In this issue:
We present an unusually comprehensive survey of
Mcintosh stereo and home-theater components.
The long-promised in-depth coverage of FM
technology, tuners, and indoor antennas begins, with
a classic "tutorial" by Dr. Rich as the main feature.
An extended analysis of the different approaches to
surround sound is followed by critical evaluations
of multichannel audio-visual electronics.
We bring you our expected variety of loudspeaker
reviews, in the price range from modest to insane.
Plus many other test reports, all our
regular columns, letters to the Editor,
and a full load of CD reviews.
Do you need it? No. Do you want it? Yes!! Can
you afford it? Hm... (See the Mcintosh reviews.)
Retail price: U.S. $7.50, Can. $8.50
Issue No. 23
Display until arrival of
Issue No. 24.
pdf 1
ISSUE NO. 23 • WINTER 1995-96 1
Contents
Mcintosh Today (What the "High End" Was Originally Meant to Be)
By Peter Aczel, Editor and Publisher
7 Stereo Power Amplifier: Mcintosh MC500
9 Multichannel Power Amplifier: Mcintosh MC7106
11 Multichannel A V Control Unit: Mcintosh C39
12 AM/FM Tuner: Mcintosh MR7084
13 Loudspeaker System: Mcintosh XRT24
14 Home Theater Loudspeakers: Mcintosh HT-3 and HT-4
8 Sidebar: Inside the Mcintosh Electronics
By David A. Rich, Ph.D., Contributing Technical Editor
Loudspeaker Systems, from Good Value to Ultimate (and In Between)
By Peter Aczel, Editor and Publisher
15 Atlantic Technology System 250
16 Hsu Research HRSW12V
17 Mcintosh XRT24 (cross-reference to Mcintosh article)
17 Paradigm Eclipse/BP
18 Snell Acoustics Type A
38 Snell Acoustics Type MC LCR 2800 and SUR 2800
Catching Up on Analog and Digital Electronics for Stereo
By Peter Aczel, Editor and Publisher
21 Outboard D/A Converter with HDCD: EAD DSP-7000 Series III
22 Stereo Power Amplifier: Mcintosh MC500 (cross-reference to Mcintosh article)
22 Mono Power Amplifier: Sentec PA9
23 Line-Level Stereo Preamplifier: Sentec SC9
24 Outboard Phono Preamplifier: Sentec PP9 RIAA
24 Compact Disc Player: Sony CDP-XA7ES
25 Stereo Power Amplier: Sunfire (follow-up)
Twice Shy: On Reencountering Multichannel Music Formats
By Daniel C. Sweeney, Ph.D., Freelance Contributor to The Audio Critic
AV/Ambience/Surround Equipment:
Our First Look at Multichannel Processors
By Peter Aczel, Editor and Publisher
39 B&K AVP2000
40 Lexicon CP-3PLUS
41 Marantz AV600
42 Mcintosh C39 (cross-reference to Mcintosh article)
42 Multichannel "Receiver:" Onkyo TX-SV909PRO
FM Tuners: The Present State of the Art of FM Reception
By David A. Rich, Ph.D., Contributing Technical Editor
59 Denon TU-680NAB
60 Harman Kardon TU9600
61 Mcintosh MR7084 (cross-reference to Mcintosh article)
61 Onkyo T-9090II
62 RotelRHT-10
58 Sidebar: How I Evaluate FM Tuners
By Richard T. Modafferi, Technical Consultant to The Audio Critic
Indoor Antennas and Boosters for FM AudioPrism, Magnum Dynalab, et al.
By David A. Rich, Ph.D., Contributing Technical Editor
Commodities and Revivals (It Don't Need Revivin' Unless It's Already Dead)
By Tom Nousaine
Hip Boots Wading through the Mire of Misinformation in the Audio Press
Nine commentaries by the Editor
Recorded Music
75 Mostly Wagner and Mahler: CD Releases of Importance by David Ranada
78 Capsule CD Reviews by Peter Aczel
Box 978: Letters to the Editor
6
15
21
27
39
43
65
70
71
75
3
pdf 2
2 THE AUDIO CRITIC
From the Editor/Publisher:
The elapsed time between this issue and
the last one has been the longest in our history. Two relevant facts must be noted.
One is obvious: this is, in effect, a double
issue. I could have split it down the middle,
made a few editorial adjustments and additions, and published it as two issues, twice
as fast. You would have been charged
twice. This way you 're getting (almost) two
issues for the price of one. The other fact,
which you couldn't have known, is that we
were about to be plugged into a large,
prosperous, and well-staffed publishing
company, supposedly without losing any of
our editorial autonomy but with all their
forces mobilized to help us to publish on
schedule. About six months were wasted
before I faced the fact that they were just
talking the talk but not walking the walk,
and I was back to square one. The idea,
however, was basically sound and should
sooner or later be realizable with other
partners. As I wrote last time, one way or
the other, we 're here to stay.
Address all subscriptions to The Audio Critic, P.O. Box 978,
Quakertown, PA 18951-0978. VISA/MasterCard: (215) 538-9555.
Fax: (215) 538-5432.
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, by telephone or by fax.
Secondly, we have only two subscription rates. If you live in the
U.S., Canada, or Mexico, you pay $24 for four consecutive issues
(scheduled to be mailed at approximately quarterly intervals). 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, and 16 through 22 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.
Subscription Information and Rates
Contents of this issue copyright © 1995 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
Technical Consultant (RF) Richard Modafferi
Columnist Tom Nousaine
Cartoonist and Illustrator Tom Aczel
Business Manager Bodil Aczel
Issue No. 23 Winter 1995-96
pdf 3
Box 978
Letters to the Editor
Our letters column is short this time, not because of a change in policy but as a result of preemption
by other editorial material. We should be back to our usual six to eight pages next time. 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 18951-0978.
The Audio Critic:
For some reason I don't understand,
when we attack a moral wrong, we usually focus on the supply side of the problem. For instance, we fault the prostitute
and drug dealer more than the john (or
jane) and drug user. And yet, if the mar-
ket side dries up, the supply side usually
follows suit.
In your ongoing war with Stereo-
phile and TAS and the myths they preach,
your focus is on the suppliers and supply
of untruths, with the hope that as the audiophile reads this discourse, he or she
will recognize the truth and will be set
free. And you're to be commended for
trying because sometimes it works.
But the greater truth is that the ma-
jority of audiophiles don't want to hear
the truth. If someone suddenly were to
discover a technology which was simple
but had the capability of reproducing mu-
sic indistinguishable from a live performance (by A/B blindfold comparison
testing, of course), and particularly if this
technology were inexpensive so that system price fell in the "consumer electronics" category, the majority of audiophiles
would reject it outright. Why? Because it
would eliminate their excuse to diddle
endlessly with their systems and spend
fortunes doing it. The majority of audiophiles don't want perfection. They want
to be constantly striving for perfection.
Their underlying motivation is not the en-
joyment of music; it's playing with hard-
ware.
A few years ago when I was an "audiophile" and subscribed to Stereophile,
there was a letter in one issue from an audio salon dealer. He had done well in the
business and had decided to rebuild his
store, and to include in it a small but
acoustically correct auditorium in which
live performances could be presented for
the enjoyment of his customers. After
completion, what he found to his dismay
was that he not only couldn't sell tickets
to most of his customers, they wouldn't
even attend for free.
What is this phenomenon? People
have certain areas in which they excel.
Men usually have an aptitude in mechani-
cal and physical pursuits, and so excel in
interests like stereo, ham radio, automobiles, sports, etc. The problem is that
where we excel, we tend to overdo or become silly. And, as is well known, men
get involved in their pursuits to the detriment of more important things—like
their families. Women do the same, except their aptitudes and excesses usually
fall in the areas of their homes, children,
and mouths.
Harley, Atkinson, Olsher, Pearson,
etc., may believe the stuff they write, or
perhaps they know better but realize
there's a bundle to be made by telling the
"market" what they want to hear. Paul in
his second letter to Timothy said, "For
the time will come when men will not put
up with sound doctrine [no pun intended,
I'm sure]. Instead, to suit their own desires, they will gather around them a
great number of teachers to say what
their itching ears want to hear. They will
turn their ears away from the truth and
turn aside to myths" (II Tim. 4:3 & 4).
He was talking about something far more
important than music reproduction, of
course, but the truth applies nevertheless.
Hartley Anderson
Long Beach, CA
P.S. I have no way of knowing this
but am inclined to think that J. Gordon
Holt, in his desire to publish regularly,
reach a larger audience, and be less on a
shoestring, wound up giving his publication to the Devil. As much as I get impatient for your next ever-late issue, please
don't succumb to any temptations that
might adulterate your publication.
I wish you were wrong about what
drives the audiophile, but you're right.
—Ed.
The Audio Critic:
Can it be true? On page 11 of The
ISSUE NO. 23 • WINTER 1995-96 3
pdf 4
Audio Critic, No. 22, it's stated that pow-
er-output specification must take the form
of "200 watts rms into 8 ohms..." and it's
implied that this is required by the Federal Trade Commission.
It's the "must" that's surprising. I'm
no longer surprised by "watts rms" in advertising or even in owner's manuals, but
it is depressing to think that it might be
embodied in a government requirement.
Not that I'm surprised by government
mistakes (it's much more surprising
when the government gets something
right), but somehow this doesn't seem
like the kind of nonsense which would
become official, considering the amount
of participation by knowledgeable people
from industry which must have gone into
the creation of such a requirement.
Say it isn't so.
Sincerely,
Dick Sabroff
Lake Mills, WI
It probably isn't so, although I can't
swear without looking up the regulation,
and that shouldn't be necessary after the
following explanation:
You are right, of course—there is no
such thing, strictly speaking, as rms power. There is rms voltage and rms current,
but voltage times current—which is power—cannot be expressed in rms form. I
said "strictly speaking," but in the early
days of audio we didn't speak strictly,
and that's where my sloppy locution originated. I tend to lapse into it from time to
time as my dotage approaches. The correct expression is "continuous power,"
and after some severe raps in the early
1970s from Paul Klipsch and other leading practitioners, the audio engineering
community pretty much dropped the incorrect nomenclature. Whether the FTC
originally used "rms" or "continuous" is
academic at this point; the latter is correct, the former is wrong, you were right,
I was careless, and that's that.
—Ed.
The Audio Critic:
...One thing that has always puzzled
me about tweako reviews of electronic
gear is the way in which a reviewer has
the remarkable ability to hear the particular sonic signature of a component under
review through all the presumed "colorations" imposed by the intervening equipment. For instance, how can a reviewer
accurately characterize the sound of, say,
a D/A converter if its signal must subse-
quently pass through two line-level interconnects, a preamplifier, a power amplifier, and speaker cable—all of which presumably superimpose their own sonic
signatures as well—before reaching the
speakers? One would think that there
would be significant masking effects
which would make reviewing an upstream component especially difficult. It
would seem that the tweaks cannot have
it both ways, with each component hav-
ing a distinct, audible sound quality while
one is able to isolate and characterize that
particular sound quality through the sonic
characters of all the other intervening
components....
Sincerely,
Jeffrey B. Plies
Watertown, MA
When you enter cloud-cuckoo-land,
the rules of clear reasoning are suspended.
You somehow expect the fantasy world of
the tweako reviewer to have some kind of
internal logic—if my D/A converter
sounds "dark" how can I be sure it isn't
my preamp?—but that's not the way the
tweako mind works. If the D/A converter
was the last thing inserted into the system, it now dominates the sound!
—Ed.
The Audio Critic:
Your comment in Issue No. 22
about tweakos and the "audio performance" of light switches suggests you
may have dealt with this matter somewhere in Issues 1-10, but just in case...
My experience in the Navy with
some of our (then, 1970/80s) technology
indicated that reliable electrical power
sources made an enormous difference in
the performance of our gear. As you may
know, we used 440V/400-cycle wellregulated electrical power on shipboard.
In port, we did not even try to keep our
sensors and weapons aligned with each
other because the supplied 117V/60-cycle
power never reliably was that. Superficially, then, there does seem to be some
merit in various tweaks such as better
connectors (e.g., hospital-grade outlets),
better house electrical-service grounding,
maybe even better power cords, and so
on. In these cases, ABX testing is mostly
impractical. I don't think any such fixes
can hurt, only unnecessarily impoverish.
Is this an area you have covered or might
consider covering?
Michael T. Corgan, Ph.D.
North Falmouth, MA
A few comments on power-line conditioners appear on p. 60 of Issue No. 16.
There are residences (not too many) with
grungy power coming out of the wall
(low voltage, line noise, etc.); each case
needs to be dealt with individually. Good
outlets, good connector hardware, good
plugs, good cords, good grounding, etc.,
prevent intermittents, assure unimpeded
signal transmission, and hardly ever need
replacement; they are highly recommend-
ed but have nothing to do with subtle
audiophile-type differences in sound.
—Ed.
The Audio Critic:
Thank you for responding to my let-
ter regarding the review of the Magneplanar MG-1.5/QR (Issue No. 21). Your
comments are interesting and raise a
number of questions worthy of further
consideration.
...I had just begun to audition loud-
speakers in the $1000-1500 range, and so
I sought out the local Magnepan dealer in
Connecticut (sorry, I'm not a Minnesota
audio mafia member). After extensive auditioning and comparisons with a number
of other highly regarded loudspeakers, I
came to the conclusion that the Magneplanar MG-1.5/QR had superior midrange reproduction, a wide and deep
soundstage, and superb transparency.
Several competitors could reach lower in
the bass and most were more sensitive,
but none matched the overall perfor-
mance of the MG-1.5/QR....
...I believe that an audiophile journal should aspire to the highest standards
and that the loudspeaker reviews presently found in The Audio Critic fall short of
these goals.
Yours truly,
Douglas M. Hughes, M.D.
Rochester, MN
It is with great reluctance that I excerpted your three-and-a-quarter-page
letter so drastically, but there was no
room here, and I felt obligated to acknowledge, even if inadequately, the continuation of the debate I had entered into.
Your letter is devoted mainly to insistent technical arguments, some of them
repetitious and already answered by me
last time, some of them new and valid,
some of them new and invalid. It's all
quite academic at this point because you
are obviously determined to go to the
barricades to defend your choice of the
(continued on page73)
4
THE AUDIO CRITIC
pdf 5
ISSUE NO. 23 • WINTER 1995-96 5
In Your Ear
pdf 6
Mcintosh Today
(What the "High End" Was
Originally Meant to Be)
By Peter Aczel
Editor and Publisher
Would you believe it? Very high-performance (and very pricey)
audio equipment, scientifically engineered exactly as if the tweako
cultists and their reviewer gurus didn't exist.
Imagine a team of high-end audio designers who
are thoroughly up-to-date on their respective specialties
(circuit design, electroacoustics, measurement methods,
computer-aided technology, etc.) but have never heard of
Robert Harley, Martin Colloms, Dick Olsher, Harry Pearson, Michael Fremer, and their ilk. All right, maybe they
leaf through a tweako magazine every once in a while
just for laughs but then go about their work exactly as
they were taught in engineering school, totally
uninfluenced by the tweako pundits. Refreshing, don't
you think? Well, that's exactly how I see the current
Mcintosh operation in Binghamton, New York, the
slightly hicked-out "upstate" location, near the Pennsylvania state line, where they started in 1949 and never
left. Never changed their basic audio philosophy, either,
as I shall explain in a moment.
Meanwhile founder Frank Mcintosh and legendary
CEO Gordon Gow are history; the company is entirely
Japanese-owned (by Clarion, the car audio people); the
President for the first four years of the Japanese reign
was English; car audio products have been added to the
line (predictably); yet the Mcintosh logo on an audio
component means the same today as it did 46 years ago:
the most serious kind of engineering without the slightest
concession to trendiness, quality construction without
showoff spending on meaningless overdesign, unique
looks, and above all unproblematic operation under all
conditions.
For example, a Mcintosh amplifier, preamp, tuner,
or other electronic component will turn on and off without a sound out of the speakers—no thump, no pop, no
sizzle, not even the tiniest transient. It was always so and
still is. Even some very prestigious high-end brands are
unable to make that claim. Not that it's the most important thing in the world, but such a Calvinist insistence on
avoiding even the appearance of evil is highly character-
istic of the company. Integrity is what they're into.
That's what earned them their fanatically loyal dealer
base, which has been the principal reason for their stability over the years. They could afford to ignore the whims
and fetishes of the volatile audiophile market of the '70s
and early '80s; they could afford to have quite minimal
dealings with the tweako audiophile journals; their dealer
network kept them afloat through thick and thin. The basic philosophy was that if a manufacturer is exceptionally
supportive of the dealers and the dealers are exceptionally supportive of the customers, then a quality product is
bound to sell whether it represents the latest-and-greatest
trend or not. Yes, from time to time a certain stodginess
crept into the product line as a consequence, but the
present management appears to be capable of reconciling
the "Mcintosh tradition" with advanced engineering. A
recent visit to the Binghamton plant provided ample evidence of that.
Under one roof, which extends over workshops
dating from 1949 and new wings with fully automated
machinery, Mcintosh convincingly melds the old and the
new in manufacturing operations. Here are glassworkers
making the half-inch-thick glass plates for the retro front
panels of Mcintosh amplifiers—they seem to exercise the
craftsmanly care of old-time Venetian glassblowers but
use the latest precision equipment. In another work area,
computer geeks are polishing an incredibly sophisticated
simulation program for multiple tweeters in line-source
arrays. In yet another place, they are assembling just a
few special-order antediluvian vacuum-tube power amplifiers with huge output transformers, wound in-house,
for the nostalgia market. And over there, that's one of the
largest and best-equipped anechoic chambers for loudspeaker measurement I have seen on either side of the
Atlantic. Then there's the PC board facility... I could go
on, but there's no end to the old/new contrasts. Had the
6
THE AUDIO CRITIC
pdf 7
new Japanese owners built a spanking new, 1990s-style
plant, it wouldn't have nearly as much character as this
organically evolved hodgepodge and couldn't possibly
command the same kind of esprit de corps as currently
prevails at Mcintosh. Being out in the boonies also helps;
isolation from the alienated plastic industrial culture promotes clear, uncorrupted thinking, which is then reflected
in the product line—at least as I see it.
That product line incorporates close to a hundred
models today if one counts all categories; here I shall review just a few of the most representative examples.
Please note that, as a departure from our usual editorial
practice, these reviews do not appear within the articles
devoted to other analog electronic components and other
loudspeakers in this issue, although cross-references are
included in those articles.
Stereo Power Amplifier
Mcintosh MC500
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
13903-2699. Voice: (607) 723-3512. Fax: (607) 724-0549.
MC500 stereo power amplifier, 500 watts per channel, with me-
ters, $6500.00. Tested sample on loan from manufacturer.
This drop-dead, knockout power amplifier has the
Mcintosh signature quality, probably more than any of
the other Mcintosh products reviewed here. If an inani-
mate piece of electronics can ever be called sexy, this has
to be the one. For once, I can actually see $6500's worth
of good stuff built into the equipment—taking into consideration the dealer's markup, of course. Do you, as a
discerning audiophile, need an MC500? No. Do you
want one? Flash reaction: yes! After researching it and
weighing the alternatives: maybe.
The Mcintosh MC500 is a solid block of huge pas-
sive components and active circuit assemblies neatly
butted together, without even a sheet-metal cover, and it
weighs (grunt!) 114 pounds after it's out of the shipping
container. Two huge brushed-metal handles stick out of
the front, but that's not what defines the MC500's look.
Its principal mark of distinction is the pair of 5½-inch analog meters behind the half-inch-thick glass front panel.
They are that special shade of Mcintosh blue and they
stare right back at you as you watch the bouncing needles—but that's not the point; there are other power amps
with big meters. Ah, but these are genuine wattmeters,
not just voltmeters calibrated for watts into 8 ohms or 4
ohms. They actually measure the voltage and the current,
multiply the two measurements, rectify and average the
product, and display true watts with peak hold. Even
without a world-class amplifier behind such an instrument, it would cost a mint. That's better value for your
money than Wonder Caps, let me tell you.
The other major distinguishing mark that's immediately visible is the pair of large output transformers—
no, no, autoformers is the correct term. Vacuum-tube
amplifiers have output transformers, matching the highimpedance output stage to the low-impedance load. In
the MC500, the low-impedance output stage is matched
more precisely to the low-impedance load by means of
the autoformer. In contrast to an output transformer, an
autoformer is an exceedingly simple device, imposing
virtually no signal-altering influences on the output signal path. It merely helps the power supply satisfy the current demands of any load closest to 8 or 4 or 2 ohms, as
selected by the user. Instead of putting the money into a
monster power supply which can deliver the same voltage into any load and thus double the power as the load
is halved (that's the Krell approach), Mcintosh gives you
the same high power into any load by means of impedance matching. The power supply is still very large and
would undoubtedly do a decent job without the autoformer, but with the latter the amplifier becomes totally loudspeaker-friendly and has money left for other goodies
(such as the wattmeters). I am not saying that this is "better" than the spendthrift brute-force approach, but it is
equally good (at least for the vast majority of applications) and makes a lot of engineering sense—which is
what Mcintosh is all about. In terms of bulletproof protection it makes even more engineering sense because
shorts across the output of the autoformer are of no consequence whatsoever, even without all the other protec-
tion features of the amplifier. I'm a minimal-signal-path
purist but I still like it! Only if the loudspeaker fluctuates
over an improbably wide range of impedances—say from
1 ohm to 24 ohms or something crazy like that—will
there be any doubt about the adequacy of autoformer
coupling.
Aside from the wattmeters and autoformers, the
MC500 operates pretty much like any other very highpowered amplifier. Both unbalanced (RCA) and balanced
(XLR) inputs are available. David Rich's sidebar explains the circuit details. Build quality is outstandingly
good, but don't expect exotic-brand capacitors and resistors on the circuit boards because Mcintosh engineers are
into function, not fetish.
My measurements ran into a slight difficulty that
proved to be purely academic in the end. On the Audio
Precision test for THD + N versus level, the autoranging
function of the instrument can generate transients that are
apparently sensed by the Power Guard circuit (see the
sidebar), which then activates prematurely. As a result, I
was unable to obtain accurate THD + N curves up to full
rated output and beyond. It made little difference, however, because (1) I was able to ascertain with less sophisticated instruments on my lab bench that continuous power
output well in excess of 500 watts per channel was ob-
tainable at each autoformer tap into the matching load,
and because (2) the THD + N had dipped to -95 dB and
even lower at the point where the test started to stumble,
with every sign of moving in the same direction to even
ISSUE NO. 23 • WINTER 1995-96
7
pdf 8
Inside the Mcintosh Electronics
The following are my conclusions
after taking off the cover of each unit,
examining the PC boards and other
component parts, walking through the
circuit schematics in the service manual, and observing the unit in operation.
The measurements and overall evalua-
tions were left to the Editor.
Mcintosh MC7106
Nothing really interesting in the
main circuit. The design makes no con-
cessions at all to "audiophile" priorities
and belief systems.
The input of each channel goes di-
rectly into the leveladjust pot. It is a
very small control but it is sealed. The
switch for the bridge operation is not as
good, being a cheap open affair. A unity buffer formed by the NE5532N op
amp follows this. Electrolytics are used
for dc blocking at the input and output
of the circuit. Another electrolytic is in
the ground leg of the feedback loop of
the main amplifier (C2 as schematized
in Issue No. 20, p. 28, Fig. 6) to reduce
the main amplifier's dc gain to unity.
Complementary differential pairs follow. They are biased with a twotransistor current source and are resistively
loaded. No degeneration is used. No
follower stage is used. The second gain
stage is a degenerated commonemitter
amplifier that is dynamically cascoded.
The dominant pole is set by a capacitor
connected from the output of this gain
stage to ground. The low distortion of
the amplifier can probably be explained
as due to cascodes in this stage. The
By David A. Rich, Ph.D.
Contributing Technical Editor
only unusual topological feature of the
amplifier is the connection of the collector of the differential pair that is not
driving the second gain stage's base to
the emitter of the cascode device. This
folded cascode configuration increases
the amplifier's openloop gain somewhat. The output stage is a triple emitter follower. Three output devices are
on each supply rail. The total transistor
count exclusive of protection is 24 bipolar devices per channel.
The power supply for the entire
amplifier consists of only one transformer, one diode bridge, and a pair of
36,000 µF capacitors. The capacitors
are PCboardmounted; thus a cheap
strapping wire for the ground is not required. The Marantz MA500, costing
about half as much per channel, is
clearly going to have better channel
separation, especially under load, given
its separate power supply in each mono
block. The Mcintosh has a builtin
relay for remote poweron through a
proprietary control system.
Protection is extraordinarily com-
prehensive. The voltage difference of
the main differential pair of the main
amplifier is monitored. Under normal
operation this is the summing junction
and it should be very small. If the amplifier has reached it limits, then the
difference voltage will get larger. This
error signal is fullwave rectified and
averaged. It then drives a light
dependent resistor (LDR). The resistive
end of the LDR is connected back to
the input of the unitygain buffer IC de-
scribed above. During normal operation the resistance of the LDR is high
and it is effectively out of the circuit,
although the tweako crowd would find
its presence unacceptable. As the amplifier starts to become nonlinear, the
resistance of the LDR drops and the input signal is attenuated. A light called
Power Guard goes on when this function is activated. The output of the
amplifier goes through a relay that activates for thermal shutdown and the
presence of dc current on the output.
The relay also is delayed in operation
on poweron and opens on poweroff.
This prevents thumps and clicks at the
speaker terminal. Traditional single
transistor overcurrent foldback protection is used and it is set to trip early.
This circuit is responsible for the
amplifier's unimpressive PowerCube.
Powersupply inrush current limiters
were said to be in this amplifier but not
observed in the schematic.
A discrete "flash" A/D converter
is used to form the LED voltage meter.
The input to this is the speaker
terminal signal that has been fullwave
rectified and averaged. As many as 66
opamp sections plus a bunch of passive components and discrete semiconductors make for a very full board.
Why cheaper monolithic ICs designed
specifically for the purpose were not
used is unclear. LED voltage level can
be adjusted to change brightness, not
something your average audiophile
asks for.
Build quality is good but not ex
lower distortion levels before clipping. That puts the
MC500 into the uppermost category of power amplifiers
in terms of low distortion, regardless of the vagaries of
the Audio Precision. I saw virtually no difference in distortion at different frequencies, including 20 kHz, so that
dynamic distortion can be said to be nonexistent.
Just in case the autoformer introduced any frequency aberrations, I wanted to make exceptionally careful
and accurate frequencyresponse measurements. Using
the unbalanced input and the 2ohm output tap, at a level
of 1 watt into the load, I measured 0.4 dB at 10 Hz, 0.1
dB at 20 Hz, 0.0 dB up to 20 kHz, +0.1 dB at 50 kHz,
+0.25 dB at 100 kHz, and +0.05 dB at 200 kHz. Thus
there is a quarterdB "resonance" (if that's what it is) at
100 kHz and absolutely flat response in the audio range.
The picture did not change as I moved the level up to 100
watts. Conclusion: forget about the autoformer as an extra circuit element affecting the frequency response.
The PowerCube test (see Issue No. 20, pp. 1617,
for a complete explanation and illustrations) proved to be
illuminating. This is basically a test of current capability.
Since the MC500, with its autoformer output, trades
maximum voltage against maximum current—more V,
less I at the 8Ω tap and more I, less V at the 2Ω tap than
the active circuit's capability—it follows that the best
possible results on this test will be obtainable at the 2Ω
tap. The amplifier actually exceeded my expectations by
drawing a truly nice PowerCube with that particular
hookup. The cube had the predictable slight downward
slant from the 8Ω tier to the 1Ω tier, but at every test
8
THE AUDIO CRITIC
pdf 9
traordinary. PC boards are double
sided. At the amplifier's price you get a
fair but not very good deal.
Given the disappointing Power
Cube of the Mcintosh MC7106, I lean
toward the Bryston alternatives, although six equivalent (>100 W) Bryston channels will cost a little more. The
Bryston is a more advanced circuit design and equally wellengineered for a
long life. The added protection in the
Mcintosh adds belts to the Bryston's
suspenders; for some this may add to a
sense of security, but those seeking
stateoftheart electrical performance
should consider Bryston. On a practical
level, the Marantz MA500 monoblock
amplifier is the device of choice for the
vast majority of users looking for a
>100watt/channel multichannel system.
Mcintosh MC500
Basically the same circuit as the
MC7106 but output devices per rail go
from 3 to 10. Balanced inputs are provided. The opamp that was the buffer
in the MC7106 is used for this function. A turnon delay muting circuit is
added at the input. The dynamic biasing circuit for the cascodes in the second gain stage uses a current source instead of a resistor. There are separate
bridge rectifiers and filter capacitors
(22,000 µF) for each channel. The
feedback loop is much more complex
because of the autoformer. The main
feedback loop is connected to the 2
ohm tap. The main amplifier drive
point is somewhere between the 8ohm
and 4ohm taps.
The true wattmeters are nice to
watch, but a pair of the Bryston 7BST
monoblocks (500 W per channel) will
cost you $2605 less and give you com-
parable performance (and then some).
Again we see the Mcintosh priorities.
It is much harder to damage a power
amplifier under a short or other fault
condition if it has an autoformer, but
we pay for this in cash as well as reduced output into a possibly mismatched speaker at a particular tap.
Mcintosh C39
Once you have a handle on the basic design of this highly elaborate AV
control unit, you'll have no trouble understanding all the others, as they are
all quite similar. (See the rest of the
AV/surround equipment reviews elsewhere in this issue.)
The phono stage is the standard
noninverting topology, with electrolyt
ics used for C1 and C2 (see Issue No.
18, p. 18, Fig. 2). An NE5534AN is the
active element. The problems with this
topology were discussed in detail in Issue No. 18. A clever plug arrangement
on an auxiliary input allows for one
more linelevel input (the 12th line input!!) if phono is unwanted. When a
plug is put in this auxiliary input jack,
the phono signal is removed from the
bus and the auxiliary signal appears instead.
Input selection is through relays.
The output of the relay block is buffered by a unitygain inverting opamp
with electrolytics on the input and output for dc blocking. Mcintosh wants no
pops or clicks under any circumstances, and as we shall see dc blocking capacitors appear throughout the circuit
chain. The active element is a Motorola
MC33178P dual opamp, one section
of which is used for each channel.
Mono mix is engaged at the input of
this stage with a relay.
The record bus is also selected
through relays. All 12 inputs can be selected. A record processor can optionally be put in series with this bus. Plugging the processor in automatically
opens the bus. The stereo bus signal is
buffered with an MC33178P, again
with input and output dc blocking capacitors. Four separate output jacks can
be connected to the bus (VCR1, VCR2,
Tapel, and Tape2). These are enabled
by relays to prevent selfloops that
could oscillate. The C39 has more
flexibility for the serious tapist than
any other preamp known to me, with
12 inputs and four tape monitor loops.
Area B (remote room) outputs are
taken directly from the record bus.
Each channel is sent to an electronic
volume control made from a Micro
Power Systems MP7529 dual 8bit
multiplying DAC (the part is a second
source of the Analog Devices AD7528,
with improved distortion specifications) and the two sections of the
MC33178P. The MP7529 is a better
part than the Sanyo and Toshiba devices that we have seen used by B&K,
Onkyo, and Marantz. Blocking capacitors are at the input and output of the
circuit. After going through a bipolar
based mute circuit (everywhere else
this is done with a relay—do not ask
me why the difference here), the signal
passes to the RCA plugs.
Video and SVideo switching are
identical. CMOS switches are used that
are then buffered by a twotransistor
discrete circuit before being routed to
the output jacks. The unit has six video
inputs and four outputs for the video
section. The outputs are for two TV
sets and two VCRs. A video modulator
is also included for TV sets that do not
point with a reactive load the voltage was higher than
into a pure resistance, the way it should be. Into a perfectly matched 2Ω resistive load, the dynamic power was
812 W (i.e., 40.3 V). At no test point was the output lower than 35.7 V or higher than 44.5 V. The absolute current limits of the amplifier do not emerge from this test,
but proper execution of the design concept is evident.
Regular readers of The Audio Critic no longer expect me to discuss the "sound" of a properly designed
amplifier, but for the sake of newcomers let me repeat for
the nth time that such an amplifier has no sonic signature.
As for value, other megaamplifiers of comparably low
distortion and high reliability can be had for less money
than the Mcintosh MC500, but they will not have its almost redundant set of failsafe protection features.
Multichannel Power Amplifier
Mcintosh MC7106
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
139032699. Voice: (607) 7233512. Fax: (607) 7240549.
MC7106 sixchannel power amplifier, $3500.00. Tested sample
on loan from manufacturer.
Take one channel of the MC500, reduce the power
a great deal, take away the autoformer and the wattmeter,
mutiply by six, and you have in essence the Mcintosh
MC7106. It weighs less than half as much (53 pounds)
and has a knobless, smooth, glass front panel. David
Rich's sidebar explains the design in full detail, so I'll go
straight to my measurements.
ISSUE NO. 23 • WINTER 1995-96
9
pdf 10
have video inputs. One assumes this is
envisioned for a TV in the B zone,
since it is hard to imagine a main system controlled by the C39 with a TV
that does not have direct video inputs.
The audio for the modulator can be selected from the Listen (A) or Record
(B) bus. A passthrough for cable TV is
provided, so cable signals can go directly to the TV.
The main audio signal is routed or
bypassed from the Dolby Pro Logic decoder by CMOS switches. In Mcintosh
designs, costlier relays are used at any
location where the user can introduce a
fault condition, such as the input or
output jacks. A fault condition that
takes the signal outside the supply rail
will cause a CMOS switch to latch up.
To eliminate this condition the power
supply must be cycled, and in a worst
case situation the CMOS switch could
be destroyed. Mcintosh does use cost
saving CMOS switches internally in
the C39, where it is not possible for the
switch to see an over or undervoltage.
The Dolby Pro Logic decoder is
the Analog Devices SSM2125, and the
CMOS switches are the Analog Devices SSM2404. The SSM2125 is surrounded by numerous passive components required by the part. In addition,
an opampbased fourthorder bandpass filter is included for the noise
source used for setting up the Dolby
Pro Logic decoder. The Mcintosh C39
has nothing out of the ordinary in this
functional block.
The signals to the Dolby Pro Log-
ic decoder are accoupled with electro
lytics. The subwoofer channel is
formed by summing the left, right, and
(if the decoder is activated) center
channel. This signal is filtered by a
fourthorder (cascade of two identical
secondorder sections) lowpass filter.
All filter sections in the C39, including
this one, are formed with the Sallen
andKey circuit. When this subwoofer
signal is activated, CMOS switches put
secondorder highpass filters in the
left, right, and centerchannel signal
paths. Again, sections of the MC33178P
are used as the active elements for the
filters, with electrolytics at the front
and back for dc blocking. Why it is expected that one corner frequency and a
fixed set of filter orders will work with
all mainspeaker/subwoofer combinations is beyond me, but this appears to
be standard practice in AV land.
The surround output (S) of the
Dolby Pro Logic decoder or the
L+R+C signal (the choice depends on
the surround mode and is made by
CMOS switches) is bandpassfiltered
(7 kHz highpass limit) and then sent to
an LV1000N Dolby Pro Logic Delay
Unit used by almost everybody. This
device has an A/D, a D/A, and a digital
delay circuit. Dolby B is built into the
LV1000N and is used as a singleended
noisereduction processor for the time
delay block as required for Dolby Pro
Logic. This bandlimited mono signal
can be used directly or it can be sent to
the optional THX module. This DSP
based unit, for which we have no schematic, creates stereo rear channels. It
uses a DSP chip and thus requires A/D
and D/A converters. The THX block
also takes in the left, right, and center
channels. What comes out is left, right,
center, and stereo surround. Holman
equalization may also be applied to the
front channels, but this information is
not available. The subwoofer output
bypasses the THX board. After all this
processing I do not find myself wanting
to rush out and purchase surround
sound speakers. [Be tolerant, folks, of
the village atheist.—Ed.]
The good news is that true discrete
5.1channel surround sound makes this
whole board obsolete. The other good
news is that the whole section of active
electronics is bypassed by the CMOS
switches when in stereo mode. The bad
news is, if you really want Dolby Pro
Logic or THX (why?), then the analog
approach used here is not the best.
What you want to do is digitize the in-
coming stereo channels with high
quality A/D converters like those found
in DAT players, or take digital data directly off the laser disc or CD and have
all the surroundsound signal processing done in the digital domain. After
all the processing is done, digital data
for the 5+1 channels is converted back
to analog, using D/A converters of the
quality found in good CD players. Now
Onkyo does just this and gives you a
whole integrated receiver (seven—
count 'em, 7—power amplifiers, including three 110watt/channel discrete
units, an onscreen TV display, and a
tuner) for $1880.00, so why do we not
get DSP in the C39?
The only downside to this digital
approach is that the peak level of the
analog signals must be set not to overload the A/D converter. On the other
hand, if the signal level is too low in
amplitude, high noise levels will result
because the full range of converter gain
is not being used. For digital signal inputs there is no downside at all.
Once the 5+1 signals leave the
surroundprocessing section, they travel to a threestage volume control.
Three stages are necessary to get the
The Mcintosh insistence on low distortion is clearly in evidence here too. THD + N with an 8ohm load
and a 20 Hz or 1 kHz input reached a minimum of103
dB before the onset of clipping, which occurred at approximately 150 watts. The 20 kHz THD + N bottomed
out at 96 dB; call it dynamic distortion, but who cares
when the numbers are that good? With a 4ohm load the
minima were 103 dB at 20 Hz, 100 dB at 1 kHz, and
94 dB at 20 kHz; clipping occurred at 300 watts. All of
these figures were obtained with only one channel driven.
With two channels bridged and a load of 8 ohms, the
minima were 102 dB at 20 Hz, 100 dB at 1 kHz, and
87 dB at 20 kHz; the clipping point appeared to be just
a hair under 500 watts. Dynamic distortion is unarguable
in the bridged mode, since the 20 kHz curve started to
part company with the 20 Hz and 1 kHz curves at 1 watt
(80 dB). Thus a bridged MC7106 is almost the stripped
down equal of an MC500 at half price, with two channels
left over—but not if you look at the finer points of performance and, especially, at the PowerCube.
The MC7106 did well on the PowerCube test only
at 8Ω and 4Ω there was severe current limiting at 2Ω;
especially into ±60°, and total collapse at 1Ω. Dynamic
output voltage was 38.8 V (188 W) into 8Ω/0° and 36.5
V (333 W) into 4Ω/0°, with slightly higher output into all
8Ω and 4Ω reactive loads. Into 2Ω/0° the output dropped
to 32.3 V (521 W), rose slightly into ±30°, and then went
kaputt at the remaining test points Dynamic headroom
appeared to be nil when checked against the continuous
power output curves.
10
THE AUDIO CRITIC
pdf 11
required range and resolution. Each
stage is formed with one half of an
MP7529BJN multiplying DAC and one
half of an MC33178P op-amp, with the
exception of the volume control for the
left and right channels, which gets the
lower-noise 5532N. The 5532N devices are used when low noise is required.
They are not used all the time because
of their high power consumption, according to Mcintosh. The first MDAC
is set to have twice the step-height gain
of the remaining two MDACs. In the
process of flowing through this electronic volume control, the signal has
seen three more electrolytics for dc
blocking. A microcontroller is on the
surround-sound board to control all the
Dolby Pro Logic electronics and the
three-stage volume control. One interesting thing is that the volume-control
knob is not connected to a rotary shaft
encoder. Instead, the volume control
generates a dc level dependent on the
position of the knob. This presumably
allows you to use volume-control position as a guide to the system's level.
Since the C39 has a nifty LED digital
readout of level, this may appear redundant—but wait, the knob has no index mark! Figure that one out. The said
dc level tells the micro what data to
send to the multiplying DACs. The
C39 even puts a motor on the pot to
implement remote operation, instead of
just changing the digital control signals
to the electronic volume control directly. Different levels between the main
and surround channels require the use
of keypad controls (ugh). The microprocessor sends different data to the
multiplying DACs as the keypad
switches are depressed.
At this point Mcintosh must have
stopped designing and grabbed some
old boards from the parts bin. The L,
R, C, and subwoofer signals are routed
to an analog board with a loudness circuit. NE5532 op-amps are the active
stage, one section per channel, and an
unsealed Alps pot does the honors for
control. They have another dc-blocking
cap on the board also. No, there is no
bypass for this board at all. Even
stranger is the fact that L and R go into
an analog balance control, another unsealed Alps pot. Now, the balance
function could be implemented with
the multiplying DACs used for the volume control. Software would then
mimic the balance control as in other
AV products, so what's the pot doing
here? Other than creating a reliability
problem, I have no idea. I do know one
thing: you cannot adjust this control
with the remote—and it is the one control function you would actually want
to run on the remote!
And then the L, R, C, and sub-
woofer signals go to (are you ready for
this?) the nondefeatable tone control
block!?! Each pair of signals out of the
four uses an NE5532 for the active
summing block and an MC33178P for
the bass control circuit. You get another dc-blocking electrolytic cap also.
Does this circuit block and the loudness circuit block affect the frequency
response of the C39, and will the situation get worse as the unsealed Alps pots
age? Yes and yes. Why did Mcintosh
do something so wrongheaded? Do not
ask me. Simple CMOS defeat switches
would have made the whole issue
moot, but they are missing on the C39.
After passing through muting re-
lays, the 5+1 signals appear on the
RCA jacks. With the surround-sound
decoder and subwoofer filter defeated,
we have passed through six active stages and seven blocking caps to get to the
Listen (Area A) unbalanced line output. L and R are also available as balanced outputs. The dual MC33178P is
used for the single-ended-to-balanced
converter of each channel, and yes,
there is another dc-blocking capacitor
at the circuit's output. Another
MC33178P is used for the headphone
amplifier. The center channel, if active,
is summed into both channels just before the (you guessed it) dc-blocking
capacitor.
As you would expect, a lot of digital logic controls the whole thing, including separate microcontrollers for
the Listen (A) bus and Record (B) bus.
Another microcontroller is used to control the hardwired CD and tuner remote
links. That brings the total to four.
Power for the C39 comes from two
transformers. But it is no audiophile
special. There are 3300 (µF capacitors
on the unregulated rails. And the regulated rails are set to ±12 V by one pair
of 7812/79M12 regulators. That's
right, all the electronics discussed
above run on just this pair of regula-
tors. The digital electronics, of course,
have their own regulated supply.
Opening the cover of the large
chassis shows where the money went.
At the back of the unit the PC boards
are stacked four levels deep. In the
front they are only two layers deep.
The downside of all the boards is that
the wiring harness is quite complex and
with its interconnectors represents a reliability problem.
Plastic knobs and a front-panel
plastic door are not what you expect at
this price. 0
David Rich makes a techie wisecrack about the
channel separation in his sidebar, but what I measured
was nothing worse than 52 dB (at 20 kHz) and went as
high as 98 dB (at 200 Hz). Furthermore, the noise floor
of the amplifier is very low, -100 dB or better at all frequencies across the audio band (e.g., -112 dB at 1 kHz),
as referenced to a level of 1 watt into 8 ohms.
Please read my comment in the MC500 review
above anent sonic characteristics. In actual hands-on use
the MC7106 behaved as agreeably as the MC500. I share
David Rich's opinion that from a purely practical point
of view six Marantz MA500 monoblocks, costing $1800,
will do the same job in a home theater system as an
MC7106 at almost twice that price. But there will be those
for whom a Japanese amplifier can never be a Mcintosh.
Multichannel AV Control Unit
Mcintosh C39
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
13903-2699. Voice: (607) 723-3512. Fax: (607) 724-0549. C39
Audio/Video Control Center, $3500.00. THX-M optional THX
processing card for C39, $500.00. Tested sample on loan from
manufacturer.
This is such an incredibly complex and featureladen preamplifier and multichannel control center that
it's much easier to tell you what isn't in it than what is.
(David Rich in his sidebar attempts to cover the latter
question.) Well, there is no provision for on-screen displays of any kind (but the front-panel knobs pretty much
ISSUE NO. 23 • WINTER 1995-96
11
pdf 12
tell it all), no AC-3 (but that's very new, and very few
have it), no Ambisonics (but only Nimbus CD collectors
will miss it), no moving-coil phono (but who needs it in a
home theater system?), and in general no signal processing in the digital domain a la Lexicon or Onkyo. So you
could say this is a somewhat retro piece of gear, but what
an impressive monster!
The quality of construction is what you would expect of Mcintosh; everything is rock solid and beautifully
put together, but again don't look for insanely costly
"audiophile-brand" capacitors or resistors where they are
not needed because Mcintosh engineers see no point in
that. The glass front panel is very handsome, but that's
where all the knobs are, so it's always covered with highly visible fingerprints. Hooking up a home-theater system
to the inputs in the rear is not as daunting a task as the
complexity of the unit would suggest because everything
is logically laid out and intelligently labeled. I left the nitty-gritty details to David Rich.
On the lab bench I was unable to duplicate some of
the specifications published in the manual. The frequency
response at full gain through a line-level input and Area
A front output had a 0.35 dB bump at around 23 Hz with
the unbypassed bass control at its detent; the specs don't
admit that. Through the same signal path, at full gain,
THD + N was -79 dB regardless of frequency at 2 V out
and climbed to -71 dB before the onset of clipping at just
over 6 V. The specs say I should have measured -86 dB
or better but don't say at what gain or output. It is possible, since the volume control is much further downstream
in the signal path than in typical stereo preamps, that I
was amplifying noise from the early stages at full gain,
but the 8 dB rise in THD + N from 2 V to 6 V does not
indicate a noise-dominated measurement. Again through
the same signal path, at full gain and 2 V out, left/right
crosstalk was -54 dB at nearly all frequencies, rising to
-48 dB at 20 Hz. That's OK but far from great. Better results would surely be obtainable if the bass, treble, and
balance controls were bypassed or digitally implemented.
The phono performance was disappointing. With
Phono in, Tapel out, 2 mV input, 40 dB gain, the RIAA
equalization error was +0.34/-0.55 dB (worst case). The
largest error was at 20 Hz, but the entire response curve
was skewed. Frequency response deviations of that magnitude are audible. Through that same signal path the
THD + N reached a minimum of -75 dB at 6 V out, at all
frequencies, this time completely noise-dominated. The
simpler phono signal path yielded better crosstalk
figures: between -70 and -80 dB at any frequency above
500 Hz, with a worst-case reading of -48 dB at the 60 Hz
hum frequency (2 mV in, RIAA-equalized).
Using the C39 was a pleasant experience, although
the lack of on-screen displays creates minor learningcurve problems for your average remote-control jockey.
Even so, there is a reassuring feel of solid quality to this
control unit as you live with it and operate it.
AM/FM Tuner
Mcintosh MR7084
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
13903-2699. Voice: (607) 723-3512. Fax: (607) 724-0549.
MR7084 AM/FM stereo tuner, $1500.00. Tested sample on loan
from manufacturer.
This quite new tuner arrived with a considerable
delay behind all the other Mcintosh pieces reviewed here
and was not measured in time for David Rich's FM arti-
cle and tuner reviews in this issue. The following early
impressions are based on a look-see under the chassis
cover, a circuit schematic as interpreted by David, and a
brief stint of the MR7084 in my reference system.
Mcintosh tuners of the past have been RF masterpieces (e.g., the fabled MR78, still a big item on the
used-equipment market), but it would appear that the
MR7084 isn't one of them. Indeed, I'm inclined to think
—although I have no proof, only circumstantial evidence—that this is the exception to the generalization I
made above regarding Clarion's noninterference with
Mcintosh product design. I somehow suspect that Larry
Fish, Mcintosh's veteran chief engineer and one of the
best RF men in the business, was told by the front office
that he must use some existing circuit boards, possibly
from the (gulp) car audio line, to lower the cost of tool-
ing up for the new model. If I'm wrong, I apologize in
advance, but here are the clues:
The MR7084 is a $1500 tuner, yet it uses the Sanyo
LA3401 multiplex decoder IC, which is spec'd by Sanyo
"for stereo systems and portable hi-fis," instead of the
top-of-the-line Sanyo LA3450, spec'd "for high-quality
stereo systems" and better by at least a factor of two in
noise, distortion, channel separation, etc. The LA3401 requires a notch filter because it lacks a pilot-tone canceler;
the LA3450 incorporates the pilot-tone canceler, which is
the preferable solution. Now why would Mcintosh
choose the less good chip, at a negligible saving, for their
expensive tuner? The chassis layout reveals the answer.
Inside the handsome box there's mostly air; the actual
tuner is crammed into a tiny circuit board. The LA3450
has a larger footprint than the LA3401 and requires more
external components—it wouldn't fit on that weensy
board, which couldn't possibly have been designed for
the MR7084 from scratch, as I see it, but borrowed from
a preexisting design.
The Sanyo LA1 175, which forms the mixer, VCO,
1st IF, and AGC, is another obvious space saver; these
stages are discrete on other expensive tuners. Sanyo lists
this particular IC as designed "for car radios, stereo system." Nor is there room on that tiny board for two sets of
filters in the IF strip, so you get no wide/narrow selection
on the front panel. Very high quality is claimed for the
single set of three ceramic filters. The quadrature detector
used for the FM demodulator, on the other hand, is not of
12
THE AUDIO CRITIC
pdf 13
high quality; it's part of the Sanyo LA1235 chip that everybody uses, but the best tuners don't use the detector
part. Nor do the best tuners use the signal meter part;
they have separate circuits for that, whereas the Mcintosh
just uses what's on the LA1235.
Overall, the Mcintosh MR7084 is still much better
built than the equivalent cheaper Japanese tuners; the PC
boards are double-sided, and the parts quality is generally
higher, but deluxe details and packaging are no substitute
for advanced design. In actual use, the FM reception and
typical signal quality at my location were satisfactory but
strictly average; I much preferred, for example, the Rotel
RHT-10 at exactly the same price (see David Rich's re-
view). Measurements of the Mcintosh and further evaluation will be published in the next issue.
Loudspeaker System
Mcintosh XRT24
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
13903-2699. Voice: (607) 723-3512. Fax: (607) 724-0549.
XRT24 floor-standing 3-way column loudspeaker system,
$7500.00 the pair. Tested samples on loan from manufacturer.
I have a problem positioning this basically very
well-engineered speaker system against others in its cate-
gory for a fair review, perhaps because there are no oth-
ers in its category. Is there another speaker between, say,
$6000 and $9000 that uses a large number of tweeters in
a line array? I'm not aware of one. Other than Infinity's
older IRS models with their zillion vertically stacked
EMIT-type tweeters and the newer Genesis five-figure
oil-sheik models by the same designer, Mcintosh appears
to be the standard-bearer of the whole concept.
Indeed, I'm almost certain that Mcintosh designer
David L. Smith's marching orders were not to create the
best $7500 speaker he knew how to design, period, but
the best $7500 speaker using a discrete-element line
array, his specialty. (He is the author of a 1995 AES
paper on the subject.) There were already two older, larger, and much costlier speakers in the Mcintosh line based
on the same design principle, but they were—shall we
say— slightly on the dinosaurian side, indicating a neces-
sity for a more up-to-date, cost-effective, and polished
design. That is exactly what the XRT24 is, and as such it
must be called a complete success. If I needed a new
speaker and had $7500 to spend, would the XRT24 be
my choice? Frankly, no, but not because of the engineering quality, which is the usual problem. My reasons will
be apparent as we proceed.
A pair of XRT24's incorporates 32 tweeters of a
high-quality 1-inch aluminum-dome type, 16 per side,
deployed in a vertical line above the woofer/midrange
enclosure. The idea is to synthesize a pair of line sources,
with the added benefit of tremendous power-handling ca-
pability. Now, 16 discrete elements in a line array, close-
ly butted together and driven equally, do not oblige the
designer by producing a smooth cylindrical wavefront in
accordance with the idealized model of a line source. Au
contraire, they yield an unacceptable radiation pattern
with very ugly lobes. The drive has to be "tapered;" i.e.,
each element or group of elements has to be driven differently with respect to level and frequency range to
achieve the desired summed response. I have personally
seen and observed in action the highly ingenious proprietary computer program Mcintosh has developed to calculate the most effective tapering. The results are excellent
in terms of summed response; the trouble is that 32
tweeters still possess some of the inherent limitations of a
single tweeter. It would undoubtedly be affordable to
choose the world's best tweeter and put two of them into
a $7500 pair of conventional speakers, but 32 of them?
As I said above, the tweeters are of high quality, but they
aren't among the world's best. The Win (proprietary) and
the Accuton (OEM), to name only two, are among the
best; 32 such tweeters would raise the in-house partsand-labor cost of a pair of XRT24's to more than $7500
without any other changes, never mind the wholesale and
retail prices. I'll come to the sonic consequences in a moment.
The lower part of each speaker consists of a forward-firing 8-inch midrange unit and two rearward-firing
10-inch woofers in a sealed cabinet. The whole schmear,
with tweeter column, is 7 feet tall but only 15 inches
wide and 18 inches deep at the base. The occupied airspace is quite modest from the waist up owing to the
slender tweeter column, which tapers from 10 inches
wide at the bottom to 6 inches at the top. Each speaker
weighs about 130 pounds. Since the high-frequency drive
is shared by 16 tweeters, they can be crossed over at a
relatively low 1.5 kHz; the midrange covers 250 Hz to
1.5 kHz; the woofers come in below 250 Hz. The elaborately controlled drive to the tweeter column is accom-
plished with a passive network; one amplifier channel per
side is all you need.
I measured the frequency response of the XRT24
from 300 Hz to 20 kHz with the quasi-anechoic MLS
method, positioning the calibrated microphone at a distance of 2.5 meters from the line array and at a height of
1.3 meters. That's not really a standard setup, but then
the almost ceiling-high XRT24 is far from a standard
loudspeaker, and I wanted to obtain meaningful results at
a realistic distance and some kind of realistic average ear
position. The resulting response on axis turned out to be
quite excellent: ±3 dB from 300 Hz to 18 kHz, and that
doesn't even tell the complete story because from 3.5 to
12 kHz the response was closer to ±1 dB, and if it had
not been for a marked dip at 3 kHz, the overall response
would have been pretty much within ±1.75 dB. Going
30° off axis horizontally—and that's generally the stereo
listening position—actually smoothed out the response
slightly, making it ±2 dB from 300 Hz to 12 kHz. Above
ISSUE NO. 23 • WINTER 1995-96
13
pdf 14
12 or 13 kHz the curve was a little rough regardless of
the microphone position, but this was due to interference
patterns, not tweeter misbehavior. I checked for tweeter
ringing with tone bursts of different frequencies and
found little or nothing, except at that 3 kHz dip, where
there appeared to be a slight storage problem, nothing
major.
The dual woofer, measured with the nearfield
method, exhibited a more or less classic sealed-box response, declining at 12 dB per octave below a -3 dB
point (f3) of approximately 40 Hz. Since the fundamental
resonance of the bass system, as indicated by its impedance curve, appears to be in the neighborhood of 32 Hz,
it would seem that the response is slightly overdamped,
maybe because the rearward-firing design assumes some
boost from the proximity of the rear wall (12 to 2 inches
suggested by Mcintosh for initial placement). Reservation No. 1: When I'm asked to spend $7500 for a loudspeaker system, I expect more or less flat response down
to the limits of audibility, not a missing bottom octave.
For example, a pair of Velodyne DF-661 three-way
speakers, coupled with a pair of Velodyne F-1500R subwoofers, would cost you $4885 (in black) and give you
16 Hz to 20 kHz response with ultralow distortion. No,
they wouldn't look nearly as impressive in your listening
room, doctor, as the XRT24's.
Speaking of distortion, the XRT24 is actually very
good in that respect, though no Velodyne. At a 1-meter
SPL of 95 dB (normalized to 50 Hz), the nearfield THD
+ N of the woofer rose gradually from 0.4% at 200 Hz to
3% at 20 Hz, staying under 1% down to 45 Hz. At 90 dB
the rise was from 0.3% to 2%, staying under 1% down to
40 Hz. Respectable figures, all in all. Midrange THD +
N, at a 1 meter SPL of 95 dB (normalized to 600 Hz),
fluctuated between 0.23% and 0.73%, with no clear relation to frequency; at 90 dB the readings were barely lower, between 0.2% and 0.6%. Again, not bad at all. The
distortion of the tweeter array was hard to measure because the drive is not the same to all 16 units. After a lot
of messing and some educated guessing I determined that
the distortion generally stays in the 0.13% to 0.4% range
at 90 to 95 dB. Thus I can't really fault the XRT24 on
distortion; these measurements fall barely short of the
Snell Type A, for example. I have already mentioned the
tone-burst response of the tweeter array; the woofers and
the midrange driver gave no evidence of storage when
tested with tone bursts.
The impedance curve of the speaker is a roller
coaster; the magnitude fluctuates between 2.6 ohms and
16 ohms, the phase between -50° and +45°. A power
amplifier with a good PowerCube performance (i.e.,
load-insensitive) is highly advisable. Nominal impedance
is 4 ohms. Separate low, mid, and high terminals are
available for optional bi- or triamplification (passive).
Now then, how does a pair of XRT24's sound?
Basically neutral balance. Excellent dynamics and power
handling. Big soundstage, which is what you'd expect
with those tall line sources. Not very deep bass. Slightly
coarse-textured highs, lacking the airy/silky quality of
the most sophisticated tweeters (such as those in the Win
SM-10, MACH 1 Acoustics DM-10, Snell Type A, etc.)
and causing my Reservation No. 2.
I could be cruel and declare the basic design strategy of using 32 plain-vanilla aluminum-dome tweeters in
a very expensive pair of speakers to be a mistake, but I
don't want to be cruel because the engineering thinking
behind the XRT24 is intellectually on a high plane as
well as innovative in many respects, while free of tweako
nonsense. To some well-heeled audiophiles the virtues of
the speaker will be more important than its two shortcomings, especially since its relatively small footprint
and high-quality furniture finish make it one of the few
very tall speakers without the probability of over-mydead-body spousal confrontations. I resolved my mental
conflicts about it by moving it out of my main listening
room/laboratory, placing it into my home theater system
in the family room, and extending it on the bottom with
the Hsu Research HRSW12V powered subwoofer.
There, in combination with Mcintosh center-channel and
surround speakers, the XRT24 is giving me total satisfaction. Is it because I'm less demanding of Bela Lugosi
than of Béla Bartók?
Home Theater Loudspeakers
Mcintosh HT-3 and HT-4
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
13903-2699. Voice: (607) 723-3512. Fax: (607) 724-0549. HT3 THX 2-way, antiphase dipole, mirror-image, left/right surround loudspeakers, $2000.00 the pair. HT-4 THX 2-way center-channel loudspeaker system, $900.00 each. Review samples
on loan from manufacturer.
These are the speakers referred to above, supplementing the XRT24 in my home theater system. They
were sent to me several months later than the XRT24 and
have not been measured yet as I am writing this. I can
vouch, however, that their construction is of high quality
and that they do their job very proficiently. The HT-4
center-channel speaker has three 1-inch aluminum-dome
tweeters arrayed in a vertical line, which is pretty much a
standard THX formation; not surprisingly this mini line
source has a sonic signature very similar to that of the big
line source in the XRT24, subtle but discernible. The
HT-3 surround speakers appear to have no signature at
typical surround-signal levels. My tentative impression is
that the HT-3 and HT-4 are somewhat overpriced, but I
am reserving final judgment until all the usual tests are
completed. •
14
THE AUDIO CRITIC
pdf 15
Loudspeaker Systems, from
Good Value to Ultimate (and
In Between)
By Peter Aczel
Editor and Publisher
Is the loudspeaker still the weakest link in the audio chain? Yes,
undoubtedly, but the evidence of these tests is that the weak link is
getting stronger all the time, in all price brackets.
Here's something to think about before you sink
your teeth into the reviews. Loudspeakers are all imperfect to a greater or lesser degree; the imperfections are all
different, so all the speakers sound at least somewhat different; therefore loudspeaker evaluation leaves at least
some room for purely subjective opinion and speculation,
much more so than any discussion of the incredibly perfect analog and digital signal paths of modern audio electronics. You'd expect most of the subjectivistic voodoo
and tweako cultism in highend audio to be about speakers—but no! Other than the wire/cable nonsense, which
arguably relates to electronic events before the speaker
terminals, most speaker discussions in tweako circles are
relatively sober, certainly not comparable on the B.S.
scale to, say, singleended triodes. The inevitable conclusion: audiophiles retain the faculty of clear thinking only
as long as the equipment is imperfect and there are real
problems to be solved; when there aren't any, they panic
and create imaginary problems.
Let's look at some solutions, then, from the real
world of loudspeaker engineering.
Atlantic Technology System 250
Atlantic Technology International, 343 Vanderbilt Avenue, Norwood, MA 02062. Voice: (617) 7626300. Fax: (617) 7626868.
System 250 sixpiece hometheater speaker system, $1446.00
(plus minor options, stands, etc.). Tested samples on loan from
manufacturer.
Home theater has somewhat different requirements
when it comes to speakers than twochannel stereo for
music. I am of the school that favors the most accurate
stereo speakers for front left and right—never mind the
"specialized" hometheater speakers that don't sound so
great on music—but for the center and surround channels
the rules are different (at least until AC3 with its five
fullrange discrete channels, or something simlar, becomes the standard). The center and surround speakers
are processors to some degree; they do not necessarily
have to be "accurate" in the linear input/output sense.
The hometheater subwoofer, on the other hand, must
meet pretty much the same requirements as a stereo
system subwoofer. The Atlantic Technology System 250
is a remarkably costeffective solution to the problem of
excellent hometheater surroundsound performance
without significant musical compromises. I never expected anything nearly as good for $1500.
Let's start with the costliest part of the system, the
252 PBM Powered Bass Module ($569.00). It incorporates a 12inch, papercone, longthrow woofer; an
amplifier that offers the choice of three 40watt (4Ω)
channels, two of them for satellites, or 90 watts of power
into the woofer alone; lowpass filter options of 80 Hz or
120 Hz; and various inputs and controls. It's quite nicely
built, not too big—approximately 2 cubic feet in internal
volume—and uses a ventedbox design with two ports.
The system is tuned to approximately 37 Hz; there is solid output down to 30 Hz, and even at 20 Hz the two ports
put out some energy. Corner placement can further support the bass, so that in most rooms quite realistic theaterlevel lowfrequency pressures are obtainable. I was
impressed. Distortion certainly isn't extra low—this is
not a motionalfeedback design a la Velodyne—but the
combined nonlinearities of the amplifier and the woofer
are well within conventional limits.
The front left and right speakers, Model 251 LR
($299.00 the pair), are of the socalled D'Appolito configuration: two 4inch polypropylene midrange/woofers,
one on top, one on the bottom, with a small plasticdome
tweeter in the middle. The sealed box is very small,
ISSUE NO. 23 • WINTER 1995-96
15
pdf 16
about one third of a cubic foot internally; the optional
metal stand tilts it slightly upward at seated listening
height. Nominal impedance is 8 ohms. I found the frequency response of the 251 LR unusually interesting: up
to 2 kHz it is dead flat, and I mean ±0.5 dB; from 8 to 18
kHz it is almost equally flat, perhaps ±1.5 dB; but between 2 and 8 kHz there is an almost perfectly symmetrical hollow, about 8 dB deep at its bottom! I have never
seen anything like it and can virtually guarantee that it is
not a design error but entirely deliberate. Somebody at
Atlantic Technology doesn't like that "presence" range,
that's for sure. (Maybe it's Peter Tribeman, their president—remember him from the old NAD days? He is
known to have very strong opinions on sound.) I must
say that the speaker sounds very sweet and creamysmooth on all program material; maybe the frequencyresponse tailoring is in anticipation of harshness on most
cinematic sound tracks. Even so, I prefer to have my
equalizer elsewhere in the audio chain, just in case some-
thing is beautifully recorded in the first place. I find it re-
markable that other reviewers of this 1994-vintage prod-
uct never really got a handle on this fascinating
peculiarity.
The center-channel unit, Model 253 C ($279.00),
uses exactly the same drivers in a box of very similar
size, except that the speaker is horizontally deployed on
top of the TV and rests on a very simple but ingeniously
designed base that permits about a 25° range of up-anddown tilting. In addition, there are two controls in the
back marked "midrange timbre" and "hi-freq. level,"
which boost/cut the midrange/woofers and the tweeter,
respectively, to permit matching the 253 C to front left
and right speakers other than the 251 LR. A dot marked
"251" shows the position of each control recommended
for matching to the 251 LR; another dot marked "cinema
eq." is the position supposed to "compensate for the natural brightness of many movie soundtracks," although the
251 position already suppresses the same two "brightness" octaves as the 251 LR. Thus doth the addition of
two cheap passive controls an almost double-priced
"Timbre Adjusting Center Channel Speaker" make—but
why not, when the total six-piece system sounds better
than it has a right to for the money?
The equally small Model 254 SR surround speak-
ers ($299.00 the pair) each have one 4-inch polypropylene driver on top, identical to those in the other units,
plus two 3.5-inch "mid-tweeters" below, angled to fire
frontward and rearward. The 3.5-inchers are phased to
produce the "decorrelated" output needed for ambience
channels. There is no need for dead flat response—
indeed, for any significant degree of input/output accura-
cy—in ambience speakers, so I did not bother to make
MLS measurements on the 254 SR, but it is fairly obvi-
ous that it has no top-octave (10 to 20 kHz) output. It
does the job, and that's it.
I already mentioned in my Mitsubishi TV review in
Issue No. 22 that the Atlantic Technology System 250,
driven by Marantz MA500 power amplifiers, produced
better movie sound than you'll hear in your typical small
neighborhood theater. To be more specific, the sound is
quite complete from T. Rex footsteps to violin harmonics;
nothing is missing either on the bottom or the top; the dynamic range leaves very little to be desired, even in a
fairly large room; no harshness or edgy quality intrudes
on any kind of program material; in fact, you could do a
lot worse than to use the System 250 as your main transducer for music, quite independently of home theater. I
played a few of the Delos 20-bit CDs that sport the Dolby Surround logo and was quite genuinely satisfied, barely aware that I was listening to a tourist-class multichannel system, until I took the same CDs to my main
listening room and replayed them through incomparably
costlier speakers. Yes, they sounded quite a bit better that
way, but the System 250 was not humiliated by the com-
parison. It is good enough not to have to apologize to
much higher-priced home-theater speaker systems.
Such a conclusion brings up a cynical thought.
Does it take no more than adding a 30 Hz bottom end
and suppressing the brightness range to make a very
modest loudspeaker array sound like a high-end system?
(I never pretended I don't have a dirty mind.)
Hsu Research HRSW12V
Hsu Research, 14946 Shoemaker Avenue, Unit L, Santa Fe
Springs, CA 90670. Voice/Fax: (310) 404-3848. HRSW12V
powered subwoofer, $850.00 each (factory-direct, including
shipping/handling). Tested sample on loan from manufacturer.
My review of the Hsu Research HRSW10 subwoofer appeared in Issue No. 19. The HRSW12V is quite
similar conceptually but considerably upgraded in a number of ways, including the considerably higher price. The
larger driver (a 12-inch unit instead of a 10-incher) and
the dedicated outboard amplifier are the most obvious
differences. The basic configuration remains the same:
cylindrical enclosure made of lightweight but extremely
strong and inert paper tubing, downward-facing driver,
long duct tuning the vent; however, everything is bigger
and beefier than in the 10-inch predecessor (which has
meanwhile been dropped from the line to be reintroduced
in powered form).
The HRSW12V cannot be evaluated from quite the
same perspective as the HRSW10. The latter was incredibly low in price for a world-class subwoofer, $375.00
each (without amplification, to be sure), so that its performance exceeded all expectations. The factory-direct price
of the HRSW12V, on the other hand, is only 22% lower
than the retail price of the Velodyne Servo F-1200R, for
example, which is also a 12-inch powered model but
with accelerometer, motional feedback, and remote control. That means the HRSW12V is out of the bargain
16
THE AUDIO CRITIC
pdf 17
basement and on the main floor, competitionwise. Our review sample did not get to us early enough to be wrung
out properly in the laboratory before this was written (we
seldom get the early samples but prefer to have the last
word anyway), so I am restricted to listening impressions
here, which are much the same—in spades!—as told in
my detailed and enthusiastic review of the HRSW10.
This is a serious contender in the subwoofer Olympics
and deserves to be judged against the high-priced entries.
I wanted that much to be said even before further testing.
Dr. Poh Ser Hsu, the designer (see the aforementioned review in Issue No. 19 for his background), claims
that his subwoofers outperform all others at 20 to 30 Hz
when the SPL is very high. I personally believe that low
distortion and good damping at 80, 90, 95, maybe 100 dB
are more important than whatever happens at 115 dB;
more about that in the next issue. I also intend to report
on the recommendation by Hsu Research that the
HRSW12V be placed in the listening area rather than
near the main speakers—another subject on which I
don't want to pass quick judgment. Meanwhile the
HRSW12V is woofing prodigiously in my home-theater
system.
Mcintosh XRT24
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
13903-2699. Voice: (607) 723-3512. Fax: (607) 724-0549.
XRT24 floor-standing 3-way column loudspeaker system,
$7500.00 the pair. Tested samples on loan from manufacturer.
The review of this quite unorthodox and very interesting speaker system appears elsewhere in this issue as
part of the feature article on Mcintosh Laboratory.
Paradigm Eclipse/BP
In Canada: Paradigm Electronics Inc., 101 Hanlan Road,
Woodbridge, ON L4L 3P5. Voice: (905) 850-2889. Fax: (905)
850-2960. In the U.S.: AudioStream, Division of Bavan Corporation, MPO Box 2410, Niagara Falls, NY 14302. Voice: (905)
632-0180. Fax: (905) 632-0183. Eclipse/BP floor-standing bi-
polar 2-way loudspeaker system, U.S. $1799.00 the pair. Tested
samples on loan from U.S. distributor.
The Janus-faced, i.e., bipolar, loudspeaker is popular among Canadian speaker designers—and I don't
mean they favor two-faced deceit, just a bidirectional
wave launch. Bipolar speakers are like two forwardfiring speakers placed back to back and driven inphase; a
positive-going signal makes both faces push; a negativegoing signal makes both faces pull (assuming no internal
phase inversion). The Paradigm Eclipse/BP is a representative upper-medium-priced Canadian specimen of the
breed. It uses an 8-inch polypropylene-cone woofer and a
1-inch aluminum-dome tweeter on each face; the two
woofers exhaust into the same columnar enclosure with a
ducted port. Woofer-to-tweeter crossover is at 1.5 kHz.
The whole affair stands a little over four feet high and is
quite handsome, one might even say expensive-looking,
in black gloss with runaround black grille cloth (other
finishes available).
Contrary to my wonted procedure, I shall discuss
the sound of the speaker first because it raises basic issues that the measurements alone do not. The soundstage
is huge, undoubtedly as a result of the delayed/reflected
rearward radiation. Certain audiophiles will make the
Eclipse/BP their choice in this price range just for that. I
feel about soundstage enlargement as I do about silicone
implants—up to a point the effect may be glamorous but
beyond that it's unnatural. The Eclipse/BP does some-
thing other than what the producers of my favorite CDs
had in mind, that's for sure. I must admit, however, that
regardless of the soundstage issue the speaker always
sounds sweet and agreeable—indeed, too sweet and
agreeable, even when the music is not. That would also
make it certain audiophiles' automatic choice. To my
ears the Eclipse/BP always sounds like a speaker, a very
likable speaker to be sure, but not like an accurate replica
of the program source. It has a bit more sonic "signature"
than I like.
Now then, why? The edgeless, bland quality is
quickly explained by the frequency response: the incredibly smooth tweeter, which stays within a ±1 dB strip
from 2 kHz to 15 kHz, is set approximately 4 dB below
the level of the woofer. I determined that with a 1-meter
MLS measurement on the forward-firing axis; it is possible that the designers had in mind some kind of summed
power response forward and rearward, but such things
generally don't work out very precisely, and certainly not
in my big room with its deadened rear wall. Interestingly,
the axial response from 15 kHz to 20 kHz rises right back
to the woofer level, obviously a "tailored" profile but of
greater concern to my dogs than to me. The phase re-
sponse is very well-behaved.
On the bottom end the vented system appears to be
tuned, not very sharply, to approximately 23 Hz; the re-
sponse is esentially flat down to 28 Hz or so. With four
8-inchers pumping, a stereo pair produces some pretty
authoritative bass, making subwoofers superfluous except to incurable bass addicts like me. The impedance
curve of the Eclipse/BP fluctuates between 3.7 and 15
ohms in magnitude, and between -20° and +35° in phase,
presenting a relatively unchallenging load to the amplifier. Distortion at all levels is typical of nonexotic 8-inch
woofers and 1-inch dome tweeters; nothing remarkable
there. The crossover network drives the woofer and
tweeter inphase, but a square pulse is not reproduced co-
herently at any acoustical summing junction. (No big
deal; I've given up on the audibility of that.)
Wanting to isolate what I perceived to be that elu-
sive signature of the Eclipse/BP, I excited both woofer
ISSUE NO. 23 • WINTER 1995-96
17
pdf 18
and tweeter with tone bursts of all frequencies. Everything was squeaky-clean; no ringing, not even a trace.
Tapping the cabinet with a padded mallet produced fairly
dead sounds off the front and rear but more of a drumlike
whomp off the larger side panels, even though high-tech
bracing is claimed in the literature. Maybe that's what I
heard as a signature, or maybe a combination of that and
the tailored response and the bipolar launch. Maybe I'm
strictly a monopole man. Maybe you'll love it. I certainly
won't argue with you because a lot of very good engineering went into this speaker. It makes a statement.
Snell Acoustics Type A
Snell Acoustics, Inc., 143 Essex Street, Haverhill, MA 01832.
Voice: (508) 373-6114. Fax: (508) 373-6172. Type A Music
Reference System, $18,9999.00. Tested samples on loan from
manufacturer.
What I wrote in my preview of the Snell Type A in
Issue No. 22 still stands: this is the best loudspeaker system, overall, that I have tested and lived with. Please note
that I said "overall," which is not the same as "in every
way." For example, the woofer of the Type A is very
good but not my favorite. From 80 Hz on up, however, I
don't know of another speaker that satifies both my listening and technical requirements quite as completely.
The system—and I mean the stereo system, not the
multichannel cinema version (see below)—comes in 7
pieces: two 63-inch high towers, 9 by 12 inches in cross
section; two outboard passive crossovers, each as big as a
bookshelf speaker; two refrigerator-sized subwoofers;
and one cute little 80-Hz electronic crossover (made for
Snell by Audio Alchemy). The latter is a very simple,
stripped-down unit, called EC-200; the recommended
electronic crossover for those who want everything to be
"high end" is the Bryston l0B-PRO/Snell, a dedicated
version of Bryston's $1195.00 super crossover. I tested
both the EC-200 and the Bryston/Snell. (See Issue No. 16
for an early review of the basic Bryston model.)
Let's get the subwoofers out of the way first because, well, they're just not as interesting as the rest of
the system. They are big, though: each enclosure meas-
ures 45 by 21.5 by 17 inches and houses an 18-inch driver loaded with a huge ducted port. The mouth of the port
is 5 by 5 inches and the duct is long. Unfortunately, the
center of the woofer is approximately 26 inches above
the center of the port, making it extremely difficult to
find a nearfield summing junction to measure the combined response of the two. I did the best I could and
found essentially flat small-signal response down to the
subsonic region, with the -3 dB frequency (f3) at approximately 12 Hz. The box appeared to be sharply tuned to
15 Hz. Can't ask for much better.
The distortion measurements, taken at the flattest
nearfield summing junction, were less impressive. With a
1-meter SPL of 90 dB, THD plus noise fluctuated be-
tween 0.5% and 5% from 50 Hz down to 10 Hz. The
0.5% minimum was at 20 Hz, the 5% maximum at 40
Hz. With the 1-meter SPL raised to 95 dB, the THDplus-noise range was 0.3% to 5%, but this time the minimum was at 34 Hz and the maximum at 12 Hz. I would
estimate the typical distortion at a random low frequency
to be 2% at 90 dB and 1% at 95 dB, but that includes the
inescapable residual room noise (probably the reason for
the higher reading at the lower level). The motionalfeedback Velodyne Servo F-1500R is at least 6 dB better
at any frequency at the same SPLs and a whole order of
magnitude (20 dB) better at most frequencies. (Other
things being equal, the closed-loop system has to be
cleaner than the open-loop system—the laws of physics
say so.) The Velodyne is also somewhat better damped
than the Snell; the latter exhibits just a bit of hangover
when excited with low-frequency tone bursts (again, not
unexpectedly in view of the fourth-order tuning of the
vented box).
Please don't misunderstand me; the Snell subwoofer is no slouch; it goes all the way down to the T. Rex frequencies, can play really loud, and has no more distortion
than is reasonable in an open-loop design. Don't refuse it
as a birthday present. It's just that the Velodyne represents a newer and more sophisticated generation of subwoofer design, in a mere 20-inch cube and at a lower
price, with amplification thrown in. (A single unampli-
fied Snell SUB 1800, if purchased separately, costs
$2499.00; that's $904.00 more than the amplified Velo-
dyne.) I removed the subs from the Type A system, sub-
stituted a pair of Velodyne Servo F-1500R's, and after
some crossover fussing (more about that in a moment)
was a thoroughly happy camper. Will Snell sell you a
subwooferless Type A Music Reference System for
$14,001 (= 18,999 - 2 x 2499)? I see no reason why not,
as long as you don't involve them in the substitution.
About the electronic crossovers—decisions, decisions. I found the EC-200 and Bryston l0B-PRO/Snell to
be equally low in distortion (meeting the spec of 0.005%,
i.e. -86 dB, at 1.5 V output, 20 kHz highpass and 20 Hz
lowpass) and equally accurate in lowpass and highpass
filter contours. The Bryston is better built and a little
more convenient to set for level; the EC-200 once gave
me a frighteningly loud pop as I was turning off various
parts of the system, but I was unable to duplicate the
same circumstances again. (I suspected dc offset, but the
measured amount was minimal. Weird.) Is it worth the
additional expenditure to get the Bryston? Some people
will undoubtedly want it. (No, there is no difference in
sound—you ought to know better than to ask such a
question.) The electronic crossover frequency is set at 80
Hz; the highpass filter is second-order (12 dB per octave),
the lowpass filter fourth-order (24 dB per octave). With
the Velodynes I left the lowpass section of the electronic
crossover unconnected, using the lowpass and volume
18
THE AUDIO CRITIC
pdf 19
controls on the back of each subwoofer instead; it's not
an exact match to the Snell lowpass slope but, after a bit
of fussing and measuring, "close enough for government
work" as the saying goes. Crossovers at 80 Hz are not
sensitive to exquisitely subtle adjustments.
You've been waiting for the main course all this
time, and with the mundane hors d'oeuvres and soup out
of the way I'm ready to serve it to you: yes, the towers!
In my not-so-humble opinion they're a truly superior
piece of engineering. Each tower houses four 6.5-inch
drivers (call them upper woofers or lower midranges),
two 5-inch midrange drivers, a single 1-inch textile-dome
tweeter, and a single rearward-firing 1-inch metal-dome
tweeter. The seven forward-firing drivers are deployed
symmetrically in a vertical array: tweeter in the center,
one 5-inch midrange above and one below the tweeter,
two 6.5-inch drivers on top of the array and two on the
bottom. The huge outboard passive crossover directs the
signal traffic to the drivers via a massive 8-cable umbili-
cal cord; the crossover frequencies are 350 Hz and 2.8
kHz. Four very precise tweeter-level settings are provided (+1 dB, flat, -1 dB, -2 dB); the rear tweeter can be
switched on or off and trimmed for level (by ear) with a
continuous control.
The amazing thing is that this complex structure
acts as a monolithic supersmooth transducer, with seamless and almost amplifier-flat response from 80 Hz to 20
kHz. My measurements (MLS from 300 Hz up, taken at
2 meters as well as 3 meters; quasi-nearfield from 80 Hz
to 300 Hz; everything set flat) corresponded very closely
to the manufacturer's spec of ±1.5 dB over that entire
range. In the crucial range from 1.5 kHz to 11 kHz, the
response is better than ±1 dB—and that with the grille in
place! (Unfortunately, a snap-on grille frame proved to
be too much of a manufacturing problem, so the grille
cloth is unremovable.) At 30° off axis, all that happens is
that the forward-firing tweeter rolls off above 11 kHz
with a slope of about 12 dB per octave. Below 11 kHz
there's hardly any difference between on-axis and offaxis response. Some speaker.
And that's not all. The distortion generated by the
towers—as distinct from the Snell subwoofers—is almost Velodyne-low. At a 1-meter SPL of 90 dB, as referenced to 150 Hz, the nearfield THD + N of the upper
6.5-inch driver hovered between 0.1% and 0.2% from
300 Hz down to 125 Hz, rising to 1.5% at 80 Hz. The
upper 5-inch driver, measured the same way, with the 1meter SPL at 90 dB as referenced to 1 kHz, yielded 0.2%
to 0.5% in its range of 350 Hz to 2.8 kHz, averaging
about 0.3%. The tweeter, again at 90 dB SPL at 1 meter
as referenced to 5 kHz, remained near 0.3% over most of
its range, dropping to 0.2% above 6 kHz. I didn't even
bother to take my usual 95 or 96 dB measurements because, as already mentioned above, the distortion almost
invariably reads lower at the higher SPL on account of
the constant room noise. These figures nudge the perfor-
ISSUE NO. 23 • WINTER 1995-96
mance of the "distortion-free" Velodyne DF-661 without
quite equaling it—but the Snell is flatter (indeed, flatter
than just about anything).
As for ringing or storage—nothing. All the tone
bursts I tried looked squeaky-clean. (That textile-dome
tweeter is especially nice—and especially interesting
coming after all the Snell models that use metal-dome
tweeters.) The impedance curve of the tower shows rather wide swings in both magnitude (3.3 ohms to 20 ohms)
and phase (-55° to +30°), not surprising in view of the
complex crossover. A high-quality (nontweako) solidstate amplifier should have no trouble driving that kind of
load. Incidentally, the terminal configuration on the outboard crossover also permits biamping or—if you'll pardon the expression—biwiring. (For the benefit of new
readers: biwiring is hogwash but harmless.)
Another interesting aspect of the design is that neither the towers nor even the SUB 1800's are so heavy
that you need piano movers to experiment with speaker
placement. Designer Kevin Voecks believes that 1.5-inch
walls in a speaker box yield no significant sonic benefits
but make life miserable for the audiophile. (Of course,
that's what certain techie masochists want.)
All right, all right, I'll tell you about the sound. The
speaker has virtually no sound of its own—that, of
course, is what makes it great. It adds no signature to the
program material, so the latter's own signature (hall
acoustics, microphone characteristics, recording setup,
etc.) emerges with unprecedented clarity. The wave
launch is medium-sized, and the soundstage is not gigantic on typical program material—but is "Wow, what a
huge soundstage!" the right impression, or even a desirable impression? The Type A tower sounds more like an
idealized 14- or 15-inch coaxial speaker, say like the old
Win SM-10, only bigger—and even that is an inadequate
description. There is certainly soundstage depth and
width but not of the in-your-face, check-it-out-dude kind.
What's more, the sound is never harsh, from ppp to fff,
even though the high frequencies are certainly not rolled
off. I guess what I'm trying to say, without resorting to
the tweako reviewers' quasi-pornographic vocabulary, is
that the speaker sounds beautiful—more like the real
thing than like a speaker. (It's easier to be highly articulate and specific about something bad than something
this good.)
I first met Kevin Voecks in 1977, when he was
barely out of his teens. (Maybe not even.) At that time he
was a classic audio tweak, fondly indulgent of his own
subjectivity. Now, 18 years later, he is one of the most
scientific, practical, self-critical, and verifiably effective
speaker designers, with more than a few world-class designs to his name. To say that I am impressed would be a
serious understatement. I only hope that he will soon do
an "A Minor," incorporating the most important aspects
of the Type A design in a more compactly packaged and
(continued on page 38)
19
pdf 20
Catching Up on
Analog and Digital
Electronics for Stereo
By Peter Aczel
Editor and Publisher
Two-channel stereo is alive and well and living in the living rooms
and family rooms of music lovers. Not all sales in the audio world
are home theater, at least not yet.
Some of the reviews below should have appeared
in the last issue but were held over; others are of more recently received equipment.
Outboard D/A Converter with HDCD
EAD (Enlightened Audio
Designs) DSP-7000 Series III
Enlightened Audio Designs Corp., 300 West Lowe, Fairfield, 1A
52556. Voice: (515) 472-4312. Fax: (515) 472-3566. DSP-7000
Series 1I1 outboard D/A converter, $2495.00 (optional balanced
outputs $399.00 extra). Tested sample on loan from manufacturer.
This is basically the same processor as the DSP7000 Series II reviewed (and recommended with minor
reservations) in Issue No. 20, with one important change.
EAD has made a questionable marketing move and incorporated the Pacific Microsonics HDCD decoder chip
in all of their latest digital processors, including this one.
I say questionable because the HDCD (High Definition
Compatible Digital) encode/decode process has so far
been endorsed only by the tweako/mystico element in the
high-end community and is regarded with considerable
suspicion by the best brains in the business. No major
label is currently releasing HDCD-encoded CDs, nor will
they as far as I know (except perhaps in isolated cases
where a recording artist absolutely insists on it). Reference Recordings is the showcase label for the system,
and the hardware customers for the decoder chip are the
familiar roster of boutique manufacturers catering to the
high-end ghetto.
One reason for the suspicion of the best-educated
practitioners is that Pacific Microsonics has never made a
full disclosure of the exact operation of the system, from
ISSUE NO. 23 • WINTER 1995-96
the input of the encoder to the output of the decoder.
That's not the way Dolby became a standard. Yes,
there's a patent, but as this is being written no one has
seen it in print yet; meanwhile the Pacific Microsonics
people have become experts at saying a lot and revealing
nothing. The only really good source of information on
HDCD so far has been the European patent application,
which is not a privileged document over there and has
been read by a number of authorities whom I trust. What
I hear from them is that HDCD involves all sorts of unnecessary and unjustified overprocessing that, in their
opinion, no "enlightened audio designer" would want to
subject the music to.
This is not the time and the place for a critique of
HDCD, which will have to wait until the most detailed
information is available and the subject is crystal clear;
maybe then a more favorable picture will emerge. The
above remarks are simply a reflection of my present misgivings. I don't believe that 16-bit linear PCM needs
"fixing," especially when the new 20-bit masters make it
even easier to end up with all 16 bits on the CD.
To get back to the DSP-7000 unit itself, the dis-
cernible changes from Series II to Series III are these:
• The HDCD decoder chip (PMD100) has replaced
the NPC SM5813 digital filter, allegedly improving the
playback of non-HDCD-encoded discs as well.
• On the front panel a blue light turns on when
HDCD decoding is automatically activated by an encoded disc. (A very old and slightly nasty ethnic joke from
New York pops into my strictly non-PC and non-HDCDimpressed mind: "Abie, turn on the blue light. The customer wants a blue suit.")
• The reclocking circuitry has been upgraded following David Rich's free advice in Issue No. 20, p. 50: it
is now a phase-locked loop, utilizing a voltage-controlled
crystal oscillator. (Gee, why don't the EAD people use
21
pdf 21
Robert Harley as their digital consultant?)
• The analog output level in response to a full-scale
(0 dB) digital input without encoding has been reduced
by 6 dB to 1 V. Apparently this is a nonnegotiable re-
quirement for HDCD licensing by Pacific Microsonics.
They must be afraid that non-HDCD playback would
sound better than HDCD because it is louder. (The
official justification is of course the convenience of equal
levels from all sources, but audiophiles know that slight
volume-control adjustments are always needed, regardless, depending on how "hot" the recording is—so what's
wrong with trimming the volume manually for HDCD?)
I find this feature particularly obnoxious because it reduces the 16-bit system to 15 bits and skews the full-scale
THD + N versus frequency measurements; luckily, if you
have an Allen wrench of the right size to open up the
unit, a single DIP-switch setting on the circuit board will
defeat the modification and restore the normal 2 V output. That's what I did (with the EAD chief engineer's
vigorous consent, believe it or not), and that's how I
made my measurements.
Lo and behold, the performance of Series III with
our usual digital test signals was not as good as that of
the pre-HDCD Series II. Now, it is an outside possibility
that our sample had slightly substandard DACs, although
Burr-Brown's quality control of their flagship PCM63P-K
is known to be virtually flawless. Perhaps EAD had
tweaked around with some circuit adjustments, but I
doubt it. In a few other published test reports the Pacific
Microsonics PMD100 chip shows no tendency to degrade the non-HDCD performance. Bottom line: I have
no explanation, but here's what I found.
Full-scale THD + N measured -92 dB to -91 dB at
the lower frequencies (i.e., 6 dB to 7 dB in excess of the
theoretical ideal of -98.08 dB), dropping above 1 kHz to
a minimum of -94.4 dB at 4 kHz and then rising again.
Series II had been approximately 3 dB better below 1
kHz. With a -20 dB digital input, Series III still showed
1 dB to 2 dB excess distortion, whereas Series II had
shown virtually none (maybe 0.5 dB). Most dramatic was
the difference in gain-linearity error: Series III was off by
+2.3 dB at -100 dB and by +0.7 dB at -80 dB in its less
good channel; the "better" channel measured +1.25 dB at
-100 dB and +0.3 dB at -80 dB. Series II had displayed
virtual perfection on this test, with no more than 0.1 dB
error down to -100 dB. Here comes the mindblower.
Gain linearity below -70 dB is always better with dither;
indeed, it shouldn't even be measured without dither—
but Series III is the exception! Down to -90 dB it was
just a hair better without dither! Figure that one out. I
couldn't; I kept muttering something about the dither in
the PMD100, but what do I know? I also did what I now
call the Rob Watts test, an FFT plot of a dithered 1 kHz
tone at -60 dB (see Issue No. 22, p. 36), which should
show a super clean -130 dB noise floor without any har-
monics. Series III showed 2nd and 3rd harmonics 18 dB
and 20 dB above the noise floor, and there were lots of
higher harmonics as well. Not as clean as the best. The
noise spectrum with digital zero input was also 5 dB to 7
dB worse at many frequencies than that of Series II, although it was still very good. One thing that hasn't
changed from Series II to Series III is the radiofrequency
interference (RFI). Even in the standby mode, just
plugged in but not operating, the unit is absolute death on
FM reception. Leave it unplugged when not in use, or
have your FM tuner several rooms away from it.
I listened to the Reference Recordings HDCD Sam-
pler Volume 2 demo CD to see if I would come under the
spell of decoded HDCD sound, but my impressions were
inconclusive. Keith O. Johnson makes good recordings,
as we all know, so most of this stuff sounds pretty nice. I
think he has been leaning toward a "wetter" sound lately
than I consider ideal. In any event, the hall and the microphone setup have a much greater influence on the sonic
end result than the use or nonuse of HDCD. The three
pairs of tracks demonstrating use versus nonuse merely
prove that there is a difference. (If there weren't a difference, after all the hype, it would be the ultimate tweako
exercise in unreality, but that's not the case.) But is the
HDCD encoded/decoded sound clearly better? To give a
definitive answer, I would have to go to a live recording
session where there's a switchabe HDCD encode/decode
loop in the direct microphone feed. I have no way of
knowing exactly what RR did to those tracks. HDCD
seems to soften the attacks and somehow defocus or airbrush or pastel-color the details, as it were—"better"
sound to some, a "processed" sound to others, and vague
perceptions in any event. Controlled listening tests are in
order, but they must be part of a total, in-depth evaluation of HDCD.
Stereo Power Amplifier
Mcintosh MC500
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
13903-2699. Voice: (607) 723-3512. Fax: (607) 724-0549.
MC500 stereo power amplifier, 500 watts per channel, with meters, $6500.00. Tested sample on loan from manufacturer.
The review of this no-holds-barred super amplifier
appears elsewhere in this issue as part of the feature article on Mcintosh Laboratory.
Mono Power Amplifier
Sentec PA9
Sentec America, Route 9, Stockport, NY 12172-0329. Voice/Fax:
(518) 828-8490. PA9 class AB monaural power amplifier,
$600.00 each. Tested samples on loan from manufacturer.
Made in Sweden and reflecting the Scandinavian
taste for simplicity, moderation, and quality in all aspects
22
THE AUDIO CRITIC
pdf 22
of design, this small mono amplifier costs twice as much
as the Marantz MA500 (see Issue No. 22, pp. 2728), has
half as much power output before clipping, about the
same distortion at low power but much higher distortion
on moderate peaks and beyond; it is, however, small and
cute (8 in. by 3 in. by 10 in. deep) and much better constructed. Its slightly exotic appearance and provenance
will make some audiophiles believe it "sounds better."
David Rich examined the circuit schematic and
sent me a memo that I might as well quote:
"Your basic fully complementary amplifier. Complementary degenerated differential pairs are driven by
singletransistor current sources. The current sources
have LED voltage references. The differential pairs are
resistively loaded. The second stage is also a degenerated
differential pair. Each side of the second differential pair
inputs is connected to the first differential pair outputs.
The use of two sets of differential pairs significantly improves the commonmode rejection ratio (CMRR) and
hence the distortion that results from commonmode signals (see the Harman Kardon PA2400 review in Issue
No. 21). A single transistor is used to bias the output
stage. The unit has no predriver stage. The output stage
has three paralleled devices. The current limiter is the
standard onetransistor design. No C1, but C2 is present
(see Issue No. 20, p. 28, Fig. 6). The mono channel uses
21 transistors and 6 diodes."
The THD + N versus watts measurements were
quite unusual in that the distortion isn't noisedominated.
Probably because of the lack of a predriver stage, there is
significant nonlinearity rising with level, long before
clipping. Minimum distortion is reached in the neighborhood of 1 watt, where it is 90 dB with an 8ohm load
and 86 dB with a 4ohm load. Clipping occurs at 72
watts into 8 ohms, with 68 dB distortion before the
breakpoint, and at 120 watts into 4 ohms, with 63 dB
distortion. These numbers apply to 20 Hz and/or 1 kHz;
the 20 kHz measurement is worse by 7 dB to 10 dB at all
levels except the lowest—dynamic distortion at its most
obvious.
The PowerCube test (see Issue No. 20, pp. 1617)
drew a fairly decent picture at 8Ω/4Ω/2Ω magnitude and
60°/30°/0°/+30°/+60° phase, with the exception of 2Ω
at 60°/+60°, where the beginning of serious current limiting was visible. The current limiting becomes severe
with a 1Ω load regardless of phase, except 0°. With the
onetransistor current limiter that was predictable. Maximum dynamic output voltage was just a little over 27 V
into the five 8Ω test loads, declining with a linear slope
to 18 V into lΩ/0° and much more severely into the reactive loads. The calculated dynamic headroom at 8Ω is
0.65 dB. Peak current output measured 13 amps. Thus
the overall PowerCube performance is unimpressive but
still better than that of the average small amplifier.
I see little reason to favor the Sentec PA9 over the
Marantz MA500, except for the PA9's quality/size ratio.
ISSUE NO. 23 • WINTER 1995-96
Line-Level Stereo Preamplifier
Sentec SC9
Sentec America, Route 9, Stockport, NY 121720329. Voice/Fax:
(518) 8288490. SC9 linelevel stereo control amplifier,
$700.00. Tested sample on loan from manufacturer.
This is the preamp sold by Sentec to complement a
pair of PA9's. It is also small and cute, about half the
size of a normal preamp and rather nicely finished. It has
six linelevel inputs, two tape monitor loops, a headphone jack, and separate balance and level controls. The
the two tape monitor loops are switched in such a way
that you can't put a tape recorder into a selfconnected
loop and cause destructive oscillations—a quality feature. If you wish, you can listen to one source and record
another. A relay with time delay at the output prevents
turnon thumps.
The circuitry is discrete and somewhat complex,
with many similarities to the PA9. David Rich examined
the schematic and summarized the design as follows:
"JFET input with current source used for the tail and re-
sistive load. The output goes differentially to a bipolar
differential pair with degeneration. The load is an active
bipolar current mirror. A lightemitting diode (LED) is
placed in series with one side of the differential pair's
collector and the current mirror output. This provides the
bias voltage for a class A bipolar emitterfollower output.
Directcoupled; dc offset is controlled with a pot. Medium feedback and good design techniques; it should have
low distortion." (It does; see below.)
The design features David didn't like he noted
thusly: "A single power supply for both channels. Low
rail voltage, ±10 V. Zener diodes are placed across the
regulated rails. The unregulated rails go to this point
through a resistor. No active regulator circuits. (I have
seen the same thing in $200 Pioneer receivers, where
cost cutting made a lot of sense; why in a $700 preamp?)
Expect proneness to line hum. On the positive side, the
preamp has RFT filtering in series with the line; I think
Sweden has tough RFI suppression requirements, now
coming to all of Europe."
Now David may have his faults but being wrong
isn't one of them. Indeed, before I found a way to ground
the unit optimally in my test hookup, the sensitivity to input lead dress was astonishing—moving the input cables
a couple of inches changed the THD + N by a factor of 6,
or even more! That was all N, namely hum. After I had
that gremlin figured out and under control, the results
were truly excellent: 95 dB to 97 dB distortion at all
frequencies between 1 V and 2 V output; slight dynamic
distortion discernible at 20 kHz beyond 2 V; clipping at
just over 5 V. Even at 20 mV out, which is 40 dB below
the 2 V linelevel reference, the THD + N at any frequen-
cy was 61 dB or lower. Changing to a 600ohm load
23
pdf 23
had negligible effect. Can't ask for much better.
Crosstalk was measured with 1 V in, the level control set for unity gain (1 V out), and the balance control
at its detent. The results were excellent; right leakage
into left was 5 dB to 14 dB lower than left into right, but
even the latter was as low as -68 dB at 20 kHz, falling to
-100 dB at 200 Hz and below. The better channel went
as low as -105 dB at the lower frequencies and did not
rise higher than -82 dB even at 20 kHz. These numbers
are right up there with the best.
One must conclude that the minor circuit compromises in the Sentec SC9 are academic because they have
little or no effect on the actual performance, which is
comparable to that of just about any high-end preamplifier. Highly recommended.
Outboard Phono Preamplifier
Sentec PP9 RIAA
Sentec America, Route 9, Stockport, NY 12172-0329. Voice/Fax:
(518) 828-8490. PP9 RIAA phono preamplifier, with MM9 moving-magnet input board, $400.00; MC9 moving-coil add-on
board, $100.00. Tested sample on loan from manufacturer.
To expand the SC9 for phono playback, you plug
this little brick-shaped black box, only slightly larger
than the average modem, into one of the line-level inputs.
An external transformer that plugs into the wall powers
the unit. There is no power switch—it's the just-leave-iton-it-hardly-draws-any-current philosophy—and no relay
to prevent turn-on thumps should there be a power interruption/resumption.
David Rich's notes on the circuit: "A two-stage
design. Topologies are the same as in the SC9, except
that the bipolar follower stage is missing . Direct-coupled
with dc servo in second stage. Better power supply, with
±11 V rails and an open-loop bipolar follower after the
zener diode. Uses the floating ground arrangement also
used in the Sentec DiAna outboard DAC. (The power
transformer has no center tap.) RFI line filter and surge
protection on secondary side of transformer."
The bench measurement results were a mixture of
excellent and merely good. The RIAA equalization was
truly excellent: ±0.075 dB from 30 Hz to 20 kHz; at 20
Hz it was off by -0.2 dB. Crosstalk was also excellent;
with a 5 mV input (RIAA pre-emphasized), it fluctuated
between -65 dB and -96 dB. No phono cartridge comes
even close to that in channel separation. Input-referred
noise measured 0.6 µV, and the spectrum of the noise
floor as measured at the output hovered around the 5 µV
level at most frequencies. I would call that very good but
not great. THD + N at all but the lowest frequencies was
-80 dB just before the onset of clipping, which occurred
at 6 V. The 20 Hz measurement tracked with the higher
frequencies up to 2 V out, after which it worsened by a
maximum of 10 dB—purely academic. I have seen some
-90 dB phono preamps, but this is pretty good THD + N
on the whole.
All of the above results apply to the MM9 board;
the MC9 board was not supplied, but David Rich had this
to say about the MC9 schematic: "The moving-coil am-
plification stage is a 4-transistor open-loop affair. Because
it has no feedback it uses a complementary topology that
gives a 3 dB noise penalty."
Bottom line: if you go with the Sentec SC9 and
need phono, by all means get the PP9 as well.
Compact Disc Player
Sony CDP-XA7ES
Sony Electronics, Inc., 1 Sony Drive, Park Ridge, NJ 07656.
Voice: (201) 930-1000. Fax: (201) 930-4748. CDP-XA7ES
compact disc player with RM-D921 remote control, $3000.00.
Tested sample on loan from manufacturer.
This is now the top of the Sony ES line of CD
players and a direct continuation of the design progression started with the CDP-X779ES (see Issue No. 18)
and continued with the CDP-X707ES (see Issue No. 21).
All three models are essentially the same electronically,
so there is no need to explain the basic design all over
again.
The outstanding feature of the circuit remains the
CXD2562 pulse D/A converter, still unsurpassed and
possibly even unequaled by any other in several areas of
performance. In the XA7ES, the DAC is further improved by coupling it to a new chip, the CXA8042,
whose ultimate function is to improve the PSRR and
whose mode of operation allows the Sony marketing peo-
ple to put "Current Pulse D/A Convert System" on the
front panel of the XA7ES to distinguish it from the very
similar X779ES and X707ES, which by the same argument have voltage-pulse D/A conversion. Pretty subtle
difference.
The analog filter at the single-ended output is also a
little different in the XA7ES; the change is from a generalized impedance converter (GIC) structure to a more
standard Sallen-and-Key filter built around the output
stage. The filter at the balanced output remains Sallenand-Key, as before. (Is this the end of Sony's advocacy
of the GIC solution?)
Much was made of the debut of the so-called Score
digital filter IC in the X707ES, but the XA7ES uses yet
another ostensibly improved filter chip, the CXD8504M,
which is said to have a wider data path than its predecessor. What hasn't changed, on the other hand, is that the
single-ended output still uses a more sophisticated circuit
than the balanced output. Actually, very little has
changed in terms of circuitry.
The mechanical and cosmetic changes are greater.
The unit is now a very professional-looking black and
has a keypad on the front panel, not just on the remote.
24
THE AUDIO CRITIC
pdf 24
Most significantly, the laser tracking assembly has been
changed to a fixed pickup design. The rigidly mounted
optical block remains stationary as the rotating disc
moves past it on a stable sled. A brass stabilizer, reminiscent of the phono era, holds the rotating disc in place,
providing damping and precise centering. If you mislay
the stabilizer—tough. The machine won't play. If you
drop it, you can break or dent something with it, as it
weighs well over a quarter of a pound. That's the price
you pay (not to mention the $3000) for a mechanism copied directly from Sony's CD mastering equipment and
claimed to increase the quality of the data stream picked
up by the laser, with less dependence on interpolation or
averaging. I couldn't measure any improvements over
the already flawless transport of the X707ES but was
duly impressed by the massive deluxeness of the whole
machine.
Actually, I couldn't measure any improvements,
period. That was no surprise because both the X779ES
and the X707ES measured within a hairsbreadth of theoretical perfection, and so does the XA7ES. Sony has
achieved the highest level of CD playback performance
known to me. All other designs, CD players as well as
outboard D/A converters, show some gain-related analog
distortion as the digital signal level is raised from -30 dB
or -20 dB to full scale (0 dB). Only the Sony units approach the theoretical 16-bit limit of -98.08 dB on fullscale THD + N. The excess distortion produced by the
XA7ES at full scale is of the order of 1 to 2 dB across the
audio spectrum, pretty much the same as in the case of
the earlier Sony units but no others (not even the costliest
outboard processors, which never do better than 4 dB and
often do worse). At the -24 dB level the excess distortion
is about 0.5 dB. The single-ended and balanced outputs
measure about the same.
Other measurements are of comparable perfection.
Quantization noise: -97.6 dB. Dynamic range: 97.5 dB.
Channel separation: 115 dB at 16 kHz, increasing to 132
dB at low frequencies. De-emphasis error: unmeasurable
(less than 0.02 dB). Low-level linearity error: 0 dB down
to -100 dB and beyond. Spectrum of 1 kHz tone at -90
dB, with dither: no trace of harmonics. No wonder they
have to reach to come up with a new top-of-the-line model almost every year.
In operation the XA7ES is smooth as silk, just like
its predecessors; I like the extra convenience of the keypad on the front panel and grumble occasionally about
the brass stabilizer when I can't find it instantly. Everything considered, this is CD-player heaven. Of course,
less expensive players sound just as good, as our regular
readers very well know, but some audiophiles will want
what is provably best in every way, and this is it. Note,
however, that the original X779ES cost only $1900;
we're now up to $3000, a 58% increase. That's not just
the yen exchange rate and not just the new features;
that's high-end marketing.
ISSUE NO. 23 • WINTER 1995-96
Stereo Power Amplifier
Sunfire (follow-up)
Sunfire Corporation, P.O. Box 1589, Snohomish, WA 98291-
1589. Voice: (206) 335-4748. Fax: (206) 335-4746. Sunfire
"Load Invariant High Fidelity Stereo Power Amplifier," $2175.00.
Tested sample on loan from manufacturer.
The preview of Bob Carver's truly different Sunfire
amplifier in Issue No. 22 was based on a preproduction
(or very early production) sample that we had our hands
on for just one day. This follow-up review is the result of
testing a completely current production unit. My description of the basic features of the Sunfire and David Rich's
analysis of the design concept and circuitry are still entirely valid as published in the preview and need not be
repeated here.
As the cliche goes, there's good news and there's
bad news (no, this is not a joke). The good news is that
the 43-pound Sunfire, at $2175, does basically the same
thing as the 165-pound Krell KSA-300S at $9500. Both
deliver at least 49 volts per channel continuously into
any load. That means at least 300 watts into 8 ohms, 600
watts into 4 ohms, etc., doubling the power into half the
load impedance. (It's possible that the Krell holds up a
little better into 1 ohm or less, but by then who's counting?) That, essentially, is the Bob Carver advantage: no
limit to cheap watts; no load too tough; no hernia. The
bad news is that the Sunfire does all that at rather high
distortion.
Now, it should be added that Bob Carver doesn't
consider the amount of distortion produced by the Sunfire to be bad news at all. Ultralow distortion isn't one of
his priorities, and there is good support for his position.
The threshold of THD audibility on pure sine waves, under ideal test conditions, is believed to be in the neighborhood of 0.2% (-54 dB); on music it's much higher.
Furthermore, above 10 kHz fundamentals all harmonics
are inaudible, whatever their level. The Sunfire doesn't
break the assumed thresholds, although it comes close.
The position of The Audio Critic has always been that
distortion is the measure of accuracy and should be kept
as low as possible within the limits of budgetary sanity.
When we have 16-bit clean digital equipment like the
Sony CD player reviewed above, do amplifiers need to
be only 9-bit clean in terms of distortion at full output?
Amplifiers that are 16-bit clean just before clipping do
exist. Bob Carver says that's nice but unimportant.
Having been told that a lot of the measurable THD
+ N in the output of the Sunfire was ultrasonic switching
noise (i.e., inaudible N), I used a measurement bandwidth of 22 kHz instead of the usual 80 kHz for most of
my bench tests. With a 22 kHz cutoff, my highest test
frequency was 6.5 kHz, in order to include the second
and third harmonic (13 kHz and 19.5 kHz) with some
lowpass margin to spare. Later I found out that in my up-
25
pdf 25
todate review sample the switching noise had been reduced to a fairly low level and was well above 80 kHz in
any event, so that the THD + N readings at high frequencies represented mainly dynamic distortion, not noise. I
did a 0to80kHz FFT analysis on a 19.5 kHz test signal
at 200 watts into 8 ohms and saw 2nd/3rd/4th harmonics
at 57/53/72 dB (plus three harmonically unrelated
blips in the 85 to 93 range) but found the overall noise
floor to be very low, centering on the 110 dB line. According to David Rich (who has familiarized himself
with the Sunfire circuit and admires many of its features),
slightly costlier input opamps would have kept the distortion down, but as I said that wasn't one of the designer's priorities.
To get deeper into the numbers, the amplifier clips
at 310 watts into 8 ohms and 620 watts into 4 ohms.
That's true, with very small corrections, for all frequencies and either channel. The THD + N readings show a
generally falling tendency with rising output, up to clipping, but not nearly as regularly and predictably as seen
in conventional amplifiers with purely noisedominated
distortion. In the Sunfire curves there is evidence of level
related nonlinearity here and there. Worstcase distortion
at high output: 47 dB at 20 kHz just before clipping,
with the 80 kHz measurement bandwidth. But even with
a 6.5 kHz input and the 22 kHz measurement bandwidth,
the distortion is no better than 57 dB from 1 watt up to
clipping in the less good channel, regardless of load impedance. At lower frequencies the picture improves; the
20 Hz and 1 kHz readings are 10 to 15 dB lower at high
output with the 22 kHz cutoff, slightly varying with load
impedance and not quite the same in both channels. Absolute bestcase distortion using the 22 kHz cutoff: 81
dB at 50 watts into 8 ohms, at the low point of an irregular dip in the 20 Hz distortion curve of one channel; the
other channel hits 80 dB just before clipping but not at
50 watts, and neither channel goes quite as low with a 4
ohm load. By comparison, using the 80 kHz cutoff, the
lowest reading for 20 Hz or 1 kHz is 64 dB, just before
clipping into 8 ohms. Compare these results with any of
those reported in our poweramplifier survey in the last
three issues and draw your own conclusions.
(It should be noted that the 22 kHz and 80 kHz
filters referred to above have flat response in their pass
band; they should not be confused with the humpbacked
Aweighting filter Bob Carver uses for some of his tests.)
The Sunfire's squarewave response shows normal
rise times, flat tops, and a bit of overshoot in the corners.
In Issue No. 20 we introduced the PowerCube test;
the reader is referred to pp. 1617 of that issue for a complete explanation and illustrated examples. The Sunfire
drew a virtually perfect PowerCube at 8Ω/4Ω/2Ω magnitude and 60°/30°/0°/+30°/+60° phase. It delivered approximately 56 V of the standard test signal (1 kHz
bursts of 20 ms duration) into those 15 test loads at the
targeted 1% (40 dB) distortion. At 1Ω magnitude and
the same five phase points the voltage declined to values
between 40 V and 29 V, either because 56 V (= 3136 W
into 1Ω) would have yielded more than the targeted 1%
distortion, or because the PowerCube hardware/software
got nervous at that level (I've seen that happen before),
or because of some other unidentified glitch. I don't quite
trust the 1Ω readings and am willing, on a benefitofthe
doubt basis, to declare the amplifier to be indeed "load
invariant" as it says on the front panel. It is interesting to
note that a comparison of the PowerCube numbers and
the continuous power curves (at 1% distortion, not at the
breakpoints) shows that the circuit design allows no dynamic headroom as such—and who needs it with this
kind of continuous output capability?
Very new readers of The Audio Critic may now
ask how the Sunfire sounds, but the regulars know better.
The amplifier has dead flat response within the audio
range, produces monstrous output levels without clipping,
and doesn't distort above the threshold of audibility, so
how could it have a sound? That applies, however, only
to the "voltage source" output terminals (see Issue No.
22, p. 31); at the "current source" terminals there is a big
1ohm resistor in series with the speaker load, and that
definitely changes the sound, in exactly the same manner
as a tube amplifier with high output impedance. The 1
ohm source impedance interacts with the loudspeaker impedance to skew the frequency response (see Issue No.
16, p. 55, Fig. 8, and the discussion on p. 56). You can
do that yourself with any amplifier and a 1ohm resistor,
but Bob Carver gives it to you built in, with the option
not to use it. Don't tell him, though, if you're not going
to use it; a tag that hangs on every newly unpacked
Sunfire amplifier says: "Stop! Stop! Stop! For best sound
and largest soundstage, biwire woofer to voltage source
(transistor amp output) and highfrequency drivers to
current source (vacuum tube output)." Tweako dealers
and their customers find the resulting softening of the
highs and slight "warming up" of the mids/lows to be
magical; to me it's just an uncontrollable tone control,
sometimes for the better, sometimes for the worse, depending on the speaker and the program material. I favor
the accurate and predictable voltagesource outputs be-
cause I can diddle with the frequency response in other
ways if I so desire.
That's the long and the short of it. The Sunfire is a
brilliantly designed, extremely powerful, highly cost
effective, quite idiosyncratic, and decidedly controversial
amplifier. It reflects Bob Carver's priorities, which are
not everybody's. It should really be compared with the
Carver Corporation's higherpriced Lightstar amplifier,
which was started by Bob before he left and finished by
others. Unfortunately, our repeatedly promised review
sample never arrived; meanwhile the Carverless Carver
Corporation (Hamlet without the Prince of Denmark, in
Sir Walter Scott's immortal quip) isn't exactly setting the
audio world on fire. I can wait. •
26
THE AUDIO CRITIC
pdf 26
Twice Shy:
On Reencountering
Multichannel Music Formats
By Daniel C. Sweeney, Ph.D.
Freelance Contributor to The Audio Critic
It appears that, despite the clinging of diehard audiophiles to pure
two-channel stereo, the industry has opted for multichannel sound;
therefore a hard-nosed examination of the givens is in order.
Editor's Note: Furtwängler was no Nazi even though he
regularly conducted his orchestra under Hitler, and Dan
Sweeney is no tweako cultist even though he is a former
contributor to Harry Pearson's magazines (The Absolute
Sound, The Perfect Vision). A freelance writer, especial-
ly a freelance technical writer, can't afford to regard
anybody's dollar bill as tainted, not even when that bill
comes from (yuck) Harry's back pocket. I have known
Dan for a good many years and consider him scholarly,
thorough, tuned-in, and sincere—one of the few audio
journalists I respect. His knowledge of the multichannel
scene is obviously wide-ranging.
* * *
One expects that the readership of The Audio Critic
is more than ordinarily well-informed on emerging trends
in consumer high fidelity. Thus I believe I shall surprise
no one by stating that the industry will very shortly be
asking us to augment or replace our music systems in order to accommodate at least one and perhaps several multichannel playback formats.
Whether the reestablishment of music-oriented surround sound will be any more successful than its initial
appearance in the '70s is less certain. And less certain yet
is the path to be followed in adapting our carefully contrived stereo playback systems to the as yet undefined recording practices that will characterize Quad Redux or
whatever one chooses to call the multichannel Restora-
tion. This article, it is to be hoped, will provide some ba-
sis for the management of these uncertainties and for determining how, if possible, one might derive satisfactory
listening experiences from multichannel music playback.
A Watershed
Heretofore this publication has had relatively little
to say about multichannel playback systems. Such systems have been selling briskly, to be sure, but only as ad-
juncts to home-theater video installations. The music in-
dustry, for the most part, has chosen not to issue recordings in the Dolby motion-picture matrix format that is the
current standard for prerecorded video software, and so
multichannel could safely be ignored by a journal with an
emphasis on music, and on serious music at that.
Now, with the reappearance of true, multichannel
discrete formats and the announcements by a number of
recording engineers that they will release music software
in these formats, multichannel can no longer be dismissed—especially not in the light of the well-nigh ubiquitous perception in high-end hardware circles that multichannel formats are poised to obsolesce two-channel
stereo. Thus we shall have to confront the new order, and
we shall have to reexamine the basic philosophies of
sound reproduction that underpin stereo and serve to
define the limits of what is possible in creating the illusion of live sound.
Why Multichannel?
Except—predictably enough—for those same individuals who cling to analog records and prefer vacuum
tubes, most industry professionals, whether press, manufacturers, or retailers, accept the proposition that multichannel systems are unquestionably superior to twochannel stereo in the same way that the latter surpasses
mono. While relatively few industry persons have any
deep understanding of human sound-localization mechanisms and how playback systems may be made to support them, most justify the assumption of multichannel
superiority on the ground that adding channels somehow
supplies a third dimension of depth to the reproduction.
In actuality such a rationale is just as naive as the
countervailing defense of two-channel stereo on the
ground that we have only two ears, and I mention it here
only to steer the discussion firmly away from the fatuities
ISSUE NO. 23 • WINTER 1995-96 27
pdf 27
which have characterized it thus far in the audio press.
Neither explanation shows any real understanding of how
a stereo system, regardless of its number of channels,
creates an illusion of space, and neither is worth even the
briefest consideration by a serious individual.
To gain such any real understanding of how a recording/playback system might create a more or less convincing simulation of an acoustical space in a real space,
one must explore the field of psychoacoustics, and this I
shall attempt to do—albeit cursorily—in a later installment, but before I review relevant research findings in
this area, I'd like to list the perceived effects that channel
multiplication is likely to provide, and then to measure
these against both the claims being advanced for the new
formats and the performance potential of the established
two-channel format. In marketing parlance these effects
are the "benefits" of multichannel, the reasons you and I
will invoke should we liquidate our savings to propel
ourselves firmly into the new era of perfect sound for all
eternity.
The Benefits
Before we discuss the listening enhancements mul-
tichannel will supposedly provide, it might be well to examine how these formats are configured.
The new actual and proposed multichannel discrete
video-sound formats (of which more presently) all provide for four or five full-frequency-range channels. The
older Dolby Surround format provides for four matrixderived channels, while the barely alive Ambisonics format can be made to yield any number of matrix-derived
speaker feeds, but is usually decoded into four channels.
In video applications, the speaker arrangement for
multichannel playback is invariably two or three channels across the front and one or two channels in back.
For music a variety of setups has been proposed, and no
real standardization is in evidence. The received opinion
among manufacturers and industry pundits is that the video configuration is likely to prevail for music, however.
While generalizing about so many different formats and playback arrays can be dangerous, one can
speak with a good degree of certainty about many of the
basic capabilities and competences of any four- or fivechannel system.
Adding a center channel between the stereo left
and right will usually—depending on loudspeaker characteristics—increase the precision of imaging in the frontal soundstage. There are, however, exceptions, and I'll
discuss those in a later article.
Placing an additional channel or two channels of
information toward the back of the room will provide a
source for ambient effects and—given the proper speaker
type and placement—will allow for the precise placement of one or two sources at the rear speaker locations,
though arguably nowhere else—the argument being that
panning between speakers cannot be in the side and rear
quadrants. Rear channels will also permit the system to
maintain precise localization along the frontal soundstage
while simultaneously providing for some sense of
"hall"—the elusive and illusive combination of "air" and
image specificity.
In addition, some evidence exists that adding channels makes the system less room-dependent. Indeed some
researchers have suggested that room acoustics recede in
importance as channels are multiplied, permitting the recreation of the recording space in an ordinary domestic listening space. This, however, is disputed by other authorities.
These then are the principal benefits, and there
aren't so very many of them. Now to the limitations.
Four or five channels probably won't permit the extension of a sound stage to the sides and the back of the
listener, at least not through simple amplitude and phase
panning techniques. Although a number of individuals
advocating a return to multichannel claim the opposite,
published experiments don't appear to support their position.
Four or five channels won't eliminate basic problems of speaker/room interaction, such as colorations and
disturbances in localization caused by early reflections
and low-frequency imbalances arising from boundary effects.
Four or five channels won't provide for stable
phantom images between speakers, particularly between
speakers defining the back and side quadrants of the listening space. Again this assertion has been disputed by
multichannel boosters, but the extant experimental evidence points the other way.
Four or five channels will not eliminate interaural
crosstalk between the loudspeakers in the array.
Four or five channels will not provide an absolutely
convincing illusion of depth extending behind the frontal
soundstage.
Four or five channels will provide only a partial illusion of hall acoustics generally. That's because two
rear or side speakers will fail to generate a completely
convincing simulation of hall reverberation. In fact, according to one experiment, even six rear channels are not
enough!
Four- or five-channel systems will not enable listeners to localize height.
Finally, four or five channels will not eliminate
audible defects at any point along the signal chain.
You'll still hear amplifiers clip and speakers resonate and
whatever other forms of misbehavior may be present in
the basic components.
On the other hand, four- or five-channel setups will
impose much greater space demands than stereo systems,
especially if identical speakers are used all round as is
currently advocated by many high-end speaker manufacturers. Not only will the speakers themselves consume
floor space, especially if properly situated well away
28
THE AUDIO CRITIC
pdf 28
from walls, but the need for unobstructed sight lines from
speakers to listening positions would impose strict con-
straints on room decor. Of course, one could ignore such
constraints, but is the perfectionist listener going to double his expenditures for speakers and electronics, and
then forgo experiencing the full performance potential of
his prohibitively costly system?
In the light of these very real limitations one would
think that the industry, at least the perfectionist wing of
it, would view the reintroduction of multichannel with
some degree of skepticism, but such is not the case. In
the course of a normal work week, I engage in at least
three or four conversations with manufacturers' representatives pertaining to multichannel—the subject is that
timely right now—and the normal industry response is
not just uncritical advocacy but what I can only describe
as delirious enthusiasm. Everybody is taking deep pulls
from the same industry hookah, and they're all succumbing to the same unbridled euphoria. Objections such as
I've raised above are blithely brushed aside with
confident assertions that the experience of multichannel
will be so overwhelming, even with haphazardly arranged systems, that listeners will be overawed.
Perhaps so, but my own fairly extensive experience
with multichannel tends to temper my enthusiasm. During the early '80s I was friendly with a recording engineer who had an extensive library of quad mixes on fourtrack reel to reel—fully discrete and of very high fidelity.
I heard a lot of this material, most of it consisting of recordings of small jazz ensembles where one instrument
was apportioned to each speaker. The recordings were
certainly vivid and undoubtedly interesting, but the illusion of a real recording space was not particularly convincing. Of course, more sophisticated recording techniques might have yielded a better approximation of
reality, but that begs the question of what recording techniques are indeed appropriate. I might add that this individual was a seasoned professional with a very strong
commitment to quad and a thorough knowledge of quad
miking practice. Most of the individuals who will be recording in 5.1 during the next few years will almost surely be less knowledgeable.
But this is all an opinion, a matter of personal taste,
and my blase reaction may be anomalous. Ultimately,
you'll have to decide for yourself. And, it is to be hoped,
the information that follows will assist you in doing so.
Formats
To experience multichannel playback, you obviously have to have multichannel sources. Several currently exist, though there isn't a tremendous amount of multichannel music software out there yet.
The first format might be termed incidental or accidental multichannel. Recordings made according to the
MS or Blumlein techniques can be made to produce derived center and rear channels of information by means
of a simple combining matrix, where stereo left and right
are summed and differenced (that is, common elements
are canceled out by phase inversion). Dematrixing for the
purpose of deriving the four channels may be done at line
level with such devices as Fosgate-Audionics surround
sound decoders, or at speaker level with Panor's revival
of the old Dynaco QD-1, more or less a recreation of the
old Dynaquad Hafler matrix decoder.
For some persons the Dynaquad version of multichannel is all they have ever needed, but it must be noted
that the approach has never won anything approaching
widespread acceptance, and for at least two very good
reasons.
First, there's the dearth of suitable recordings.
Blumlein may be every audiophile's theoretical ideal, but
most recording engineers appreciate a few more options
than Blumlein techniques provide. Then there's the fact
that simple Hafler decoding provides for only 3 dB of
separation between adjacent channels, which means that
the frontal soundstage tends to sound disturbingly monophonic as a result of the addition of an L + R center channel. Some listeners like that, but this one doesn't, nor
does the listening public at large apparently.
Ambisonics
To discuss Ambisonics at all is to open an institu-
tional-sized can of worms which I have no desire whatsoever to plumb. Nevertheless, Ambisonics still survives
(to about the same extent as the California condor), and
since at one time or another it has enlisted the support of
some of the most eminent thinkers among audio academicians, it deserves at least a few words.
Ambisonics is the most recent incarnation of a for-
mat initially developed in Japan in the '70s and dubbed
UMX in its homeland. As such, the format is in a real
sense the last survivor of the quadraphonic experiment.
Never really embraced by the Japanese music lover, UMX eventually found a home in Britain where it
permutated into Ambisonics (horrible name!), and such
luminaries as Michael Gerzon and P. B. Fellgett contributed to its theoretical foundations and promoted it in
scholarly publications. Basically an extension of MS
techniques, Ambisonics used a matrixed three- or fourmike coincidental setup where an omni was juxtaposed
with horizontally crossed figure eights and occasionally a
single vertically oriented figure eight as well. Each mike
fed a separate track on a tape recorder, and on playback
the tracks were differentially combined in a matrix to
produce separate speaker feeds. In the consumer B format, additional matrixing was done in the recording so as
to combine the three or four tracks on the master into a
stereo pair.
Ambisonics was normally played back over four
similar speakers arranged in a diamond, but additional
speakers could be used, and the diamond arrangement
could be modified as well, thanks to various mix options
ISSUE NO. 23 • WINTER 1995-96 33
pdf 29
on the matrix decoder.
Many technical papers have been published purporting to show that Ambisonics provides an adequate
sampling of a soundfield to allow for nearly perfect recreation in a playback system. But all such analyses I have
seen ignore two significant factors in the recording/
playback process.
The first is arrival-time differences from ear to ear,
a major sound localization cue for humans. True, Ambisonic microphone setups do encode side-to-side phase
differentials, but those are not precisely the same thing,
and because the setup is essentially single-point, there's
no distance offset to mimic the disposition of human
ears. The second phenomenon is the crosstalk components produced by pairs of loudspeaker, whereby each
ear hears both speakers simultaneously. Ambisonics, like
normal stereo and most multichannel formats, makes no
provisions for addressing such components.
In any case, all existing consumer Ambisonic recordings have been issued only in the B format with its
badly degraded separation, and thus actual playback has
tended to bear the earmarks of the passive Hafler matrix—plenty of ambience and envelopment but imprecise
localization.
Discrete Ambisonics could easily be accommodated on compact disc by the new AC-3 or DTS compression algorithms, but no one in the recording industry is
promising anything. Currently Ambisonic recording activity appears to be quiescent, and none of the major labels ever showed any real interest in the format. Small
English labels such as Nimbus were the chief software
supporters, while Meridian and, very briefly, Onkyo have
been the only major hardware manufacturers to offer decoders. [See the Onkyo review on page 42.—Ed.]
I haven't cared for the few B format recordings
I've heard, but opinions differ. I would imagine that fully
discrete stuff might be very convincing—much like a
Blumlein recording stretched out in two dimensions.
Much Ambisonic material has been reissued on
compact disc. Meridian still makes a decoder, the $3600
Model 565, which also does Dolby Pro, THX, and much
else. If you feel compelled, you can still get into Ambisonics and join the worldwide fraternity of True Believers who feel it's the most significant development in the
history of recorded sound. Just don't expect much new
material to appear.
The Dolby Motion Picture Matrix
Here at last we're firmly in mass-market territory.
As most of us are aware, the Dolby motion picture fourinto-two-into-four matrix has become the standard format both for theatrical feature films and for home video
releases, and indeed so popular is the format that most receivers sold today include onboard surround-sound processors for decoding Dolby matrix materials.
Less well-known is the fact that a considerable
34
number of music recordings exist in the Dolby matrix
format. BMG/RCA have released scattered titles, and Delos has put out a fair portion of its catalogue in Dolby
Surround. ["Encoded naturally...through the use of mi-
crophone techniques," not with a matrix encoder, according to Delos.—Ed.] And because the Dolby matrix
uses similar encode/decode equations as the old Hafler
matrix, it's entirely possible to send Blumlein recordings
through a video surround-sound processor and get pretty
acceptable results.
Still, I don't see that the Dolby matrix offers any
compelling inducements to the music lover to expand his
system to multichannel if music reproduction is the sole
function of the system.
For one thing, the amount of music software, even
with the inclusion of purist Blumlein recordings, isn't
sufficient to provide clear justification for a multichannel
makeover. But beyond that, the operation of the active
matrices used in all current Dolby processors is so problematic as to invite rejection by any truly discerning listener.
At the heart of the problem is the matrix multiplier,
the logic circuit used to increase adjacent channel separation in the output and thus to create a semblance of discrete multichannel sound. Such an active matrix decod-
er's logic works by detecting channel dominance on a
momentary basis and then applying cancellation signals
to the channels adjacent to that carrying the dominant
signal. These cancellation signals consist of right total or
left total in either normal or inverted polarity, and they
serve to null crosstalk components to which they always
bear an antiphase relationship. The cancellation strategy
does work impressively in terms of meeting its stated
goal of heightening separation, but it is not without side
effects, and these side effects are of such severity as to
call into question the whole notion of using the matrix
multiplier for music reproduction. The cancellation signals themselves are usually generated by devices known
as VCAs (voltage-controlled amplifiers), which are commonly used in signal processors of many sorts and which
do just what the name implies, control an output voltage
by means of an input voltage. Unfortunately, while the
signal passing through the VCA is ramping up in response to the input, it is being amplitude-modulated and
thus distorted. All of the distortion components thus
created are passed on to the next stage, in this case to the
output channels of the matrix decoder. Furthermore,
since the distortion components do not bear an antiphase
relationship to anything in the crosstalk, they will not be
subject to cancellation.
Ever notice why distortion specs are almost never
listed for surround-sound processors? Because they're
terrible, at least when the device is measured on a dynamic basis, which is the only way it should be measured.
Cancellation-signal-borne distortion does not ex-
THE AUDIO CRITIC
pdf 30
haust the list of decoder aberrations, however. Gain riding concurrent with cancellation is another problematic
aspect of high-separation decoding. Gain riding is done
to ensure constant power in a channel as crosstalk components are being canceled; naturally the cancellation
process reduces the signal amplitude in the processed
channel and so a boost is required. Unfortunately, like
the cancellation process itself, the gain riding process
creates unavoidable degradations in the output signal.
Because gain riding amplitude-modulates the signal, it produces the familiar amplitude-modulation distortions, i.e., harmonic and intermodulation distortion. But
as is the case with single-band companders, which themselves are essentially automatic gain riders, the gainriding circuits in high-separation decoders are prone to
breathing effects—effects that are not always masked by
the signal.
Pumping and breathing can also occur when cancellation signals suddenly stop, allowing crosstalk com-
ponents to resurface. This particular embarrassment re-
sults from the fact that the logic can only address a single
pair of adjacent channels at any one instant; in other
words, crosstalk components are only removed selectively on a channel-by-channel basis, so that the system is
not highly separated all round. Essentially the logic cir-
cuit works only on one quadrant at time, and when it is
tending to one pair of channels, crosstalk is occurring
elsewhere. Such crosstalk creates a distracting, level-
dependent background—though in present-day decoders
this problem has been much ameliorated over the norm in
decoders of the past.
Of course, there are digital decoders on the market,
and these dispense with the VCAs—prime sources of the
distortion problem in analog implementation. Unfortunately, the digital logic does emulate the performance of
the VCAs and the amplitude modulation still occurs.
There are certainly advantages to digital matrix decoding—you don't have to worry about circuits drifting—
but they don't eliminate the fundamental problem of the
matrix, which is distortion due to gain modulation and
noise due to gain riding.
Obviously, the more work the logic circuit is do-
ing, the more breathing and distortion it is producing, and
for this reason among others Dolby specifies long time
constants for the logic, causing it to generate cancellation
signals at fixed levels for periods on the order of a sec-
ond, except in the presence of very marked shifts in chan-
nel dominance. This means that a lot of fairly low-level
information is escaping the logic and is making its way
through the outputs, accompanied by a great deal of
crosstalk.
It must be said that when the Dolby Pro Logic pro-
cessor is decoding movie sound tracks, this liability is
not as bad as one might think, simply because movie recording techniques have grown up around the matrix and
its limitations, and more than likely Dolby personnel
have supervised the mix to avoid taxing the capabilities
of the circuit. Most movie-sound recordists just don't
place important information in all channels simultaneously, so the crosstalk, when it occurs, isn't ordinarily going
to compromise imaging very much. But with complex
musical scores one can argue that the Dolby Pro Logic
time constants aren't ideal. From my experience, the Fosgate decoder circuits based on variable time constants do
a better job on music.
But all this is very much beside the point, really,
because the music industry shows no signs whatsoever of
adopting the Dolby matrix. No major artist or recording
engineer supports it, and all attempts to build a niche
market for surround-music playback among high-end
buyers have failed. The matrix will undoubtedly remain
the norm for movies for some time to come, but its pros-
pects in the music business remain dim.
THX
THX is not a multichannel format as such. I mention it here simply because it is often misidentified as a
format and because the THX program does play a very
significant role in the multichannel universe.
THX began as a motion-picture theater certification
program and as a means of promoting the ideas of Tomlinson Holman, founder of the program at Lucasfilms, for
improving theater sound systems. The consumer version
of THX, which was developed much later, is essentially a
licensing program for components and, latterly, software
to be used in multichannel home-theater systems.
THX standards cover laserdisc players, laserdiscs,
surround-sound processors, amplifiers, speakers, screens,
and even interconnect cabling. The program endorses
both the Dolby matrix and Dolby AC-3 discrete 5.1channel formats, and all THX-licensed processors must
meet Dolby specifications.
It is not my intention to discuss the THX standards
here except in regard to the processor. (A fuller consideration of the program will be reserved for a possible later
article.) At least one aspect of the processor standards is
very germane to the present discussion, however, that being the TXH decorrelation circuit, which has a direct
bearing on the role envisioned for the back channel by
Tom Holman.
Normally the surround channel in the Dolby matrix
is reproduced over two loudspeakers placed in the rear or
along the sides of the listening space, a single speaker
having been found to produce undesirable localization effects in the presence of diffuse, ambient information.
Holman feels that the use of two speakers, absent appropriate processing strategies, merely ameliorates the problem—in other words, one still tends to to localize sounds
to the speaker location, only now one perceives two
speaker locations rather than one.
Since Holman has always maintained that the re-
production of ambience constitutes the paramount pur-
ISSUE NO. 23 • WINTER 1995-96 35
pdf 31
pose of the rear channel, he finds localization to the
speaker position to be a significant problem. He has proposed solving that problem by altering the signals so as
to muddle directional cues. The best solution, Holman
feels, is to make the signals reproduced by the two sur-
round speakers slightly different from each other, and to
that end he developed a technique he calls, somewhat
confusingly, pitch shifting, whereby different equalization curves are imposed upon either channel. This type of
decorrelation was present in all THX-approved proces-
sors initially, though later phase-shifting circuits were approved as well.
In a full THX system, the decorrelated signals are
always reproduced by dipolar rear speakers, where the
listener sits in the null of either dipole, that is, sits directly beside the speaker so as to perceive almost no direct
sound. The reflected, decorrelated sound he does perceive is supposed to reproduce ambient effects more successfully than the output of conventional systems, such
output consisting of unprocessed direct sound.
Assessing the ultimate success of the THX treatment of the rear channels is not easy. From my experience it is indeed difficult to localize to the rear speakers
in a THX system—and I've reviewed several for other
publications—but I've always felt that there was some-
thing not quite kosher about the effect. It's not an effect
you'd likely hear in any other setting, and so it seems
vaguely unnatural—at least to me it does. Lucasfilms
have performed many listening tests purporting to prove
the efficacy of the THX methods, but there isn't a lot of
independent research to back up the THX positions. I'd
say the jury is still out.
I'd note here that THX notions concerning decorrelation and diffuse dipolar rear speakers were conceived
to support the Dolby matrix with its monophonic surround channel, which conveys difference information.
Many questions have arisen regarding the suitability of
the THX speaker configuration for the 5.1 discrete formats. (Decorrelation is obviously unsuitable for two discrete rear channels.) The answers to those questions will
come in time when 5.1 software becomes more widely
available, but at this time the evidence to form firm judgments is simply insufficient.
The DTS 5.1 Format
DTS (Digital Theater Sound) is a SouthernCalifornia-based company owned in part by MCAUniversal. The company's chief business is the licensing
of their proprietary data-compression technology and the
sale of audio playback equipment to the motion-picture
exhibitor industry, but the firm also has eyes on the consumer audio and video markets, where it hopes to establish its six-channel system as the new standard.
DTS is hardly a household word, but the company
itself has to be considered a comer. The DTS digital
sound system for movie theaters, which was first used in
Steven Spielberg's Jurassic Park, is today employed in
many more theaters than the rival Sony SDDS or Dolby
SR-Digital systems—over 5000 screens at the time of
this writing, and the company appears poised for eventual dominance in this market.
Nevertheless, in spite of DTS's remarkable success
in pro audio, the good industry position within the consumer electronics press and among most of the audio hardware manufacturers is that DTS is a negligible and laugh-
able presence in the consumer electronics industry and
that, conversely, Dolby's rival AC-3 consumer system
has already emerged as the clear victor in the multichannel format war.
Conventional wisdom on competing technologies
is nearly always correct because it reflects industry consensus and thus forms the basis of a self-fulfilling prophecy, but it seems to me that in this instance it cannot be
entirely trusted. A careful consideration of the current positions of DTS and Dolby simply does not warrant the assumption that DTS can be dismissed out of hand.
MCA's resources exceed those of Dolby by a wide margin, and DTS's growing dominance within the film in-
dustry has to be counted as a factor, particularly when
one considers that Dolby's 5.1-channel digital theater
sound system is dead last behind Sony SDDS. Bear in
mind also that Dolby's celebrated triumphs, i.e., winning
the endorsement of the Laserdisc Association and the
HDTV Grand Alliance may not be decisive, since the issue of what system will be adopted for the mass-market
digital video disc is still outstanding. Toshiba-Warner is
offering lukewarm support for the Dolby system while
providing room on the discs for simultaneous DTS audio
tracks, while the Sony-Philips consortium has yet to announce a standard. Until DVD standards are completely
resolved, DTS is still in contention. [This was written be-
fore the Toshiba-Warner and Sony-Philips factions final-
ly agreed to support a single DVD standard.—Ed. J
Concerning the format itself, the DTS consumer
system has been configured to produce a "5.1 channel"
output, meaning that five fully discrete full-range channels are specified along with one subwoofer channel.
This configuration is already standard for digital movie
theater systems, and since the initial market for discrete
multichannel is expected to be in the area of video software, movie-theater practice will prevail in the home as
well.
Because DTS must fit six channels of information
into the space occupied by two channels of compactdisc-video audio, the signals for each of these six channels must necessarily be compressed; that is, data must
be selectively removed—hopefully in such a manner that
deletions will always go unnoticed because of the masking effects of the program material. However, compression in the DTS system intended for laserdisc and the
compact disc is relatively gentle as compression goes, occurring at a 4:1 ratio, which means that one quarter of the
36
THE AUDIO CRITIC
pdf 32
initial quantity of data is preserved. While that may seem
pretty severe, be aware that the Dolby AC-3 consumer
system operates at an 11:1 compression ratio, while the
Dolby SR-Digital theater system is even more severely
compressed.
The issue of compression is the focus of much con-
troversy in the audio press, with subjectivists—not unex-
pectedly—being against it, and the Stereo Review crowd
being very much for it. Since nobody in the press has
much, if any, experience in comparing compressed with
uncompressed material, the positions are essentially ideological.
Still, no one can argue convincingly that compression can remain transparent at any degree of severity. At
some point data losses will become audible, but at what
ratio? Twenty to one? Ten to one? Four to one? Two to
one?
Moreover, the issue is vastly complicated by the
existence of several competing compression algorithms,
all of which are deemed superior by their respective
backers. "We can get away with 10:1, while they sound
lousy at 4:1" is representative of manufacturers' claims
in this regard.
So is DTS at 4:1 superior to Dolby AC-3 at 11:1?
Not necessarily. If Dolby's algorithm enjoys the edge its
backers claim for it, then the Dolby system could outperform the DTS system with less than half the data. But we
don't know if Dolby's algorithm really is better because
we lack the evidence to make such determinations.
Now it would seem that a listening demonstration,
where signals from master recordings were submitted to
either scheme and the results compared, would go a long
way toward resolving the issue, and Terry Beard, the
president of DTS, has told me personally that he would
submit his system for such comparative evaluation at any
time. However, Dolby has allegedly declined all such in-
vitations and will not submit its system for comparison
even to the companies backing digital video discs. Dolby
representatives have indicated that the company has performed internal listening evaluations, and Dolby management is satisfied with the results.
In any event, the Dolby format is already available
on laserdisc, and numerous titles have been issued as of
this writing. But thus far not one piece of music programming has been issued on compact disc.
On the other hand, no video material has been released in the DTS consumer format, but two compact
discs have been issued, on the dmp label and on Brad
Miller's Mobile Fidelity International. Both Miller and
dmp's Tom Jung have indicated that they will releasing
more material in 5.1 in the future.
Dolby AC-3 5.1
Like DTS, the Dolby AC-3 format is a compres-
sion algorithm, not a channel allocation per se. Unlike
the DTS algorithm, Dolby's has been described in detail
ISSUE NO. 23 • WINTER 1995-96
in white papers available for the asking from Dolby Laboratories and in technical articles in the consumer press.
If you haven't already steeped yourself in the background material surrounding the Dolby vs. DTS controversy, I'm sure you're asking yourself why Dolby would
opt for a higher compression ratio since, at best, maintaining transparency then becomes more difficult. The
answer is that Dolby did so at the behest of two industry
committees, the one controlling the destiny of highdefinition television in the U.S. and the other established
to set standards for the 12-inch video laserdisc.
In both cases the capacity for wideband multichannel audio was limited. In the case of HDTV, the limits
were set by the video requirements and the spectrum allo-
cation—and there wasn't much way around either. But in
the case of the laserdisc, the problem was a bit more
complex.
The Japanese wise men who set the course for the
consumer mass media decided that backward compatibility with the existing laserdisc must be maintained as
much as possible. That meant preserving one of the analog tracks and both of the stereo digital tracks. Which in
turn meant that all of the data for the six new audio channels had to go in the space taken from a single analog
track. And that, my friends, is a tight squeeze.
DTS proposed a different solution, one unacceptable to the Laserdisc Association. Why not scrap the two
stereo digital channels and leave the two analog channels? All laserdisc players can play FM analog, so the old
players can play the new discs, albeit in analog mode
only.
Association leaders were quick to object that the
analog sound was inferior and that customers would not
be happy. They also could have argued that in some parts
of the world, such as the U.K., the newer players cannot
in fact play analog at all, and thus if DTS were introduced as a worldwide standard, it would require that foreign retailers maintain a double inventory of discs.
The issue is further complicated by the fact that the
Japanese are nurturing hopes that the new 5.25-inch video disc will do what the laserdisc never could and win a
mass market. Since compatibility is not a problem with
the little disc, severe compression is scarcely required.
And of course if the little disc becomes a big seller, the
continued allegiance of a few Japanese firms to the laserdisc format will surely be reexamined. And if laserdisc is
on the way out anyway, why push so hard for heavy
compression? Reserve it for HDTV broadcasting and
leave it at that.
Of course, it might be objected that severe compression is highly desirable, if for no other reason than it
gives an opportunity to the rationalists in the industry to
stick it to audio tweakdom bigtime. You didn't like the
compact disc with 20 kHz bandwidth and 16 bits? Well,
how about an average bit rate of 8 bits? Put that in your
pipe and smoke it. Whether such a motivation is going to
37
pdf 33
figure in industry deliberations is somewhat doubtful,
however.
Describing Dolby's adaptive subband coding technique in any detail would constitute a lengthy article by
itself, so I'll confine myself to the bare essentials here.
Two basic principles are embodied in the coding—
selective reduction of resolution on a sample-by-sample
basis reflecting spectral distribution of signal content, using bandpass filters to control granulation noise attendant
upon bit reduction, and assignment of bit allocations
based on masking principles. Like the analog spectral recording circuit to which it is somewhat related in principle, the Dolby AC-3 encoder looks at the signal from moment to moment, decides where noise will be masked by
the signal content in contiguous and adjacent bands, and
allows the noise to occur where the signal will bury it. In
addition to subband coding, the Dolby AC-3 uses delta
modulation to effect a further considerable reduction in
bit rate. High-frequency channel separation is also selec-
tively reduced where it is deemed to be inaudible.
Unlike DTS's Terry Beard, Dolby representatives
have never claimed that their system is completely transparent—at least not at the specified 11:1 compression ratio. They simply maintain that it represents an improvement over the Dolby matrix—a position that would
prompt little disagreement in the industry.
Dolby AC-3 can be decoded into four, two, or one
channel as well as in the 5.1 format, but it will probably
appear in music software in 5.1. Thus, as with DTS, music recording will mostly likely hew to a pattern best suit-
ed to public address applications within commercial mo-
vie theaters.
[This introductory article does not cover the specifics of recording for multichannel and of optimizing the
listening environment—speakers, rooms, etc.—for multi-
channel playback. Further articles by Dan Sweeney on
these and related subjects are in the pipeline and will be
seen in future issues.—Ed.]
Loudspeaker Systems
(continued from page 19)
less forbiddingly priced design. Snell Acoustics would be
doing all of us a favor by making such a speaker happen.
[Flash! The above was on its way to the printer
when I learned that Kevin had left Snell Acoustics and
joined Harman International, where his mandate will be
to create an "ultimate" speaker system for a newly
formed division of the company. Snell has new owners.
End of
an
era.]
Snell Acoustics Type MC
LCR 2800 and SUR 2800
Snell Acoustics, Inc., 143 Essex Street, Haverhill, MA 01832.
Voice: (508) 373-6114. Fax: (508) 373-6172. Music & Cinema
Reference System: MC LCR 2800 center-channel loudspeaker
system, $4799.00 each; MC SUR 2800 dipole surround tower,
$6198.00 the pair. Review samples on loan from manufacturer.
These models are intended to complement the
Type MC Reference Towers in the THX-certified Snell
Music & Cinema Reference System, but they will work
just as well with the non-THX Type A towers. I received
them a very long time after the A's, so their testing was
inevitably delayed. You can expect detailed reviews in
the next issue. I have one important observation to make
here, however.
The SUR 2800 would be a hazard in households
with small children, dogs, or physically uncoordinated
old persons. It will almost surely come crashing down if
bumped hard because it is well over 7 feet tall, has a footprint about the size of the magazine you're reading,
weighs 110 pounds, and its center of gravity is way up
there. Yes, a pair of them will launch the surround sound
over everybody's head without any obstacles, but to my
mind they represent the triumph of engineering rigidity
over common sense.
I'm almost certain I'll have nicer things to say
about their performance in the review.
Coming:
• Reviews of some very interesting loudspeaker systems, including the new
Waveform Mach 17, the NHT Model 2.5, the AR Model 303, and others.
• The next installment of our FM tuner survey.
• A first look at Bob Carver's Sunfire "True Subwoofer," an 11-inch cube
claimed to reproduce 18 Hz at 110 dB SPL with low distortion.
• More reviews of multichannel AV electronics, including some with AC-3.
• A reevaluation of the MiniDisc in the light of third-generation Sony technology.
38 THE AUDIO CRITIC
pdf 34
AV/Ambience/Surround
Equipment: Our First Look at
Multichannel Processors
By Peter Aczel
Editor and Publisher
Is two-channel stereo obsolete? Just in case it isn't, these multichannel front ends have perfectly adequate stereo modes, but the
question is—are they already obsolete for multichannel?
Testing electronic signal paths in audio equipment
used to be a fairly obvious, even if sometimes complex,
procedure because all the tests were basically about one
thing—does the output resemble the input? The new generation of multichannel AV electronics changes that perspective totally—the output is a highly processed and
therefore completely altered version of the input. Only
the front left and front right channels follow the basic
rules of stereo and even they are almost invariably routed
through the processing stages without a bypass, processed or not. That makes testing from the point of view
of the audio purist not only highly problematic but, in the
final analysis, irrelevant.
One could, of course, try to verify whether or not
the processing follows accurately the prescribed protocols of Dolby, THX, or what have you, but even that
isn't necessarily a measure of quality when the processing is implemented with standard ICs that everybody
uses. Furthermore, new formats with new protocols are
coming—AC-3 has quietly arrived already (without
much AC-3-encoded software, to be sure)—so that the
models reviewed here may have a short life expectancy
at this point, although they and others like them are
what's currently out there. Their input/output accuracy
appears to be limited in nearly every case by all the extra
silicon in the signal path; forget about anything even approaching 0.001% (-100 dB) distortion; -80 dB is more
like it.
For all of the above reasons, the reviews that follow are not our usual fine-tuned critiques of electronic
performance but merely brief evaluations of engineering
thinking, circuit design, parts quality, and user interface.
I recommend that you first read David Rich's description
of the Mcintosh C39's design details in his sidebar to the
feature article on Mcintosh Laboratory, further back in
this issue. Then come back here and continue. If you are
totally unfamiliar with, or even just slightly confused
about, the various multichannel formats implemented by
these procesors, then the Dan Sweeney article on the preceding pages is required reading before you get involved
in the reviews.
B&K AVP2000
B&K Components, Ltd., 2100 Old Union Road, Buffalo, NY
14227-2725. Voice: (716) 656-0026 and (800) 543-5252. Fax:
(716) 656-1291. AVP2000 Multi-Zone Remote Controlled AW
Preamplifier with Dolby Pro Logic, $998.00. Tested sample on
loan from manufacturer.
This is both an amazing and a frustrating piece of
equipment. Amazing because its circuitry is engineered
with the utmost sophistication and elegance to make it an
up-to-the-minute, ultrareliable, full-featured AV control
unit; frustrating because the user interface is so unfriendly as to evoke a "forget about this damn thing" reaction
from the average user. Luckily the control settings in a
typical home-theater installation are left pretty much
alone, or at least changed infrequently and even then not
radically. Memory presets also help to some degree.
No on-screen displays are available. All control
functions depend on the 5-inch fluorescent display on the
front panel, which can display 16 alphanumeric characters in a single line. You step through menus with subdirectories, one line at a time, select the function you
want, and press Enter. You do this by means of the remote control or, with fewer options, a row of buttons on
the front panel. The displays are cryptic and too small to
be read from a normal viewing/listening distance, unless
your vision is far better than 20/20. If you misplace the
remote, you lose some of the functions (such as, for
ISSUE NO. 23 • WINTER 1995-96 39
pdf 35
example, balance). I became irritable using the B&K for
more than a few minutes; techie geeks probably won't
mind.
As there is a streak of techie geek in me, and more
than just a streak in David Rich, we loved, on the other
hand, the simplicity and rationality of the circuit design.
For example, there are no pots anywhere in the signal
path. "This thing will last forever," David wrote in a note
to me. Both the volume and the balance functions are implemented with the Sanyo LC7535 passive programmable attenuator. Only one stage is required for these function because the LC7535 is a logarithmic device that
changes the level in 1 dB steps. All audio switching is in
CMOS logic, using the Sanyo LC7821 chip. Zone 1 and
Zone 2 (what Mcintosh calls Area A and Area B) each
use only one op-amp, an Analog Devices OP249, in the
audio signal path, and there are no coupling capacitors.
Now that's a truly clean approach, the kind that gladdens
the heart of the audio purist; the trouble is that THD + N
measurements still run into what would appear to be the
AV-processor "stone wall" of -80 dB (0.01%); in this
case the curves for the front outputs bottom out at -76 dB
to -78 dB, almost regardless of frequency The Sanyo
chips just mentioned, which are always in the signal path,
are the most likely limiting factor here. At all but the
highest output levels there's about a 13 dB drop in noise,
though, when all the Dolby processing hardware is bypassed ("direct" mode). With the inputs shorted (zero signal), the noise floor in the direct mode is 20 to 25 dB
lower than when the Dolby circuitry is in the signal path.
Maybe it can't be done any better because all these
figures are well within the published specs. One respect
in which the B&K AVP2000 is fully the equal of topnotch stereo-only preamps is crosstalk; in the direct
mode, at full gain, front left/right out, I measured -100
dB at the lower frequencies, -95 dB at 1 kHz, and a maximum of-70 dB at 20 kHz (ref. 2.2 V out).
How does the AVP2000 compare to the Mcintosh
C39, our conservative/comprehensive high-end archetype
in this product category? The build quality is about the
same, at 28.5 cents on the dollar. Of course, there's less
stuff in the B&K—fewer inputs, fewer features (e.g., no
phono), no separate record bus (you have to record what
you're listening to), less protection, no user-friendly
knobs, as I said; on the other hand, B&K scores with an
unexpected quality touch here and there, such as for example ±15 V on the analog power rails as against ±12 V
in the Mcintosh. The surround processor board is the
same in both units, except that the B&K uses cheaper
TL071 op-amps for the subwoofer summing and all
filtering; however, filter orders and frequencies appear to
be identical. THX costs $700.00 extra (B&K AVP4000).
So, the decision regarding the purchase of a B&K
AVP2000 hinges mainly on the acceptability of its user
interface. The unit is AC-3 upgradable, so that issue
shouldn't stop you. The basic engineering is superb.
Lexicon CP-3
PLUS
Lexicon, Inc., 100 Beaver Street, Waltham, MA 02154-8425.
Voice: (617) 736-0300. Fax: (617) 891-0340. CP-3
PLUS
. Digital Surround Processor with remote control, $3200.00. Tested
sample on loan from manufacturer.
Depending on your priorities, this is either the most
desirable surround-sound processor extant or an expensive luxury you can do without. As an all-purpose AV
control center for an elaborate home-theater system, the
Lexicon CP-3PLUS is definitely underendowed. As an en-
gine for sophisticated multichannel audio processing in a
domestic chain, it has no equal. For example, it has only
four inputs for program sources, only one tape output,
and no S-video facilities; on the other hand, it can drive
seven audio channels plus a subwoofer channel with an
almost endless variety of surround/ambience/reverb effects, all of them digitally produced and far beyond the
capabilities of other processors.
Unfortunately, Lexicon is a somewhat nervous
outfit when it comes to lending equipment to reviewers;
we received no circuit schematics to be analyzed by Dr.
Rich and were pressured to return the CP-3PLUS before
we had done as much with it as we would have liked to.
Thus my comments here are necessarily brief. I call your
attention to David Ranada's exhaustive user evaluation in
the May 1995 issue of Stereo Review, I was not able to
get as far with this equipment as he did—and I trust his
critical skills.
The user interface of the Lexicon bears a definite
resemblance to that of the B&K discussed above (despairingly), but it is friendlier in several respects. There
are two lines, not just one, of 16-character alphanumeric
LCDs on the front panel, and more importantly, the processor has a character generator for a video overlay dis-
play on a TV screen. Yes, everything is menu-driven,
and the menus are labyrinthine, but the on-screen display
shows the full menu in each case, not just a fragment of
it, and that makes all the difference in the world. I still
like knobs and switches (one of my old lab instruments
from the 1970s has 54 of them), but I realize that Lexicon's implementation is the logical one when there is
such an overwhelming wealth of functions and settings.
I won't even attempt to go into the details of what
the CP-3PLUS can do—the manual is humongous—but
it's all done with DSP, which is by far the most accurate
way, not with analog chips (and that goes for Dolby Pro
Logic and THX, as well as the various Lexicon proprietary functions). When I tried the CP-3
PLUS
as the front
end for my reference stereo system, however, I wasn't
particularly pleased; I often need more than four inputs
and have little patience with long menus when I'm used
to turning a knob by two clicks for a particular function.
The sound was perfectly fine, of course. As a control
center for my home theater system, I found the Marantz
40 THE AUDIO CRITIC
pdf 36
AV600 to be more convenient, more video-friendly, easier to set up, and easier to use under standard conditions
(especially when shared with the whole family), though
not nearly as sophisticated in terms of technical features.
As for measurements, I found some peculiarities.
For one thing, I was unable to set the unit for dead-flat
frequency response. With input and output at unity gain,
the best I could get at the front left/right outputs was a
gently falling response below 1 kHz, namely -0.05 dB at
400 Hz, -0.1 dB at 100 Hz, -0.15 dB at 20 Hz, -0.34 dB
at 10 Hz. Not very significant, but a Bryston it ain't.
Above 1 kHz the response was flat up to 20 kHz and
dropped to -0.15 dB at 40 kHz. THD + N once again ran
into the -80 dB "stone wall" of the surround-sound
world; that's exactly what I measured at all frequencies
at the maximum front output of 7 V, just before clipping,
with input and output at unity gain. Changing the load to
600 ohms affected only the 20 Hz minimum disortion
(-75 dB) and reduced maximum output to between 4 V
and 5 V. With input and output at maximum gain, THD +
N increased by 10 dB to 15 dB! The noise floor of the
Lexicon with shorted inputs was comparable to that of
the B&K AVP2000, certainly not lower, perhaps even
higher here and there by a couple of dB. Front left/right
crosstalk was highest, -74 dB, at 15 kHz and dropped
steadily with decreasing frequency to -112 dB at 20 Hz;
very good indeed. That was as far as I went before I had
to return the unit.
My general assessment of the CPSPLUS, without
the benefit of a circuit analysis by David Rich and further
study, is that it is not for the typical audiophile user, even
if he/she can afford it. It is basically a piece of professional gear, albeit entirely compatible with domestic systems. That it has some awesome capabilities is without
question.
Marantz AV600
Marantz America, Inc., 440 Medinah Road, Roselle, IL 60172-
2330. Voice: (708) 307-3100. Fax: (708) 307-2687. Model
AV600 Preamplifier/Tuner, $1199.99. Tested sample on loan
from manufacturer.
"All theory, dear friend, is gray," Mephistopheles
tells Faust, and the Marantz AV600 proves it. Of all the
AV control units reviewed here, the Marantz is "theoretically" the least desirable, but it is the one that I left in my
home theater system because of its practical advantages.
A clean, elegant design is definitely not what it is; with
the surround-sound decoder and the subwoofer filter defeated, the signal still passes through 11 active stages
plus 16 blocking capacitors, 7 wiring harnesses, 4 CMOS
switches, 3 analog pots, 2 digitally controlled IC pots,
and 1 relay, in order to get from the input to the output
(David Rich's count). On the other hand, it has every
ISSUE NO. 23 • WINTER 1995-96
conceivable function, feature, and convenience you could
ask for in the front end of a home theater system, and
they all work. On top of it, the user interface is friendlier
than most.
Even an FM/AM tuner is included in this design in
its pursuit of all-inclusiveness but it's a rudimentary one;
we didn't even bother to test it for our survey of high-
quality tuners in this issue because the specs put it in a
totally different category (image rejection 50 dB, other
specs also of entry-level standard). The IF, FM demodulator, stereo demodulator, and AM radio circuits are all in
one combo chip, the Sanyo LA 1856 (strangely, they do
not use the Philips chip set). Overall, it's a mid-fi type of
tuner design.
The build quality of the AV600 is similar to that of
other Japanese equipment in its price range—single-
sided PC boards with jumpers, etc.—and not as good as
that of the U.S.-made B&K and Mcintosh units. The
preamp section has no phono, and those 11 active stages
use el cheapo JRC (New Japan Radio) NJM4558 and
NJM2058 op-amps. Input-level and tone controls are
likewise cheap analog affairs instead of solid-state controls that offer greater reliability. The tone controls are
not defeatable. One nice feature is that a stereo pair of
subwoofers can be accommodated, with 80 Hz fourthorder lowpass filtering. The switchable 80 Hz highpass
filter design for the front and center channels is secondorder. Sallen-and-Key circuits are used for all filters. The
Dolby Pro Logic decoder is also from JRC, an NJM2177.
THX is included in the base price; the decoder uses the
Analog Devices AD 1877 for A/D conversion and the
Sanyo LC78835 for D/A; the DSP engine is the Yamaha
YSS205B, which also does the delay functions for Pro
Logic. In the mono and stereo modes, the surround decoder is bypassed. The power supply is a pleasant surprise, as it provides a full ±15 V (even the Mcintosh C39
is limited to ±12 V).
The superiority of the user interface to what we
usually see in AV control centers is due to (1) very good
on-screen readouts and (2) an ergonomically superior and
very comprehensive remote control. Unfortunately, the
on-screen display capability is disabled when the S-video
connections are used, so that if you want the better color
performance available with S-video, you are reduced to
the 10-character alphanumeric display on the front panel
(which is well-designed and surprisingly informative).
This is an AV tradeoff which should disappear as more
sophisticated equipment comes on the market.
The somewhat heavy-handed circuit design doesn't
appear to affect basic performance, at least not much.
The "stone wall" of -80 dB (0.01%) THD + N I keep
coming up against in AV processors is once again evi-
dent; 20 Hz and 1 kHz just before clipping, at 5.5 V out
of front left, are at exactly that irreducible minimum value, with 20 kHz bottoming out at -75 dB and 2 V. Par for
the course, almost regardless of topology. The frequency
41
pdf 37
response suffers a bit on account of the unbypassed tone
controls; in the less good channel I measured a bump of
+0.35 dB at 80 Hz and one of +0.47 dB at 15 kHz, both
impossible to get rid of with any kind of knob twiddling.
But what's a fraction of a dB in that highly processed
sonic soup? Front left/right crosstalk is in the -77 to -69
dB range up to 1 kHz, gradually worsening (by as much
as 20 dB) on the way up to 20 kHz. That's still OK but
not great. The irreducible noise floor of the unit, with
shorted input, fluctuates spectrally between -110 and -95
dB as referred to 2 V out, which is again OK but could
be better.
Bottom line: at $1200, with THX and all other
goodies included (except AC-3), this is good value and
unquestionably does the job. Furthermore, the AV600 is
designed to mate perfectly with the Marantz MA500
monoblocks, still our recommendation for top value.
Maybe Marantz's Japanese engineers are right—put in
every possible feature, make sure everything works,
make it easy for the user, and let the David Riches of the
E.E. world grumble about the circuitry. Even so, check
out the Onkyo below before you make up your mind.
Mcintosh C39
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
13903-2699. Voice: (607) 723-3512. Fax: (607) 724-0549. C39
Audio/Video Control Center, $3500.00. THX-M optional THX
processing card for C39, $500.00. Tested sample on loan from
manufacturer.
The review of this highly elaborate but not quite
up-to-the-minute unit appears elsewhere in this issue as
part of the feature article on Mcintosh Laboratory.
Multichannel "Receiver"
Onkyo TX-SV909PRO
Onkyo USA Corporation, 200 Williams Drive, Ramsey, NJ
07446. Voice: (201) 825-7950. Fax: (201) 825-8150. Integra
TX-SV909PRO Audio/Video Control Tuner/Amplifier with re-
mote control, $1880.00. Tested sample on loan from manufacturer.
This remarkable piece of equipment throws in seven (7) perfectly good power-amplifier channels with the
control/processing functions offered by the units reviewed above and hardly charges you extra (well, very
little extra). What's more, Onkyo does it all with DSP, in
the manner of the pricey, ultrahigh-tech Lexicon. Hard to
resist, wouldn't you say? It looks like a trendsetter.
Now, it should be made very clear that this is no
longer the current model. For a while it coexisted with
the newer TX-SV919THX ($2000.00) and was then
dropped from the line, probably because there was little
or no interest in its Ambisonic decoding capability (see
42
the Dan Sweeney article in this issue), whereas the THX
capability of the newer model became paramount. The
two models are so similar in concept and architecture,
however, that you can consider this review to apply to
both in all essentials. There may even be some samples
of the older model languishing in Onkyo dealers' stockrooms. Since an AC-3 version is a reasonable expectation, both models may soon be superseded; the hometheater industry is still in its infancy and quite volatile.
The TX-SV909PRO is probably the most complex
piece of electronic gear for home use we have encoun-
tered so far. One could spend months "deconstructing"
and testing it; Onkyo must have worked on it forever.
The basic picture is this:
The control section is a digital labyrinth, with DSP
chips, microprocessors, A/D and D/A converters, CMOS
logic, etc., ad infinitum. David Rich can parse it and calls
it "a remarkable achievement;" my eyes glaze over. The
build quality is upper-middle-class Japanese, better than
Marantz, all neatly mating plug-in boards, no wiring harness. The power-amplifier channels, on the other hand,
are quite rudimentary albeit well-designed, with only a
single output device per rail. The tuner is again of the
upper-middle Japanese receiver grade, better than the one
in the Marantz AV600, but not interesting enough to be
included in our FM tuner survey. The phono section is a
single-stage circuit using a low-cost op-amp. The power
supplies for the low-level analog circuits operate at ±12
V. One baffling aspect of the design is that the available
line-level outputs are potted-down versions of the poweramplifier outputs (why not their inputs?). That mixes the
power-amp distortion and noise into the line output. Go
figure. From input to output, the signal path in the Onkyo
comprises 5 low-level active signal blocks, plus the A/D,
D/A, and power amp, plus 7 electrolytic coupling capacitors end to end, plus a 7-gang analog pot. That's still a lot
less congested than the Marantz (see above).
The user interface is far from simple (how can it
be?) but is greatly enhanced by comprehensive on-screen
displays—not disabled when S-video is used, in contrast
to the Marantz—and by an economically well-designed
remote control, most of whose functions are duplicated
by a large assortment of buttons on the main unit.
As for the -80 dB "stone wall" of distortion, this is
the first piece of AV surround equipment I have tested
that breaks it; THD + N bottoms out in the -82 to -85 dB
range, regardless of frequency, with the power amplifier
always in the loop! Clipping occurs around 140 watts
into 8 ohms and 220 watts into 4 ohms. The D/A conver-
sion (NEC µPD6376 dual-channel DACs, fed with NPC
SM5840EP digital filters) appears to be limited in resolution by the analog distortion. Gain linearity errors exist
from -60 dB down; the worst is -2.7 dB at -83 dB.
None of the above changes our basic conclusion
that what we have here is an early blueprint of the future.
THE AUDIO CRITIC
pdf 38
FM Tuners:
The Present State of the
Art of FM Reception
By David A. Rich, Ph.D.
Contributing Technical Editor
FM appears to be the orphan subject of audio reviewers, probably
because most audio reviewers have no grounding in radio-
frequency (RF) technology. This article is intended as a remedy.
Editor's Note: Once again, Dr. Rich has come up with a
tutorial type of survey article that may be a bit too technical in spots for some of our readers. Once again I say,
don't worry about it. The consumer knowledge he imparts is clear and simple at all times. If you follow the
circuit discussions, fine; if you don't, you'll still know
what are the corner-cutting solutions, what a quality design entails, what to look for when you pay a lot of money, how to be an enlightened purchaser of FM tuners. It
would be wrong to start running for cover as soon as you
see an engineering-oriented paragraph. No other audio
publication gives you comparable insights, so take advantage of them. Technical concepts have a way of sink-
ing in, even when you find them bewildering at first. All it
takes is the desire to know. Of course, as I've pointed out
before, The Audio Critic is not "My First Book of Electricity. " If you know the difference between a volt and an
ohm, between ac and dc, you'll get something out of David Rich's techie marathons.
* * *
The importance of FM radio.
Why are we running a major article on FM tuners?
FM tuners deliver a signal that is not as good as a CD, so
who needs them? Music lovers need them because they
offer a world of music not available from a personal CD
and record collection, no mater how large. Each week,
commercial and noncommercial FM stations present
"live on tape" performances of this country's symphony
orchestras. They also offer other live-on-tape perfor-
mance series of opera, chamber music, and acoustic jazz
not available from any other source but FM radio. Com-
mercial recordings are made only of the best-known per-
forming organizations, and they do not capture the re-
markable quality and range of the musical groups
throughout the nation.
Each weekday, NPR's Performance Today fills
part of the gap left by commercial recording companies.
Live performances recorded throughout this country are
presented, along with insightful commentary. The quality
of the programs' production and the music itself are unmatched by anything to come from the BBC or European
radio. FM radio also brings educational programming,
produced by NPR, PRI, and independent sources, that
helps the young and not so young understand fine music.
These programs include Adventures in Good Music,
Pipedreams, The Record Shelf, Saint Paul Sunday, Sunday Opera (with Wayne Conner), and Schickele Mix.
Other examples of quality music and spoken-word
programs that come through a tuner include Jazz from
Lincoln Center, New Sounds, A Prairie Home Companion, Selected Shorts, Car Talk, Wade in the Water, Music
from the Hearts of Space, and Thistle and Shamrock.
Yes, a lot of trash is on FM, but the preset switches on
your tuner take care of that problem
Unfortunately, this is not the best of times for ra-
dio. Commercial pressures are pushing fine classical and
jazz stations off the air. Noncommercial radio stations
have started to play rock music (they call it "world music" and "adult alternative") in an attempt to boost listenership and thus increase contributions. The problem here
is not that noncommercial radio systems are starved for
cash but that the present distribution method, where each
town has its own NPR station, is so inefficient that little
cash gets used to produce programs and broadcast them.
The money instead flows into the pockets of those who
run the local station.
All these upheavals on the FM dial may cause you
to go in search of a new tuner. For example, in Philadel-
phia the NPR station switched to all-news to gain listen-
ers. The music programs are now only available on a station 40 miles away, and to receive that station well
Philadelphians suddenly need a good antenna and tuner.
ISSUE NO. 23 • WINTER 1995-96
43
pdf 39
FM has some fundamental limitations that could be
overcome by newer technologies. For example, digital ra-
dio could be transmitted across the country using a satel-
lite. It would be similar to the DSS television system.
Hundreds of different channels would be available. This
is a much more efficient method than the present system.
Narrowcasting is possible because the signal goes to the
whole country. The satellite dish could be as small as two
inches and it could even fit on a car. The technology exists today, but commercial radio stations are blocking its
implementation. They fear that this technology would su-
persede them. Since they have been given their licenses
to the FM band for free, it makes no sense that they have
any right to block a better technology. Unfortunately
sense is not something reliably found in the halls of government.
The fundamentals.
Enough of this philosophical discourse on radio;
let's get on with the business at hand. How does an FM
receiver work, and how do you know if you have a good
one?
The first thing to understand is that the tuner has to
do two distinct jobs. One is to extract the desired signal
from all the signals and background noise it is receiving.
The other is to demodulate the signal so the two channels
of audio information can be recovered. The circuitry
should reject as much of the noise and undesired signals
as possible, so that the tuner can perform its second function as well as possible.
Different signal conditions put different requirements on a tuner. If we are trying to receive a strong signal from a directional outdoor antenna which is free from
multipath (reflected signals) as well as from strong adjacent (200 kHz away) and strong alternate (400 kHz
away) signals, then the tuner's first job is easy. How
good the station sounds will then depend on how well the
tuner demodulates the signal (Job Two). Traditional au-
dio measurements (signal-to-noise ratio, distortion, and
frequency response) tell us how well the tuner is doing
its second job. The only distinction with tuners is that the
measurements should be across a range of antenna signal
levels, since the tuner's performance degrades as the sig-
nal level decreases. Often specs are given only at unrealistically high antenna signal levels. Accuphase, a Japanese high-end manufacturer, is unique in presenting the
complete information, and they even guarantee the specifications! Of course, the Accuphase tuner costs
$2995.00, so it had better perform well, and the company
should have no reason to hide anything by giving only
minimal specifications.
Now consider the case where we are trying to receive a weak signal accompanied with strong adjacent
and alternate signals. This signal is being received on an
indoor antenna having a small gain and poor front-toback ratio (see antenna article this issue), so the signal is
noisy and corrupted with multipath. Now the ability of
the tuner to do Job One is most important. As we shall
see below, removing the adjacent- and alternate-channel
signals is going to distort the signal. The low signal level
will also cause the audio signal-to-noise ratio to be low,
regardless of how well the tuner has been designed to do
Job Two, because of the theoretical limits of FM recep-
tion. The job of the tuner under these conditions is to reproduce the signal with a performance level as close to
the theoretical limits as possible.
As you would expect, some tuners do one job better than the other. For this reason, as we shall see, there is
no "best" tuner. Further, it is important to understand that
the tuner's specifications can tell you a lot about how
well it will do the job of receiving the signal, but they are
by no means complete. The signal conditions that are
present at the input of a tuner are very complex, and
there is no substitute for connecting the tuner into your
signal environment to see how well it performs. (Just to
make sure even the far-gone tweaks understand this, we
are talking about noise and distortion levels here, not
"slam" or "pace" or some other tweako thing.)
So why do you need to read this article? Well, if
you have the need for a super tuner to do Job One well,
the best thing would be to convince all the dealers in
town to loan you all their best tuners. You could then
pick the one that gave the best results at your home location. Unfortunately, we have not encountered such
friendly dealers. So we'll try to give you some signposts
regarding the small group of tuners that may be best for
your signal conditions. That way you'll only have to try a
couple of tuners at home. But please do not run down to
the store on just our advice and purchase something you
cannot return because what worked great for us may not
work great for you.
You should also note that tuners have dozens of internal adjustments. If they are not set correctly, then all
bets are off. We have encountered many tuners that were
not adjusted properly. No doubt you may also end up
with a misadjusted sample. Thus you might have the per-
fect tuner and never know it because it is not performing
properly. Getting a tuner properly adjusted is not an easy
job. Proper adjustment of tuners often requires equip-
ment that is not in your average TV repair shop, such as a
low-distortion stereo FM signal generator and a high-
precision distortion analyzer. In addition, the adjustments
often interact, and thus the process of adjusting a tuner
properly requires a large amount of time. If you have a
dealer who has the equipment to do the job (ask to see
it!), you are better off purchasing the unit from him than
saving 5% on mail order.
It is a lot easier to assess how well a tuner will do
Job Two. As we shall see below, many manufacturers
make big cost-cutting moves that are easily identifiable
and measurable.
That said, we should also point out that designing a
44
THE AUDIO CRITIC
pdf 40
tuner is a much bigger problem than designing an amp or
preamp. Most designers of tweako audio stuff would not
know where to start.
The ins and outs of demodulation.
I am going to reverse things and look at Job Two
before Job One. Before we discuss how a signal is decoded by the tuner, we need to know how the signal is encoded (modulated). Frequency Modulation involves assigning to the transmitted signal an instantaneous
frequency (for techie folks this is the derivative of the
signal's time-varying phase) that represents the amplitude of the information signal. As the amplitude of the incoming signal changes so does the instantaneous frequency. A voltage-controlled oscillator, as the name indicates,
is a device that will do frequency modulation, since a
voltage controls the frequency of oscillation. Now, the
VCO must be very low-noise or the noise will also be
transmitted. In addition, the VCO must be very linear in
mapping the signal's amplitude to the instantaneous frequency. Any nonlinearity will result in distortion. You
have no control of this, since the VCO is on the site of
the FM transmitter. So, if the station you want to receive
has poor equipment, it does not matter how good your
tuner is.
Now, it should be clear that the louder the signal
gets, the further the frequency deviates from the zeroinput-signal frequency (the carrier). The FCC restricts
how far a signal can deviate, lest a signal from one station should wind up in the space of another. The FCC
thus sets the maximum frequency deviation from the carrier frequency that a transmitted signal can have (75 kHz
for FM). Since the process of Frequency Modulation is
nonlinear, the spectrum of the signal at the output of the
VCO can be much more than twice the frequency deviation. To understand this we need to use the Bessel function [Cook 1968].
[This section contains too much math and has been
censored by the Editor.—DAR]
We can thus see that the typical FM spectrum occupies from 150 to 200 kHz. To prevent exceeding the FCC
limits, stations put limiters in the audio signal path. Because loud stations attract more listeners, many stations
compress the audio signal and then let the limiter come
on often. Nothing can be done to fix the signal once it is
mangled in this manner. Classical, jazz, and some NPR
stations do try to produce a low-distortion, low-noise signal with a wide dynamic range, worthy of the tuners covered in this article. But it must be noted that even these
stations will use some compression to prevent the signal
from becoming inaudible on cheap equipment or in a car.
So how should we demodulate an FM signal? The
optimal receiver in communication systems often involves placing the encoder in a feedback loop. For the
case of FM, we would have the decoder's output connected to the input of a VCO (the modulator for FM sig-
Figure 1:
Phase-locked
loop FM
demodulator
with high-
performance
VCO.
Audio
Signal
Dynamic Twin Drive V.C.0
nals). We then need a mechanism to compare the output
of the VCO and the incoming signal, so that the VCO's
instantaneous frequency matches the instantaneous frequency of the incoming signal. An appropriate mechanism is a phase detector which compares the phases of
two signals. (Recall that the instantaneous frequency of a
signal is related to the phase of the signal.) The error signal at the output of the phase detector is filtered to stabilize the loop and returned to the VCO input to close the
loop. The circuit that has just been described is a phaselooked loop (Figure 1). It has been shown that the PLL is
indeed close to the optimum demodulator for FM [Viter-
bi 1966]. As we shall see, many high-end FM tuners use
a PLL demodulator. This a relatively recent development
because, as stated above, the VCO must be very linear
and have very low noise. Only recently have the Japa-
nese designers been able to make such a highperformance VCO available at low cost. An interesting
property of a PLL is that it can demodulate a signal with
a lower signal-to-noise ratio at the antenna terminal than
other types of demodulators [Panter 1965].
One advantage of a PLL is that the bandwidth of
the demodulator can be easily changed. For the most
faithful demodulation of the signal, the loop should have
a wide bandwidth. This is the condition we want for a
good, clean signal, but if the incoming signal is noisy,
then we want to reduce the loop bandwidth to reduce the
effect of the noise [Gardner 1979]. Another key advantage of the PLL is that the output signal level is independent of the amplitude of the incoming signal (the reduced
signal level may reduce loop bandwidth, however; see
[Gardner 1979]). Since noise and interfering signals will
cause the incoming signal to have significant amplitude
variation, we want the detector to ignore these variations.
The specification that tells us how well a tuner rejects
AM signals is the AM rejection. AM suppression is
defined in terms of the relative disturbance caused by amplitude modulation when the carrier is simultaneously
amplitude- and frequency-modulated [IEEE 1975]. This
is a key specification to determine how well the tuner re-
jects interference such as fading, multipath, airplane
flutter, lightning, electrical equipment noise, etc. Super
tuners should be in the 80 dB range. Care must be taken
with this specification because it gets better at higher RF
signal levels, and most data sheets do not give the level
at which the test was performed.
The pulse-count demodulator is another highperformance circuit (Figure 2 shows an implementation
by Accuphase). Based on theoretical analysis this circuit
ISSUE NO. 23 • WINTER 1995-96 45
FM
Signal
pdf 41
jacent and alternate channels and by RF intermodulation.
Adding to the fun is the fact that subcarriers containing
Muzak or digital data such as paging signals are also sent
along with the FM signal you are listening to. [The politi-
cally corrupt and cynically compromised FM stereo standard bulled through the FCC bureaucracy in 1961 is the
reason for this.—Ed.] These adjacent-channel frequen-
cies, when demodulated, cause beat interference called
"birdies." (If you listen to a weak stereo signal on your
tuner during a quiet passage you will know why they call
it birdies.) Removing the interfering signals with a sharp
filter before the switches will clean up the birdies (they
call it an antibirdie filter) but will also cause phase shifts
that prevent synchronization of the switches to the transmitted left and right channels. That results in distortion
and reduced channel separation. A specification called
"SCA rejection" indicates how quiet the tuner is when
the Muzak subcarrier is present.
A better approach is to use a sinusoidal 38 kHz signal and an analog multiplier, instead of the square-wavedriven switch. No demodulation of out-of-band signals
can occur in such an arrangement. The problem is that
you need to have a high-performance analog multiplier.
These are open-loop devices; they have no feedback loop
in the multiplier core to improve the accuracy of multiplication. Multiplication accuracy to 0.1% is thus very
difficult to achieve. Obviously, any inaccuracy in the
multiplication will give rise to distortion. It is difficult to
design analog multipliers to have good noise performance. The Rotel RHT-10 tuner reviewed below is the
first tuner to my knowledge to do this. (Something did
not turn out quite right with the design because an antibirdie filter is still in the signal path.)
Pioneer also uses a sinusoidal 38 kHz signal. They
combined this with a pulse-count detector (see explanation above), since the output of the pulse-count detector
is binary. They connect the input of the switch to the 38
kHz signal and switch it with the binary signal from the
pulse-count detector [Ishida 1984]. The problem with
this approach is that it only works with a pulse-count demodulator. Pioneer appears to have extended the ap-
proach to the use of PLL and quadrature detectors in
their current product, but how it works using these demodulators has not been disclosed. (I tried to get an answer from the engineers at Pioneer, but their response
was to refer me back to the original patent they had filed,
which discusses only the use of pulse-count detectors. Either my question did not get translated into Japanese correctly, or the engineers just do not want to disclose what
they are up to.)
Sansui uses another approach that allows for full
integration of the demultiplexer (another term commonly
used for the stereo demodulator or stereo decoder). Sansui attempts to approximate a sine wave using Walsh
functions [McGillem 1974]. The Walsh functions are a
set of periodic rectangular pulses which, when combined
at different amplitudes, can approximate an arbitrary periodic waveform. The Fourier series does the same thing
but with a set of periodic sine waves. What makes the
Walsh functions ideal for integrated systems is that
summed scaled rectangular pulses are easy to generate on
an IC. It turns out that two scaled Walsh functions are all
that is needed to eliminate all the harmonics in the PLL
output, up to the 8th harmonic [Takahashi 1985]. To generate the correct Walsh waveforms, the VCO in the Sansui chip runs at 304 kHz. Since Sansui has fallen on hard
times, you cannot purchase this interesting tuner (TUX701) with the above chip and a nice PLL FM detector
anymore.
Chips by Allegro, Sanyo, and Sony use a similar
scheme, although they have not published the detail that
Sansui has. Antibirdie filters are typically not used in
front of these chips. The team from Delco Electronics
[Manlove 1992] also uses a similar idea, although the
complete circuit implementation is significantly different.
Delco uses more periodic rectangular functions to get the
job done. This removes even more harmonics.
You may have noticed that this is the third time I
am mentioning this Delco chip, so you might want one in
your tuner. Unfortunately, to my knowledge, Delco does
not sell chips, so this chip can be found only in GM car
radios. Yes, the FM demultiplexer chip in your car is
quite possibly better than the chip in your tuner.
Noise in an FM system rises with the audio fre-
quency [Taub 1971]. For this reason the signal transmit-
ted at the FM station is pre-emphasized, so that the signal
rises 6 dB per octave above 2 kHz. A matching deemphasis circuit is in the receiver. This de-emphasis circuit counteracts the rise in the noise. The problem with
this scheme is that the signal levels at frequencies above
2 kHz must remain low, or else the station will overmodulate. Acoustic music and speech luckily have spectral
densities that decrease with frequency above 2 kHz.
In a stereo transmission, the mono signal occupies
the same spectral band as it would if a mono-only transmission were taking place (see explanation above). That
was an FCC requirement in order to insure that a mono
receiver could receive a stereo signal. The information
about the stereo signal is contained higher up in the spectrum (Figure 4). These new sidebands come from the
sampling action of the switches at the transmitter, which
run at 38 kHz. It can be shown (but not here, says the
Ed.) that these sidebands contain the signal L-R. When
you add L-R to the mono signal, you get the left channel.
Subtract it and you get the right channel. It can be shown
that the sampling action of the switching circuit in the
48
THE AUDIO CRITIC
Figure 4:
Demodulated
spectrum of
a stereo
FM signal.
pdf 42
Figure 5: Simplified block diagram
multiplex demodulator does this function, but do not
worry if this is not clear. What you need to know is that a
mono FM signal has a bandwidth to 15 kHz, but a stereo
FM signal goes out to 53 kHz. Now recall that noise increases with frequency in FM. So the L-R information in
the 23 to 53 kHz portion of the composite signal has a lot
more noise. How much poorer is the signal-to noise-ratio
of the stereo information? An order of magnitude poorer, 22 dB worse to be exact [Taub 1971].
If the incoming signal is strong and clean, the theoretical signal-to-noise ratio is very high and this 22 dB
performance deficit is not significant, since other noise
sources dominate. The signal-to-noise ratio then should,
ideally, become the value produced by the tuner when it
is receiving a mono signal. If the signal is weak, then we
can hear every bit of the 22 dB noise penalty when we
switch from stereo to mono. It turns out that the ear tolerates fairly low channel separation at higher frequencies, so a blend circuit is often used to reduce the channel
separation (and hence the noise) in the upper frequencies.
Pioneer takes this one step further. They have a set of
blend circuits, each in a specific passband. A circuit
looks at the noise in each passband and adjusts the
amount of blend up or down in response to the noise in
that band. This dynamic adjustment gives a better stereo
image, with less noise on weak signals. The downside of
this system is that it is very complex.
Carver (the man and the company) varies the
amount of L-R signal used, depending on how different
in level and content L is from R [Feldman 1982]. Carver
creates two signals, L/R and (L+R)/(L-R), and uses
them to sense how different the levels of L and R are.
Carver also uses more L-R when the leading edge (fast,
short-term information) of a signal is detected. Carver
says he uses a psychoacoustic phenomenon known as the
precedence effect. When leading-edge information occurs, it is critical to the localization process. At other
times, Carver uses a phony L-R signal concocted from
the low-noise L+R. He does this with what he calls a
phase randomizer and a spectral shaping circuit. It would
of a superheterodyne receiver.
appear that all this should not sound at all like the original stereo signal, but in practice it works very well, at
least to my ears. Is the Pioneer approach better than the
Carver approach? Stay tuned (no pun intended); we intend to tell you when we have both tuners in our laboratory back to back.
Pioneer has one more trick up their sleeve. They
noticed that most of the noise and interference occurs
above 38 kHz. They also determined that they only need
the information from 23 kHz to 38 kHz (for the techie
crowd, this is the lower sideband of the L-R as seen in
Figure 4) because the upper and lower sidebands contain
the same information. Now Pioneer uses only the information in this band (for the techie folks, they use a single-sideband demodulator). Unfortunately, this process
introduces some distortion in actual practice, and the THD
of the decoder is in the 2% range. That is very listenable,
however, in the context of a weak FM radio signal.
The front-end circuitry.
OK, it is now time to look at the front end of the
tuner, since we started in the middle of the signal path.
The job of the front end is to provide the sensitivity to detect the presence of the desired signal and the selectivity
to accept the selected station and reject all other stations.
Figure 5 shows a superheterodyne receiver. The RF amplifier, the local oscillator, and the mixer are called the
front end. In a superheterodyne receiver the incoming RF
signal is mixed with a signal from the local oscillator,
changing for each station, in such a way that the output
has the same center frequency for all stations. This approach allows for excellent selectivity because the tuned
circuits in the IF strip that reject interference are fixed
tuned circuits. This results in much sharper cutoffs than
if the filters had to be variable, as they do in the RF stages. In addition, it is possible to have higher-gain amplifiers at lower frequencies.
The RF stage amplifies the weak signals to a level
at which the mixer can work properly; thus it increases
sensitivity. If the signal were strong you would think you
ISSUE NO. 23 • WINTER 1995-96 53
Radiofrequency
amplifier
Mixer
Intermediate
frequency
amplifier
Demodulator
Demodulated output
Output
Amplifiers
Local
oscillator
Simultaneous
tuning
pdf 43
could bypass the RF stage, but you cannot for two important reasons. The first is local oscillator radiation. If the
RF stage did not exist, the signals from the local oscillator could get back into the antenna through capacitive
and inductive coupling. Since the local oscillator runs between 98.6 MHz and 118.6 MHz, this is a good way to
cause interference right in the FM band.
The second thing the RF stage does is improving
image rejection. In a superheterodyne receiver, a problem
at the mixer occurs because the mixer can translate signals both above and below the local oscillator frequency,
and it thus is possible for an undesired station to get
translated to the IF frequency. In most FM receivers, the
local oscillator is set above the incoming RF signal; thus
the difference frequency contains the desired signal at the
output of the mixer. The image frequency is above the local oscillator frequency. The image frequency is thus the
desired RF signal's frequency plus twice the IF frequency. Now, if the IF frequency is 10.7 MHz, the image frequency can never be in a broadcast FM signal. That is
not an accident; we do not want an FM broadcast signal
to become an image signal. The FM band is 20 MHz
wide (87.9 MHz to 107.9 MHz), and since the image signal is at the RF signal frequency plus twice the IF frequency, any IF frequency greater than half the width of
the FM band will work. They put the IF frequency as
close as possible to this limit, because the lower the IF
frequency the easier it is to build a selective filter and
create high-gain amplifiers. (For homework show why
the IF frequency of an AM receiver is 455 kHz). So we
are concerned with what is on the air between 109.3
MHz and 129.3MHz. What is there usually consists of
not very large signals because these are frequencies for
aviation activities. Note that if the local oscillator had
been chosen to be below the RF frequency, the images—
in this case the incoming signal frequency minus 21.4
Mhz—would come from TV channels 3 to 6, at 60 MHz
to 88 MHz, each TV channel being 6 MHz wide [Cook
1968]. So now you know why that is not done. The specification for a broadcast tuner that tells you how well it
rejects images is called the "image-response ratio" or
"image-rejection ratio" [IEEE 1975].
Unfortunately, other image interference is possible,
because the mixing signal often contains harmonics.
These harmonics also generate sum and difference prod-
ucts that can move undesired signals to the IF strip. TV
channels 9 to 13 (186 MHz to 216 MHz) are sources
from which second-order images, caused by the second
harmonics of the oscillator, can get into the IF strip. Filtering the RF signal beforehand and attenuating the unde-
sired one will prevent these images from contaminating
the desired signal. Third- or higher-order images are rarely a source of interference because they are very highly
attenuated by the tuned RF amplifier [Cook 1968].
Another interference is caused by spurious response. Spurious response is caused by nonlinearity of
the RF amplifier or mixer. This can cause intermodulation products which again can fall into the IF band. Under worst-case conditions, with a heavily overloaded RF
front end, a station may appear at a number of places on
the dial. A common mechanism producing a spurious signal is the second harmonic of the local oscillator beating
against the second harmonic of an RF signal. The IEEE
test for this [IEEE 1975] is called the "characteristicfrequency test." In this test the incoming frequency is set
to 103.35 MHz and the receiver is tuned to 98 MHz. The
tuned circuit before the first RF stage helps here also by
filtering out undesired signals before they get to the tuner.
Other scenarios involve two incoming signals and
the local oscillator together forming a spurious signal
[Cook 1968]. The IEEE standard test for RF intermodula-
tion involves testing for the condition where the second
harmonic of one interfering signal at 98.8 MHz is mixed
with another interfering signal at 99.6 MHz. The mixed
signal is at 98 MHz. The test is reported as the "twosignal spurious-response ratio." The frequencies of the
two signals are too close together for the RF stage to pro-
vide any filtering. The test thus checks for how well the
RF and mixer stages are designed by checking to see if
they generate intermodulation distortion.
The IEEE standard allows the worst-case result of
the tests for characteristic frequency and two-signal spurious response ratio to be reported as a single number
called the "spurious response ratio," representing the
poorest of these measurements. Unfortunately, this is often not followed in specification sheets, and it appears
that the characteristic-frequency spurious response ratio
is then published as the spurious response ratio, period.
In the case of a very strong station, it is important
that some method of signal attenuation be supplied at the
Figure 6: RF stage of the now defunct
Technics "Professional Series" ST-9030
FM stereo tuner (from the good old days when
front ends were front ends).
54
THE AUDIO CRITIC
Front End Block Diagram
pdf 44
input of the RF stage to prevent overload. In some tuners
this is automatically controlled by an automatic gain circuit. Other tuners have manual controls that enable the
attenuation. The manual controls are nice because sometimes the AGC gets things wrong. Consider the case of a
relatively weak-signal channel adjacent to a very strong
signal. The AGC would turn off the attenuation to gain
up the signal, but in doing so it would cause the strong
adjacent-channel signal to overload the RF stage. Smart
AGC circuits, such as used in the National Semiconductor LM1865, can be designed to look for strong adjacentchannel interferennce before increasing RF gain.
In the good old days, super tuners had super front
ends, For example, the Technics ST-9030 had an eightgang tuning capacitor and two RF amplifier stages (Figure 6). The ST-9030 had an image and spurious-response
rejection of 135 dB because of this complex RF stage.
The eight gangs added up the following way: One was
for the local oscillator, and one was for a filter after the
local oscillator to remove harmonics. Each of the RF
stages had a double-tuned filter preceding it, as did the
mixer. Each double-tuned section needed two gangs. A
double-tuned section provides a flatter passband and a
steeper rolloff rate than a single-tuned section. Today
you get one RF amplifier and the equivalent of three or
five gangs. The result is that some of the best tuners today have characteristic-frequency spurious-response and
image-rejection ratios of only 80 dB. In addition, they
may lack sufficient RF gain, especially when operated
from an indoor antenna. The Magnum Dynalab "Signal
Sleuth" is an extra stage of tuned RF gain which can be
added to any tuner. This device can solve some difficult
signal problems that some of the modern super tuners
cannot handle. The Signal Sleuth is reviewed in this issue.
Another important parameter that the RF amplifier
should satisfy is constant input impedance, independent
of frequency. The need for this is explained in the indoor
antenna reviews found elsewhere in this issue. As in the
case of antennas, the deviation from the ideal input impedance can be given as a voltage standing-wave ratio
(VSWR). This is how Accuphase gives it (as usual they
have the most complete set of specifications). Most manufactures do not give the specification at all.
In modern receivers, all the tuned circuits are tuned
with voltage-controlled capacitors that use a diode assembly. This is called a varactor. The voltage range to
tune the varactor should be large, so that the timevarying information signals present at one end of the varactor cannot significantly change its capacitance. Signaldependent changes in capacitance can give rise to modu-
lation components. Early varactors had this problem, but
voltage swings for tuning over the FM band are now 20
V for modern varactors, and the information-signal
swing is much smaller than this.
The voltage for the varactors comes from the loop
filter that forms yet another PLL. The local oscillator is
the VCO of this PLL. A set of digital dividers after the
VCO is used to tune the tuner. The output of the dividers
is connected to a phase detector. The other input of the
phase detector comes from another digital divider that
has as its input a crystal reference oscillator. Under microprocessor control, the dividers are set so that the local
oscillator runs at the correct frequency to move the desired signal to 10.7 MHz. The voltage at the output of the
PLL loop filter also adjusts the tuned circuit's RF stage
and mixer to the correct position to receive the desired
station. Clearly, no noise should be on the voltage that is
connected to the varactor in the VCO because this will
give rise to phase noise in the local oscillator. Since the
PLL is a closed-loop system, the loop filter cannot be de-
signed to provide an arbitrary amount of filtering because
this will cause the PLL to go unstable [Gardner 1979].
The best way to insure a quiet varactor input voltage is to
run the phase comparator at as high a frequency as possible. The undesired signals at the output of the phase comparator can then be more easily removed by the loop
filter because they are high in frequency.
Early digital tuners with slow IC technology had a
problem running at high phase-comparator speeds and so
they acquired a bad reputation. The phase comparators
also had problems with dead bands that made things
worse. (A dead band is a region where the phase detector
gives no change in output even when the phase difference of the incoming signal changes.) Modern ICs run at
higher frequencies and use high-performance phase comparators; thus modern tuners do not suffer a performance
penalty when they use digital tuning. RF noise from the
microprocessor and other digital circuits also caused
noise problems in early frequency-synthesized tuners.
These problems do not occur in modern designs because
the problem is understood and techniques to minimize
the interference are known. Linearity of the VCO is not a
problem in digital tuning because only one frequency has
to be synthesized to receive a given station.
The problem with digital tuning is that the station
may not be at the exactly correct frequency. This is especially true of stations retranslated to other spots on the
dial on FM cable systems. It is thus desirable that a finetuning mode be available to detune the receiver.
The mixer is a relatively simple circuit but very
hard to optimize, according to Richard Modafferi, who
has done state-of-the-art mixer designs. Obtaining sufficient dynamic range is the really tough design challenge.
Two design parameters must be optimized to have good
dynamic range: (1) noise levels in the mixer must be very
low so that weak signals are received; (2) the mixer stage
must be able to handle strong signals in a manner that
does not allow intermodulation distortion to occur, which
is a more difficult requirement here than in the RF stage
because the mixer is not just an amplifier; it is also performing frequency translation. A lot of other design chal-
ISSUE NO. 23 • WINTER 1995-96
55
pdf 45
lenges also exist in mixer design, such as making sure the
mixer does not load down the RF stage or interfere with
the proper operation of the local oscillator. A balanced
mixer will usually perform better than a single-ended
one. In a balanced mixer, even-order harmonic distortion
components from the local oscillator do not become involved in the mixing process. This reduces spurious in-
terference. The downside of a balanced mixer is that it is
more complex. Also, it must be designed very carefully if
is to be truly balanced at RF frequencies.
The IF section amplifies the desired signal and removes the interfering signals before the FM detector.
One thing the IF strip should not do is directly pick up
any signals at its center frequency. This can happen because the IF strip has a lot of gain. 10.7 MHz happens to
be kept relatively quiet by the FCC because of this problem. Correctly shielding the IF stage also helps. The IF
rejection specification is a measure of how well the IF
strip rejects this signal. In the IEEE test called the "IF response ratio" [IEEE 1975], an FM-modulated signal with
a carrier at 10.7 MHz is applied to the antenna terminals.
The gain requirements for the IF strip are quite
high. At the antenna input a signal may be as small as 1
µV. The FM detector may require a signal of 1 V. That
adds up to 120 dB of gain. The RF section is going to
have about 30 dB of gain, so the IF strip needs to have a
gain of 90 dB. Fortunately the stage does not have to be
linear, so feedback is not used and the gain is achieved
by cascading individual stages of open-loop gain. This
does not prevent the possibility of oscillation in the IF
strip through parasitic coupling between the output and
input of the strip. Because of all the gain in a tuner, the
designer must always be on guard against such problems.
In between each stage is a tuned filter. This is im-
portant because, as we gain up the desired signal, we also
gain up the undesired signals. The filters reduce the am-
plitude of the interference before it is sent on to the next
stage of gain. In this manner the desired signal amplitude
becomes larger than the interfering signal.
For the FM signal to pass through the IF section
without picking up distortion, it is important that none of
the sideband components of the FM signal be shifted in
phase relative to one other. (No, all you tweaks, this has
nothing to do with the ear's sensitivity to phase. In FM
we have modulated the signal so that the instantaneous
frequency, and thus the phase, contain the information
about the baseband signal. That is why phase integrity is
important in FM signals.) Now we have a problem, since
a sharp IF filter will reject the interfering signals, but it is
also going to affect the phase. This is because you cannot
have a sharp transition band in a filter without significant
phase delay [McGillem 1974]. Consequently, a good tuner offers selectable IF filters. Wide filters give low distortion in the demodulated audio signal. Narrow filters give
good rejection of interfering signals but have poorer audio performance. Some tuners even have an extra narrow
56
mode for bringing in very difficult signals, at the penalty
of even more audio signal distortion.
Modern filters in a tuner are mechanical resonant
elements whose input is driven by, and whose output is
picked up by, an electromechanical transducer, such as a
piezoelectric element. The quartz crystal is a simple mechanical filter. In the IF strip multipole elements, often
called ceramic filters, are used to give flat passbands and
sharp rolloffs. These filters perform much better than the
old double-tuned circuits they replace. Another advantage is that these filters do not have to be tuned. In the
wide mode two filters may be in the signal path; in the
narrow mode four filters. In extra narrow we can see five
or more filters, and the passband of each of the filters will
be decreased at the cost of phase distortion. The Pioneer
Elite F-93 wins the contest of the most mechanical filters
in the IF, with eight filters. The old Mcintosh MR-78 had
a filter that placed transmission zeros at the alternate- and
adjacent-signal positions. That appears to be a good idea.
I cannot figure out why it is not used in other tuners. (The
high cost of the filters is the likely reason.)
Pioneer came up with a very innovative solution to
the "narrow filters distort the signal" problem. They
used this solution in the F-91 tuner. It is such a good idea
that it could have gotten someone a Ph.D. (Indeed, a related concept did get someone a Ph.D.—see [Rich
1991].) What the engineers at Pioneer observed was that,
although the spectrum of an FM signal is wide when av-
eraged over time, at any instant in time the signal has just
one spectral line at the instantaneous frequency. So, if
you build a time-varying filter with a very narrow bandwidth that moves with the instantaneous frequency of the
desired signal, then you can eat your cake and have it
too. The very narrow bandwidth removes the interferers,
but the time-varying nature of the filter insures that the
FM signal is not distorted. To move the filter around, a
complete auxiliary IF strip designed in the conventional
manner and a PLL FM detector are required. We now
have a chicken and egg problem, since the conventional
receiver has to place the passband of the time-varying
filter in the correct place. This is what limits the performance of the system.
One very interesting (at least to this author) variation on this is to replace the bandpass filter with a notch
filter. Then the stronger signal is removed, but a weaker
cochannel interfering signal may now be revealed. Further information on this is in—you guessed it—[Rich
1991].
Surprisingly, the F-91's tracking filter was dropped
from the newer Pioneer Elite F-93. The reason given by
Pioneer was that the circuit needed very high-precision
(read high-cost) parts and that performance nearly as
good was achievable without them in the revised tuner.
The F-93 reduces the distortion caused by the fixed narrowband IF filter by compensating for it in the FM demodulator.
THE AUDIO CRITIC
pdf 46
Sony uses a variation on the tracking-filter theme
that involves dynamically varying the position of the RF
and mixer filters in response to the signal level at the output of the FM demodulator (which represents how far the
instantaneous frequency has moved from the carrier frequency). This is easier to implement in the RF and mixer
stages than in the IF because these filters are already tunable. One would think the large signal delay in the IF
strip and detector would prevent this from working well,
but Sony is shipping tuners with this thing, so it must do
something positive.
One wonderful property of FM is that two cochannel signals can be very close in amplitude and yet only
the stronger will emerge from the FM detector. This is
the capture effect in FM systems. It does not exist in AM
systems. A specification that tells you how much stronger
a signal has to be to capture the receiver is the capture ratio. In a super tuner it is as small as 1 dB. The capture ratio is also important for a situation where multipath exists, since a multipath signal looks like a cochannel
signal to the receiver [Panter 1965].
When multipath or a significant interferer is
present, the amplitude of the IF signal can vary sig-
nificantly [Panter 1965]. This does not affect the performance of the FM detector provided it has good AM re-
jection—strike one for the quadrature detector. When
multipath or a significant interferer is present, the com-
posite instantaneous frequency variation can be very
wide; thus the FM detector must have a very wide linear
range to accommodate these excursions—strike two for
the quadrature detector. The widebanding works because
the inband signal at the output of the FM demodulator
turns out to be the desired signal. The interfering signal
gives rise to distortion products that are not present inband. If the discriminator is not wideband enough, then
inband distortion products may appear in addition to the
out-of-band products [Panter 1965]. The bandwidth requirements rise steeply as the power ratio of the desired
to the interfering signal closely approaches 1. Thus a detector which does not have wide bandwidth will have a
poor capture ratio.
The quadrature detector does not strike out because
of the limiter stage. The ideal limiter takes an input that
is varying in both frequency and amplitude and provides
an output of constant amplitude that retains the instantaneous frequency information of the transmitted signal
[Cook 1968]. In essence it is a high-gain amplifier that is
driven to clip. At some small signal level, the limiter will
not have enough gain to clip, and the circuit will just amplify the signal and pass on the amplitude variations.
Cascading multiple limiters reduces the signal level that
is required to saturate the limiter. So the limiter removes
the amplitude variations that the quadrature detector is
sensitive to and gives the receiver its AM rejection, usually 50 to 60 dB. Of course, if the detector has some AM
rejection, like the PLL or pulse-count detector, then the
limiter is icing on the cake and the AM rejection goes to
80 dB.
The limiter can also help to reduce the bandwidth
requirements at the detector in the presence of an interferer. This is done by the use of a cascaded narrowband
limiter [Baghdady 1955]. What Baghdady found was that
when the desired and interfering signal enter the limiter,
the energy of the interfering signal spreads out in its
spectrum. If you then send the signal through a narrowband filter, some of the interferer's energy will be removed. Continuing to do this a number of times will reduce the energy of the interfering signal relative to the
stronger signal at the detector's input. This then eases the
bandwidth requirements of the detector.
We have now arrived at the input of the FM detector, which is where we started at the beginning of this article. With a basic understanding of how an FM tuner operates, we can move on to the reviews. It should be noted
that we did not hook all the fish in this first survey. Promising tuners not included are the Pioneer Elite F-93 and
Sony ST-SA5ES referred to above. Also promising are
the very expensive Accuphase T-109 that was also mentioned above and the bargain-priced JVC FX1010TN
and Yamaha TX-950. Unfortunately, one of the great
tuner manufactures of the past, Kenwood (great because
they also design ham radio equipment and thus have a
large and experienced design team) no longer brings any
super tuners into this country.
References
Baghdady, E. "Frequency Modulation Interference
Rejection with Narrow-Band Limiters." Proceedings of
the IRE, January 1955.
Clarke, C. and D. Hess. Communication Circuits
Analysis and Design. Addison Wesley, 1971.
Cook, A. and A. Liff. Frequency Modulation Re-
ceivers. Prentice-Hall, 1968.
Feldman, L. "Carver TX-11 Stereo FM Tuner."
Audio 66.12 (December 1982).
Carson, R. Radio Communication Concepts: Ana-
log. Wiley-Interscience, 1990.
Gardner, M. Phase Lock Techniques. 2nd ed. Wi-
ley, 1979.
IEEE Std. 185-1975. "Standard Methods of Testing
Frequency Modulation Broadcast Receivers." Copyright
1975. Approved February 1977, American National Stan-
dards Institute. (Also listed as Institute of High Fidelity
IHF-T-200 1975.)
Ishida, K. and T. Numata. "FM Direct Stereo De-
coder." Journal of the Audio Engineering Society 32.6
(June 1984).
Manlove, G., J. Marrah, and R. Kennedy. "A Fully
Integrated High-Performance FM Stereo Decoder." IEEE
Journal of Solid-State Circuits 27.3 (March 1992).
McGillem, C. and G. Cooper. Continuous and Dis-
crete Signal and Systems Analysis. Holt, Rinehart and
ISSUE NO. 23 • WINTER 1995-96
57
pdf 47
How I Evaluate FM Tuners
By Richard T. Modafferi
Technical Consultant to The Audio Critic
I've designed and built RF devices
(transmitters, tuners, shortwave radios,
cable TV amplifiers, etc.) for nearly 50
years, and thus one may assume in this
regard that
(1) I know how to design and test
RF devices, or
(2) I'm good at fooling people that
I'm good at (1) above, or
(3) we're all ignorant and it doesn't
matter.
Perhaps a combination of the three
items above is closest to the truth. Regardless, the Editor has roped me into
the job of testing, but not reviewing,
tuners for The Audio Critic. An explanation of my test procedures is thus
necessary.
An FM tuner is really just a radio.
By this I mean that a tuner should work
like a radio and receive as well as possible all signals present at its antenna
terminals. Reception problems and the
solution thereof are the largest challenge facing an FM tuner designer. Radio circuits in an FM tuner must work
over a very large dynamic range, up to
120 dB in the best designs, as for example a lower limit of 1 µV and an
upper limit of 1 V. In addition, FM tuner circuits have an astonishing amount
of voltage gain, up to 130140 dB in
some designs. My MR78, for exam-
ple, has a voltage gain in its RFIF circuits of about 137 dB! One µV at the
75 Ω antenna input produces 7 V at the
input to the detector bridge, at an impedance of 52Ω. This is gain, folks. I'm
not impressed by the bragging that goes
on about about the dynamic range in
the latest digital stuff. Tuner designers
play with bigger numbers! The challenge of making tuner circuits work
well over these big numbers involves a
lot of study, experiment, and sleepless
nights. The reward at the end, however,
is a super tuner and good FM reception.
I share a mountaintop with two
broadcast stations, an AM on 1430 kHz
600 feet away, and an FM on 92.1 MHz
only 138 feet away. Big brother on 92.1
measures 1 V at my tuner antenna terminals. Another station on 105.7, across
the valley, comes in at 0.25 V. Weak
signals exist on 91.3, 91.7, 105.3, and
106.3 MHz. I try reception of the afore-
mentioned weak signals in one of my
tests for a tuner's reception capability.*
Tuner RF Performance Tests
(1)
Outdoor towermounted high
gain antenna system. The tuner is con-
nected to this antenna, and its ability to
receive weak signals on 91.3, 91.7,
105.3, and 106.3 MHz is noted. The
tuner must cope with a 1 V signal on
92.1 and a 0.25 V signal on 105.7.
(2) Selectivity test. I attempt recep-
tion of a weak NPR station on 91.3 (75
miles away), adjacentchannel to a local NPR station on 91.5 MHz (4 miles
away). So far, I've found only three
tuners that can receive 91.3, and these
are the Onkyo T9090II, my MR78,
and the Accuphase T109.
(3) Indoor antenna reception test.
The tuner is connected to a three
quarter wavelength "gammamatched"
vertical antenna, and reception quality
on local stations is noted. Reception
should be free of spurious responses. In
some cases reception of fairly weak
distant signals is possible.
(4) FM generator strongsignal test.
The tuner is connected to an FM gener-
ator, and 1 kHz monophonic harmonic
distortion is observed as signal is increased from 1 µV to 1 V. Distortion
plus noise should drop from the initial
value of 35% (30.5 dB to 26 dB) at
1 µV to below 0.25% (52 dB) at 10
15 µV and remain low up to the 1 V
antenna input level.
(5) 2f1±f2 test. Two stations close
in frequency will produce a pair of spurious signals whose frequencies are
given by this formula. For example,
105.1 and 105.7 MHz combine to produce a "spurious" at 106.3. I look for
it. This is the classic "stations coming
in all over the dial" or "at wrong places
on the dial" syndrome. The 2f1±f2 test
is by far the toughest of the spurious
response tests. The front end in Figure
6 may do well on the other spurious
response tests and yet do very poorly
on this one because the extra RF stage
could cause intermodulation to get
worse.
Baseband IM Distortion Test
The tuner is connected to a Sound
Technology 1020A FM signal genera-
tor. A stereo signal of 1000 µV is applied to the left channel, modulated
100% at 10 kHz. Output from the right
channel is observed. A spurious output
tone of 1 kHz will appear, and in good
tuners this tone should be 60 dB or more
below 100% 1 kHz modulation (0 dB
reference level). I devised this tough
test during development of the MR78
25 years ago. In 1968, only one tuner,
the Marantz 10B, could pass this test.
The MR78 passes. Today, almost any
decent modern tuner passes also.
Residual-Junk Listening Test
This time I play music into the tuner from the Sound Technology 1020A. I
modulate the left channel only and listen to the right. A perfect tuner would
produce silence. Good tuners yield a
clean lowlevel output. Even a small
amount of distortion is audible because
you listen to the "residual" channel, already down 40 to 50 dB because of
separation. Here, 0.1 % (60 dB) distortion, as referred to the other channel's
output level, will be only 10 to 20 dB
below the residual and hence clearly
audible! Surprisingly, some tuners do
well on this test, implying that their
stereo decoders have very low distortion, and also that the tuner's entire circuitry has very good linearity.
This ends my regular test regime. I
also perform some additional distortion
and separation measurements, generally in order to verify the manufacturer's
specs and to check tuner alignment. I
do a touchup alignment if needed.
*David Rich notes:
Here in Audio Critic country, in Eastern
Pennsylvania, reception problems abound.
The principal classicalmusic station is 40 to
60 miles away (depending on the test site). Un-
fortunately, a 50,000watt rock station is alter-
nate channel to this station and about 5 to 15
miles away. This is the type of condition that
separates the real tuners from the pretenders.
Over at the commercialfree low end of
the dial, adjacentchannel stations are coming
from all around the area. Some come from lo-
cal 100watt stations or 10watt translators.
The local NPR and college stations are quite
clean but more often than not they play rock.
Only when they play what a government
supported station should, not Classic B Sides,
can we find out how well the tuners do Job
Two. The marvelous Mercer County Classical
Network comes in on a new translator. Before
that it was 70 miles away and one of my principal worstcase test signals for this survey.
58
THE AUDIO CRITIC
pdf 48
Figure 7: Block diagram
of the ubiquitous
Sanyo LA1235 chip
(with the iniquitous
quadrature detector).
Winston, 1974.
Panter, P. F. Modulation, Noise, and Spectral Anal-
ysis. McGraw-Hill, 1965.
Rich, D. "Cochannel FM Interference Cancellation
Using Adaptive Notch Filters." Polytechnic University,
Ph.D. Dissertation, January 1991. UMI Order Number
9117490.
Takahashi, S. and H. Numata. "Applications of
Walsh Functions to a FM Stereo Demodulator." Journal
of the Audio Engineering Society 33.9 (September 1985).
Taub, H. and D. Schilling. Principles of Communi-
cation Systems. McGraw-Hill, 1971.
Viterbi, A. J. Principles of Coherent Communica-
tions. McGraw-Hill, 1966.
Denon TU-680NAB
Denon Electronics, a division of Denon Corporation (USA),
222 New Road, Parsippany, NJ 07054. Voice: (201) 575-7810.
Fax: (201) 575-2532. TU-680NAB AM/FM stereo tuner with
remote control, $600.00. Tested sample on loan from manufacturer.
The RF section of this tuner is reasonably welldone with the equivalent of five gangs, one at the RF
input and two at the mixer section. The mixer is fully balanced—but to no avail, as we shall see below. Another
pair of tuned stages is in the local oscillator. One is used
to set the local oscillator's frequency; the other filters the
signal from the local oscillator as part of a buffer stage.
The IF strip starts with a single discrete stage and
uses two integrated amplifiers between the ceramic
filters. Two filters are used in the wide mode. Four filters
are used in the narrow mode, with an additional integrated IF amplifier also switched in when narrow mode is engaged.
The final stage of the IF strip and the FM demodu-
lator are in the ubiquitous Sanyo LA1235. This chip is
the most popular (it's real cheap!) IF detector. It uses a
quadrature detector design, as shown in Figure 7. Now,
Denon knows that a quadrature detector just won't cut it
in a high-end tuner. They used an advanced PLL detector
in the last generation of products that did not cost more.
Two tank circuits in the quadrature detector must be ad-
justed for the tuning to be on channel center and for mini-
mum distortion (try to get that one done by your local TV
repairman), and the unit came to us incorrectly adjusted.
Another adjustment involving a distortion measurement
must be performed on the tank circuit connected between
the mixer and the IF stage. The remaining adjustments
should be within the scope of a competent repair person.
The multiplex decoder is another IC, the Allegro
ULN3827. Allegro is a U.S. company that designed this
chip primarily for car audio, and most of its specifications
are not state-of-the-art for monolithic multiplexer chips.
It uses a 608 kHz mechanical resonator in the VCO and
has a 19 kHz pilot canceler that requires no adjustments.
Walsh functions are used to improve rejection of spurious signals like the SCA carrier. Surprisingly, Denon still
chooses to use an antibirdie filter. A dynamic blend circuit is included, which activates when the noise ratio becomes high. Most autoblend circuits (clearly a function
you need more in a car than at home, where you can flick
the blend switch) activate the blend on signal level, not
noise level, and can blend a weak but clean signal accidentally. The PLL loop bandwidth in this chip is made
wide for fast acquisition when no stereo signal is present
and then is made narrow to increase noise immunity
when the stereo signal is captured.
Another nice feature of the Denon is that the channel-separation adjustments are made individually for the
narrow and wide modes to insure the best possible performance in the narrow mode. A pair of passive filters
follows the multiplex decoder to remove the pilot tone
and other higher-frequency products. The outputs are
ISSUE NO. 23 • WINTER 1995-96
59
pdf 49
buffered with JRC NJM4558 devices.
The AM section is quite complex and performs
much better than most AM tuners. It has a wider audio
bandwidth and lower distortion than traditional AM tuners, as well as sophisticated noisereduction circuits. It
also has excellent sensitivity. (We can confirm that Rush
Limbaugh sounded better on the Denon, but we performed no controlled listening tests.) If you have an AM
station in your area that broadcasts something of interest,
this may be the tuner for you.
A single 12 V supply powers the analog stages.
Two l0kΩ resistors tied to the 12 V supply and a 100 µF
filter capacitor form the analog ground path. You cannot
get any cheaper than this. General construction quality is
that of a massmarket product.
The FM RF front end has a very low noise figure.
The 50 dB RF quieting occurred at 10.5 dBf. Most tuners
are 20 dB noisier at this signal level. RF intermodulation
performance was not so good, however. Strong stations
prevented reception of nearby weaker stations that could
be received by other tuners in this survey. This indicates
that the RF stage (and/or mixer stage—remember that
mixer stages are very hard to design) does not have adequate dynamic range, although the cause could not be
identified by me from the schematic. Poor dynamic range
can cause intermodulation of signals at the RF stage's input, since the stage becomes nonlinear.
Measured 1 kHz THD out of the box was 57dB
for a 30,000 µV signal at 91.1 MHz. This fell short of the
specified 60 dB. In addition, minimum distortion was
not achieved when the unit was tuned dead on. The Mod
afferi 10 kHz stereo IM test gave a result of 56 dB on
our test sample; on another sample tested by Modafferi
the result was 69 dB. Clearly our sample was misaligned as delivered, and that includes the quadrature detector. Channel separation from 50 Hz to 15 kHz was
>39 dB in wide mode and >33 dB in narrow mode. It
never met the specified 50 dB at 1 kHz, measuring 47 dB
in wide mode, but the fullband results are very good.
Frequency response in stereo just made the +0.5 dB, 1.0
dB strip given in the manufacturer's specification sheet.
That could be audible and should be tighter in a $600
tuner.
The Denon proved to be an average performer
when presented with good signals and did poorly under
difficult signal conditions. The AM performance is truly
exemplary, and the circuitry to achieve this must have
added to the cost of the unit. If AM performance is im-
portant to you, consider this unit. Others should pass it
by, as I believe the less expensive Harman Kardon
TU9600, JVC FX1100BK, Sony STS550ES, and Yamaha TX950 should at least match its FM performance
at much less cost. The slightly more expensive the Rotel
RT990BX and Onkyo T9090MKII will blow it away
on FM. (They have no AM, so again if that is important,
look into the Denon.)
Harman Kardon TU9600
Harman Kardon Incorporated, a Harman International Com-
pany, 80 Crossways Park West, Woodbury, NY 11797. Voice:
(516) 4963400. Fax: (516) 4964868. TU9600 "active tracking" AM/FM stereo tuner with remote control, $449.00. Tested
sample on loan from manufacturer.
We have had this not very new but still current
model inhouse for some time, but for some mysterious
"organizational" reason it never got to be measured by
Rich Modafferi. [Mea culpa.—Ed.] I don't want to wait,
however, to tell you what you can get for $449.
The front end and IF sections are pretty typical of
units in this price range, but the FM demodulator and
stereo decoder are anything but. The tuner has a stateof
theart PLL FM demodulator, but to make it work at this
price they could not design a highlinearity VCO. So
they got real smart (yes, it is patented) and realized that
the output of the VCO in a PLL must always track the incoming signal accurately, even if the transfer function
from voltage to frequency of the VCO is not very linear.
This is because of the phase detector used in the PLL.
What Harman Kardon does is to take the output of
the VCO and send it through a Sanyo LA 1235 with its
quadrature detector. No amplitudemodulation problem
occurs in the quadrature detector because the VCO output is a constant and it's big. This combination does not
allow the lowdistortion properties of the PLL FM detector to be exploited but it does allow its excellent performance under poor signal conditions (including large
amounts of AM) to be taken advantage of.
The loop bandwidth of the PLL is made narrow to
exploit the circuit's special signaldemodulation properties under poor signal conditions; as stated in the main article, this may cause some distortion. The VCO output is
thus an FMmodulated version of the desired signal
cleaned up by the PLL. The quadrature detector then demodulates the cleanedup signal.
OK, what more could you want for $449? Well,
how about the topoftheline Sanyo multiplex decoder
called the LA3450? It looks like a second source for the
Sony CXA1064S chip used in the topoftheline Sony
tuners. No other multiplex decoder has better specs than
the Sanyo LA3450; it some cases it is significantly better.
Yes, it has a pilottone canceler. Yes, the VCO runs at
456 kHz and requires no adjustment because it uses a
mechanical resonator. Yes, it has no antibirdie filter because of a birdie noisereduction system that appears to
be similar to the Sansui approach. The decoded signal
then goes, not to some cheap opamp, but to a discrete
amplifier designed in the Krell style. Other highend
touches can be found, although there are big limits at this
price.
Measurements of the TU9600, plus more design
details, will be published in the next issue.
60
THE AUDIO CRITIC
pdf 50
Mcintosh MR7084
Mcintosh Laboratory, Inc., 2 Chambers Street, Binghamton, NY
13903-2699. Voice: (607) 723-3512. Fax: (607) 724-0549.
MR7084 AM/FM stereo tuner, $1500.00. Tested sample on loan
from manufacturer.
This somewhat baffling tuner was sent to us much
later than the other units reviewed here and will therefore
have to wait until the next issue for full coverage; meanwhile the Editor discusses it briefly elsewhere in this issue as part of the feature article on Mcintosh Laboratory.
Onkyo T-9090n
Onkyo USA Corporation, 200 Williams Drive, Ramsey, NJ
07446. Voice: (201) 825-7950. Fax: (201) 825-8150. T-9090II
Quartz Synthesized FM Stereo Tuner, $790.00. Tested sample
on loan from manufacturer.
This tuner has been with us for 11 years, with the
Mark II revision appearing in 1988. When it first appeared it was the DX (distant reception) engine for the
masses, with Richard Modafferi's Mcintosh MR-78 being the only thing that could outperform it. Today the T9090II may be the ultimate super tuner for bad signal
conditions, as the MR-78 is history and Mcintosh is unwilling to do an update on it. (Engineering has done a
couple of designs but marketing has canceled full-scale
development. The marketing folks say they would sell
exactly six of them because the price would be in the
middle five figures.) The only challenger in current production is the Accuphase T-109. On the difficult realworld signal tests presented to the tuner by Richard Modafferi, the Accuphase has slightly better spuriousresponse characteristics, but its selectivity in not quite as
good. Since the Accuphase sells for almost four times the
price of the Onkyo, we can definitively say that the Onkyo is the cost-effective DX engine.
The Onkyo has a number of nifty operational features. They include a variable gain stage that can even be
operated by remote control. Two antenna inputs are
switchable on the front panel. I have one on my FM
cable; the other goes to the indoor antenna (cable FM
does not carry many local stations in my area). Because
crosstalk between the two antenna inputs is only 40 dB,
this may not work in all cases if you are trying to receive
a weak signal off the indoor antenna at a place on the dial
where the cable company has placed a strong signal.
Speaking of cable, the tuner shows center-tuned conditions with three LEDs. Since the tuning increment is 25
kHz, you can tune to the station's center even if it is not
being transmitted correctly. (Richard Modafferi points
out that he has seen frequency-synthesized tuners with 10
kHz tuning increments. That is the increment needed to
solve in every case the problem of an off-center station.)
Off-center stations often occur on FM cable, when stations are moved around the dial to a different place than
the over-the-air broadcast position. Provisions for connecting an oscilloscope to the tuner to check for multipath are also included. If you want to get depressed, try
using this to view the quality of your cable FM on a
scope. The antenna inputs are some nonstandard things
that will take push-on connectors but not the screw-on
type. This is a royal pain in the tochis.
The tuner automatically selects the stronger signal
of the two antenna inputs. Signal strength is displayed in
dBf. The measurements do not appear to be to extraordinarily accurate; although the signal-strength circuit is
quite sophisticated, as explained below, our sample expanded changes for high-level signals and compressed
changes for low-level signals. Even so, this is a very
good way to get a feel for the relative strength of the signals. The tuner will automatically decide the control parameter, such as IF bandwidth, RF attenuation, blend activation, etc. It appears to do a reasonably good job, but
you can override the tuner's choices and store your own
choices if you wish. One nice feature that is not present
on this tuner would be a recording calibration tone indicating the 50% modulation level. The record companies
have apparently put enough of a scare into the hardware
companies for the latter to drop any feature that would
make things easy for the home tapist.
The RF section is the equivalent of six gangs. One
is at the RF input and two are in the mixer section. The
mixer is fully balanced. Another pair of tuned stages is in
the local oscillator; one is used to set the oscillator's frequency, the other filters the oscillator as part of a buffer
stage. The RF section can be bypassed through a separate
tuning element that is coupled directly to the mixer stage.
Local oscillator reradiation would appear to be a problem
with this arrangement, but FCC regulations test for such
things, and this unit could not be offered for sale if it
failed the test. The ability to bypass the RF stage when
hot signals are present is nice because RF overload problems due to inadequate dynamic range cannot occur if the
stage is not in the circuit.
The IF stage for the wide and narrow modes starts
with a single discrete stage and uses an integrated NEC
µPC1163H amplifier between the ceramic filters. The
final IF amplifier and limiter stage is the Sanyo LA1235,
but the FM quadrature detector stage of the chip is used
only for muting circuits. Two filters are used in the wide
mode and four filters in the narrow mode. When the narrow mode is switched in, another µPCl 163H IF amplifier
is put in the signal path. The supernarrow mode switches
in a completely separate IF strip. One double-tuned LC
stage, five ceramic filters, two discrete amplifiers, four
µPC1163H integrated amplifiers, plus another LA1235
form the supernarrow IF strip. A diode switch selects
which IF strip's output will be sent to the FM detector.
Another µPCl 163H is used after this switch.
ISSUE NO. 23 • WINTER 1995-96
61
pdf 51
The signal-strength circuitry is quite complex. The
Sanyo LA1235 has a signal-strength indicator output that
is derived by measuring when its internal IF sections limit (it has six stages of amplification—see Figure 7). The
earlier the stages limit, the stronger the signal. Unfortunately, even a relatively weak signal may limit the first
stage because it has been through a significant amount of
amplification in the IF strip that precedes the LA 1235.
For that reason, signal-strength meters using the LA 1235
meter output are of limited usefulness. In the Onkyo, ad-
ditional circuitry is included to test if earlier stages in the
IF strip have gone into limiting. Additional places where
limiting is tested for are at the output of the first amplifier
in the IF strip, at an intermediate point in the supernarrow IF strip, and at the end of the supernarrow IF strip.
The meter circuit has four—count 'em, 4—trim pots.
Other tuners give you one. These are set for the correct
indication at 5 dBf, 45 dBf, 85 dBf, and 105 dBf. Levels
in between are less accurately displayed. A total of 50 active or passive components are used in signal-strength
meter circuit. The muting circuit is also quite complex to
prevent false mutes. This circuit monitors outputs from
the LA 1235 chips in both the standard and supernarrow
IF strips, as well from the PLL detector. A total of four
adjustment pots are used in the circuit to insure it works
reliably.
The FM detector is a PLL. This is the most nearly
optimum FM demodulator, and that's what Onkyo needs
to build a super tuner. The phase detector is the standard
double balanced mixer topology made from a four-diode
bridge and a pair of broadband transformers. Dual varactors and two MOSFETs form the VCO. The loop filter
uses an NJM4560 op-amp. The other section of the opamp also buffers the loop filter output and does the deemphasis equalization. This composite audio signal goes
directly to the output stage in mono! It never sees the
multiplex decoder. It is a sign that this tuner is intended
for DXing deep fringe stuff, where the chance of getting
anything usable in stereo is very unlikely. When stereo
decoding is used, the loop filter output goes through a
separate buffer circuit that includes the antibirdie filter.
Stereo decoding does not get as much attention as
the rest of the design. An NEC µPC1223C multiplex decoder is used. The chip requires an antibirdie filter, as we
saw above, and uses a low-end VCO design, namely an
RC oscillator at 76 kHz that must be adjusted. An openloop pilot-tone canceler is included, but no adjustments
are provided to insure optimal cancellation. The age of
the design really shows here, as much better multiplex
decoder chips are available now. Individual channelseparation adjustments are provided for all three IF
modes. Passive lowpass filters follow the decoder chip.
The output buffer is an NJM4560. Two electrolytic ca-
pacitors are in the composite audio signal path in mono
and four are used in stereo.
Construction quality of the unit is at the level of
Japanese mass-market equipment; given all the stuff in
the unit and the reasonable price, you would not expect
anything else. The power supply to the audio section is
±15 V. A total of 8 voltage regulators is used. Adjustment of this tuner should be doable by any competent
service technician because no distortion measurements
are needed. The PLL detector does not need such adjustments.
The first unit sent to us by Onkyo performed very
badly because it was either defective or totally misaligned. We sent it back. A second unit appeared to be
working correctly and yielded the following test results:
THD at 1 kHz in stereo was -62 dB in wide mode. That
is very good, but it misses Onkyo's spec by 12 dB. The
THD at 1 kHz was -36 dB in supernarrow mode. The
Modafferi 10 kHz stereo IM test came out at -60 dB in
wide mode, -64 dB in narrow mode, and -41 dB in supernarrow mode. Channel separation was >37 dB across
the band in the wide mode. At 1 kHz it was 50 dB. That
is 5 dB short of the specification given by Onkyo; the
more important across-the-band separation figure is, on
the other hand, 4 dB better than the specification. With
the blend function enabled, channel separation is reduced
to 18dB at 15 kHz. Frequency response in stereo fits into
a ±0.4 dB window, which is better than the spec.
If you don't want to spend more than $1000 on a
tuner and have difficult signal conditions, the Onkyo T9090II is more likely to get you a listenable signal than
any other tuner we have tested.
Rotel RHT-10
Rotel of America, P.O. Box 8, North Reading, MA 01864-0008.
Voice; (800) 370-3741. Fax: (508) 664-4109. RHT-10 FM ster-
eo tuner with remote control, $1499.90. Tested sample on loan
from manufacturer.
Rotel and Harman Kardon are the only companies I
know of that are using audiophile circuit-design attributes in the design of tuners. The Rotel RHT-10 takes
things further than Harman Kardon has attempted. As we
shall see, Rotel dropped the ball a couple of times, but I
know of no other tuner that attempts to do what Rotel has
done here. The RHT-10 is not designed to be a super tuner. Instead, the design is for signal conditions that are at
least good.
In contrast to a spaceship like the Onkyo tuner, the
Rotel has very few control buttons on the panel. Unfortunately, the presets are available only on the remote. Because of this I found this tuner to be a real pain in the
neck to use (where did I put the remote??!!). Audio output is very high. It is fixed and must have been set here
so the tuner would work with the Rotel RHC-10 passive
control unit. It is a pain because the thing is 6 dB higher
than CD standard level. You switch to tuner and then
dive for the volume control on the preamp.
62
THE AUDIO CRITIC
pdf 52
The RF section is the equivalent of six gangs, one
at the RF input and two in the mixer section. The mixer
is, surprisingly, not fully balanced. Another pair of tuned
stages is in the local oscillator. One is used to set the oscillator's frequency, the other filters the oscillator as part
of a buffer stage. As in the Onkyo, the RF section can be
bypassed through a separate tuning element. How local
oscillator reradiation is suppressed is unclear, but the
FCC would not allow the unit to be sold if this problem
existed. The IF strip starts with a single discrete stage
and uses two integrated amplifiers between the ceramic
filters. Two filters are used in the wide mode and four
filters in the narrow mode. No super narrow mode is included. As I said above, this is not designed to be a super
tuner. No instructions on how to adjust the tank circuit at
the output of the mixer are included in the service manual. This adjustment typically involves a distortion measurement. Adjustment of components in the RF and IF
stages can be done by any competent technician.
The last stage of the IF strip and the FM decoder is
the el cheapo Sanyo LA 1235. For $1500 I thought I was
going to get something other than a quadrature detector.
Two adjustments are required for the detector, one involving a distortion measurement. The presence of the
quadrature detector in this design is like a large zit on a
pretty face because everything else is done so well for
the reception of good-to-excellent incoming signals. The
output of the detector passes through an antibirdie filter,
and then the high-end fun begins.
At first blush things look normal, with a Sanyo
LA3433 multiplex decoder chip on the board. It uses a
456 kHz ceramic filter in the VCO for narrow PLL lock
range and low phase noise. But this chip is used only to
generate the 19 kHz pilot tone for the pilot-tone canceler
and the 38 kHz square-wave signal for the multiplex decoder. The rest of the circuits on the chip are not used at
all. Instead, high-quality op-amps and discrete circuits
are used wherever the audio signals actually travel.
First the 38 kHz output of the LA3433 is converted
to a sine wave by means of a tuned circuit. This circuit is
outside the PLL feedback loop, so adjustment of the tank
is critical if the phase relationship between the composite
FM signal and the 38 kHz sine wave are to be maintained. Two inductor adjustments do this. Pioneer puts
the tuned circuit in the PLL loop when they generate a 38
kHz sine wave for their proprietary decoder. That eliminates the need to trim, but a custom IC design is required
to do this. Rotel decided to do it with standard parts to
save the cost of designing a special IC. Luckily, a simple
level-peaking adjustment sets the inductors accurately,
but they can drift after adjustment.
An NE5532AN chip (which is a lot better than
what you will usually find in tuners' audio sections) is
the active circuitry used to filter the 38 kHz sine wave.
Another NE5532AN sums the incoming composite sig-
nal and the 19 kHz pilot tone together for pilot-tone can-
ISSUE NO. 23 • WINTER 1995-96
cellation. An inductor forming part of a bandpass filter
that filters the pilot-tone canceler signal is adjustable The
inductor sets the phase of the pilot tone, and a pot sets its
amplitude. These are simple adjustments that can be
made with a voltmeter, but again they can drift.
So now we have a composite signal with a pilot
tone canceled and a 38 kHz sine wave. Multiply them together using a pair of analog multipliers and you would
have birdie-free stereo. Unfortunately, low-distortion analog multiplier chips are expensive, so nobody has used
this approach until now. Rotel does not use the analog
multiplier chips, but they do take the plunge by making a
discrete version of an analog multiplier instead. First the
composite signal is converted into a pair of balanced currents. This is done with a clever two-transistor discrete
circuit that uses feedback. These currents (Iee) each enter
the tails of two differential pairs. The 38 kHz sine wave
signal (Vid) is applied to the bases of each differential
pair. Now, the differential output current of a bipolar differential pair can be written as
Iod = k1Ieetanh(k2Vid)
where k1 and k2 are constants. (I'll buy Bob Harley dinner
at CES if he can tell me what k1 and k2 are.) For small
values of Vid the hyperbolic tangent operation can be
dropped (remember from senior math in high school
something called the Taylor series), and now we have
Iod = k1k2IeeV
id
Just what we want—an analog multiplier. Rotel
uses an NE5532AN as a differential-to-single-ended converter at the multiplier core's output. The 5532 also acts
as the current-to-voltage converter for the analog multiplier. Rotel uses the fully balanced structure to take care
of even-order distortion terms in the multiplier core, like
the errors caused by the presence of the hyperbolic tangent function. (OK, I hear you. This is a lot harder to understand than "the midrange is woolly," but you've got to
admit it is more interesting [...get a life, Dave!—Ed.],
and it could actually affect the sound quality!) Channelseparation adjustment pots make sure the correct amount
of L+R gets added to the correct amount of L-R to reproduce the left channel (or subtracted, for the right channel). Separate pots are supplied for the wide and narrow
modes. The modes are switched by relays, not cheap
transistors! The blend circuit that follows uses a CMOS
switch, but the output muting is also done with a relay. It
all sounds so very good, but something must be wrong
under the surface because no antibirdie filter should be
required in this setup, but it is in the signal path.
To conclude the signal path, audiophile-designstyle AD847 op-amps are the output buffers in the Rotel.
Electrolytics are used for the dc blocks at the front and
rear of the buffer. The antibirdie filter circuitry adds three
more electrolytics into the composite FM signal path. A
second-order lowpass filter with a 15 kHz passband is
formed around the AD847. Normally a higher-order filter
would be used, and more high-frequency energy was ob-
63
pdf 53
served at the output of this tuner than usual.
High-end design practice continues in the power
supply. Separate secondaries of the large transformer go
to separate bridges for the analog and digital sections.
The analog side uses 4700 µF capacitors on the unregulated rails. A two-transistor open-loop regulator is driven
by a reference formed with a current source and a zener
diode to generate the regulated ±14 V rails for the stereo
decoder. The LA 1235 gets a separate two-transistor dis-
crete regulator that uses feedback. The other five regulators for the tuner are integrated units.
A very nice feature of this tuner is the inclusion of
a separate IF strip (two ceramic filters and a discrete
amplifier) plus an LA 1235 chip for the tuning meter.
This is done because the main IF filter strip is designed to
limit even on weak signals. Once this happens, the
strength of the signal cannot be determined. That's why
most signal-strength meters pin even when the tuner is
receiving a relatively week signal. With this specially designed low-gain IF strip made just for the tuning meter,
the relative signal strength of strong signals can be determined.
Construction and parts quality of the Rotel RHT-
10, both inside and out, are a step above the Japanese
mass market, but they are not quite up to the level of the
Mcintosh MR7084 at the same price. For example, although a double-sided board is used, the top side is only
a ground plane. The board has no through holes, and
jumpers need to be used instead. Despite the build quali-
ty, some of the internal adjustments were way off. Channel separation was running 20 dB as the unit was delivered. The pots were noisy and had to be cleaned before
they could be adjusted.
Although it is not designed as a super tuner—recall
it has no supernarrow mode in the IF and that el cheapo
quadrature detector causes the AM suppression to be no
better than 60 dB—it sure performed like one in many
ways. Real-world 2f1±f2 spurious signals were present
only under the most challenging test conditions that
Richard Modafferi could throw at it. Only the Accuphase
beat it here, and performance was similar to that of the
Onkyo T-9090II.
Selectivity was not as good as on tuners with a su-
pernarrow IF strip but it comes very close. The Rotel can
bring in things in narrow mode that require suppernarrow
mode on the Onkyo. That means the signal will be cleaner on the Rotel, since the narrow filter distorts the phase
less. But it must be noted that under worst-case condi-
tions (40 dB differences in signal level between adjacent
channels) the Onkyo could cleanly reproduce signals that
were barely listenable on the Rotel.
THD at 1 kHz in stereo was -60 dB in wide mode
and -44 dB in narrow mode. Better results in wide mode
would have been possible with something a little better
than the quadrature detector because the stereo decoder is
designed to have very low distortion, as explained above.
The Modafferi 10 kHz stereo IM test result was a superb
-75 dB in the wide mode (no surprise, given the sophistication of the MPX stage). The IM result was -44 dB in
the narrow mode. Channel separation after readjustment
was 41 dB across the band in wide mode and 30 dB in
narrow mode. Frequency response just made it inside the
±0.5 dB window given in the manufacturer's spe-
cification sheet.
At $1500 this is a good tuner. But at $750 it would
be a runaway bargain for those of you who do not have
worst-case signal conditions. But it is not $750—or is it?
Rotel has a tuner called the RT-990BX that sells for just
that price and is said to be almost identical to the RHT-
10. We would like to tell you for sure, but Rotel of
America appears to be no longer willing to lend equipment or even send a service manual to The Audio Critic.
It would seem that any criticism of the company's products, no matter how slight and how well documented, is
unacceptable to them. This once again proves that great
engineering and production people can work for companies that have marketing employees with small-minded
agendas and petty resentments. So if you have the cash
and decent signal conditions, try the Rotel RHT-10. If
you do not have money coming out of your ears but want
to take a chance, try the RT-990BX.
* * *
The author wishes to thank Angelo Mastrocola
and Richard Modafferi for reviewing the manuscript of
the above article.
AES Technical Paper by Rich and Aczel
On October 6, 1995, at the 99th Convention of the Audio Engineering Society in
New York, David Rich presented to the Analog Electronics session his paper,
coauthored by Peter Aczel, "Topological Analysis of Consumer Audio
Electronics: Another Approach to Show that Modern Audio Electronics Are
Acoustically Transparent." The preprint number of the paper is 4053. Copies are
obtainable from the Audio Engineering Society, 60 East 42nd Street, Room 2520,
New York, NY 10165-2520. Voice: (212) 661-8528. Fax: (212) 682-0477.
64
THE AUDIO CRITIC
pdf 54
Indoor Antennas and
Boosters for FM
By David A. Rich, Ph.D.
Contributing Technical Editor
A little bit of this material may overlap the main article on FM
tuners but is repeated here in context for the sake of clarity.
Old-timers who are into FM may recall Leonard
Feldman's test of 11 outdoor antennas in the January
1983 issue of Audio. The tests were run at the CBS Technology Center. A testing tower 75 feet above ground level and 400 feet away from the source antenna was erected. The relevant specifications of the antennas were then
measured. Needless to say, we do not have such a tower
set up here at The Audio Critic, nor do we have the RF
test equipment to conduct the tests even if we did have
the tower. So this report on indoor antennas is going to
be mostly a use test.
OK, let's start with the traditional disclaimer that if
you want the best performance from your FM receiver
you want a roof antenna. First, it is obvious that on the
roof you have the advantage of additional signal strength
through height. In addition, you have the advantage of
preventing the antenna from interacting with building
materials, stuff in the room, or people. Needless to say,
aluminum siding makes a great Faraday shield. But I
hear you whine that installing an outdoor antenna is expensive. Just think of it this way: a $300 antenna installation driving a $300 tuner is going to work a lot better
than a $600 tuner connected to a cheap indoor antenna.
The reason is that the fundamental limitations of an FM
receiver require that the signal at the antenna input be
free of noise and multipath if the tuner is to operate at
maximum performance. So why are reviewing all these
indoor antennas? In a single word, apartments.
Let's start by discussing what we should have
measured if we had that tall tower. The first important
test is the antenna gain. Gain is measured relative to an
omnidirectional antenna of proper size for the test signal's transmission frequency. Gain is measured at the
main lobe of the polar pattern of the antenna's reception.
To reject interfering adjacent or cochannel signals that
originate from other specially separated transmission
towers, or to reject echo signals (multipath), we want the
antenna to attenuate signals not at the main lobe. Two
specifications summarize the characteristics of the reception polar pattern. One is the beam width, which is the
angle around the main lobe for which the gain of the antenna stays above -3 dB (since we are working with signal power, -3 dB is the half-power point) relative to the
gain at the main lobe. We want the antenna to be as di-
ISSUE NO. 23 • WINTER 1995-96
rectional as possible, so we want the beam width to be as
small as possible to reject interference. Of course, if the
beam width is very narrow, it might become impossible
to set the antenna's main lobe in the direction of the station even with a rotor. The other specification that de-
scribes the polar pattern is the front-to-back gain. The
name says it all, since it's the ratio of the antenna's gain
at its main lobe to the gain 180 degrees from the main
lobe. This again represents a figure of merit as to how
well an undesired signal will be rejected. An interesting
phenomenon that results when an antenna has a nonuniform reception pattern is passive gain. Passive gain is
best explained by an example supplied by AudioPrism's
Victor Tiscareno. Imagine a light bulb in the middle of
the room. All areas of the room will receive light of
equal intensity. Now put a reflector in the back of the
bulb. Now the light from the front of the bulb is increased, but at the expense of the light in the back of the
bulb. Similarly, an antenna with a narrow beam width
will have gain enhancement at the center of the main
lobe relative to an omnidirectional antenna.
The next important specification is the antenna's
output impedance. For maximum power transfer, we
want the antenna to present a purely restive load equal to
the input impedance of the tuner (75 or 300 ohms). Uhoh, now I did it. The audio cable manufacturers have all
risen to their feet to tell us about the characteristic impedance of their cable. Please sit down, folks; this maximum power transfer stuff is for RF systems, and I do not
care at all about the transmission-line characteristics of
your audio cable.
Real antennas do not present a purely restive load
at all frequencies, and the impedance that the antennas do
present varies with frequency. When the antenna does
not present the ideal termination impedance, some of the
energy received by the antenna is reradiated out of the
antenna (the audio cable folks are rumbling again—
forget it, the wavelengths of the signals in the audio band
are so much larger than at RF that this effect is irrelevant
in the audio band). The measure of this phenomenon is
called the voltage standing-wave ratio (VSWR) because
it is the ratio of the maximum voltage of a signal in a
transmission line to the minimum signal level in the line.
If the line is perfectly terminated , the VSWR is unity.
65
pdf 55
The worse the termination, the higher the VSWR. The
worst-case value of the VSWR over the band of interest
(88 MHz to 108 MHz) thus presents in a single number
the accuracy of the impedance match of a given antenna.
So what is wrong with the two-dollar dipole that
came with your tuner? It is actually not as bad as it is
made out to be if it is aimed properly. One problem is the
that the dipole is bidirectional, so it does not reject a multipath signal coming from the rear. (The nonuniform po-
lar pattern does give the antenna 3 dB of gain.) But the
really big problem with the dipole is that, unless your
wall just happens to face in the exact direction to maximize the signal pickup of the station you want, the dipole
will not do a good job. Even if you luck out with one sta-
tion, if you want multiple stations you are almost certainly out of luck. That is were the Magnum Dynalab Silver
Ribbon comes in.
Essentially the Silver Ribbon is a rabbit-ear anten-
na. The difference is that the single rod on each side of
the standard rabbit-ear antenna is replaced by a U-shaped
element formed by parallel metal bands connected at the
end. The size of the bands can be adjusted. The instruction manual claims that maximum reception results for
the FM broadcast band with the elements fully extended,
but reducing their size will sometimes improve the impedance match of the antenna. The adjustment of the anten-
na is made while watching the signal strength meter. The
elements are adjusted for maximum signal strength. Unlike a dipole, this antenna can be easily rotated for optimal signal reception. In an area of high or moderate signal strength, it may be all the antenna you need. My
experience with it showed it performed well in bringing
in strong signals when properly aimed. Like most rabbitear designs, it is somewhat sensitive to the presence of
the person fiddling with it. I was unable to figure out how
to rotate it without being near it. On weaker stations a
properly aimed dipole could outperform the Silver Ribbon, probably because of the height advantage gained by
mounting the dipole close to the ceiling. In addition, the
rabbit-ear type of antenna is usually near metal objects,
which can affect its performance. Unfortunately, a properly aimed dipole is a mater of luck.
Moving up a step in price, and with significantly
improved cosmetics, we have the AudioPrism 6500. Inside the small 9.5 by 9.5 by 2.5-inch box is a loop antenna. The use of the loop antenna allows the size reduction
from the standard dipole, but this type of antenna has a
high Q. As Richard Modafferi points out, it is basically a
big LC tank circuit in a box. This means that the antenna
is sensitive to only a small range of signal frequencies,
and thus it must be tuned. A small knob on the front of
the box is used for the tuning. That is good news if you
have a selectivity problem with your tuner. Such a problem is most likely to occur when the station you are trying to tune in is weak and a strong, relatively close signal
is present. The 6500 even allows you to try and capture a
weak signal by placing a cheap Radio Shack amplifier after the antenna. Without a high Q antenna, the Radio
Shack unit is usually overdriven by the strongest signal
on the dial and makes things worse, not better. The same
problem occurs with the teeny active antennas that are often sold by dealers instead of the better passive antennas,
such as I am reviewing here. (Have you ever seen a good
review of one of these weensy boosted antennas? Apparently they sell because of the cute plastic case or the fact
they are plugged in—the unwashed masses think if it has
active electronics it must be better—or the very high
markup on this class of products.) Do not think you have
built an el cheapo substitute for the Magnum Dynalab
"Signal Sleuth" (reviewed in this issue) by combining
the 6500 and a Radio Shack amplifier this way. The
bandwidth of the 6500 is pretty broad, since the -3 dB
points are 1.2 MHz apart and the transition band is shal-
low because this is a single-tuned system.
The downside of the 6500 is that if you do not need
the selectivity you are stuck fiddling with this antenna every time you change a station. The VSWR of the anten-
na when tuned is clamed to be 2.1:1 or less.
An interesting feature of this antenna is that it is
omnidirectional when placed flat on a table or shelf, but
when it is set upright on its narrow side it becomes directional. When it is stood up on its side it is clamed that the
radiation polar pattern is cardioid in shape. Front to back
gain is clamed to be 6 dB. Since it is now directional, it
picks up about 3 dB of gain. If you are in an area with
strong, interference-free signals, you can leave the antenna flat and do not have to worry about moving it around
when changing stations (you still have to tune it—drat).
If you have multipath or a strong undesired signal, the
6500 may yield better reception than a dipole or rabbit
ears because it does have some front-to-back rejection.
Things get a little more unpredictable, though, because
the antenna receives the horizontally polarized component of the signal when it is flat but the vertical signal
component when it is on its side.
When the signal strength was high enough, the
6500 worked very well in my setup. It almost always
equaled or outperformed a dipole or rabbit ears. Best results occurred most often with the antenna on its side.
Again, a properly aimed dipole near the ceiling performed better on weaker signals, for the reasons cited
above. Adding the Radio Shack amplifier or even the
Signal Sleuth did not help. AudioPrism never claimed the
6500 was designed for a fringe-area signal; that is why
they make other antennas.
If you want to bring in weak stations, you need
something bigger with more gain, like the AudioPrism
7500. Did I say bigger? Well, this thing is seven and a
half feet tall! That is how big the enclosure has to be to
house a full-size half-wavelength monopole antenna on a
quarter-wavelength matching stub. The antenna is covered with a four-inch diameter tube. Its base is a 13-inch
66
THE AUDIO CRITIC
pdf 56
diameter wooden disk, which also contains a foil ground
plane. The ground plane attempts to isolate the coaxial
cable exiting from the antenna, so it will not interact with
the antenna's radiation pattern. For this to be completely
effective, the ground plane would have to be a half wavelength long, but that is clearly impractical.
Metal spikes attached to the base keep the unit
stable. The unit has a low WAF (Wife Acceptance Factor—perhaps the PC crowd would want to change that to
SAF for Spouse Acceptance Factor), which could get
even lower if the spiked feet scratch an expensive hardwood floor. AudioPrism will apply custom fabric to the
7500 if this is required to bring about spouse acceptance.
Please note that the 7500 also has the potential to increase the MAF (Mortgage Accumulation Factor), since
your spouse can bring up the removal of the 7500 and its
replacement by a roof antenna as one more reason to
leave the apartment and purchase a house.
Now, the first thing that you ask about this antenna
is how does it work, since it is a vertical pole, not the
horizontal rod that we associate with FM and TV antennas. The answer comes from the fact that electromagnetic
waves can be polarized, just like light waves. Originally,
FM stations broadcasted only a horizontally polarized
signal, hence the need for a horizontally oriented antenna. With the advent of car FM radios, things had to
change, since a car antenna is a vertical device. Radio
stations started to broadcast circularly polarized signals
in response to this need. The 7500 takes advantage of
this, and it is thus designed to receive the vertically polarized component of the transmitted signal.
The 7500 is claimed to have a suppressed vertical
radiation pattern. This is achieved by adding a quarterwave matching stub to the half-wave element. This apparently is called a J-pole in amateur radio circles. (Unlike modern audiophiles, the amateur radio crowd has
had real problems to deal with. The origins of this design
are said to date back to developments during World War
II, according to Victor Tiscareno.) The radiation patterns
of each of the elements differ, and using both elements in
this configuration sets the radiation pattern of the antenna to see more energy coming from the horizon. This
modification of the radiation pattern also allows the antenna to have a gain of 5 dB. The radiation pattern also
helps reduce the antenna's sensitivity to multipath
reflections in the vertical plane, such as those created by
the overflights of airplanes. Of course, since this an omnidirectional antenna, it is much more sensitive to multipath than a directional unit. It is also unable to reject
strong adjacent- or alternate-channel stations that are spatially separated from the desired station. The VSWR of
the antenna is clamed to have a worst-case figure of
1.9:1. This figure, which is as good as that of almost any
outdoor antenna, is still not as good as claimed for the
Day Sequerra FM Urban Antenna, which is very similar to but twice the price of the 7500. The VSWR of the
ISSUE NO. 23 • WINTER 1995-96
Urban Antenna is claimed to be 1.4:1. Apparently the
1.9:1 is a guard-banded figure, and Mr. Tiscareno informs us the antenna typically has a VSWR of 1.5:1. We
did not get to test the Urban Antenna, but we should note
that it must be wall-mounted, whereas the 7500 is freestanding, and the Urban Antenna does not have the
ground plane.
I expected good things from the 7500 but I did not
get them. I tried it out in a number of test locations, even
though lugging this seven-and-a-half-foot thing around is
not a lot of fun. (You keep bumping into things, or it gets
tipped over. Needless to say, the antenna had a few
bruise marks on it when we returned it.) In every test location the 7500 was outperformed by the 6500 or the
APPA-8500 (see below), and even the lowly dipole. My
guess is that the antenna's omnidirectional characteristic
was the problem. I can confirm this thing has gain, since
it could produce some wild DXs (I got stations broadcasting over 100 miles away—most likely atmospheric scatter), especially when the Signal Sleuth was in the signal
path. Unfortunately, there was a lot of fading on the
weak signals; I would not say any of these distant stations was clean enough for serious music listening.
Please understand that none of the above should reflect
badly on this antenna. Others in different areas of the
country have reported very satisfactory results with the
7500. (For example, see Fred Rosenberg's column in "Is-
sue Number Five" of the now defunct Sounds Like.... I
must add that Rosenberg's columns in Sounds Like...
were the only thing there worth reading.) Victor Tiscareno reports that the 7500 works best in flat areas that do
not have multipath problems. Kansas residents, please
take note. But what I needed was gain and directionality.
What I needed was an AudioPrism APPA-8500.
AudioPrism calls the APPA-8500 "the ultimate
FM antenna system...the first multi-element, directional
and full-sized indoor FM antenna on the market today..."
It's hard to argue with this; the antenna has three selecta-
ble lobes with 120° beam width, a front-to-back ratio of
10 dB, a gain of 8 dB, and a VSWR of 1.9:1. A good out-
door antenna would have better specs, with less than half
the beam width and a front-to-back ratio of better than 16
dB, but the APPA-8500 comes closer to replicating the
performance of an outdoor antenna than anything else
known to me. The problem is that it is inside the building
and not up on the roof, and that makes all the difference
in the world.
The APPA-8500 is not easy to hide: it is 63 inches
tall and 12 inches in diameter. Inside are three halfwavelength dipole antennas. Directionality is achieved
by using two of these antennas as reflectors for the third
active one. In my opinion, the 8500 has a higher WAF
than the 7500 (its proportions make it look a lot less
weird), but it retains the same spiked feet that could increase the DPF (Divorce Potential Factor) or LBF (Lease
Breakage Factor—remember, if you live in a house you
67
pdf 57
want an antenna on the roof) if they end up doing damage to a wood floor. Optional oak end caps are offered to
increase the WAF.
A wired remote control with a 20-foot cable allows
you to select one of the three antenna lobes, or you can
select all the antennas in an omnidirectional mode. The
box does this by determining which antennas act as
reflectors and which is the active element. In the omni
mode all three are active. The big advantage of this remote approach is that you are not near the antenna; thus
you are not affecting it by your proximity to it. The remote box also allows for selectable attenuation of the sig-
nal by as much as 18 dB. The box is line powered, but
this is only to drive the relays which select the antennas.
No electronic amplification is used.
All this high tech did great things for me. The an-
tenna's gain brought in the weak signals, and the directional design of the unit kept out interference sources.
The omni position always resulted in increased background noise. This showed why the 7500 did not work in
my location. I wrote out a check to the manufacturer to
purchase the 8500 the second day it was in my apart-
ment. After that it never left the apartment. I was afraid
that if the Editor got hold of it he would never give it
back. [You forget, Dave: I live in a house.—Ed.] That
said, this does not mean you should rush out and purchase one. If signal strength is not a problem, the less expensive directional antennas may give as good results as
this five-foot 25-pound monster.
The key thing I want you to come away with is that
all these indoor antennas are intelligent designs and good
values, but because of their diverse design goals the only
way to find out which will work best for you is to try
one. I hope the information above will help you narrow
down the choices. Never purchase an antenna without a
money-back guarantee. Your dealer should be able to
help you choose an antenna, but most dealers want to
have nothing to do with antennas because of the strong
possibility that a given antenna will be returned. Selling
overpriced audio cable is much easier and more
profitable.
Finally, I restate that none of these antennas will
ever come close to matching the performance of a good
outdoor antenna. If you can install such an antenna, by
all means do so. It will be the cheapest way to upgrade
AudioPrism, 1420 NW Gilman Boulevard, Suite 2593, Issaquah,
antenna, $299. APPA-8500 indoor antenna, $499 (w. remote).
All three AudioPrism antenna models are reviewed
in the article above.
Magnum Dynalab 205
'Signal Sleuth'
Magnum Dynalab Ltd., 8 Strathearn Avenue, Unit 9, Brampton,
Ont., Canada L6T4L9. Voice: (905) 791-5888. Fax: (905) 791-
5583. Magnum Dynalab Corp., 1237 East Main Street, Bldg.
#2, Rochester, NY 14609. Voice: (716) 654-6340. Fax: (716)
482-8859. 'Signal Sleuth' 205 FM antenna signal amplifier,
$279. ('Silver Ribbon' indoor antenna reviewed with related
products in article above.) Tested samples on loan from manu-
facturer.
The world of high-end audio is a strange place.
Dozens of companies produce nothing but overpriced interconnect cables. The audiophile can choose among
1000 different versions of these useless products. On the
other hand, if you need a tunable FM antenna amplifer, a
genuinely useful product, you have a choice of one.
Luckily it turns out to be a good one.
But why would you need one? Its primary use is to
amplify low-level signals so that the signal level at the
FM detector will be high enough to insure low-noise reception. So why can't you just buy a Radio Shack amplifier for $30.00? First, the Radio Shack unit is not tunable. Most weak stations are surrounded by stronger
signals. With an untuned amplifer you raise the signal
level of both the weak and the adjacent strong stations. If
the strong station does not cause the amplifier to overload, it will cause the tuner to overload. The Signal
Sleuth has a tunable 600 kHz bandpass filter. Thus even
alternate channels (400 kHz away from the carrier frequency of the desired station) are reduced in amplitude
by the Signal Sleuth. (This is not to be confused with an
adjacent channel, which is 200 kHz away from the carri-
er frequency. Note that a station occupies a spectrum
with a width of 200 kHz about its center frequency.
Thus, to attenuate an adjacent channel, significant
filtering must occur 100 kHz away from the center frequency of the desired station.) Second, the Radio Shack
unit has significant self-noise and is relatively nonlinear,
which can cause spurious response problems (see below).
The result is that the signal coming out of the unit may
be more corrupted than the signal coming in. Note that
small amplified antennas have exactly the same problem.
It is important to understand that the filtering function of the Signal Sleuth does more than just reject some
of the alternate channel. It also improves image rejection.
In a superheterodyne receiver, the incoming RF signal is
mixed with a signal from the local oscillator. By changing the local oscillator frequency for each station, the
output of the mixer becomes a signal which has the same
center frequency for all stations. A problem with this is
that the mixer can translate signals both above and below
the local oscillator frequency. It is thus possible for an
undesired signal to get translated to the IF frequency. For
68
THE AUDIO CRITIC
the quality of your FM signal.
WA 98027. Voice: (206) 641-7439. Fax: (206) 644-5485. High-Q
6500 indoor antenna, $125 (wood), $99 (vinyl). Model 7500 indoor
pdf 58
most FM receivers the local oscillator is set above the incoming RF signal; thus the image frequency is the desired RF signal's frequency plus two times the IF frequency. The IF frequency of most FM receivers is 10.7
MHz. Note that the image signal is never a broadcast FM
signal. Other image interference is possible because the
mixing signal often contains harmonics. These harmonics
also generate sum and difference products that can move
an undesired signal to the IF frequency. Filtering the RF
signal prevents these image signals from contaminating
the desired signal by attenuating the undesired signal before it gets to mixer.
A second problem, resulting in other spurious re-
sponses, is the nonlinearity of the RF amplifer or mixer.
This can cause intermodulation products, which can
again fall into the IF band. Under worst-case conditions,
with a heavily overloaded RF front end, a station may ap-
pear at a number of places on the dial. A common mech-
anism for a spurious signal is the second harmonic of the
local oscillator beating against the second harmonic of an
RF signal. The Signal Sleuth again helps by filtering out
undesired signals before they get to the input of the tuner.
In the case of a desired signal that is very strong, the
Sleuth provides attenuation to prevent overload of the
tuner. In the good old days when super tuners had super
RF stages, you would not have needed the Signal Sleuth.
For example, the Technics ST-9030 had an eight-gang
tuning capacitor and two RF amplifier stages. It had an
image and spurious-response rejection of 135 dB. Today
you get one RF amplifer and the equivalent of three or
five gangs. The result is that some of today's best tuners
have image and spurious-response ratios of only 80 dB.
Unfortunately, if you have two stations close together
that cause a spurious signal, then the Signal Sleuth cannot help. Such a spurious response is caused by the RF
stage or mixer stage in the front end. These circuits generate the intermodulation-distortion products that cause
the spurious response. Adding more stages of RF as the
Signal Sleuth does may actually make things worse.
The Signal Sleuth cabinet is the size of a small
preamp. The unit looks right at home in an equipment
cabinet with other high-end stuff, with its thick, black,
anodized metal faceplate and aluminum knobs. Two
switches are on the left of the front panel, one for power,
the other for bypassing the unit. Of the two knobs on the
right of the front panel, one controls the gain or loss of
the unit, the other tunes the center frequency of the bandpass filter. The gain control sets the gain from -30 dB to
+30 dB. It has a click position when rotated fully clock-
wise; this is used to tune the unit. With the gain control
in the click position the tuning knob is adjusted for maximum signal level as read on the tuner. Then the gain con-
trol is adjusted to a point where the received signal quali-
ty does not improve with increasing rotation. Putting
more gain in the path than is necessary can cause over-
load problems in the tuner's front end, especially if
strong signals are near the weak signal you are trying to
receive.
Inside the unit is a single-sided PC board, a small
transformer, and the front-panel-mounted controls. On
the PC board is a three-stage RF amplifer. A tuned LC
circuit is used in each RF stage. The stage is tuned using
a varactor diode in place of the capacitor. Varying the reverse bias voltage of a varactor diode causes its capacitance to change in a predicable fashion. The front-panel
control sets the voltage on the diode, and this is how the
unit is tuned. One would think that Magnum Dynalab
would provide an external dc input so that their tuners
could send the dc control voltage to the Signal Sleuth.
That way Magnum Dynalab tuners could control the Signal Sleuth and it would not have to be tuned separately.
Such an option does not exist on this model. The front-
panel switches and tuning knob are sealed units of good
quality. The gain pot is a much lower-quality unit. The
gain adjustment as well as the bypass function are also
accomplished by dc control. The RF signals never leave
the PC board.
Some tuners have circuitry which selects IF bandwidth and stereo blending functions automatically, based
on signal strength. Often a clean but weak signal is degraded by this process. An interesting thing you can do
with the Signal Sleuth is to gain a signal up to the point
where the tuner goes into wide-bandwidth mode and
turns the blend off. Of course, it does not make sense to
put a $250 RF amplifier in front of a $200 tuner so you
can defeat some of its automatic switching circuits. A
better approach would be to sell the tuner and purchase
one with manual override.
An important thing to understand about an RF amplifer is that it is really effective only if a weak signal is
not embedded in noise. If the signal is in the deep fringe
zone, or if you are using a small indoor antenna, this may
not be the case. If the signal is embedded in noise, all
you are going to do is to gain up the signal and the noise.
This will accomplish nothing. To see if the Signal Sleuth
will help you, it is necessary to try it. Your dealer should
be willing to loan you a unit. (See why they would rather
sell wire!) The Signal Sleuth probably performs at its
best when connected to a good high-gain roof antenna
with a rotor. This setup would give the highest signal-tonoise ratio at the input of the Signal Sleuth. Using a
small indoor antenna is probably not going to work so
well. I tested the Signal Sleuth with a setup midway between these two, using the AudioPrism 8500. In some
cases the Signal Sleuth did nothing to improve the signal,
but it was able to deliver one previously unlistenable
NPR station in decent mono and it improved another
NPR station from strictly mono to somewhat noisy stereo. It worked well enough for me to make me decide to
buy it. •
ISSUE NO. 23 • WINTER 1995-96
69
pdf 59
"LPs Making a Comeback" and
"Tubes Breathe with New Life" are
headlines we see with a fair degree of
regularity these days. Is it true? Do
these ancient technologies have for-
gotten performance advantages that
all the world's top designers and scientists have managed to overlook in
the race to digital and microelectronics?
Of course not. Why the hoopla
then? Why are we reading about revivals in the consumer and underground press? The answer involves
economics. Ever heard of the Product
Life Cycle? It is a forecasting tool
that predicts sales of a product (and
the underlying technology) over a
growth-and-decline cycle with several definable stages. Understanding
the PLC will help us understand the
market dynamics that enable revivals.
A new technology must poten-
tially offer the means to produce
goods an order of magnitude better
than existing products to capture
financing for startup. Existing products often have legions of loyal users,
so a technology with zero market
share must promise fundamentally
superior performance to warrant con-
sideration.
To complicate things, first generations of the new technology often
need considerable debugging. For example, early transistors were pretty
unreliable. Furthermore, the "old"
technology often suddenly gets much
better when a successor appears on
the horizon. Remember how, after
being "too expensive" for 15 years,
linear-tracking tone arms miraculously showed up on even close-'n' -play
turntables?
After the Introduction Phase,
sales and revenues skyrocket as the
newer technology continues developing and the market recognizes the innate performance/cost advantages.
People stand in line—they stood in
line to buy CDs and players in the
mid-80s during the Rapid Growth
phase of the life cycle. They are
standing in line to buy DSS now.
Eventually, growth slows as the
market Matures and Saturates. Products take on the characteristics of a
commodity as performance becomes
optimized. Promotion, style, features,
and price become the major competitive forces in a Mature market.
Sooner or later, someone finds
an even better way to provide the
function and, as a new contender enters the Introduction and Rapid
Growth stages, the now "old" technology slides into Market Decline,
although sales stagger along for a
while as prices plummet and inventories are "blown out."
So what happens after the Mar-
ket Decline? Because the production
facitities have long since been paid
for and are generally still in pretty
good working order, producers work
like hell to squeeze a few more dollars out of them. Products fall into
specialty market niches, where demand from a small number of people
who have strong attachments, often
sentimental, to the older technology
remains strong.
In this Residual Use phase, mar-
keting becomes paramount. Manu-
facturers find increasingly clever
ways to "resell" output to a small,
dwindling base of customers. "Revivals" and press announcements proclaimimg the "Second Coming" are
major promotional tools used to help
squeeze the remaining life out of an
obsolete technology.
Prices often increase dramatically, even though fixed costs of production are low because the plants were
fully amortized during the Growth
and Maturation phases. Variable
costs inevitably increase as smaller
and smaller lots of raw materials are
being bought and each promotional
dollar sells fewer units. However,
loyal followers display price-inelastic
behavior—they continue to buy even
when prices rise because their main
attachment to the products is senti-
ment- and not performance-driven.
The producers' main job becomes maintaining contact with the
old users and attractively promoting
the product. Pitches aimed at "exclusivity" and "connoisseurship" are ex-
tremely effective at this stage.
Sometimes seemingly unrelated
economic events help out producers.
For example, following the Cold War,
Soviet military cutbacks left many
overseas vacuum-tube producers with
excess capacity and a corresponding
need to find new customers. Viva!
Revival in North America!!
It is easy to forget that it was the
superior performance of the new
technology that sucked out demand
for the old technology in the first
place. Repeat: the new stuff had to be
radically superior to the old to enable
displacement of an entrenched market.
There are false starts too. Quad,
for example—a new technology that
was not well enough developed to
displace the existing products. Furthermore, most major audio advancements are tied to developments in
other markets. For example, the development of digital audio products
coattailed advancements in telecommunications and computers. Of
course, we were all dragged kicking
and screaming into multichannel audio (including stereo) by the movie
people.
Solid-state electronics and CDs
have already reached the Market Saturation stage, where they have basically become commodities. Power
amplifiers and CDs (excuse me while
I duck) are so competent that they are
basically differentiated by Price, Features, Style, and Promotion. Even the
goldenest of ears cannot tell an el
cheapo transistor piece of junk from
a pair of the best monoblock power
amplifiers ever built when the blind-
folds come out. Same with preamps
and CD players.
Other signs of Market Saturation
are obvious when you know what to
look for. Remember the Watt Races
and the Distortion Races where the
battles raged at specification levels
that had zero performance impact?
Today's Jitter Race and Bit Wars
(Super Bit Mapping, HDCD, outboard DACs, Legato Link, etc.) are
good examples of marketing strategy
based on style masquerading as per-
70
THE AUDIO CRITIC
pdf 60
Editor's Note: This is the third issue without David Rich's byline in this column but not because he
has lost interest in audio mendacities and techno-howlers. On the contrary, he remains the chief
instigator here, constantly finding new outrages and calling me up to sic me on the offenders. It's
just that the limited time he is able to spend on The Audio Critic must be allotted to circuit evalua-
tion and equipment reviewing. The steady broadening of the "Hip Boots" marshes and swamps now
makes me think that an overview of their "geography" is in order, with capsule summaries of the
principal sources of polluted information. Call it an abstract of our demonology.
As I have repeatedly discussed and explained before, there exists an unfortunate schism in today's audio
community. On one side, there are the scientific audio
practitioners, consisting of (1) members of professional
engineering societies such as the AES, the IEEE, the
ASA, etc.; (2) high-level engineering employees of large
electronics companies such as Sony, Philips, AT&T, etc.,
responsible for important product development; (3) academics in the E.E. and physics departments of universities; and (4) a few technologically disciplined audio journalists, whom you can probably count on your ten
fingers. Of course, (2) through (4) often overlap with (1).
On the other side, there are the antiscientific tweaks and
cultists, who may or may not earn their living in audio
but who never, never have graduate degrees in engineering or physics and who are never responsible for product
development anywhere above the trendy boutique level.
A lot of them write about audio. None of them give a
hoot about scientific proof. All of them influence the innocent.
Let us zero in on the principal habitats of this anti-
science faction.
The high-end dealers.
There are only a few hundred of these from coast
to coast, but they set the tone. A well-intentioned but
poorly informed music lover walks in with a bundle of
money, and with just a little bit of bad luck he/she will
end up with a 9-watt single-ended triode amplifier and
silver cable. I'm not saying there isn't a high-end dealer
in the country who will recommend sensible equipment
and good value, but there can be no doubt that the audio
salons are the main incubators of costly misinformation.
No, I'm not going blame the high-end manufactur-
formance.
In all of these cases products are
essentially differentiated with the
equivalent of chrome bumpers and
trim, while producers proclaim fantastic improvements in sound. Interestingly, some of the fiercest advocates are the very people who
mocked the big Japanese participants
during the Distortion Wars.
An audio-enthusiast performance fanatic (a.k.a. Weasel or
Geak) deploys his resources where
they return the biggest payback in
sound quality. After cable, intercon-
nect, and DAC differences evaporate
when the blinders come out, a Geak
looks to Better Recordings, New Formats, and Better (and more) Speakers
to deliver more music.
Notice I didn't say "more music
for the buck." Resources, time, as
well as money, are finite. Even if you
can afford it, wasting time auditioning cables and DACs suboptimizes
the throughput of your system. There
are more effective ways to improve
the sound of your system.
What goes around, comes
around. Be sure you keep an objec-
tive eye open when the salesman
with the Revivalist Hat appears at
your door. Not that buying old technology is necessarily a bad thing. On
the contrary, you may need a new
record player to retrieve certain program material that may never appear
on CD. Just don't kid yourself into
thinking that a new record deck or an
outboard DAC is somehow going to
magically make your system sound
better.
Is there a Tube Revival and LP
Revival? Sure, but only a dead technology needs Resurrection. 0
ISSUE NO. 23 • WINTER 1995-96
71
Hip Boots
Wading through the Mire of Misinformation in the Audio Press
pdf 61
ers, at least not much. They will make whatever sells and
stop making whatever doesn't sell. The high-end market
is driven by the beliefs and dreams of the high-end customer, who in turn is putty in the hands of a know-it-all
floor salesman. It's him that the manufacturer must impress and cultivate, not the buying public. The savviest
high-end manufacturers have always known that.
Stereophile magazine.
This is the tweako journal currently preferred by
the true believers. According to one highly intelligent
and ethical manufacturer I happened to discuss it with, it
is addressed mainly to the floor salesman, who always
needs highfalutin technical arguments for one-upmanship
and customer manipulation. I don't think there are
enough floor salesmen in audio to account for the magazine's entire circulation, but the characterization is on tar-
get. It's certainly not the engineering/academic community that supports Stereophile; I hear only snickers from
degreed professionals when the name comes up, if not
vehement contempt.
What's truly insidious about this publication is that
they use electronic measurements cosmetically, just to
create a visual aura of scientific objectivity, without the
slightest effort to link the readouts, graphs, charts, etc., to
their totally irresponsible subjective conclusions. They
are opposed to controlled (ABX) listening tests, which of
course do not serve their agenda, and come up with the
most outrageous pseudoscientific sophistries to reject the
overwhelming evidence of such tests as performed by
others. Tom Nousaine has a large collection of case histories on that subject.
Perhaps the most infuriating of their smoke-and-
mirrors bench tests are Robert Harley's pretentious and
wrongheaded jitter measurements. Bob Adams of Analog
Devices, Inc., one of the world's most highly respected
authorities on digital jitter, has repeatedly explained in
simple language (in our Issue No. 21 and elsewhere) why
you cannot measure jitter the way Harley does. Harley
and the Stereophile editors know this, yet he keeps doing
it the same old untutored way, and they keep publishing
it. Even the Catholic doctrine of the forgiveness of sins
excludes obstinate sinning; Harley's obstinate perversion
of digital theory is surely unforgivable in the secular here
and now.
Hi-Fi News & Record Review.
American culture has its oldest roots in England
and so does the tweako culture of Santa Fe's Stereophile.
HFN/RR is the leading antiscientific audio journal in the
U.K. and the former editorial demesne of John Atkinson.
That's where he developed the reviewing style and journalistic attitudes that originally attracted Stereophile publisher Larry Archibald's attention. He hasn't been with
his alma mater for almost ten years but it seems to have
made no difference; the intellectual climate there remains
the same because unaccountability is a habit that tends to
linger. Everything in HFN/RR must be read exactly as if
it appeared in Stereophile—with several grains of salt.
The Absolute Sound.
Harry Pearson, Editor and Publisher of this ultratweako journal, is the Charles Manson of audio. Not that
he instigates homicide, but consider all these similarities:
he is totally self-absorbed; he spouts muddle-headed philosophy at the drop of a hat; he has delusions of grandeur
(the "Pearson Publishing Empire," etc.); he has a small
but fanatical pack of disciples who will commit violence
(to science, logic, common sense, even human decency)
at his bidding; when provoked he rants and raves and
lashes out, completely out of control; faithful cohorts
suddenly turn against him—I could go on but I think the
point is made.
Yes, he was one of the mid wives of the High End
more than twenty years ago, but he never really understood the cause-and-effect relationship between technology and the sounds he was hearing; he believed, and still
does, that a good amplifier is made like a violin by some
kind of sensitive Stradivari-like artist/craftsman. A good
general-science teacher in 9th grade would have done
him a world of good. Today his magazine is a farrago of
tubes, vinyl, phono cartridges, and interconnects—so
deeply immersed in weirdness and unaccountability that
not even tweaks can take it seriously anymore. Every
once in a while he feels the need to run a "technical" article; since no practitioner with genuine credentials wants
to have anything to do with TAS, he has to get pseudotechies like the Canadian charlatan Gerard Rejskind (see
this column in Issue No. 19) to do their voodoo. New
"brilliant" technical writers have been announced—yeah,
right, and Ralph Nader is joining General Motors.
Audio magazine.
This is the original audio publication, the grand-
daddy of them all, founded as Audio Engineering in
1947. The "engineering" part, eventually dropped from
the name, was completely dominant for many years, al-
though the Journal of the Audio Engineering Society,
launched in 1953, then became the official publication of
the audio engineering profession. Sometime in the mid-
'80s, long after Gene Pitts had become the editor, Audio
changed its editorial course and began to give "equal
time" to science and fantasy. Alongside such thoroughly
solid engineer-reviewers as Don Keele and David Clark,
there appeared unaccountable golden-eared dilletantes
like Anthony Cordesman (blithely commuting from TAS)
and Jekyll-and-Hyde techie/tweaky split personalities
like Bascom King. Even the late Len Feldman, a rock
solid laboratory practitioner for decades, began to hear
things that didn't exist. Before you could believe a re-
view in this strange new something-for-all-tastes Audio,
you had to know your players.
72
THE AUDIO CRITIC
pdf 62
The irony is that Gene Pitts, who had presided over
this bastardization process from its inception—I don't
know exactly how willingly or reluctantly—has recently
been let go and replaced by Michael Riggs, former executive editor of Stereo Review, whose audio philosophy is
thoroughly objectivistic, just about the same as ours.
What will happen next? Will the tweaks be terminated or
just reined in? By the time this is in print, we might have
the answer.
Other audio and video publications.
Only the Journal of the Audio Engineering Society,
Stereo Review, and The Audio Critic have so far made an
unequivocal stand on double-blind listening comparisons
at matched levels. All other audio and video publications
are either fence-sitters or avoiders of the subject. (I'm not
counting independent columnists like Larry Klein and
Bob Pease.) For example, The Sensible Sound, a pleasant
little audio journal on the whole, lists highly credible
contributors like Tom Nousaine and Rich Modafferi (our
new RF expert) on the masthead, but also prints gushing
descriptions by lightweight reviewers of the nonexistent
sonic personalities of various amplifiers and preamps.
(Again, "equal time" for truth and nonsense.) Or take the
superglossy, somewhat trade-flavored Home Theater
Technology, whose number two executive is an avowed
admirer of The Audio Critic. That doesn't prevent them
from having Corey Greenberg as their technical editor,
the same gonzo reviewer who learned at Stereophile how
to talk the talk without having to walk the walk—how to
write about big differences in sound between, say, two
AV preamps, without double-blind listening tests, without measurements, without any kind of accountability.
Check it out, Beavis—this preamp is cool, the other one
sucks.
I could have picked other examples, in audio and
video, current or recently defunct, but they all have basically the same approach: plug it in, listen casually, write
whatever pops into your head, pretend you never heard
of more accurate methods of testing.
The New York Times.
The Western world's newspaper of record, right?
All the news that's fit to print, right? Well, when it comes
to audio, the Times has not taken the scientific high
ground. Far from it. Hans Fantel used to do their Sunday
audio column and was not without some undisciplined
dilettantish tendencies, but their new man, Lawrence B.
Johnson, is a major disaster. He is a Stereophile editor,
for crying out loud! He edits the Stereophile Guide to
Home Theater, and his Times pieces fit perfectly the
track gauge of the Atkinson, Archibald, and Santa Fe.
Analysis of nonexistent sonic differences between electronic components, self-indulgent subjectivism, inaccurate technical explanations, assorted audiophile myths,
the whole bit. A postscript by Larry Archibald is all
that's missing. (I have even heard facetious/paranoid suggestions to the effect that Johnson is a mole, controlled
by Archibald, cleverly infiltrating the U.S. daily press.)
It is difficult for me to believe that nobody at the almighty Times is aware of the schism I discussed at the
beginning of this column—the schism between science
and tweako cultism in audio—but that must be the case
because no serious newspaper, especially not one with a
superb science section like the Times, would knowingly
side with the tweako faction and choose, of all possible
choices, Lawrence B. Johnson.
Hey, hey, LB J, how many myths did you shill today?
Business Week.
At least one of the senior editors at this McGrawHill publication needs a new Rolodex. He is the editor
that correspondent Tim Smart reports to. The Rolodex
should have names like Stanley Lipshitz, Dick Greiner,
Floyd Toole, Mark Davis, David Clark, Bob Adams, etc.,
on it. Why? Because Tim Smart, the magazine's usual reporter on audio matters, obviously talks only to tweaks
and reads only tweako publications. He believes and tries
to disseminate the untutored agenda of his sources (see
Issues No. 16 and 22) and needs some serious factual
correction from headquarters. The authorities on the suggested Rolodex could impart to his boss the scientifically
correct information that nobody at McGraw-Hill appears
to know how to get.
Question: Is it possible that Business Week is just
as shaky on subjects I know less about?
Forbes magazine.
Ditto. They seem to be doing the same thing as
Business Week (maybe not as often) and need the same
Rolodex. It occurs to me that the two business magazines
may be calling each other for audio information. Yikes!
Letters to the Editor (continued from page 4)
MG-1.5/QR, whereas our small group
here didn't think the speaker was even
close to being recommendable to our
readers. Superb transparency? Not to my
ears or those of my associates. Indeed, I
spent more time on this speaker trying to
find a correlation between the sound and
the measurements than I usually do when
we didn't like the sound in the first
place—and you don't think I did enough!
Dr. Rich liked the sound of the speaker
even less than I did; he impatiently dis-
missed the few small virtues I had found
in it—and he goes to at least one live
concert a week, sometimes two.
Maybe if you went to hear the Min-
nesota Orchestra once a week...
—Ed.
ISSUE NO. 23 • WINTER 1995-96 73
pdf 63
Recorded Music
Editor's Note: David Ranada was so late with his reviews for this issue (late even for me, the grand
master of the late game) that I felt compelled, while waiting, to come up with my usual quota of
capsule reviews, in the usual combination of new and not-so-new releases. I keep repeating that this
is not really the way I want to run this column, that we need a degreed musicologist here (which
David is), that we also need credentialed nonclassical reviewers, but my protests keep turning into
"famous last words." Anyway, our readers appear to like my slight reviews more than I do. Hey...
Mostly Wagner and Mahler:
CD Releases of Importance
By David Ranada
Contributing Editor at Large
Malcolm Arnold: A Grand, Grand Over-
ture; Concerto for 2 Pianos (3 Hands);
Carnival of Animals; Symphony No. 2.
Royal Philharmonic Orchestra, Vernon
Handley, conductor. Conifer 75605
51240 2.
With instrumentation including
full orchestra, pipe organ, three vacuum cleaners, a floor polisher and four
rifles, you can't fail to be entertained
by Malcolm Arnold's A Grand,
Grand Overture, originally written
for the one of the famous Hoffnung
comedic concert music extravaganzas. Also Hoffnung-inspired is the
Carnival of Animals, containing a
movement (Chiroptera) inspired by
bats, in which the orchestra apparently plays like mad, but silently, for
about half a minute. (Of course, we'll
have to wait for an HDCD recording
of the work before we'll be able to
"hear" the immense amounts of ultrasonic energy such batlike activity
will generate!) The 1953 Symphony
is most immediately fascinating for
its occasional pre-echoes of textures,
instrumentation, and chord progres-
sions of music for the original Star
Trek TV series. Engineering is by the
ever-reliable Trigg vi Tryggvason,
and the recording locale is the All
Saints' Church, Petersham, Surrey,
which probably accounts for the presence of the magnificent pipe organ
sound at the parodistically extended
ending of the Overture. If only all recordings of Strauss's Also sprach
Zarathustra were so organ-ically
blessed.
* * *
Richard Wagner: Tristan und Isolde.
Berliner Philharmoniker, Daniel Barenboim, conductor. Teldec 4509-94568-2.
Richard Wagner: Die Meistersinger von
Nürnberg. Bavarian Staatsorchester,
Wolfgang Sawallisch, conductor. EMI
CDCD 55142.
Tristan und Isolde should be an
easy opera to record. There are only
a handful of main characters, no passages in which more than two of
them sing simultaneously, and only a
few, noncomplex passages for chorus
in the first act. Yet there have been
sonic problems with every single
commercial stereo recording of the
work, and the new Teldec is disap-
pointingly no exception.
At least the orchestral pickup is
magnificent. That you can seemingly
hear everything in the score is a tribute as much to the Teldec engineers
working in the Berlin Philharmonie
as it is to Barenboim's skill in balancing orchestral sections. The basic
sound of the orchestra is also magnificent, with a firm low end to the
brass and a supremely smooth and
blended string sound, probably the
best Tristan playing on disc.
This recording's problem is with
the voices, most of which are only
barely able to get through Wagner's
difficult writing without totally
breaking down. Even the crucial secondary roles such as Brangäne and
Kurnewal, which have been reliably
cast on previous recordings, are here
given to singers with basically unattractive instruments.
The engineering serves none of
them well. The soloists are miked in
such a way that there is very little
ISSUE NO. 23 • WINTER 1995-96
75
pdf 64
"leakage" of the voice into the orchestral pickup. While you can hear a
voice reverbing around the hall when
the orchestral textures are thin and
the dynamic scale is toward the quiet
side, as textures thicken and the volume rises, the singers sound more
and more as if they were in isolated
sound booths, even though they
weren't. As a result, the "size" of the
voice, which is gauged by its relationship to the orchestral sound, is
artificially reduced, especially if you
play the recording too softly. While
you never lose the words beneath the
orchestra, the final result sounds
faked, as if these voices needed electronic assistance to be heard at all.
Besides, in every live performance of
the work I've heard, there are passages when the orchestra does swallow
the voices. If it weren't for the singing, this production would probably
be the most recommendable of the
stereo Tristans.
Beside the inadequacies of the
vocal sound, the fadeout at the end of
the first disc and overlapping fadein
at the beginning of the second (the
start of Act I, Scene V) is only a minor irritation. A musically appropriate and dramatically less distracting
fade point occurs only a few mo-
ments earlier, on a solo timpani roll.
(Disc 2, containing Act I, Scene V
complete, is by far the most successful single passage in the perfor-
mance, mainly because the orchestral
textures are thin enough, most of the
time, for the voices to sound naturally arrayed against them.) The translation in the libretto contains what
seems to be virtually the standard
version by Lionel Salter (it's included with the Polygram-label recordings by Solti, Bernstein, Kleiber, and
Böhm). While accurate, it leaves out
much of Wagner's scene-setting
stage directions, turning the whole
enterprise even more into an oratorio.
EMI at least has an excuse for
the multiple engineering inadequacies of its new Meistersinger record-
ing: the work is Wagner's most com-
plex. The riot scene in Act II, for
example, has all the major characters
singing different simultaneous lines,
along with a subdivided chorus and
full orchestra playing multiple leitmotivs in the complex counterpoint
that distinguishes Meistersinger tex-
tures from all of Wagner's other music.
Those textures are not clarified
particularly well by EMI's recording,
which among other things often has
part of the first violin section leaking
into the right channel (obviously a violin microphone panned to the wrong
side in the mixer). While this fault is
most obvious on headphones while
following a score, other sonic problems are clear even over speakers.
For instance, at the end of Act II, the
Night Watchman disappears to the
left while the horn he is supposed to
be playing disappears to the right.
And I leave it for the listener to find
the weird two-second multiple-take
overlap that was allowed to pass.
These are only a few examples of the
sloppy engineering characterizing
this release.
The basic orchestral image is
that of a multimike production, with
certain orchestral sections becoming
unpleasantly mono-sounding at
times.
The bass drum during the entrances of the trade guilds in Act
III—Wagner's only use of the instrument in his mature operas—is anemic. It would normally produce a
pleasant surprise in an opera house.
On the whole, the engineering in this
production is a far cry from the stun-
ning clarity EMI achieved in their
Bavarian recording of Humperdinck's Hansel und Gretel (EMI
CDCB 54022), easily the best-
sounding recording of that Wagnerian work. It is even inferior, in basic
orchestral and voice pickup, though
not in distortion or noise, to the older
Kubelik performance (Myto MCD
925.69), a recording that reportedly
was to be released on Deutsche
Grammophon.
Die Meistersinger still needs a
clearly superior stereo recording. The
sonic inadequacies of this EMI production balance out its many vocal
felicities and conductor Sawallisch's
superb pacing, to add up to a rendition equal to but not better than Solti's (London 417 497-2). The upcoming second Solti Meistersinger,
to be recorded live in Chicago's Orchestra Hall, will probably not fill the
bill, to go by the Solti live-recorded
rendition of Verdi's Otello (London
433 669-2), a performance notable
for its lack of dramatic tension. I
hope to be proven wrong.
* * *
Gustav Mahler: Symphony No. 9. Berlin-
er Philharmoniker, Herbert von Karajan,
conductor. Deutsche Grammophon 439
024-2.
Gustav Mahler: Symphony No. 9. Colum-
bia Symphony Orchestra, Bruno Walter,
conductor. Sony Classical SM2K 64452.
Gustav Mahler: Symphony No. 9. Polish
National Radio Symphony Orchestra, Mi-
chael Halász, conductor. Naxos
8.550535/36.
Readers of my previous reviews
will know of my obsession with orchestral seating plans, specifically the
use of the standard turn-of-thecentury string arrangement plan—
first violins and cellos left, second violins and violas right—with music
written for it, which basically means
all orchestral music written between
around 1850 and 1925. Among all
the orchestral pieces of this period,
which make up the core of the symphonic repertory, in no other work is
the musical importance of this arrangement more significant than in
Mahler's 9th symphony.
Throughout his career Mahler
granted the second violins an unusual
amount of independence; they had
previously been an orchestral section
that had mostly subsidiary or backup
material to play. In Mahler not only
do the "seconds" perform their standard task of reinforcing the "firsts,"
but he often gives them prominent
autonomous melodic lines. Mahler
also bases many of his special effects, such as sudden transferals of
background textures or melodies
from side to side, on the seconds'
placement opposite the firsts. It is
one of the joys of those who can read
an orchestral score to be able to follow such devices, even with monaural recordings.
The 9th symphony is outstanding even among Mahler's works for
having second-violin lines of absolutely critical prominence. You need
only play the first couple of minutes
of the first two movements to realize
that the second violins are carrying
the main melodic burden, the firsts
being silent until quite a bit later each
time. Throughout the rest of the piece
you'll often find the two violin sections' melodic lines continuously in-
76
THE AUDIO CRITIC
pdf 65
tertwining, so much so that without a
left/right split of the two sections
you'll have a hard time deciding
which notes belong to which line.
Such activity continues until the final
pages of the score, where the melodic
lines disintegrate into single notes
traded across the width of the sonic
stage. The highest note of the final
chord belongs to the seconds, the
firsts are silent. Without following a
score you can hear these things only
if you are listening to a recording
having a separated left/right violin arrangement, the sonic differences between the two sections in an allviolins-left recording being usually
nonexistent.
Many conductors seem to have
recognized the importance of separated left/right violins for Mahler's 9th
symphony, if not his others. The 9th
is probably the Mahler symphony
that has been most often recorded
with left/right violins. The commercial stereo recordings by Abbado
(DG), Morris (IMP Classics), Kubelik (DG), and Klemperer (EMI) are
the ones I know about; there may be
others. A notable new one is that
conducted by Michael Halász on
Naxos, the only digitally recorded
budget-priced Mahler 9th. He uses
what might be called a "compromise" orchestral layout: cellos remain on the right, but the violins are
split and the violas are on the left.
While not absolutely authentic
historically, the arrangement works
extremely well here, aided by Naxos's clear, well-balanced and lifelike
sound, perhaps the best-sounding,
least overproduced Mahler 9th since
Inbal's on Denon—and at a budget
price, no less! The sound lacks only a
little firmness in the bottom two octaves (20 to 80 Hz) that would have
given the double basses and the rare
but crucial bass drum passages more
solidity.
The performance itself is excel-
lently paced, especially the last
movement, which has too frequently
been stretched to nearly a half hour's
duration to the detriment of the
movement's dramatic proportions
(it's 24:23 here). Unfortunately,
ISSUE NO. 23 • WINTER 1995-96
Halász makes very little of Mahler's
copious, even obsessive, markings
for short-term dynamics (loudness
variations within a single note) and
articulation (defining the attack and
release of each note). Without close,
even exaggerated, observance of
these markings, the music will sound
grand but bland, as it does with
Halász and does not under such noted Mahlerians as Tennstedt and
Bernstein. Still, the Naxos release is
the least expensive Mahler 9th with
left/right violins and as such is a lowrisk investment for any serious Mahlerphile.
Another Mahlerphile must-have
is Sony's latest remastering of Bruno Walter's classic stereo recording
of Mahler's 9th with the Columbia
Symphony Orchestra (here, a pickup
ensemble made up of Los Angeles
musicians). This is not so much for
the performance, which is only a little less bland than Halász's and as
such quite different from Walter's
gripping 1938 live performance with
the Vienna Philharmonic (on EMI
CDH 63029), the first recording of
the piece and one that Walter always
considered unsatisfactory. Nor is it
for the sound, which was superb for
its era (1961)—and much less harsh
here as well as slightly lower in noise
than in its previous Columbia CD incarnation (averaged spectra show a
rolled-off treble response in the newer release compared with the old
one)—but no match for any of the
more modern digital recordings,
much less Denon's or Naxos's.
The reason why every Mahler
lover should have this album is that
there are two "bonus" tracks on the
first disc. One is an interview with
Walter on his relationship with Mahler as pupil and musical protege
(Walter conducted the world premiere of Mahler's 9th in 1912, the
year after the composer's death). The
other bonus track is a fascinating
glimpse into Walter's rehearsal technique as he explores the 9th with musicians who had probably never
played the piece before. You'll hear
Walter memorably explaining some
of Mahler's orchestration tricks. The
program notes by Andreas Kluge are
unusually clearheaded and nonidolatrous on the subject of Walter's relationship to Mahler and his music.
Walter's 9th is part of a mostly
complete set of Sony remasterings of
his Columbia recordings, of which
the discs of Mahler's music are musically the most important. After the
9th, I'd rank in importance the recording of Das Lied von der Erde
(still one of the better sung versions)
and the 1st Symphony. Less critical
are the poorly recorded 2nd Symphony and the mono-era 4th and 5th
Symphonies. In the 5th the remasterers have let stand an irritatingly large
amount of disc noise from their original source materials, putting sonic
values above musical ones.
Karajan's live DG recording
with the Berlin Philharmonic of the
Mahler 9th has been considered one
of the great ones since its first release
(like Bernstein's with the same orchestra on the same label). It has also
been refurbished recently as part of
Deutsche Grammophon's Karajan
Gold series. The remix job has produced a closer view of the orchestra,
with far less hall reverberation than
before. As a result, the short-term dynamic inflections that Halász missed
emerge even more vividly under Karajan. There also seems to be more
bass firmness in the new mix, though
long-term spectral averaging shows
that the overall balance has not been
changed by equalization. Passages
containing bass-drum rolls send the
waveform into clipping more often
than in the first release, though the
clipping is inaudible as such. If you
have a copy of the earlier release, the
changes are not large enough to warrant buying the remastering. But the
performance, inauthentic seating plan
and all, remains magnificent—and
this from a reviewer who, in general,
hasn't liked Karajan's interpretive
approach. The Karajan Gold remasterings do raise some interesting
"ethical" issues, however. Can Karajan truly be said to be the conductor
of this performance when the resulting orchestral balances were not
heard nor approved by him? •
77
pdf 66
first-rate; the conductor is
quite famous locally and
obviously excellent. As for
the sound, the Great Hall
of the Moscow Conservatory has superior acoustics,
the dynamic range of the
recording is outstanding,
low-level detail is particularly fine, and an overall
naturalness is very much
in evidence. Gene Pope
claims an utter breakthrough
in audio quality; I'm inclined to wait for more
PopeMusic releases before
seconding that.
Reference Recordings
This is the only label
I'm aware of that has completely bought into HDCD,
the digital audio enhancement technology claiming
to correct for the "shortcomings" of ordinary linear
PCM without disclosing
how it's done. (See also
my review of the EAD
DSP-7000 Series III in this
issue. That's the HDCDequipped processor I used
to audition this recording.)
•
George W. Chadwick: Sym-
phonic Sketches (Jubilee,
Noël, Hobgoblin, A Vag-
rom Ballad); Melpomene
Overture; Tarn O'Shanter
(symphonic poem). Czech
State Philharmonic, Jose
Serebrier, conductor. RR64CD (1995).
George Whitefield Chadwick (1854-1931) is a seriously neglected but far
from negligible American
composer in the general
stylistic mold of Brahms
and Dvorak. These are
highly listenable pieces,
beautifully played and
conducted here. Good music, good orchestra, good
conductor—those are the
reasons why this CD is
worth having. As the first
HDCD release, ever, of
symphonic music with
strings, it is on the other
hand very disappointing.
The HDCD-decoded sound
is distant, hollow, poorly
balanced, blunted on top,
and strangely restricted in
dynamics—far from natural. One is tempted to use
the word processed. What
could be the weak link?
Keith Johnson, although
not as experienced in recording the full orchestra
as, say, John Eargle or
Jack Renner, has certainly
made some superb symphonic recordings (e.g.,
the Arnold overtures on
RR-48CD), so why not in
this instance? Maybe he
was totally unfamiliar with
the hall (Stadion Hall in
the Czech city of Brno)
and was forced to work
under pressure. But maybe
it's that hush-hush, tweako
HDCD process? Maybe it
can't sound natural? I'll
reserve judgment on
that—but not forever.
Telarc
Slowly Telarc is running out of standard repertory pieces to record in
great digital sound and
may have to turn to more
esoteric music to keep the
label going. Or will it be
Brahms and Tchaikovsky
all over again in 20-bit
versions?
•
W. A. Mozart: Le nozze di
Figaro. Alastair Miles,
Figaro; Nuccia Focile, Su-
sanna; Alessandro Corbelli, Count Almaviva; Carol
Vaness, Countess Almaviva; Susanne Mentzer,
Cherubino. Scottish Chamber Orchestra & Chorus,
Sir Charles Mackerras,
conductor. CD-80388 (3
CDs, 1994).
Mackerras seems to be
intent on repeating the
mid-1930s "miracle of
Glyndebourne" and record
the digital era's model version of each major Mozart
opera. This is No. 3, after
Die Zauberflöte and Così
fan tutte. He has the musi-
cal culture, the conductorial ear, and the orchestra to
go all the way; I don't
think he has the voices.
This Figaro is effervescent, authoritative, complete, and wonderful in 17
different ways, but the
singing in it is musical and
spirited rather than beautiful. I was spoiled in my
youth Figaro-Wist (Siepi
as Figaro, Steber as the
Countess, etc.) and am unable to be thrilled as I listen to this. Too bad, because the next more
excitingly sung version
will undoubtedly lack the
musicianship of this one.
I'll leave the details to the
more scholarly reviewers.
Jack Renner's 20-bit recording is all one could
ask for (same venue, same
equipment as in the Cosi
of 1993).
•
Gioacchino Rossini: "Overtures." Guillaume Tell, La
gazza ladra, L'italiana in
Algeri, Semiramide, La
scala di seta, Tancredi, II
barbiere di Siviglia. Atlanta Symphony Orchestra,
Yoel Levi, conductor. CD-
80334 (1992-93).
Recorded in two sessions almost a year and a
half apart, this collection
was then delayed more
than a year before being
released. I don't know
why Telarc wasn't more
excited about it because, in
terms of balancing performance against sound quality, this is arguably the
Rossini-overture CD of
choice for the discriminating audiophile. The playing is crisp, buoyant, accurate; the surprise sforzandi
are delightful; yet corniness is happily avoided at
all times. The Semiramide
overture receives one of
the best performances
known to me (shades of
Toscanini!); the others are
not quite as good but
close. The Atlanta orchestra sounds like one of the
Big Five here, especially
the woodwinds, and Michael Bishop's stunning
demo-quality recording
gives them an almost unfair advantage over their
compeers. Good stuff!
•
Arnold Schönberg: Ver-
klärte Nacht, Op. 4; Pelleas und Melisande, Op. 5.
Atlanta Symphony Orches-
tra, Yoel Levi, conductor.
CD-80372 (1993-94).
This is the postromantic Schönberg, not the later
dodecaphonist. That means
the music is enjoyable, not
just mandatorily admira-
ble. Wagnerians like me
eat it up, of course. The
main course here is not the
superb but smaller-scaled
Verklärte Nacht, of which
many better versions exist,
but Pelleas und Meli-
sande, which is scored for
a huge post-Wagnerian orchestra and gives the ex-
cellent Atlanta forces, not
to mention recording engi-
neer Michael Bishop, the
chance to produce some
wonderful sounds. Not a
masterpiece, not for everybody, but highly recom-
mended if you dig it.
•
Franz Schubert: Sonata in
A Minor, D. 959; Moments
Musicaux, Op. 94, D. 780.
John O'Conor, piano. CD80369(1993).
The sonata is one of
the posthumous three from
Schubert's miraculous last
summer (1828). It is sublime, but O'Conor's playing is too laid-back and
monochrome (albeit fluent
and polished) to project
adequately the sprawling
magnificence of the work.
The lighter, more casual
Moments fare better, and
Jack Renner's B&K 4006
recording is exemplary,
exactly as close-up as I
like it.
•
Richard Strauss: Salome's
Dance; Suite from Der Rosenkavalier; Burleske, Op.
11; Festival Prelude for
Organ & Orchestra, Op.
61. Cincinnati Symphony
Orchestra, Jesús LópezCobos, conductor; Jeffrey
Kahane, piano (in Burleske). CD-80371 (1994).
Another 20-bit sonic
blockbuster engineered by
Michael Bishop. Only the
Rosenkavalier music is Al
Richard Strauss here, but
everything the man ever
composed is hugely enjoy-
able as sheer sound, and
this CD indulges you with
that near-guilty pleasure as
few others. Want to show
off your $50,000 system to
a visitor? The lush strings,
big bass, and open soundstage of this recording will
do the job. The Cincinnati
forces play better than I recall them in any previous
effort, and López-Cobos
conducts them like a dedicated Straussian. Kahane
is suitably flashy in the
Burleske. I loved every
minute of the 65 on the disc.
Teldec
It's nice to observe
that the supposedly ultracommercial Warner empire is making recordings
that often rival any audio-
phile boutique label in
sound quality.
•
Aaron Copland: Old
American Songs (Orches-
tral Version); Down a
Country Lane; Eight Poems of Emily Dickinson;
Billy the Kid (Selection).
Dawn Upshaw, soprano;
Thomas Hampson, baritone. The Saint Paul
Chamber Orchestra, Hugh
Wolff, conductor. 9031-
77310-2 (1992-93).
Apple pie? This is as
American as pumpkin pie
at Thanksgiving! Two of
America's finest singers
perform one of America's
finest composer's settings
of some of the finest ailAmerican texts. The Emily
Dickinson poems are
definitely the centerpiece
here, but there's also my
favorite nonsense song, "I
Bought Me a Cat" ("My
cat says fiddle eye fee"),
and all sorts of other goodies. Great singing, good
playing by the half-sized
orchestra, very acceptable
sound (but not Teldec's
best).
•
Engelbert Humperdinck:
Hansel und Gretel. Jenni-
fer Larmore; Ruth Ziesak;
Hildegard Behrens; Bernd
Weikl; Hanna Schwarz;
Symphonieorchester des
Bayerischen Rundfunks,
Donald Runnicles, conductor. 4509-94549-2 (2 CDs,
1994).
The Bavarian radio orchestra must be able to
play this in their sleep be-
cause they recorded it only
four and a half years earli-
er for EMI, in the same
hall yet, with some overlaps in the cast and Jeffrey
Tate conducting. This new
version is a little lighter in
touch, less "operatic," better balancing the childlike
aspects of the work against
the Wagnerian harmonies
and orchestration. The recorded sound is also better, state-of-the-art I'd say;
more transparent, freer on
the dynamic peaks, firmer
in the bass than the already
excellent EMI. Of the versions I know—and I've
been listening to this since
the age of eight—this is as
good as the best. Of
course, to appreciate this
music to the fullest extent,
you must dig the whimsical quasi-Wagnerisms of
the score.
•
Richard Wagner: Tristan
und Isolde. Siegfried Jerusalem, Tristan; Waltraud
Meier, Isolde; Matti Sal-
minen, König Marke;
Falck Struckmann, Kurwenal; Marjana Lipovsek,
Brangäne; Berliner Philharmoniker, Daniel Baren-
boim, conductor. 4509-
94568-2 (4 CDs, 1994).
I just want to add a
few comments to David
Ranada's longer review
above. I agree with his
overall view of the recording, but the singers are a
little better than he avers.
Meier is a musically very
credible Isolde, if no Flagstad, and Jerusalem, likewise no Melchior, at least
doesn't shout here as is his
wont and sings intelligently. Barenboim's terrific
conducting took me by
surprise; he has advanced
several pegs in my rankings.
Troy
This label is a division
of Albany Records.
•
Jacques Ibert: "Jacques
Around the Clock" (Cham-
ber Music for Flute). Sue
Ann Kahn, flutist, et al.
Troy 145 (1991-93).
This collection is interesting for a number of reasons. These short, light,
1920s/' 30s/French-modern
pieces are unavailable
elsewhere. Sue Ann Kahn
is an excelent flutist who
has the idiom down pat.
Her collaborating colleagues are also fine musicians. The recording is by
Max Wilcox in the American Academy of Arts and
Letters, meaning it's lucid,
natural, lovely-sounding.
Most important of all, the
music is sheer fun. Try it.
80
THE AUDIO CRITIC
pdf 67
Loading...