reviews of power amplifiers in various price and power categories, with
circuit critiques by Dr. Rich.

More loudspeaker reviews, including tests of recent models from Snell, DCM,
ACI, Monitor Audio, and others.

An original report presenting first-time insights into delta-sigma ("1-bit") converters,

including a review of a new state-of-the-art DAC design.
Our first look at the still controversial Digital Compact Cassette system, including a

review of the top-of-the-line Philips/Marantz DCC deck.
Test reports on reference-quality Pioneer Elite rear-projection TV and

laser disc equipment.
A probing evaluation of FM tuners and indoor antennas, both in

theory and in practice.
More "Hip Boots" (because you can always rely on the audio press for misinformation),

more letters (because the customers always write), and of course all

our other regular features.

pdf 61

Subscription Information

and Rates

First of all, you don't absolutely need one of our printed subscrip­tion blanks. If you wish, simply write your name and address as legibly
as possible on any piece of paper. Preferably print or type. Enclose with
payment. That's all. Or, if you prefer, use VISA or MasterCard, either
by mail or by telephone.

Secondly, we have only two subscription rates. If you live in the
U.S., Canada, or Mexico, you pay $24 for four consecutive issues
(mailed at approximately quarterly intervals, barring unscheduled
delays). If you live in any other country, you pay $38 for a four-issue
subscription by airmail. All payments from abroad, including Canada,
must be in U.S. funds, collectable in the U.S. without a service charge.

You may start your subscription with any issue, although we feel
that new subscribers should have a few back issues to gain a better un­derstanding of what The Audio Critic is all about. We still have Issues
No. 11, 13, 14, 16, 17, and 18 in stock. Issues earlier than No. 11 are
now out of print, as are No. 12 and No. 15. Please specify which issues
you want (at $24 per four).

One more thing. We don't sell single issues by mail. You'll find
those at somewhat higher cost at selected newsdealers, bookstores, and
audio stores.

Address all subscriptions to The Audio Critic, P.O. Box 978,

Quakertown, PA 18951-0978. VISA/MasterCard: [phone].

pdf 62