Analog Devices AN557 Application Notes

AN-557

a

One Technology Way • P.O. Box 9106 • Norwood, MA 02062-9106 • 781/329-4700 • World Wide Web Site: http://www.analog.com

APPLICATION NOTE

An Experimenter’s Project for Incorporating the AD9850 Complete-DDS Device as a

Digital LO Function in an Amateur Radio Transceiver*

PIC “N” MIX DIGITAL INJECTION SYSTEM

By Peter Rhodes, BSc, G3XJP (email [e-mail])

PART 1 OF 5

This construction project brings together a number of
themes which I have been kicking around for some time.
But first, why PIC “N” MIX?

TWO ESSENTIAL TERMS

PIC—A range of microcontrollers produced by Arizona
Microchip Inc. In this application, the PIC16C84.

DDS—Direct Digital Synthesis. The technique of digi­tally generating the output frequency directly (as opposed
to typically mixing the output of a VFO with a crystal
oscillator—or employing phase-locked loop techniques).
In this application the Analog Devices AD9850 “com­plete DDS synthesizer” chip is used.

IN BRIEF . . .

PIC “N” MIX provides PIC controlled direct generation
of the required injection frequencies into the signal fre­quency mixer in your transceiver.

PIC “N” MIX also in the sense that you can pick and
choose which functional elements you build; and in
the sense that there are by design a number of differ­ent mechanical configurations to best suit your
circumstances.

You are also presented with the radical choice of using
the software I have designed—or writing your own.

The PIC microcontroller (and about 400 hours of soft­ware development) provides control and operational
flexibility while the DDS chip is used to synthesize the
RF output giving stability and low-phase noise.

*This five-part article is reprinted in its entirety by permission of RadCom

Magazine, a ham radio magazine publication in the U.K.(website
www.rsgb.com), and the author. All international copyrights are reserved.

CONVERGING THEMES

Discounting the value of your time, I would argue that
for years it has been viable to build multiband HF trans­ceivers which outperform their commercial counter­parts at any point on the price versus performance
graph—from the cheap and cheerful through to the truly
exotic. Except, that is, for one critical element—the
injection oscillator.

I have been building VFOs for years that for all practical
purposes didn’t drift. Almost all were based on the
Vackar running somewhere between 5 MHz–10 MHz.
Besides some time consuming temperature compensa­tion, I never gave them a second thought.

But they need about eight x’tals, a mixer and switched
bandpass filters before they can feed both the signal fre­quency mixer—and a frequency counter which gives a
natural display of exactly not quite the frequency you
are on! It can all be made to work, but only at substantial
cost in time, money and space. And the only incremen­tal feature easily obtained is IRT.

Then in February 1996, Technical Topics reported the
results of some phase noise measurements made by
Colin Horrabin, G3SBI and Jack Hardcastle, G3JIR on a
stable Vackar as “rather disappointing.” This set me
thinking. Most of us ignore oscillator phase noise
because we can’t measure it. Myself included. Does it
really matter in practice?

The ARRL handbook has an excellent section on the
subject which concludes “. . . far-out phase noise can
significantly reduce the dynamic range of a receiver. Far­out phase noise performance has effects just as critical as
blocking dynamic range and two-tone dynamic range
performance of receivers.” Yes, but does it really matter
in practice? I mean, am I truly going to fail to copy real
signals on a significant number of occasions because of
poor phase noise performance?

REV. 0

AN-557

I determined to find out by adopting the simple expedient
of fitting a changeover switch between my traditional
VFO and a phase-quiet alternative of the same power
output. Then, under a variety of practical conditions,
could I tell the difference? The problem, of course, was
to find this alternative without spending impracticable
sums of money.

Technical Topics came to the rescue again by first bring­ing to my notice the Analog Devices AD9850 DDS chip.
A few minutes on the Internet produced the data sheet—
and it all looked too good to be true.

So, I set about designing some traditional TTL to control
it and actually got as far as building some of the boards
before giving up. Because although I have no doubt
it would have worked, 28 TTL chips to control one
DDS chip—and provide a modest range of useful
features—was ignoring any reasonable definition of the
“in practice” imperative.

It was obvious from the outset that some form of micro­controller would provide the solution to the control
problem and at the same time offering the ability to
provide a range of operational features. What put me off
for months was the costs of acquiring the development
environment and the hardware to program the chip. A
glance in the larger catalogues suggested little change
from a £200 investment for PIC development—totally
unacceptable.

The bottom line is this. Arizona Microchip provide on
their website their complete development environment
at no cost—as well as copious application material. And
there are numerous circuits for PIC Programmers pub­lished on the Internet which you can build for less than
£5. The project was born.

. . . AND THE CONCLUSION?

Phase noise does matter in practice. On a substantial
number of occasions it makes the difference between R2
and R5 on SSB signals.

For example, the home-brew net convenes daily around
lunch time on 80 m just down from the SSTV calling
frequency and just up from a prominent French coastal
station. A convenient source of large adjacent channel
signals.

If the band is flat and quiet, it makes no difference. If
conditions are lively—using the DDS source—then I can
often copy Ed, EI9GQ at only just R5. Switch over to the
VFO and the readability instantly degrades to near hope­less if and only if there is significant adjacent channel
activity. The effect is insidious. Its not that Ed’s signal
goes down. Its that the base level of band background
noise appears to go up. It doesn’t of course.

What is happening is that the noise sidebands on my
VFO are mixing with adjacent signals to produce
incremental noise in the passband. A very salutary
experience because this noise is totally indistinguish­able from band noise and you could operate for years
without realizing what was happening.

It would seem that there is a basic conflict in VFO
design. The traditional view is that you drive the oscilla­tor gently to keep the heat (and, therefore, drift) down
and follow it with an appropriate buffer to get the power
up to the required level. This approach also maximizes
phase noise.

Conversely, if you drive it hard then it becomes increas­ingly difficult (in my experience, next to impossible) to
maintain acceptable frequency stability.

With the DDS approach, phase noise and drift are
intrinsically small. The topic is covered shortly.

PIC “N” MIX SUMMARY

Before covering the essential theory these are the
features on offer should you adopt my software:

GENERAL SUMMARY

• PIC “N” MIX replaces the functions of the crystal
oscillator bank, VFO, mixer, bandpass filters, power
driver and frequency counter associated with a con­ventional HF transceiver with significantly enhanced
features and lower cost. Not merely a VFO!

• Alternatively, it acts as a programmable and/or tunable
signal source with output from audio to 40 MHz in
10 Hz steps.

• All functions are controlled by either a multifunction
tuning knob—or by a simple telephone keypad with
65 discrete key combinations recognized by the
software.

• A large six-digit seven-segment display with auto­ranging gives a resolution of 10 Hz.

• Two independent VFOs provide IRT, ITT and cross­band operation.

• A variety of tuning and scanning modes provides
operational flexibility.

• Any desired frequency may be entered directly from
the keypad.

• The switch-on frequency and nine band initialization
frequencies are user programmable.

• As are 10 frequency memories.

• Any three IF offsets (USB, LSB and CW separately) in
the HF range may be entered.

• USB/LSB/CW selection outputs—and band switching
outputs to the host transceiver are provided as a
hardware option.

–2–

REV. 0

AN-557

• Front panel LEDs provide status information and
double as a bar-graph to show tuning rate.

• Finally, there are a number of possible physical
layouts providing flexible outboard or integrated
configurations.

ADMINISTRATIVE FEATURES

• The frequency accuracy is determined by a reference
oscillator in the VHF range. You may use any crystal
in the range 100 MHz–125 MHz and program the
actual frequency into the software yourself.

• Final calibration and any subsequent correction for
crystal ageing are achieved using the tuning knob to
drive a trimmer in software. A physical trimmer
which would inevitably introduce drift and phase
noise is neither required nor provided.

• IF offsets may be entered from the keypad and/or
trimmed to zero beat with the host transceiver carrier
crystals.

• As an injection oscillator, the output frequency is the
selected IF frequency plus or minus the desired fre­quency. The choice of high-side or low-side injection
may be made “on the fly” with the sideband selection
outputs to the host being switched to correspond.

OPERATIONAL FEATURES

• Intelligent tuning continuously monitors the speed
and duration of tuning knob rotation to vary the tuning
rate dynamically. Thus the longer and faster you turn
the knob, the greater the tuning increments.

• A software flywheel engages automatically at high
tuning speeds for rapid and/or large frequency
excursions—and is disengaged by the slightest turn
of the knob in the opposite direction.

• As opposed to traditional tuning where rotation of
the knob alters frequency, a tuning rate option is pro­vided whereby rotation of the knob alters the rate of
frequency change—from zero to very fast.

This is particularly useful for casually scanning around a
band without having to continuously turn the knob.

• Guard channel operation provides normal tuning, but
with a brief switch to another chosen spot frequency
about every 20 seconds.

• Up to ten memories may be programmed with
frequency. As opposed to merely providing spot
frequencies, they are also jumping off points for
further tuning.

• Memory scanning mode cycles between the ten
memory frequencies at a speed determined by the
tuning knob.

• Spot scanning switches between two chosen spot
frequencies at a speed determined by the tuning
knob.

• Range scanning tunes up and down between two
chosen limits with frequency increments determined
by the tuning knob.

AD9850 DDS

Throughout this article, I have used the nomenclature
used by Analog Devices in their data sheet and only
mentioned the features and configuration of the chip
used in this project. There are others.

There is little you need to know about the internal workings
of this device. The most significant consideration is that
it contains the DAC—necessary to convert the digitally
generated sine wave to analogue form—on the chip. So
you neither have to worry about specifying a suitable
DAC nor interfacing it.

REFERENCE

CLOCK

CLKIN

AD9850

32-BIT TIMING WORD

(GENERATED BY PIC)

F

OUT

Figure 1. DDS Block Diagram

The basic block diagram is shown in Figure 1. There is a
simple relationship between the output frequency
FOUT, the reference clock frequency CLKIN, and the 32­bit tuning word ∆Phase:

FOUT

= (∆

Phase

×

CLKIN

)/2

32

Using a 125 MHz clock, the highest frequency permitted,
this gives us tuning increments of 0.0291 Hz, orders of
magnitude better than needed for this application. In
practice this means that using 10 Hz tuning increments
an error of 0.0291 Hz is significantly smaller than, for
example, any drift on your carrier x’tal.

Stability in a DDS system is the same (in parts per
million) as that of the reference clock x’tal oscillator. For
example, if the 125 MHz clock drifts by 10 Hz then on
80 m with 12.5 MHz injection, you will drift by 1 Hz.
Phase noise on the DDS output is better than that of the
reference clock—which contributes most of the system
phase noise. The improvement is

20 log (

CLKIN/FOUT) dB

Is it that simple? Unfortunately, not quite, for as well as
generating the required frequency, aliased or image
outputs are also present. This is inherent in any sampled
signal and the output observes Nyquist’s theorem. The
aliased images are at multiples of the reference clock,
CLKIN ± the output frequency FOUT. Thus with a clock
frequency of 125 MHz and the wanted output at 20 MHz,
the images will be at 105 MHz (first image), 145 MHz

REV. 0

–3–

AN-557

AD9850 14-PIN DIL

(second image), 230 MHz (third image), 270 MHz (fourth
image) . . . and so on.

Another consequence of Nyquist’s theorem is that the
maximum theoretical output frequency is half the refer­ence clock frequency—but in practice, one third is usually
taken as a rule-of-thumb limit—to provide a reasonable
separation between the wanted signal and significant
images.

The amplitude of the images follows a sine envelope as
shown in Figure 2. A low-pass filter is therefore inserted
in the output to reduce the image outputs; and on the
highest bands using a high IF, the Tx/Rx signal frequency
tuned circuits offer further protection. Using the highest
possible reference clock frequency obviously helps.

SIN(X)/X ENVELOPE
WHERE X = (␲) F

AMPLITUDE

FIRST

20 105 145 230 270 355

IMAGE

REFERENCE CLOCK CLKIN FREQUENCY – 125MHz

SECOND
IMAGE

THIRD

IMAGE

Figure 2. DDS Output Spectrum

There are other discrete AM spurious outputs as a result
of limitations in DAC technology. The significant ones
are few in number and appear from the user’s perspective
to be at random frequencies. Analog Devices specify
them as better than 50 dB down and the practical conse­quence of these is an occasional birdie.

The remaining AM spurs form a continuous noise floor
at about 70 dB down and these give rise to the greatest
concern. A typical double balanced mixer will furnish
about 40 dB further suppression—so if the mixer is
injected at +7 dBm, weak birdies will be heard if the
band noise is less than 2 mV at the mixer RF port. On the
LF bands with most receivers this will be academic but
on, say, 10 m a typical Rx will need to use an RF preamp
with some 25 dB net gain to both retain adequate sensi­tivity and to mask the noise floor. This topic will be
much less of an issue when 12-bit DDS is available at
affordable prices but meanwhile this 10-bit DDS may not
be suitable for all home-brew Rx topologies, particularly
if you are reluctant to alter your gain distribution.

The final challenge with the AD9850 is its size, see
Figure 3. Designed for surface mounting, it is truly
microscopic. Much effort has gone into finding repeatable
amateur methods of mounting it which do not com­promise performance. Analog Devices recommend a
4-layer board with dedicated power and ground planes.

/CLKIN

OUT

FOURTH
IMAGE

FIFTH

IMAGE

Figure 3. AD9850’s 28-Lead Shrink Small Outline

Package as Compared to a 14-Lead DIL Package

I tried it on double-sided board, both surface mounted
and let into a slot so that it sat in the thickness of the
PCB. I had no great problems hand-etching the boards—
but found substantial difficulty in soldering the chip to the
pads. The best I managed was with a medium-sized iron
and a length of sharpened copper wire bound to the bit—
and very fine solder. The propensity to bridge adjacent
leads was enormous. Worst, it seemed impossible to
maintain clean power and ground plane layouts— which
ultimately prejudices the phase-noise performance.

After obtaining a batch of 50 unmarked devices in the
same packaging at a rally and having destroyed many in
the quest, I settled on a dead-bug approach with con­tinuous power and ground planes—mounted as a sub­assembly on a DIL socket and with the input/output
leads taken out to the DIL socket on fine wires.

This method is reproducible if you have average eye­sight (or a good magnifier) and a short-term steady
hand. The process is described in detail in Part 2 of
this article.

THE WORLD OF PICS

THE 16C84 IS ONE of a large and growing range of 8-bit
microcontrollers. The devices vary according to speed,
the amount of memory, built in devices (including A-D
converters) and other features. For the latest detail, con­sult the Arizona Microchip website.

The 16C84 specifically is—in brief—an electrically
reprogrammable device with 1 k of program memory
(i.e., room for 1024 instructions), 36 bytes of working
data and 64 bytes of data EEPROM which survives
power down; and 13 input/output pins.

Also on the website you will find the integrated develop­ment environment MPLAB which was used exclusively
in developing my software. It includes an editor, assem­bler and simulator. The latter is particularly useful since
you can progressively build and test code with your target
chip simulated on the PC—no real hardware needed.

If you want to download MPLAB, watch your phone bill
because it is about 5 MB when unzipped!

You can run elements of the software under DOS, but
I used it exclusively under Windows. At first under
Windows 3.1 on a 386 and latterly under Windows ’95
on a 486. Both were entirely satisfactory. C++ compilers
are also available, but I haven’t tried any of them, all my
work being in assembler.

–4–

REV. 0

AN-557

Of the various programmers available, I built TOPIC by
David Tait
can also build ones for serial port operation and some
even need no power supply, deriving their power from
the port.

Having conducted the intellectual exercise of “designing”
some aspect of the software, the mechanics are easy
enough. After typing in the code using the editor, you
assemble it and then run it on the simulator—if necessary
one instruction at a time—looking at intermediate and
end results to see if it works. You can also check execu­tion times. When you are happy, you then download the
software onto the PIC using the programmer (say, 10
seconds) and run your code in the real world. If you are
careful, the PIC can be programmed in situ in the target
environment which speeds up the process enormously.

The assembler language itself is easy to learn with only
35 instructions. The art, it turns out, is usually not
whether you can write something that works but rather,
can you find an efficient enough way of doing it to
squeeze it into the space without unduly compromising
features, performance and ultimately maintainability?
As Eric Morecombe once said “Composing good music is
the same as composing bad music. Its just a matter of put­ting the notes in a different order.” So it is with software!

So, if you have never written any software before and
have a PC with at least temporary access to the Internet,
you can have a go with no incremental cost. (Or you
could buy a suitable secondhand PC for about £50—and
most Internet service providers offer a free trial period.)

Think of the range of applications—self-tuning ATUs,
intelligent AGC generators, keyers and readers; in fact
any application involving control or logic is a potential
candidate where one 18-pin DIL coupled with your intel­lect can replace acres of conventional hardwired logic at
trivial cost. Who says computers and amateur radio
don’t mix? In my view these microcontrollers are going
to dominate many aspects of home-brew construction
before long.

THE INPUT/OUTPUT CHALLENGE

AS JUST MENTIONED the 16C84 has 13 input/output (I/O)
pins for controlling its environment. How many are
actually needed? The following is the first-pass answer:

Inputs—total of 15 as follows:

PTT Line Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Keypad 4 × 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Shaft Encoder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

Outputs—Total of 74 as follows:

6 Digits × 7 Segments + Decimal . . . . . . . . . . . . . . . 48

Status LEDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Band Switch Outputs . . . . . . . . . . . . . . . . . . . . . . . . . 15

AD9850 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Giving a grand total (apparently) of 89.

6

which runs out of the PC parallel port. You

REV. 0

Clearly something has to give and some supplemental
hardware is needed. There is, however, one mitigating
feature. The 13 I/O pins on the PIC can be used as either
inputs or outputs—and you can change them “on the
fly” in midprogram so with cunning they can be both!

Firstly, the 12 keypad switches aren’t individually moni­tored. Each row is tested in turn looking at each column
in turn for key presses. This needs only seven I/O lines.

Next, rather than drive each display separately, each
one is driven in turn—in rapid succession; i.e., they are
multiplexed. Two low-cost decoder chips are added and
this gets the I/O count for the display segments down to
seven. And of these, three outputs are in fact the same
lines as used for the three inputs for the keypad col­umns; and the other four outputs are also multiplexed to
drive the keypad rows.

Then three serial in, parallel out latches are added to
handle status and band switching.

These have three unique data lines, a common clock line
(with all four again multiplexed with the display)—and a
latch line shared with the AD9850.

The final touch is to drive the decimal point output on
the same line as the shaft encoder direction input.

If you have kept up with this, then you will agree that the
total I/O count is now down to 13! Figure 4 shows what it
all looks like—and for good measure two lines are also
shared with in situ programming. The only other viable
approach would be a multi-PIC solution. It turns out to
be marginally more expensive and significantly more
intellectually demanding.

There now remains but one question. Can we multiplex
all this multiplexing fast enough in the software so that the
user sees “instant” response and smooth “continuous”
operation? The answer, it transpires, is that it is not even
difficult!

BUDGETS

Cost—If you were to buy all the electronic components
from new, you should allow about £75.

Time—Construction time is obviously very variable, but
a good estimate would be one day each to make the
PCBs and one and one-half days to assemble them. You
will need about two hours to build the DDS subassembly.
So this is not a “weekend project,” but it probably won’t
exceed two!

If you design your own software, times are impossible
to estimate. But you can write some software to do some
one useful thing—say, generate a fixed DDS output
frequency—very quickly. Its the integration of the whole
which takes time.

Power—you need 12 V dc at 400 mA, smoothed but not
necessarily regulated. From 10 V–13 V is acceptable.

–5–

AN-557

DDS BOARD DISPLAY BOARD

REGULATORS

5V TO

LOGIC

12V

5V, 8V

8V TO

REFERENCE

OSCILLATOR

PTT

LINE

5V

BUFFER

5V

DIRECTION

PULSES

4MHz

3-TO-8

DECODER

A

B

C

ENABLE

PIC16C84

RA3

RB7

OSC1

OSC2

RA0

RA1

RA2

RA4

RB0

RB1

RB2

RB3

RB4

RB5

RB6

LPF

5V

COMMON
ANODE

BUFFER

RF OUT

7 dBm

0

1

2

3

4

5

6

7

3. 7 27. 56

BCD TO

DECODER

110MHz

REFERENCE

CLOCK

OSCILLATOR

7-SEG

A

B

C

D

E

F

G

AD9850

CLOCK

DATA

W_CLOCK

FQ_UD

DISPLAY
BOARD

1

2

USB Rate Cal

4

5

CW Scan Mem

7

8

LSB Freq Save

0

A/B A=B Split

8-BIT SR

AND

LATCH

DATA

CLOCK

LATCH

8-BIT SR

AND

LATCH

DATA

CLOCK

LATCH

3

6

9

#

8-BIT SR

DATA

CLOCK

LATCH

AND

LATCH

USE USB IF
USE CW IF
USE LSB IF
BROADBAND
[SPARE]

1.8MHz

3.5MHz

7.0MHz

10MHz
14MHz
15MHz
18MHz
21MHz

24MHz
28MHz
29MHz

Rx = VFO A
Rx = VFO B
Tx = VFO A
Tx = VFO B
LSB

CW
USB
SIG GEN MODE

SWITCH
OUTPUTS
TO HOST
Tx/Rx

Figure 4. PIC “N” MIX block diagram, illustrating PIC input/output allocations and physical partitioning. Besides power
supply distribution and decoupling, all functional elements are shown.

–6–

REV. 0

AN-557

PART 2 OF 5

In this issue the alternatives and techniques for mechani­cal construction are explored. These include a process
for making one-off PCBs—and for mounting the DDS
chip on a DIL socket carrier.

OVERALL STRATEGY

When it comes to the gross layout of the hardware,
flexibility is a design objective. When it comes to the
mounting of the DDS chip itself, a successful outcome is
likely only if you absolutely follow the rules and allow
me to adopt a somewhat dictatorial style.

Your first decision revolves around whether you are
building an external injection source or are integrating it
mechanically with your Tx/Rx.

In either case, self-evidently, the tuning knob and key­pad need to go on the front panel with the display board
immediately behind it.

The DDS board is the same size as the display board. It is
designed for mounting parallel to and behind the display
board, or at right-angles to it, or completely remotely
from it and connected to it by ribbon cable. The last
choice is not relevant in a self-contained external source.

The tuning knob may be mounted on either side of the
display, the choice being governed simply by whether
you are right or left-handed. The keypad should be
mounted on the same side of the display as the tuning
knob. Should you mount it on the opposite side of the
display, although it may give some appearance of better
aesthetic balance, you are courting an ergonomic disas­ter. Visual feedback of your key presses is given via the
display and status LEDs and your forearm will inevitably
obscure the view.

In the photographs, you will note that my keypad is
mounted contrary to these recommendations. This is a
layout peculiar to my requirements since I am unusual
in being mostly ambidextrous, preferring twisting

3.25" NOMINAL HEIGHT FRONT PANEL

motions (e.g., screw drivers) with my right hand and
pushing motions (e.g., sawing) with my left hand. In
practice, I, therefore, use both hands, but most people
would find this uncomfortable.

The second decision is whether to build the shaft
encoder as an integral part of and mounted on the DDS
and display boards—or to split them. The choice is
yours and is governed mostly by where you are starting
from. A 12" separation between the two presents no
performance issues. If you want to take this approach,
simply cut both boards, separate them and reconnect
them using four flying leads or some ribbon cable. The
four leads are +5 V, 0 V, pulses and direction. Obviously
you could build them like this in the first place.

The final consideration is the housing for a stand-alone
unit. Those of us who have built so far have found no
need for a screened enclosure but it would obviously
represent good practice. In any event, you will need to
consider weighting or securing the box since Newton’s
Second Law applies when you press the keys—and the
last thing you want is the box skidding around.

DISPLAY BOARD MOUNTING

The display board mounts immediately behind the front
panel. You will need an aperture of 3" × 3/4" to view the
frequency readout. Having cut the aperture, you need to
back the hole with some optical filter material which
either corresponds to the color of your display (typically
red/green) or—and preferably—is circularly polarized.
The latter gives much superior performance in bright
natural light but for some reason has become expensive
in recent years.

Figure 5 is a suggested front panel template which also
shows how I have accommodated the status LEDs.
Three mm holes are drilled for these, the LEDs are
inserted in the board but not soldered. The front panel is
mounted into position, and the LEDs are adjusted in their

DISPLAY BOARD - 6.1" ⴛ 2.75"

3" ⴛ 0.75" DISPLAY CUT-OUT

SIG

Rx Tx

LSB CW USB

GEN

Figure 5. Drilling template for front panel. The position of the tuning knob shown assumes you are mounting the shaft
encoder on the display and DDS boards. It could be much further to the right or on the opposite side of the display.

REV. 0

–7–

AN-557

holes for equal protrusion. They are then tacked and
finally soldered to the display board when fully aligned.

If like me, you deprecate the idea of screw-heads showing
on the front panel, then you will need to glue some nuts
or threaded pillars to the back of the front panel to
mount the display board. I find nut rivets ideal for this
since they have a large surface area which makes for
strong and permanent adhesion using super glue.

DDS BOARD MOUNTING

Figure 6 shows the configuration for right angle mount­ing and Figure 7 illustrates parallel mounting.

DISC MOUNTING KNOB

REAR BEARING

ACETATE DISC

FLYWHEEL

To ensure full access during commissioning I would
strongly recommend that you avoid the parallel mounting
configuration to start with. If this is your target configu­ration, join the two boards with a short length of 0.1"
pitch ribbon cable. This allows access to both sides of
both boards for testing.

If you are mounting the two boards at right angles in close
proximity, then the best approach is to permanently
solder the two boards together as shown in Figure 6.
Butt the two boards to form a small “T” junction (not an
“L”), tack them lightly together, check the angle and
then run beads of solder along the full length of both

DISPLAY BOARD

FRONT BEARING

IR DETECTOR

DDS BOARD

IR DIODE

TUNING KNOB

Figure 6. DDS board mounted at right angles to and integral with the display board. Also illustrates a suggested mounting
method (not to scale) for the shaft encoder disc, IR diode and detector. Note the long lead lengths on the latter to give simple
adjustment of diode and detector positions relative to the disc. The disc needs to be mounted near enough to the display
board to clear the x’tal oscillator enclosure to be described later. The rear bearing is mounted on a piece of PCB soldered to
the DDS board and/or the rear of the x’tal oscillator enclosure.

FLYWHEEL

BEARINGS

IR DIODE

DISPLAY BOARD

IR DETECTOR

DDS BOARD

TUNING KNOB

Figure 7. Alternative mounting method (not-to-scale) where the DDS board is mounted parallel to the display board on
spacers (not shown). A small hole is drilled in the DDS board to pass the infrared, and the rear bearings are fitted to the DDS
board. The leads for the detector pass through the board to the tracks—which are cut to avoid interference with the rear
bearing. The detail will become apparent when the DDS PCB is described later.

–8–

REV. 0

AN-557

sides to intimately join the ground planes. Join the
edge-connectors with a small solder bridge and test
for shorts.

A further advantage of taking this approach is that the
display board need not be secured to the front panel.
Mounting the DDS board to a horizontal base with the
display ICs touching the rear of the optical filter provides
effective location.

MAKING THE PCBs

In my article on the Third Method Transceiver, I
described an approach to constructing boards without
etching which proved very popular. It would be perfectly
viable to use this technique for the display board in this
project, but wholly inappropriate for the DDS board. So
what follows is a technique I have used for many years
for making one-off PCBs without the expense of UV
exposure techniques. I must emphasize that this
approach is viable only for one-offs and is hopeless if
you need greater quantities. I would also be very sur­prised if these particular boards can be made using an
etch-resist pen, since some of the tracking is very fine.

The technique revolves around removal of material
where you want to remove copper—rather than apply­ing resist where you want to retain copper.

The board is firstly cut to size and then drilled. For any
surface mounting areas, the board may be gently
punched but not drilled. The idea is to give yourself
guides to draw the artwork directly onto the board.

a good room temperature—certainly never cold. The
heat from a desk light makes it even easier and helps
prevent paint chipping.

Finally, the scribing tool itself is important. It needs to be
pointed but not incredibly so. And it also wants to retain
the point. I find the best tool is to take a masonry nail—
which is hard steel—cut the head off and grip it in a
draughtsman’s clutch-pencil. Failing that, a long masonry
nail through a cork is pretty comfortable.

Sharpen the point with a rotate and drag motion on a
piece of emery and when you have got it as sharp as you
can, blunt it ever so slightly on a piece of fine wet and
dry. Try it on a piece of scrap, holding the scribe at about
45°, and you should get a clean fine line. Resist gouging
out the copper. You are only trying to remove paint!
Repeat the sharpening process every ten minutes or so.
You will feel when it is not cutting the paint cleanly. By
the way, for really fine work (you won’t need it here) a
sewing needle is excellent as is an old gramophone
needle.

When you have scribed both sides of the board and
checked it meticulously, etch the board in the conven­tional manner with ferric chloride. You will find you will
get through very little FeCl because the total amount of
copper removed is very small. Observe all the usual
safety precautions. Keep the board and FeCl solution
gently on the move all the time to get an even etch and
have the courage to overetch it slightly if anything. Make
sure both sides are fully etched before removal.

With the board clean but not polished, it is sprayed both
sides with an aerosol of car paint. Matt black is best for a
contrast color against the copper. It is important to put
on a light enough coat to just cover it, but not to get any
substantial build up of paint thickness.

Then, only after the paint is truly dry, the paint is
removed between the tracks using a scribing tool. You
use the holes, punch marks and master artwork as a
guide. You only need to remove a fine line of paint. In
fact if you stand a few feet back from the finished board,
it looks substantially like continuous copper. Note that
if, for example, you have two parallel tracks, you would
need three scribed lines to implement it.

The technique takes a little getting used to, but if you
should make a mistake, simply repaint the affected area
with a small brush and do it again—differently!

There are some important tips:

Tape the board down to a reasonable block of wood to
stop it skidding around and to prevent scratching the
paint on the reverse side. You can also use a square
against the edge of the wood if you want posh lines—
but the square needs to be transparent if you want to
avoid frustration. Use a piece of Vero board as a guide if
you need to scribe edge connectors. Scribe the board at

Wash the board thoroughly in cold water, inspect and
etch further if necessary. Finally wash the board with hot
water and then clean off all the paint using cellulose
thinners. A small paint brush helps to get the paint out
of the holes, but being a good insulator, this is not critical.
Polish the board with fine wet and dry (used wet) or a
polishing block.

Now for the important stage. Using a continuity tester
check for isolation between each and every adjacent
track. If you find any shorts that are obvious, clear then
with a sharp blade. If they are not obvious, my practice
which I hesitate to publicize is to connect two test probes to
a car battery and then blow off the short. Be careful!

The end result is an individual piece of craftsmanship—
produced with no greater effort or time than is needed
to draw the artwork onto film in the first place. And it is
home-brew! You end up with much more ground plane
than is typical with other approaches—which can only
be to the good. And there are no critical processes in the
sense that you can see what is happening all the time
and can avoid moving on until you have got it right. I
commend it to you.

REV. 0

–9–