A High-Performance
Homebrew Transceiver:
Part 1
Here is a general description of a transceiver built
without regard to size or complexity—the only
goal is optimum performance for DX and contests.
H
am radio affords an opportu-
nity for a variety of fascinat-
ing experiences. For many
hams, foremost among these is operating
with equipment built in a home work-
shop. Traditionally, hams have built
equipment that outperforms factory
gear. Today, modern methods allow the
production of miniaturized, microproces-
sor-dependent units that are difficult
for radio amateurs to duplicate. How-
ever, in regard to basic performance and
operating features, the home workshop
can often produce gear that outclasses
factory equipment.
This article begins a description of a
homebrew transceiver built for seri-
5259 Singer Rd
Las Cruces, NM 88005
k5am@roadrunner.com
16 QEX
By Mark Mandelkern, K5AM
ous DX work and contest operating. It
was designed without compromise
regarding performance and basic
operating features, but includes no
modern computer-related features.
The goal was to build a radio that
would outperform any available fac-
tory-built radio, regardless of price.
In 1948, I began homebrewing my
first station, W9ECV, in Wisconsin. By
1990, everything in the shack was
homebrew except the transceiver. Work
began on this high-performance home-
brew transceiver—it was completed
three years later. Two portions of the
radio, perhaps the most innovative,
have already been described in QEX:
The high-performance AGC system and
the non-crunching noise blanker.1, 2
A description of the completely home-
1 Notes
appear on page 24.
brew K5AM station recently appeared
in NCJ.3
Basic attitudes toward homebrew-
ing acquired in the 1940s have been
retained. Homebrew means “built-
from-scratch,” and often newly de-
signed. I’ve tried to keep up with some
of the latest devices and techniques,
but this transceiver uses no micropro-
cessor, no synthesizer and no phase-
locked loops. No phase noise, no spurs,
and no birdies! This, of course, in-
volves limitations: No memories, no
instant frequency jumps, no computer
control, no DSP.
For ordinary DX work, nonetheless,
these features are not missed. For con-
testing, however, here is the dis-
claimer: World-class contesters would
not find this radio acceptable. They
often want to operate two radios simul-
taneously, with computer control. They
want frequencies read from the radio
and entered into the computer to create
band maps, and for logging purposes.
In other words, they want bells and
whistles ad infinitum. So I must hedge
on the contesting claim in the subtitle of
this article; this radio is designed for
the sort of contesting that I do. This
includes single-operator DX contests,
domestic HF contests and VHF contests
on several bands with transverters.
Results have been gratifying. A number
of section and area awards have been
won with this homebrew radio. I have
operated the best of the current factory
radios and—in many respects—find
them lacking in comparison. The bottom
line is that this homebrew transceiver
has certain performance characteristics
that surpass those of the best factory
gear. This more than compensates for
the lack of bells and whistles, at least
for my style of operating. As for backing
up the basic performance claim, I’ll give
data in a following article about align-
ment and measurements.
This radio was built for daily use in
a station with a heavy operating
schedule. It was not built as an experi-
mental platform. Experimenting with
each circuit, trying at each stage to
obtain results surpassing all previ-
ously published circuits, would have
meant that the project would run to
decades, rather than years, and would
likely never have been completed.
Besides, this builder has no profes-
sional training and lacks the expertise
for such a project; I merely put to-
gether circuits, already optimized by
experts, to produce a complete operat-
ing unit. Individual circuits were
chosen from the available literature at
the time of construction.
This article describes only the general
plan of the transceiver, with emphasis
on design considerations and a discus-
sion of features required for serious
DX work and contesting. Subsequent
articles will give circuit details.
Conversions: Fewer are Better
The craze for multiple conversions
began in the 1950s. At that time, there
were some good reasons. A single-
conversion receiver with a 455 kHz IF
had intolerable images on the 10-meter
band. Thus, a first conversion to about
3 MHz was a great improvement. In
addition, it was difficult to obtain very
sharp selectivity at 455 kHz, so a third
conversion to about 50 kHz was helpful.
Eventually, we saw advertisements for
quadruple-conversion radios.
The advantages of multiple conver-
sions came at the cost of increased
IMD. Mixers have noise figures equal
to their conversion loss, and this loss
must be compensated by gain stages
ahead of the mixer. The result is very
high signal levels at the last mixer
before the high-selectivity filter. It
is this last mixer that mainly deter-
mines the close-in dynamic range of
the receiver. The current hype about
strong front-ends is misleading. The
crunch is at the last mixer [for nearby
signals—Ed]. Even more misleading
is the hype about strong preamps for
VHF and UHF. A preamp cannot
improve the IMD performance of a
receiver, it can only degrade it; the
more gain in the preamp, the more
trouble down the line.
The situation is different today than
in the 1950s. Excellent crystal filters
for both SSB and CW are available at
about 9 MHz. With a first up-conver-
sion and a second conversion down to
9 MHz, superb image rejection and
selectivity are easily obtained.
This transceiver tunes the bands in
1-MHz segments, with a variable first
IF tuning 40 to 39 MHz. The tuning is
“reversed” because the injection is on
the high side; eg, at 68 MHz for the 28
to 29 MHz band. The result is a very
high-performance radio with only two
conversions. The high-side injection
method results in virtually no spurious
responses; it is used on all bands. There
is no difference between the 160 and 2-
meter bands, as far as mixing scheme is
concerned. The arithmetic is easy: Sim-
ply add 40. Injection is at 41.8 MHz for
the 160-meter band and at 184 MHz for
2 meters. All the front-end oscillators
use third or fifth-overtone crystals, well
known for their low phase noise. For 2
meters, the 92-MHz oscillator is fol-
lowed by a balanced doubler. (In this
article, the author uses the term “front-
end” to indicate both the receive front-
end and the transmitting circuitry
along with a control panel for a particu-
lar frequency range: HF, 6 meters or
2 meters.—Ed.)
I didn’t invent this conversion
scheme. It is taken from the Signal
One CX7, a radio that appeared in
1969, and that had dozens of ground-
breaking innovations. Few hams
today know that the CX7 is the grand-
daddy of all present-day radios. I
operated, repaired and modified my
CX7 for 20 years before designing my
homebrew transceiver. Did I learn
anything from the CX7? Well, when it
first arrived in my shack, all I could
say was: What are those funny little
things with three legs?
Basis for Design
Many ideas were taken from the CX7
besides the frequency-mixing scheme.
Notably the RF speech clipping, the
transmitter driver circuits and the PA
bias and ALC circuit. In addition, some
surplus parts were used from basket-
case CX7 radios found at flea markets.
The crystal filters, precision-machined
mechanical parts (bearings, etc) for the
PTOs, the panel escutcheon for the
frequency counter, the conduction-
cooled 8072 PA tube, the anode clamp
and heat sink, and a few other miscel-
laneous small parts were salvaged from
various dismantled assemblies.
The goal in building my own radio
was to obtain improvements in perfor-
mance, operating convenience and new
features. The features I had used for 20
years and liked, I retained. Even the
panel layout bears some resemblance
to the CX7. This led one friend to ask:
“Is your radio a CX7 clone?” Definitely
not! Is every radio with a 455 kHz IF a
clone of the first one? Although the
frequency-mixing scheme in this radio
is the same as in the CX7, virtually
every circuit is newly designed. At the
same time, my debt to the CX7 design-
ers is enormous. It was essential to
keep the same IFs for two reasons:
First, I planned to use surplus 9-MHz
CX7 crystal filters, which are excellent,
if selected from a batch. Second,
transverters that fed directly into the
CX7 40 MHz first IF had already been
built.
Here are some features of the K5AM
homebrew transceiver:
• Balanced JFET and balanced
MOSFET mixers
• Careful gain distribution
• High dynamic range
• Non-crunching noise blanker (see
Note 2)
• High-performance, no-pop, no-click
hang AGC circuit (see Note 1)
• Super-sensitive integrating squelch
for SSB and CW (mainly for 6-meter
DX)1,4
• Complete TTL logic control. Eg, the
mode switch has seven leads, not
dozens.
• Quick, one-button PTO switching
• Relay-switched crystal filters
• A sharp CW filter at the IF output
(in addition to one at the input)
• Electronic attenuators for all audio
level controls. This reduces hum
problems.
• 60-Hz filter in receiver audio
• High-pass filter in transmitter audio
to eliminate externally induced hum.
• Pulse tuning circuit for safe, easy
external amplifier tuning; set for
33% duty cycle5
• Sharp CW shaping. No dit delay
after key closure.
Mar/Apr 1999 17
• RIT—separate knobs for each PTO
• One-button second-PTO monitoring
• Automatic (optional) transverter
switching tied to PTO switching. This
is mainly for instant 6 to 2-meter
switching during VHF contests. It is
also useful for instant switching
between 144.200 MHz and 3.818 MHz
(75-meter liaison) during meteor
showers.
• Clear, sharp audio—the audio out-
put module uses a class-A, push-pull
circuit
Tuning the First IF
This radio differs from nearly all
current designs—the first IF circuits
are tunable. One advantage applies to
VHF DX operators. The basic 40 to
39 MHz tunable transceiver forms an
excellent foundation for the attach-
ment of VHF transverters. For the
6-meter band, the LO injection is at 90,
91, 92 or 93 MHz, from four separate
crystal oscillators. This high-side in-
jection yields no detectable spurious
responses. The overall conversion total
is merely two. This is in sharp contrast
with common practice, where 50 MHz
is converted to 28 MHz; LO injection at
Fig 1—K5AM homebrew transceiver simplified block diagram. This diagram includes the basic transceiver panel, tuning 40 to
39 MHz, and the three front-end sections, for HF, 6 meters and 2 meters. Additional front-end sections, for other VHF or UHF
bands, can be easily added. The dashed lines indicate transmit signal paths.
18 QEX
22 MHz allows the possibility of un-
wanted spurs. In such an arrangement,
the overall conversion total can be as
high as five, with the resulting high
possibility of spurious responses,
spurious emissions and IMD.
Tuning the first IF means that the
circuits following the first mixer have
a bandwidth of 1 MHz and thus the
second mixer must be strong enough to
handle any signals within this range.
This is in contrast with current produc-
tion radios that use a fixed-frequency
first IF and crystal filter at about
70 MHz, to limit the spectrum of sig-
nals within the first IF strip. These fil-
ters are usually about 20 kHz wide,
with poor shape factors and ultimate-
rejection characteristics, compared to
filters at 9 MHz. During lab tests using
two-tone spacing of 20 kHz, such re-
ceivers may demonstrate excellent
dynamic range. Nevertheless—with
today’s crowded band conditions—one
cannot hope to find a clear 20-kHz-wide
segment in which to operate. Published
reviews do not address this problem.
This radio, on the other hand, while
not employing a first IF filter, does
have good dynamic range performance
with respect to adjacent-channel
signals, as well as more distant sig-
nals. This performance relates more
closely to real operating conditions.
Another problem arises in some cur-
rent production receivers. These de-
velop AGC in the first IF, ahead of the
sharp crystal filters, and apply AGC
voltage to the first IF strip and front-
end. This may prevent IMD in the first
IF strip, but gain is reduced and weak
signals may be lost. Receiver sensitiv-
ity is reduced in the presence of nearby
strong signals that lie outside the
second-IF passband. Hence, front-end
AGC and gain reduction may be unac-
ceptable for some weak-signal work.
In this radio, AGC is applied only to
the IF strip at 9 MHz, after the crystal
filters. This arrangement allows the
IF gain to be controlled with no loss of
sensitivity in the front-end. The mix-
ers are built to handle strong signals,
and the sharp 9-MHz crystal filter
following the second mixer effectively
keeps off-channel signals out of the
AGC circuits. The front-end runs wide
open at full sensitivity, the best
arrangement for weak signals. On the
lower HF bands, it is sometimes pru-
dent, although not always necessary,
to use the front-end attenuator.
Block Diagram
Fig 1 shows the main sections of the
transceiver. Its most unusual feature
is that it is built on separate panels;
one for the basic 40 to 39 MHz trans-
ceiver, and one for the 200 W HF front-
end. These two items were completed,
after three years work, in 1992 and
together form an HF transceiver in the
usual sense.
However, the transceiver covers all
ham bands from 160 meters to 2 meters,
so we need a brief description of how
the transceiver fits into the complete
station. The 6 and 2-meter front-end
sections were built a few years earlier.
These are often called transverters, but
in this case they have 40 to 39 MHz
outputs; they do not translate the VHF
bands to any amateur band. They are
integral parts of the transceiver, func-
tioning exactly like the HF front-end.
Now there are 11 ham bands from 160
through 2 meters. It takes four panels,
and we have only 2 W on 6 and 2 meters,
not a very acceptable power level for an
11-band transceiver. So, we now add
100 to 200 W 6 and 2-meter amplifiers.
One 6-meter amplifier was built in
1951, when I was still in high school.
So, I finally have a complete 160 to
2-meter homebrew transceiver at about
200 W, on only six panels, which took
only about 40 years to build! There are
also homebrew 1.5-kW amplifiers cov-
ering all bands.5,6,7 A block diagram of
the entire station is shown in the NCJ
article (see Note 3).
Compactness and miniaturization
have obviously not been prime goals for
this project. On the other hand, this
style provides a good deal of flexibility.
For example, many homebrewers con-
centrate on VHF/UHF SSB/CW DX op-
erating. Building the basic transceiver
separately allows one to add whatever
VHF/UHF front-end sections are de-
sired. The 40-MHz IF, with high-side
injection, works very well on the 6 and
2-meter bands, and should work well on
higher frequency bands.
Receiver Gain Distribution
Receiver gain distribution is a crucial
factor in obtaining high dynamic range.
In this radio, the IF strip—after the
sharp filters—was designed with the
highest practical gain. This allows the
signals at the mixer to remain at a low
level. Gain before the sharp filters
increases signal levels at the last
mixer, and so reduces dynamic range.
Gain after the filters has no detrimen-
tal effect on receiver performance, just
as turning up the audio gain in a large
room does no harm. This is one reason
why a receiver with minimal conver-
sions, and thus less need for front-end
gain, has the best potential for superior
performance. The most serious limita-
tion on putting most of the gain after
the filters is the possibility of BFO
signal leakage into the 9-MHz IF strip.
The IF strip in this radio has the un-
usually high gain of 107 dB, and oper-
ates at the unusually low signal-input
level of –119 dBm—or –128 dBm with
the 200 Hz CW filter. This requires
exceptional filtering and shielding of
the IF strip, the BFO circuits and the
power supply leads. The construction
methods are briefly described below.
Frequency Mixing Scheme
and IF Shift
Fig 2 shows the premixing scheme.
The 3 to 4-MHz VFO is converted to 31
to 30 MHz LO injection power at the
mixer, and thus 40 to 39-MHz signals
are converted to 9 MHz. Most signifi-
cant is the way the BFO frequency is
subtracted, resulting in an IF shift, an
essential operating feature. In this
way, the BFO frequency does not
affect the receive frequency.
With this premixing method, the
receiver frequency is the suppressed-
carrier frequency for SSB and the
zero-beat frequency for CW. If a
14.010 MHz CW signal is tuned for a
500-Hz audio tone, the receiver will
read 009.5. An advantage of this
method is that if tuned to zero-beat,
the receiver reads the actual signal
frequency. A more important advan-
tage appears in VHF DX work, where
operators frequently shift between
SSB and CW—without changing

Fig 2—Simplified frequency-mixing scheme. Premixing the low-frequency VFO up to the injection frequencies required by the
mixer would be simple enough. The main feature here, however, is the mixing of the BFO frequency to obtain IF-shift
operation, a necessity for any radio designed for serious DX or contest work. The result is that the BFOs can be tuned without
changing the receive frequency. Credit for this mixing scheme is due to the CX7 designers. Rather than actual frequencies,
round numbers are used in this diagram; this allows the basic idea of the mixing scheme to be shown without lots of ugly
decimals. See text for details.

Mar/Apr 1999 19
frequency—during a single contact.
With some “modern” radios, a fre-
quency shift occurs when changing
modes, or changing CW filters, often
resulting in a lost contact. With the
system used here, the CW signal is
transmitted within the SSB channel.
For transmitting CW, the master
oscillator is disabled, and the offset
oscillator is used for injection at the
PTO mixer. The BFO still provides
the signal for the CW carrier gate on
the AF board. Thus the offset oscilla-
tor can be used to vary the transmit
frequency. Using the round numbers
from Fig 2, setting the offset oscillator
to 33.9995 MHz (via the front panel CW
OFFSET knob) will result in a 500-Hz
offset. The operator will hear a 500-Hz
tone when the offset monitor button is
pressed.
This frequency mixing method is ex-
actly as in the CX7, although in Fig 2,
round numbers have been used in order
to show the basic idea without lots of
ugly decimals. The actual VFO fre-
quency range is 3.1 to 4.1 MHz, eliminat-
ing an obvious spur on the entire 80-
meter band. To accommodate this shift,
the master oscillator runs at 43.1 MHz.
The actual crystal filter center frequency
is 8815 kHz, and the actual BFO frequen-
cies are 8816.5 kHz and 8813.5 kHz. Any
crystal filters in the 9 MHz range with
corresponding BFO crystals may be
used, with no circuit change. The actual
frequencies in Fig 2 are now easily calcu-
lated; in USB and CW modes the BFO is
at 8.8165 MHz, the BFO mixer has out-
put at 34.2835 MHz, and the PTO mixer
output, for LO injection, has the range
31.1835 to 30.1835 MHz.
The functioning of the IF shift
feature can be seen in Fig 2. When
receiving USB signals, for example,
the BFO frequency is above the crys-
tal filter passband, because of the
high-side injection and the resulting
inversion in the front-end. If the BFO
frequency is decreased somewhat using
the IF SHIFT control on the front panel, it
will be closer to the crystal filter pass-
band, resulting in a lower-frequency
audio passband. At the same time, the
output frequency of the BFO mixer will
increase, causing the PTO mixer output
frequency (the injection frequency) to
also increase, in the same amount. This
causes the frequency of the signal in the
IF strip to decrease, so it mixes with the
BFO to produce the same audio tones
as before. Thus, the receiver is still
precisely tuned to the station. One can
work out the simple formula for the fre-
quency of the receiver audio output,
using the signal frequency, the trans-
mitted tone frequency and all the oscil-
20 QEX
lator frequencies. The BFO frequency
appears twice, with opposite signs, and
cancels out.
Front-Panel Controls
While not having the bells and
whistles that a microprocessor-con-
trolled radio might have, this radio does
have some features that, as least for
this operator, beat current factory ra-
dios in operating convenience. I do have
a few late-model radios up at my Horse
Mountain VHF contest station (at 7900
feet) and do know that memories and
other features are useful and fun, and
that one can become accustomed to
menu-driven controls. Nevertheless, I
find that the traditional panel controls
on my homebrew radio at home allow
quicker and easier operation in the heat
of battle, such as during a rare one-
minute DX opening on 160 meters, or
during a contest.
The front panel is shown in Fig 3.
The template in Fig 4 shows the con-
trols and labels clearly. Some of the
control features are discussed below.
The dual PTOs have separate, large
tuning knobs and separate, large RIT
knobs. This feature is very desirable;
in some factory radios, the RIT, once
activated for one VFO, also affects the
other VFO, where it is not wanted. The
separate RIT circuits are turned on by
simply pulling the knobs.
The PTO control switch (A/B) is a
three-position, black Bakelite bat
handle lever switch, directly beneath
the digital frequency display. The right
position selects PTO A, left selects PTO
B. Center position selects split; receive
on A, transmit on B. What about the
opposite split? I never use it. It’s best to
acquire a fixed habit in this regard and
stick to it; then there is less chance of
Fig 3—Front panel of the basic 40 to 39-MHz transceiver. Four of the operating
features are enabled by means not apparent in the photo. The A RIT , B RIT ,
BLANKER , and DUAL RX knobs are all attached to potentiometers with ganged push-
pull switches. Pulling the knob out actuates the function.
transmitting on the wrong frequency. If
desired, the opposite split may be ob-
tained using the monitor switch.
For monitoring the B channel (with
counter read-out), there is a momen-
tary push button directly beneath the
digital frequency display, and a lever
switch at the lower left. I thought that
having the button in the middle would
be most convenient, but toward dawn
after a long night on the 160-meter
band, the lower switch is used more
often. All the lever switches on the
front panel are old Switchcraft models
with black Bakelite levers. They’re
still found on the surplus market.
They are manufactured in either
momentary or fixed styles, but the
fixed version can be used either way:
Lean gently on the lever for momen-
tary action, push harder to make it
hold. For example, the KEY switch can
actually be used to send CW in an
emergency, or pushed fully down to
hold for a steady carrier.
When the lower B-channel monitor-
ing switch is pushed up, it lets you set
the PTO B frequency while listening
to PTO A. A panel control sets the
amount of CW OFFSET, and an audio-
monitor momentary push button to
hear the offset. The controls are ar-
ranged so that the button can be
pushed—and the CW OFFSET knob
turned—with one hand. Another knob
sets the OFFSET LEVEL of the tone in
the headphones. The transmit-
sidetone frequency is independent of
the offset, and is set internally, with
a level control on the front panel. The
DUAL RX feature is turned on and off
by simply pulling that knob; the knob
adjusts the balance.
It is universally agreed that the best
speech processing method is RF clip-
LISA MANDLEKERN
ping. This requires a second SSB crys-
tal filter, but the extra DX countries
and contest points make it well worth
the trouble. Separate MIKE-gain and
CLIPPING controls are provided on the
front panel. After changing mikes, one
calibrates the radio for the new mike
by simply turning the CLIPPING control
fully ccw and adjusting the MIKE con-
trol for minute indications on the CLIP
meter (at clipping threshold). Then
the CLIPPING control is advanced to the
desired level. Microphone calibration
ensures the best gain distribution in
the transmit audio circuits, the proper
audio level at the balanced modulator,
the best transmit-audio quality and
optimal carrier suppression. The CLIP
meter circuits are adjusted so that
full-scale on the meter represents
20 dB of RF compression; one-quarter
scale represents the normal level of
6 dB compression.
In addition to the front-panel MIKE
jack (which I never use), there is a
high-level speech input jack on the
rear panel, which receives audio from
the station audio-distribution system
and digital voice recorder.8 There is no
KEY jack on the front panel; I don’t like
a clutter of cables on the operating
bench. There are KEY, KEYER and FSK
jacks on the rear panel.
The radio has two analog meters on
the panel. In receive, one meter reads
SIGNAL level. When the squelch circuit
Fig 4—Front panel layout. This shows the transceiver controls and their labels. The digital readout counts the PTO outputs
directly, and shows only the kilohertz part of the transceiver frequency. The megahertz band is indicated by band switches on
the front-end panels. For details of the operating features, see text.
is in use, the other meter reads inte-
grator voltage (SQUELCH); this allows
quick and easy squelch-level adjust-
ment. In transmit, one meter reads
ALC voltage. In SSB mode, the other
meter reads RF-CLIP level. Constant
monitoring of these two meters en-
sures good signal quality.
Full break-in CW operation (QSK) is
included for HF: no dit shortening no
lag and break-in ability up to 50 WPM.
Semi-break-in ( SQSK) is available for
all bands; the delay is set by a knob on
the panel. A three-position panel
switch chooses QSK, SQSK or neither.
A Curtis keyer chip is included, with a
SPEED control on the panel. The CW-
waveform make and break times have
separate internal adjustments.
The TUNE switch provides PULSE
tuning (described earlier) if pushed up
or a steady carrier if pushed down.
The AF GAIN (AFG) knob on the front
panel also functions when using an ex-
ternal audio DSP filter. This is much
more convenient than dealing with the
AF-gain control on the external unit.
For this purpose, audio jacks on the
rear panel provide connections to the
line-level input/output jacks on the
DSP unit. Also included are special
amplifier and attenuator circuits to set
(and forget) the proper drive level to an
external unit, and to equalize the DSP
on/off audio levels in the radio.
In lieu of the RF gain (RFG) control
found on almost all receivers, this
transceiver uses an IF Gain (IFG) con-
trol. This IFG control lowers the gain
only of the IF strip, leaving all stages
ahead of the sharp crystal filters run-
ning at full gain. This preserves full
sensitivity for weak-signal work. The
main use of this IFG control is for
“AGC threshold” operation. For this
reason, the circuit is arranged so that
using the IFG control does not cause
the S-meter to read upwards. AGC-
threshold operation is very effective
with extremely weak 160-meter DX
signals and with EME (earth-moon-
earth, moonbounce) signals. Even the
best AGC system is not as good as a
well-trained ear.
In this radio, the AGC threshold is
about 10 dB above the MDS (mini-
mum-discernible signal) level. Avail-
able gain is sufficient so that ambient
antenna noise activates the AGC sys-
tem. For AGC-threshold operation,
the IF Gain is reduced so that weak sig-
nals and noise are a few decibels be-
low the AGC threshold. In effect, the
AGC threshold is raised; there is no
decrease in sensitivity. The ear can
now hear the weak signal in the noise.
The AGC system is prevented from
reducing the receiver gain at every
little static crash. On the other hand,
large static crashes and loud signals
will activate the AGC system and pro-
tect the operator’s ears. This is impor-
Mar/Apr 1999 21
tant for hams who wish to work DX for
many, many years. This operating
method is safe, in contrast to the all-
too-common practice of turning off the
AGC.
Most radios suffer from two RFG
control problems: Most RFG controls
cause the S-meter to read upwards,
making it very difficult to set the cor-
rect AGC-threshold operation level.
Also, most RFG controls reduce the
front-end gain; this reduces receiver
sensitivity and causes signals to be
lost. The IFG control in this radio
raises the AGC threshold; the S-meter
correctly indicates signal levels rela-
tive to the set threshold, and receiver
sensitivity is not reduced.
The STBY/OPERATE switch is com-
bined with the SQUELCH switch. OPER-
ATE is up, STBY center, and SQUELCH
down. Thus switching to SQUELCH auto-
matically switches to STBY, and disables
the transmitter. This is the simple one-
touch procedure for leaving the radio on
an HF DX operation frequency or a VHF
DX calling frequency and going back to
the workbench.
The SQUELCH level control is a sepa-
rate knob, not linked to the SQUELCH
switch, so it does not have to be reset
each time the squelch is turned on or
off.
There is one last, very important
feature: There are no concentric dual
knobs!
Construction and Layout
Fig 5 shows the basic transceiver,
with four main circuit boards. Each
board is hard-wired to the radio, with a
12-inch-long harness. This allows the
boards to be lifted out for testing while
in operation. Sub-boards are mounted
on hinged spacers, for immediate
access. The PTOs are mounted to the
front panel, which can be removed in
seconds, with no knobs or couplings to
remove. The rear panel, which carries
the power supply, receiver audio-output
module and input/output jacks, can be
removed in seconds.
All the gear in the shack is built on
standard 19-inch black rack panels.
The basic 40-MHz transceiver is built
on a 8.75-inch-high panel. The 200 W
HF front-end is on a 7-inch panel. The
2-W 6 and 2-meter front-ends are each
built on a 3.5-inch panel.
The need for exceptional filtering
and shielding was mentioned above in
soldered-in, forming a 1.5-inch-high liv-
ing space for the circuits, and a 0.5-inch-
deep space below for filtering. All power
and control leads go through soldered-
in feedthrough capacitors in the floor,
and then through additional filters in
the “basement.” Individual stages are
separated by soldered-in walls. Signal
leads pass through tiny windows in the
walls. As required, some compartments
are fitted with shielded ceilings.
Some of the surplus parts used in
this design may not be readily avail-
able. If readers want to duplicate the
radio, perhaps only in part, any of the
currently available replacement crys-
tal filters in the 9-MHz region could be
used. Because of the unique premixing
scheme, no circuit changes will be
needed; only the two BFO crystals will
need to be selected accordingly. The
VFO problem will always present an
interesting challenge.
The block diagram in Fig 1 reflects
the physical layout into four main
boards called LOGIC, RF, IF and AF.
Smaller assemblies are mounted on
the front and rear panels. The four
boards, the panels and the HF front-
end formed the six main phases of the
project. A total of three years was
scheduled, six months for each phase.
I started with a rough overall plan, but
not a complete plan for each phase; no
schematics at first, only input/output
specifications. Typically, the six
months work for each phase were di-
vided into four months of designing
and two month’s of building. General
descriptions of each of the six main
parts of the radio are given below, with
emphasis on design criteria and oper-
ating features. Subsequent articles
will deal with circuit details.
Logic Board
No one who has replaced the mode
switch in a traditional radio has fond
memories of the experience. With up
to six modes, including several CW
filter choices and numerous circuits to
control, there can be dozens of leads.
After fighting mode switches for
decades, I used modern logic ICs to
put an end to this nightmare.
There are even better reasons for
using logic control. One is crystal filter
switching. Switching filters directly at
the mode switch is problematic in that
it requires the filters to be somewhat
exposed. In contrast, logic control with
miniature relays allows careful shield-
ing and greatly improved ultimate
rejection. That is, there is reduced
“blow-by”—the signals that leak
around the filter.
Logic-mode switching also offers
mechanical advantages that are im-
portant for the experimenter. The
mechanical linkages and couplings
that connect a traditional mode switch
to the front panel are eliminated. I
LISA MANDLEKERN

the section under “Receiver Gain Distri-

bution.” To exemplify my techniques,
the three signal boards can be used.
Each measures about 7.5×15×2 inches
and is constructed using double-sided
circuit-board material. A floor is Fig 5—A view from above.

22 QEX
used no such linkages in this radio. A
further advantage is the ease with
which one can make modifications to
the various control functions, simply
by adding a few TTL gates.
Controlling six modes, the mode
switch used here has merely six TTL-
level leads, and a ground. The logic
board converts the TTL signals to
whatever levels are required to control
the various circuits. In addition to the
mode-switch signals, some panel con-
trols (PULSE / TUNE , STBY, A / B , DUAL
RX , B MONITOR and B SPOT) are con-
verted to TTL signals that are com-
bined in the logic chips for control of
various circuits. This makes it very
easy to implement any desired func-
tion. For example, one touch of the
TUNE button automatically shifts the
radio into CW, silences the mike,
closes the PTT and KEY lines, disables
the sidetone and shifts the carrier into
the clear channel you have found.
The PTOs are also controlled by the
logic board. The single, three-position
PTO lever switch has only two leads
and a ground, instead of a 16-pole,
three-push-button, 24-wire assembly,
plus an 8-pole, 16-wire concentric rotary
knob obstacle course, as in the CX7.
PTO control for flexibility is of special
concern. The radio has two PTOs and
dual-receive capability. In addition, it
provides instant one-button monitoring
of the second channel, for split fre-
quency DX operations. (Dual receive is
not suitable for extremely weak, barely
readable DX signals.) In addition, there
is provision for spotting the transmit
frequency (as for 40-meter DX SSB
work) while receiving. The logic board
must select which PTO to read on the
frequency counter in any given situa-
tion. For all this PTO and counter con-
trol, it was expedient to go beyond the
simple TTL gate chips and use the
larger data selectors, the 74151, as
logic-function generators. Each vari-
able is assigned a symbol, the logic func-
tion that gives the desired result is writ-
ten and the chip is wired accordingly.
RF Board
This board contains the balanced
JFET receive mixer that converts the 40
to 39-MHz signals to 9 MHz, and the
balanced MOSFET transmit mixer that
does the reverse. It also contains the two
9-MHz BFOs, the premixing circuits
that enable the IF-shift circuit, the dual-
receive circuits and the front-end of the
tunable noise blanker. Premixing does
allow the possibility of spurious re-
sponses or birdies, so there is very
extensive filtering and shielding.
This board includes the oscillator
that enables the adjustable CW offset.
Also included is the mixer that mixes
the offset oscillator with the normal
output of the BFO mixer, to produce
the audio tone for headphone monitor-
ing of the offset.
IF Board
The IF strip functions in both re-
ceive and transmit modes, using six
MOSFET stages, and several addi-
tional switching and buffering stages.
The board includes the crystal filters,
the RF clipping (and clipping meter-
ing) circuits and the AGC circuits.
For SSB, there are matched 2.4-kHz
filters at the input and at the output of
the IF strip. The result is a 2.0-kHz
passband. The second filter is essential,
because the RF clipping takes place
within the IF strip and produces distor-
tion products. This is an unavoidable
effect of RF clipping. Because of the
greatly increased effectiveness of the
SSB signal, it is well worth the effort
and cost of the second filter. The gain of
the IF strip is reduced greatly—to a
fixed level—during SSB transmissions.
Clipping occurs at a fixed level within
the strip. The amount of clipping is
adjusted by varying the gain of a trans-
mit stage leading to the IF strip. The
transmit drive level is set on the front
panel of each front-end module. There
is no transmit output level control on
the transceiver’s panel—it would have
to be readjusted when switching be-
tween front-ends.
RF speech clipping requires careful
monitoring. A meter provides more
definite indications than does head-
phone feedback. The meter indicates
the actual amount of clipping, and
alerts the operator to changes caused
by fatigue, over-enthusiasm or changes
in mike position. The RF clipper itself
is not expensive—two diodes at a total
cost of 10 cents—but associated cir-
cuitry is required for convenient panel
adjustment, reliable operation and ac-
curate metering. CLIP-meter calibration
is independent of mike-input levels or
the setting of the MIKE control.
For CW, narrow IF filters can be
switched in. There is a 200-Hz filter at
the IF strip input, and a matched
250-Hz filter at the output. The second
filter removes excess noise developed
in the IF strip. Many radios exhibit a
disturbing behavior when a sharp CW
filter is switched in. Their gain drops
noticeably. In this radio, an extra
stage of amplification ahead of the
sharp CW filter compensates for the
loss. When the sharp filter is switched
in, the S-meter holds steady, and there
is no change in the sound of the re-
ceived CW signal, except that the in-
terfering signals are gone, and the
noise-level is reduced.
The AGC system has been fully de-
scribed in a previous article (see Note
1). The AGC system can make or break
an otherwise good radio. Poor attack
performance—with clicks and thumps
—can cause operator fatigue, a crucial
factor in all-night DX operating or all-
weekend contesting. Poor decay perfor-
mance can cause excessive receiver re-
covery time after a strong signal ceases
transmitting, preventing a weak signal
from being heard. Even worse, poor
decay characteristics can sometimes
cause excessive receiver recovery delay
after every transmission, preventing
reception at precisely the most impor-
tant time. To avoid these problems,
hang AGC circuits with carefully con-
trolled timing are required.
AF Board
This board contains the product de-
tector, balanced modulator, sidetone
oscillator and low-level AF circuits.
The need to isolate the BFO and the IF
strip has been noted above. Hence, the
BFO signal from the RF board is
routed to the AF board at a low level,
then amplified and fed to three BFO
gates. Two of these feed the product
detector and balanced modulator; the
third is the carrier gate for CW.
The station microphone audio-distri-
bution system includes a one-stage,
high-pass filter, mainly to eliminate
hum introduced by an external digital
voice keyer (see Note 8). The AF board
includes two stages of high-pass filter-
ing, effectively eliminating any re-
sidual hum.
PTOs, Counter and Power Supply
The PTOs are straightforward, each
with four stages of buffering. To avoid
any possible spurious emissions dur-
ing split operation, the PTOs are pow-
ered on and off at each TR transition.
This permits full break-in operation
with no chirp. The counter is simple,
using 7400 series TTL chips. It reads
to 100 Hz, which I find adequate.
The power supply may seem over-
designed, but there is a reason. It re-
sults from years of experience with
radios in which transient pulses travel
between stages, and boards, by way
of the power-supply circuits, or even
a common power transformer. The
elimination of these transients is
crucial for proper AGC performance.
The result is four separate regulated
Mar/Apr 1999 23
supplies, with four separate small
transformers. These power only the
basic transceiver, with its 200-µW
exciter output. Locating the 200-W HF
power amplifier on a separate panel
keeps the rest of the transceiver cool
and stable. The four separate small
transformers, rather than being a
problem, yield advantages in acquisi-
tion, mounting, space fitting and cost
over a four-secondary transformer.
The four supplies include ±18 V for
the main boards and +8 V for the logic
board and counter. The fourth supply
is a separate +18 V supply for the re-
ceiver audio-output module, since this
stage is often a serious offender in
producing power-supply transients.
Each of the four main boards and the
counter then have individual on-board
regulators for +15, –15 or +5 V, as
needed. This double regulation avoids
all transient problems.
HF Panel
The little-known 8072 external-
anode tetrode—manufactured by RCA
among others—is a gem. Only half the
size of a 6146, it easily delivers a linear
200 W, even up on the 2-meter band,
where I also use one. It is rated to
500 MHz. Being conduction-cooled, it
needs no noisy forced-air cooling, but
does require a special anode clamp and
heat sink, salvaged from a junked CX7
found at a flea market. Although it is
not necessary, a small inaudible muffin
fan is attached to the heat sink. The fan
comes on only when transmitting and
has a timer to keep it going for one
minute after each transmission. This
means that it runs continuously during
a contest. A new 8072 is expensive. I use
a total of eight of these tubes on various
frequencies, but they were all obtained
as used surplus at very little cost. With
adequate protection circuits, they seem
to last forever.
The CX7 was the first ham trans-
ceiver produced with so-called “broad-
band PA tuning”. In fact, the PA used
a pi-L circuit on the seven 1-MHz
bands, switching banks of internally
adjustable tune and load trimmer ca-
pacitors. This is merely a fixed-tuning
arrangement. It works well enough on
the lower bands, but the poor L/C ratio
on 10 meters makes coverage of an
entire 1 MHz segment difficult. In
addition, coil-turn shorting and tor-
oid-core losses result in reduced
24 QEX
output on 15 meters. In addition, a set
of manual controls was provided for
use when desired.
This homebrew radio uses a varia-
tion on this theme. Ten 1-MHz bands
were to be covered. The lower seven
bands use seven separate fixed-tuned
pi or pi-L networks, with no provision
for manual control. For the three up-
per HF bands (24, 28 and 29 MHz)
there is one manually tuned pi-
network with front-panel controls and
no provision for fixed-tuned operation.
This arrangement suits my operating
habits ideally. It improves perfor-
mance by avoiding shorted turns on a
single tank coil, and using more-ap-
propriate components and better L/C
ratios. For contesting, the panel con-
trols are tuned for 28 MHz, and the
result is equivalent to fixed-tuned op-
eration on all bands.
A drive control on the panel of the HF
front-end is adjusted for the correct
amount of ALC compression—mea-
sured at the 8072 grid—or for proper
drive level to an external 1.5-kW ampli-
fier. ALC also runs from each driver and
each kilowatt amplifier back to the
corresponding front-end panel, with
ALC metering at the transceiver. There
has been much written about distortion
caused by ALC circuits. It is true, but
applies only to improperly designed and
improperly adjusted ALC circuits. This
radio applies ALC control voltage to
dual-gate MOSFETs, as is common. An
IMD problem can occur if too much gain
reduction is attempted by varying the
bias of a MOSFET. An extreme case
occurs when ALC is improperly used to
reduce the gain of a radio to drive a low-
input-level transverter. At the K5AM
station, the gain of each front-end panel
is adjusted to obtain 3-dB ALC com-
pression, a moderate amount. A single
MOSFET may be able to handle a 3-dB
gain reduction. To ensure the cleanest
signal possible, however, ALC voltage
is applied to three cascaded MOSFET
stages, so that each reduces the gain by
only 1 dB.
The drive control could be used to
reduce the HF panel output to 5 W for
QRP work, which presents refreshing
challenges. Using the drive control,
however, involves the usual touchy ad-
justment problem and band-change
inconvenience. The HF front-end sec-
tion has a separate QRP level knob, and
a switch to enable it. The knob adjusts
an output detector that feeds the ALC
circuits. The drive control is used to
obtain the correct amount of ALC com-
pression, read directly on the trans-
ceiver panel. It is not as critical on CW
as on SSB, but too much ALC compres-
sion will distort the CW waveform.
The QRP switch is part of the AMP
(amplifier) switch, with three positions:
LP (low power), BF (barefoot, meaning
transceiver only) and PA (external am-
plifier enabled). The 6 and 2-meter
front-ends have similar switches. The
entire station is controlled from the
front-end panels conveniently located
near the operator. On HF, the three
positions instantly provide 5 W, 200 W
or 1500 W output. The LP position can
be used with the level set for 150 W for
low-power contesting.
There is a panel switch for a second
receive antenna, which is a necessity on
160 meters and useful on other bands,
as well. The receiver circuits are fully
protected—using reed relays—against
RF energy picked-up by the auxiliary
antenna while transmitting on the
main antenna. A jack on the rear panel
allows the auxiliary receive antenna
relay to be controlled by a button on the
desk, or by a foot switch.
Summary
This article gives a general descrip-
tion of a high-performance homebrew
transceiver built with regard only to
the best performance. It shows what
can be accomplished without micro-
processors, synthesizers and PLLs.
Notes
1 M. Mandelkern, K5AM, “A High-Perfor-
mance AGC System for Home-Brew
Transceivers,” QEX , Oct 1995, pp 12-22.
2 M. Mandelkern, K5AM, “Evasive Noise
Blanking,” QEX , Aug 1993, pp 3-6.
3M. Mandelkern, K5AM, “A Homebrew
Contest Station,” NCJ, July/Aug 1998,
pp 12-13.
4 M. Mandelkern, K5AM, “A Sensitive Inte-
grating Squelch,” QST , Aug 1988,
pp. 27-29.
5M. Mandelkern, K5AM, “Design Notes for ‘A
Luxury Linear’ Amplifier,” QEX, Nov 1996,
pp 13-20.
6M. Mandelkern, K5AM, “A Low-Drive, High-
Power All-Band Tetrode Linear Amplifier,”
CQ, July 1990, pp 60-65.
7M. Mandelkern, K5AM, “A Luxury Linear,”
QEX, May 1996, pp 3-12.
8M. Mandelkern, K5AM, “The AMSAFID: An
Automatic Microphone Switcher Amplifier
Filter Integrator Distributor,” QST , Nov
1995, pp 47-49.
 A High-Performance
Homebrew Transceiver: Part 2
Let’s begin our look inside this transceiver with
the IF board. It contains the IF amplifier, AGC,
noise-blanker and RF speech clipper circuits.

By Mark Mandelkern, K5AM

P
art 1 gave a general description of
the K5AM homebrew transceiver,
built for serious DX work and
contest operating.1 This article gives
complete circuit details for the IF board.
While obtaining gain at 9 MHz is routine,
care is taken to ensure optimal AGC
performance, non-crunch noise blanking
and QRM-piercing transmit audio.
The IF board in any radio is perhaps
its most crucial component. While the
other parts of the radio are certainly
important, converting to the IF—and
later to audio—are essentially trans-
1Notes appear on page 8.
5259 Singer Rd
Las Cruces, NM 88005
k5am@roadrunner.com
lation functions. The IF board must
provide most of the gain; it must
control this gain automatically and
smoothly, provide all of the selectivity,
blank out noise and also process the
transmitted signals.
IF-Board Features
The IF board is shown in Fig 1. Relay-
switching of the SSB and CW filters
avoids diode-generated distortion.
This also reduces “blow-by,” resulting
in high ultimate attenuation. Sharp
filters are used near the IF amplifier’s
output, as well as at the input. This
improves the signal-to-noise ratio.
Other features include:
• A high-performance, no-pop, no-
click, hang-AGC circuit2
• A non-crunching noise blanker3
• A sensitive integrating squelch cir-
cuit (see Reference 2)
• RF speech clipping for transmitted
SSB with a punch
• Special operating features (discussed
in Part 1)
The Circuit
A general description of the IF board
has been given in Part 1. The IF board
consists of the IF amplifier, noise
blanker and AGC section. The block
diagram in Fig 2 shows the arrange-
ment of the three sections as well as
the individual stages of the IF amp-
lifier. An explanation of the terminal
designations is given in Table 1. The
attenuator pads at the input and
output of each crystal filter provide
proper out-of-band terminations,
Sept/Oct 1999 3
reduce distortion and improve the
skirt and ultimate selectivity. The
9-dB pad before the second SSB filter
compensates for the additional loss of
the CW filter, and assures proper
performance of the RF speech clipper
(see below). Fig 3 shows the IF-
amplifier schematic diagram. The
various control lines are provided by
the logic board.
RF Speech Clipper
A general description of the RF
speech clipper, including microphone
calibration and operating instruc-
tions, was given in Part 1. The clipper
has three sections. The DSB amplifier,
Q8, controls the amount of RF clipping
by adjusting the input level to the IF
amplifier, using the CLIPPING control
on the front panel. When receiving, or
when transmitting in the carrier
modes (CW, AM, FM), the DSB
amplifier (Q8), the clipping meter
amplifier (Q9) and the SSB output
buffer (Q10) are disabled by control
line βSSB. The clipper proper consists
of the two diodes in the drain circuit of
Q4. The third section of the RF speech-
clipping system provides metering on
the front panel. It consists of amplifier
Q9 and two op amps; see Figs 2 and 5.
Clipping-meter amplifier Q9 parallels
Q4, with a detector at the output
rather than clipping diodes. This
detector provides an output that
tracks the amount of conduction in the
clipping diodes. The detector output is
amplified by a peak-indicating circuit,
shown in Fig 5, and displayed on meter
M1. The peak-indicating circuit gives
the operator true indications of RF
speech clipping level.
One special element in the clipper
circuit is crucial for proper perform-
ance. Credit for it is due to Robert
Sherwood, NC0B, and it was commun-
icated to me by Paul Kollar, W8CXS.
This crucial element is the isolation
between the clipping diodes and the
following SSB filter. Rob conducted
extensive tests on the CX7 clipping
circuit—which had no isolation. He
found that the clipping circuit caused
distortion because it was an inad-
equate termination for the filter. The
necessary isolation may be provided by
either a pad or a buffer stage. In this
circuit, an additional 9-dB pad is
added ahead of the SSB filter. The pad
also conveniently compensates for
the extra loss of the CW filter. The
resultant 12 dB of attenuation
provides a stable load for clipping
stage Q4 as well as proper termination
for the filter.
4 QEX
Noise Blanker This precise information can then be
The noise blanker has been relayed to the local power company.
described in Reference 3. When An AGC Update
not in use, the noise blanker is
More experience on the 160-meter
switched completely out of the signal band with extremely weak DX sig-
path by relay K1, at the input to the IF
nals under high-atmospheric-noise
amplifier. This avoids any possible conditions has led to a further AGC
signal degradation by the blanker gate
improvement. Under such conditions,
diodes. The noise-channel gain is set by it is best to use zero hang time, in
the front-panel BLANKER control
the FAST mode. The modification is
through control line δN and the op-amp simple. Look at Fig 8 in Reference 2,
circuit shown in Fig 5.
disconnect the Fast Adjust trimmer
A tap on the noise amplifier provides from terminal F of the AGC HANG TIME
a signal to feed an external monitor
switch; then connect terminal F to
scope via the SCOPE (output) jack on the collector of the discharge
the rear panel. A scope is very useful
transistor, Q304.
for locating nearby power-line noise.
Attempts to locate such noise by Construction
watching the S-meter indication while The general method of construction
rotating the antenna are doomed; the was described in Part 1, where the
meter only shows the aggregate noise need for careful shielding and lead
peak. A scope, on the other hand, can filtering was emphasized. Each power
be used to determine the direction of and control lead to the IF board passes
the individual noise sources, since through a π-section filter (two 10-nF
each noise source tends to have a bypass capacitors and a 1-mH choke)
distinctive oscillograph signature. and a 1-nF feed-through capacitor.
Table 1
Terminals on the diagrams are labeled according to the function of the signal or
control line. The characteristic of each line is indicated by an initial Greek letter,
according to the following scheme:
α alpha Line keyed to ground, such as PTT and Keyline
β beta Control line that switches nominally from +15 to –15 V
δ delta dc control line
ι iota Local oscillator injection voltage for a mixer
ρ rho Rheostat or potentiometer control line
σ sigma Signal
µ mu Control line that switches nominally from 0 to –15 V
φ phi RF voltage obtained from an oscillator
Fig 1—Top view of the IF board in the K5AM homebrew transceiver. Several
shielding-compartment covers have been removed for this photo. The IF amplifier
is at the top, with signals traveling left to right. The noise blanker and AGC
sections are below, on the left and right, respectively. (Photo by Lisa Mandelkern.)
 In passing signals to and from the
board, it is important to avoid any
signal leakage. Teflon-insulated mini-
ature coax, type RG-178B, facilitates
good connections because it allows
safe use of a soldering iron. Simply
drill a hole (# 51 bit) through the
circuit board, remove the cable’s outer
jacket and solder the braid to both
sides of the board.
The board’s bottom surface is shown
in Fig 6. Each IF stage is built dead-
bug style on a sub-board mounted
inside its separate compartment.
Power and control leads connect to the
feed-through capacitors on the floor of
the main board. Coax cables run
through the floor and up to the sub-
boards. The AGC sub-board is
constructed using wire-wrap methods
on perf-board.
Alignment
The IF amplifier operates at a gain
of 90 dB overall, between terminals
σ9R and σ9R1. First, the no-signal
AGC line level is set to 2.0 V by the Rx
Gain Adjust trimmer in the AGC
section (see Fig 7 in Reference 2).
Then, with the AGC off, and an input
at σ9R of –100 dBm, adjust the IF Gain
Adjust trimmer at Q7 for an output at
σ9R1 of 200 mV (P-P, –10 dBm).
The output level of the IF amplifier
with the AGC on is set by the AGC
Adjust trimmer (see Fig 6 in Refer-
ence 2). With a –70 dBm signal at
terminal σ9R, adjust the trimmer for
200 mV (P-P) output at terminal σ9R1.
The S-Meter Adjust trimpot, shown
in Fig 9 of Reference 2, sets the full-
scale S-Meter reading at 100 dB above
the AGC threshold. The S-Meter
threshold is the same as the AGC
threshold. The S-Meter is calibrated
directly, from 0 to 100 dB above the

Fig 2—IF board block diagram. This

shows the individual stages of the
9 MHz IF amplifier. The AGC section has
been described in Reference 2, and the
noise blanker in Reference 3. Potentio-
meters labeled in all capital letters are
front panel controls; others are circuit-
board trimmers for internal adjustment.
The transistors shown are all small-
signal, dual-gate, VHF-type MOSFETs,
except Q13, a strong bipolar, and the
small bipolar emitter followers labeled
“EF.” The gain of the Q13 stage
compensates for the additional loss of
the CW filter. The triangles indicate op
amps. When receiving, the input is at
terminal σ9R; the output is at terminal
σ9R1. For transmitting SSB, the input is
at terminal σ9DSB. For transmitting in
the carrier modes (CW, AM, FM), the
input is at terminal σ9C. Terminal σ9T is
the transmitting output.

Sept/Oct 1999 5
Fig 3—IF amplifier schematic diagram. Each resistor is a 1/ 4-W carbon-film type. The diodes are all small-signal silicon types,
such as 1N4148. The unlabeled coupling and bypass capacitors are all 10-nF disc ceramic types. Also, each control and power
terminal has a bypass capacitor that is not shown. The trimmer capacitors are 5 to 18-pF miniature ceramic types. The crystal
filters, salvaged from an irreparable CX7, have an impedance of 220 Ω; the attenuators and transformers are designed
accordingly. Potentiometers labeled in all capital letters are front-panel controls; others are circuit-board trimmers for internal
adjustments. Certain other simplifications have been incorporated in this schematic to save space and improve clarity. In the
MOSFET stages, the gate-2 isolation resistors, source resistors and drain-circuit decoupling resistors are all unlabeled; each
is 100 Ω. Note that there is no bypass capacitor directly at gate 2; the resistor serves as a parasitic suppressor, which is more
effective than a ferrite bead. As a further simplification, circuit elements for stages Q2 through Q7 that are identical to those of
the preceding stage are omitted. An obvious exception: The drain circuit of Q5 is the same as that of Q1. Not shown on this
diagram is the additional filtering at each terminal; it is described in the text under “Construction.” The noise blanker was
described in Reference 3; some additional details are given in Fig 4. The AGC section was described in Reference 2; the AGC
amplifier and detector circuits, located in the IF-amplifier section, are shown in Fig 6 of Reference 2. The clipping-level circuit
requires only a single lead to the front panel, and provides a range of –0.6 to +2.0 V for gate 2 of Q8; this allows a 25-dB gain
variation. The tuned circuits at the inputs to stages Q1 and Q5 provide a voltage gain. On the other hand, the inductors in all
the drain circuits are untuned, and function simply as chokes. The R/T line disables Q5 through Q7 during transmission; it
switches from 0 to –15 V. The squelch circuit (part of the AGC section) disables Q7 by control line βSQ. The
1-µF capacitor in this muting circuit is a monolithic ceramic type.

K1—Low-loss, high-isolation RF DIP
relay; DPDT, 12-V dc. Omron #G5Y-
254P-DC12; Digi-Key #Z704 (Digi-Key
Corp, tel 800-344-4539, 218-681-6674,
fax 218-681-3380; www.digikey.com).
K2-K6—Low-loss, high-isolation RF DIP
relay; SPDT, 12-V dc. Omron #G5Y-1-
DC12; Digi-Key #Z724.
L1—0.7 µH, 15 turns #26 enameled wire
on a T-37-6 powdered-iron toroidal core.
L2-L8—14 µH, 16 turns #26 enameled
wire on an FT-37-61 ferrite toroidal core.
L2 and L6 are tapped 4 turns from the
cold end.
Q1-Q10—Small-signal VHF-type dual-
gate MOSFET. Type 3N140 is used here,
but any similar type may be substituted.
Type NTE 221 is available from Hosfelt
Electronics Inc, tel 800-524-6464, 740-
264-6464, fax 800-524-5414, fax 740-264-
5414.
Q11-Q12—2N2222A.
Q13—2N5109.
T1-T3—8 bifilar turns #26 enameled wire
on an FT-37-61 ferrite toroidal core.

6 QEX
Fig 4—Noise blanker schematic diagram
(partial). The schematic for the main
portion of the blanker is shown in Fig 5
of Reference 3; this diagram completes
it. To correlate the two diagrams,
augment Fig 5 in Reference 3 by
labeling the terminals σ9B1 (upper left),
σ9N1 (lower left) and σ9B2 (right). Also,
label the transformers T1 to T3, left to
right. For general notes on the
schematic, refer to the caption for Fig 3.
Noise pulses, from the tunable noise
channel on the RF board, enter the
blanker section at terminal σ9N. Signals
(with noise) enter at terminal σ9B and
leave (noise-free!) at terminal σ9B2. The
blanker is switched on by control line
βN; the noise-amplifier gain is set by
control line δN. The 25-kHz bandwidth
filter is salvaged from an irreparable
CX7; it has an impedance of 3200 Ω, obtain desired coupling; two turns used The following refer to Fig 5 in Reference
and the matching inductors are here. Alternatively, cover a 1-inch piece 3, as indicated at the beginning of this
designed accordingly. There is a similar of #16 bare wire with Teflon-sleeving caption.
filter in the noise-channel amplifier on and wrap #24 bare wire completely over
the RF board. These filters have poor a 1/ 2-inch length of the Teflon. Adjust by T1—8 bifilar turns #26 enameled wire on
ultimate attenuation; thus the noise sliding the Teflon, with wrapping, an FT-37-61 ferrite toroidal core.
amplifier here uses four tuned circuits partially off the larger wire. The result is T2-T3—Tapped winding, 12 bifilar turns
to narrow the response. a homebrew piston trimmer. #26 enameled wire over full core length
L1-L6—14 µH, 16 turns #26 enameled of an FT-37-61 ferrite toroidal core.
C1—Fractional-pF “gimmick” capacitor. wire on an FT-37-61 ferrite toroidal core. Untapped winding, 24 turns #26
Two 1-inch pieces of wire, one #18 with Taps from the cold end: L1, L4, L6, 2 enameled wire over the full length of the
Teflon sleeving, one #22, bare. Twist to turns; L5, 4 turns. first winding.

Sept/Oct 1999 7
 AGC threshold. The AGC threshold at
terminal σ9R is –100 dBm. The
threshold is –110 dBm at the 40-MHz
input on the RF board, and nominally
–130 dBm at the antenna. The MDS
(bandwidth = 2 kHz) at terminal σ9R
is –116 dBm. The MDS at 40 MHz is
–126 dBm, and nominally –138 dBm
at the antenna.
Setting the Clip-Meter Threshold
trimmer requires a means of precisely
determining the point at which clipping
begins. This is done with a dual-trace
scope by monitoring the IF amplifier
signal level at the input to Q4, and at
the transmit-output terminal, σ9T.
With the front-panel CLIPPING control
set to minimum, the two-tone audio test
oscillator and MIKE controls are set to
obtain 1 dB of compression at the
output; the Clip Meter Threshold
trimmer is then adjusted for a 10%
meter indication. The CLIPPING control
is now advanced to obtain a 20-dB
increase at the input to Q4, and the Clip
Meter Adjust trimmer in the control
section, Fig 5, is adjusted for a full-scale
Fig 5—Control-section schematic diagram. Each op amp is one section of an indication. Finally, a meter calibration
LM324N, powered from the ±15-V rails. The 1-µF capacitor in the peak-hold circuit chart is made, so the operator may
is a monolithic ceramic type. The blanker gain circuit, requiring only a single lead select a desired degree of RF com-
to the front panel, provides a range of –3 to +2 V for the gate-2 leads of the noise- pression. For normal use, 3 dB is
amplifier MOSFETs. The circuit also disables the noise amplifier when not in use,
by means of control line βN. For general notes on the schematic, refer to the adequate and pleasant sounding. For
caption for Fig 3. This control section occupies a portion of the AGC sub-board. extreme conditions, 10 dB may be used
effectively.
When transmitting, the DSB input
level at terminal σ9DSB is nominally
200 mV P-P; the carrier input level at
terminal σ9C is 500 mV P-P. The SSB
Level and CW Level trimmers are
adjusted to obtain 200 mV P-P output
at terminal σ9T. The AGC sub-board
reduces the gate-2 voltage of the
MOSFETS to about 1.2 V during each
transmission; this can be varied as
required by a trimmer in the AGC
section (see Fig 7 in Reference 2).

Summary
This article gives a complete de-
scription of the 9-MHz IF board in a
high-performance homebrew trans-
ceiver. The board is designed for opti-
mal performance; it includes sharp
Fig 6—Bottom view of the IF board. Effective filters are installed at each terminal, filters for SSB and CW, a high-perfor-
and coax cables are soldered directly to the double-sided circuit board (see text). mance hang-AGC circuit, a non-
To minimize connector troubles, the board is hard-wired to the radio; a 12 inch crunching noise blanker and a power-
long bundle of wires and cables allows the board to be easily lifted and serviced.
The IF amplifier has unusually high gain, and operates at an unusually low signal
fully effective RF speech clipper.
input level. This is done as a gain-distribution method to achieve high dynamic
range. The stages ahead of the sharp SSB and CW crystal filters can operate at References
relatively low gain and are not easily overloaded. The stages after the filters are 1. M. Mandelkern, K5AM, “A High-Perfor-
shielded from signals outside the filter passband. The high gain of the IF amplifier mance Homebrew Transceiver: Part 1,”
makes it vulnerable to stability problems. The low signal input level makes it QEX, March/April 1999, pp 16-24.
vulnerable to entrance of BFO energy. These vulnerabilities create the necessity 2. M. Mandelkern, K5AM, “A High-Perfor-
for exceptional filtering and shielding of the IF strip and account for the large mance AGC System for Home-Brew
number of RF chokes, bypass capacitors and feed-through capacitors on the Transceivers,” QEX, Oct 1995, pp 12-22.
bottom of this board. The careful filtering and shielding also contributes to the 3. M. Mandelkern, K5AM, “Evasive Noise
effective ultimate attenuation of the filters. (Photo by Lisa Mandelkern.) Blanking,” QEX, Aug 1993, pp 3-6.

8 QEX
 A High-Performance
Homebrew Transceiver: Part 3
Mixing, premixing, dual receiving, IF shift and
CW offset—all these topics are covered in this
description of the 40-MHz RF board.

By Mark Mandelkern, K5AM

T
o many builders, the RF board
in any radio is the most inter-
esting. It includes the receive
mixer, a prime component in deter-
mining the dynamic range of the re-
ceiver. The RF board in this radio also
contains the transmit mixer, BFOs,
premixers for LO injection, LO ampli-
fier and the tunable noise channel.
The basic radio covers 40-39 MHz;
three front-end sections covering the
ham bands from 160 to 2 meters are on
separate panels. This RF board, part
5259 Singer Rd
Las Cruces, NM 88005
k5am@roadrunner.com
of the main panel, establishes the
40 MHz to 9 MHz transitions.
Part 1 gave a general description of
the K5AM homebrew transceiver,
built for serious DX work and contest
operating. 1 Part 2 described the IF
board. 2 The RF board described in this
article is shown in Fig 1.
Features
The main features of the RF board are:
• Balanced JFET receive mixer
• Balanced MOSFET transmit
mixer
• Premixing for the IF-shift circuit
• Adjustable-waveform keying
circuit
1Notes appear on page 50.
• Offset oscillator for panel-adjust-
able CW offset
• Tunable noise channel for the
noncrunching blanker
Circuit Description
A general description of the RF board
has been given in Part 1. The block dia-
gram in Fig 2 shows the arrangement of
the various RF-board stages. Figures 3,
4, 5, 6, 7, 8, 9, and 10 show schematic dia-
grams of the different RF-board sections.
BFO
See Reference 1 for a discussion of
the mixing scheme. To provide IF shift
(IFS) operation, the BFOs are pre-
mixed along with the PTOs to obtain
Nov/Dec 1999 41
the LO injection frequencies for the
receive and transmit mixers. BFO fre-
quencies are also routed to the AF
board to provide LO injection for the
product detector and balanced modu-
lator.
The BFO circuit is shown in Fig 3.
Although the BFO is tunable, a wide-
range VXO circuit (with its inherent
drift) was rejected in favor of a simple,
adjustable crystal oscillator. A nar-
rower tuning range results, but it is
more than adequate for normal oper-
ating. For simplicity, two separate os-
cillators are used: one for USB/CW,
the other for LSB. The components are
selected to obtain the required IFS
range. For normal SSB use, ±500 Hz is
adequate.
The CW filters on the IF board have
a center frequency of 8815.7 kHz.
Thus, to provide full CW tone range for
receiving, the USB oscillator must
tune downward at least 800 Hz.
The circuit is somewhat unusual, as
the variable tuning element is in the
feedback loop. One would expect this
to cause unwanted oscillator output-
level changes as the BFO is tuned, but
level changes also occur when pulling
the crystal: The output drops with in-
creased capacity across the crystal.
With the varactor diode in the feed-
back loop, the feedback increases with
increasing diode capacity; this coun-
teracts the aforementioned effect. The
result is much less output variation
than that with the varactor across the
crystal. The output is essentially con-
stant over the range normally used in
operation.
The transmit-frequency trimpot
adjustments are critical. The BFOs
must be positioned at the proper
points on the SSB filter passband skirt
to obtain the best transmitted-audio-
frequency response. The initial set-
tings have held within 10 Hz during
the last seven years of operating; no
doubt, the choice of quality crystals
was a factor in this happy situation.
The simple IFS circuit allows BFO
tuning while receiving and automatic
return to the proper frequency when
transmitting. When receiving, the
µIFS control line is nominally –15 V.
The IF-shift control on the front panel
may then vary the voltage at terminal
IFS1 from 0 to –15; this tunes the
varactor diode VC1 in the oscillator
tank circuit. At the same time, the
transistor in the IFS circuit is cut off,
and the USB XMIT SET trimpot has no
effect. When transmitting, the µIFS
line shifts to zero. Now the panel IFS
control has no effect, but the transistor
42 QEX
Fig 1—Top view of the RF board in the K5AM homebrew transceiver. From right to
left, the sections are: the BFO, BFO mixer, PTO mixer, LO amplifier and main
mixers and tunable noise channel. Several section-shielding covers have been
removed for the photo.
is turned on, allowing the USB XMIT SET
trimpot to set the BFO to the proper
frequency. When the CW offset spot-
ting button is pressed, the µIFS control
line also shifts to the transmit state to
ensure proper offset adjustment. The
Zener diode in the IF-shift circuit is
needed because the control line shifts
to –15 V only approximately. In prac-
tice, the op amps on the logic board that
drive the control lines provide about
–14 V. According to the op-amp data
sheet, only –13 V can be assumed. The
Zener holds the transistor’s emitter
voltage below –12 V, so the control line
easily keeps the transistor turned off.
The 10-kΩ resistor at the IFS1 termi-
nal provides a load to ensure con-
duction in the diodes. Without this
resistor, one may observe floating and
drifting of the bias voltage applied to
the varactor diode.
BFO Mixer and Offset Oscillator
In this mixer, the BFO frequencies
are mixed with the fixed 43.1-MHz
master oscillator. The circuit is shown
in Fig 4. The output of this mixer is
nominally 34.285 MHz, shifting
slightly with sideband selection and IF-
shift operation. For CW-offset opera-
tion, this mixer is switched off (ignore
the BFO frequencies here), and the
panel-adjustable offset oscillator is
used instead. Considerable tempera-
ture compensation is used to cancel the
drift of the varactor diode in the offset
oscillator.3 The offset spotting mixer is
enabled by the CW SPOT push-button on
the front panel, which also enables both
the normal BFO mixer and the offset
oscillator. The resulting audio tone rep-
resents the actual offset; it is fed to the
headphones by the AF board. A panel
control sets the headphone level.
In the Signal/One CX7, the 43.1-MHz
oscillator was tuned by a varactor
diode, adjustable from the front panel.
This was used in conjunction with an
HF (100-kHz marker) calibration oscil-
lator—itself adjusted using WWV—to
calibrate the radio on each band. This
resulted in some drift, destroyed the
CW-offset settings and made band
changing very inconvenient. Here the
oscillator is fixed; the highest-quality
crystals are used in the front-end sec-
tions for each band. This results in
maximum convenience and read-out
accuracy within 100 Hz. An important
feature is the use of 10 separate oscilla-
tors on the HF panel in lieu of a crystal
switch, which can cause frequency er-
rors, instability and even total failure.
PTO Mixer and Dual-Receive Circuit
This premixer produces the final
LO-injection frequencies for the main
receive and transmit mixers. The cir-
Fig 2—RF board block diagram.
Terminal σ40 is the 40-MHz input/output
terminal from the rear panel. When
receiving, the output is at terminal σ9R.
For transmitting, the input is at terminal
σ9T. Potentiometers labeled in all-
capital letters are front-panel controls.
An explanation of the terminal
designations is given in Part 2, Table 1.
The control lines are provided by the
logic board.
Nov/Dec 1999 43
cuit is shown in Fig 5. To minimize spu-
rious responses, three band-pass filters
are used for the LO injection: two here
and a third in the LO-amplifier section.
The dual-receive feature is very
simple; it is not equal to the much more
elaborate subreceivers found in some
contemporary commercial radios. The
two PTOs each produce an LO injection
frequency, so the receiver responds to
two different frequencies at the an-
tenna. Although simple, it can be very
effective in certain DX split-frequency
situations. With appropriate switching
and combining of the external front-end
sections, it can also be used to monitor
two VHF DX calling frequencies simul-
taneously, or an HF and a VHF fre-
quency. Control line µD energizes the
DUAL RECEIVE control on the front panel.
This applies gain-control voltage to the
PTO buffers, resulting in adjustable
balance control.
This section also contains a diode
switch that routes the appropriate
PTO frequencies to the counter on the
front panel. Signals are selected by the
readout-control line βR from the logic
board. The control lines µIA and µIB
control the PTO buffers and thus the
received signal. This allows some flex-
ibility. For example, PTO B can be
read out and tuned while listening to
a signal on PTO A.
LO Amplifier
The grounded-gate balanced JFET
mixer used for the receiver has LO in-
jection applied to the JFET source ter-

Fig 3—BFO schematic diagram. Except as noted, each resistor is a 1/4-W, carbon-film type. All trimpots are one-turn miniature
types, such as Bourns type 3386; Digi-Key #3386F-nnn. (See Reference 8.) The unmarked coupling and bypass capacitors are
all disc ceramic types; 1 nF in circuits above 30 MHz, 10 nF below. Also, each control and power terminal has a bypass
capacitor that is not shown. Except as noted, the trimmer capacitors are Erie series 538, a sturdy 9-mm-diameter type that will
withstand extensive testing and adjustment. These trimmers are available on the surplus market. (See Reference 9.) Xicon
7-mm ceramic trimmers are possible substitutes. (See Reference 10.) Except as noted, other capacitors are silver-mica types.
Electrolytic capacitors are tantalum. Values of RF chokes (RFC) are given in microhenries. Potentiometers labeled in all-
capital letters are front-panel controls; others are circuit-board trimpots for internal adjustment. All coils are wound with #26
enameled wire. The control signals are provided by the logic board. Some part designators differ from QEX style so they
conform to the author’s diagrams.
MOSFETs are small-signal VHF dual-gate types. Type 3N140 is used here, but any similar type may be substituted. Type NTE
221 is available from Hosfelt (Reference 11). Except where otherwise indicated, the diodes are all small-signal silicon types,
such as 1N4148, and the bipolar transistors are type 2N2222A (NPN) or 2N2907A (PNP).
For clarity and to save space, LSB circuits (which are identical to the USB circuits) are indicated only as blocks. The only
variation concerns the IFS panel control, which is a dual control (single shaft). The LSB section is wired so that the clockwise
indicator arrow points towards ground. This is done so that the control functions the same on either sideband with respect to
received audio passband.
L1—Small, molded RFC. Select, if panel control. Used here: USB, none; Y1—Fundamental crystal, USB/CW,
needed, to adjust oscillator range; LSB, 2.2 kΩ. 8816.5 kHz, type CS-1, ICM #433375-
1.6 µH used here. VC1—Varactor diode, nominal 33 pF. 8.8165. Socket type FM-2, ICM #035007
R1—Select as needed to obtain center Motorola type MV2109. NTE type 614 (Reference 12).
BFO frequency at center position of (see Reference 11). Y2—Same as Y1, except LSB, 8813.5 kHz.

44 QEX
minals. It requires more LO power than
other configurations. The transmit
mixer also requires considerable LO
power. The circuit for the amplifier that
provides it is shown in Fig 6; the ampli-
fier is operated in the linear region. The
LO power for the two mixers is provided
through adjustable trimpots.

Receive Mixer
The second mixer in a dual-conver-
sion receiver is the most important
single stage in the radio (see the discus-
sion in Part 1). The circuit is shown in
Fig 7. The singly balanced JFET mixer
results in excellent dynamic range. The
circuit is based on ideas found in a
Siliconix manual.4 There is no balance
control; best mixer performance is
achieved with a matched pair. A pair
was selected from a batch of 10 devices.
Best IMD performance and stability
are obtained with the common-gate
(grounded-gate) configuration.5 In this
configuration, the manufacturer sug-
gests from +12 to +17 dBm of LO power
for best performance. The dc source bias
is set at 2 V, as measured at test point
TP1. This dc bias is half the –4 V gate-
cutoff bias of the selected JFETs; this
allows full LO voltage swing without
cut-off or gate conduction. The LO in-
jection level is set for a dc reading of 1.1
V at TP2 (on a 10M Ω meter). This rep-
resents about 3.1 V P-P, or +14 dBm.
Ahead of the receive mixer and fol-
lowing the transmit stages are the
main 40-MHz band-pass filter and the
TR relays. These are shown in Fig 10.
Mixer gain is only about 3 dB. A strong

Fig 4—BFO mixer schematic diagram.

For general notes on the schematics,
refer to the caption for Fig 3. DBMs and
MMICs may be obtained in small
quantities directly from the
manufacturer (Reference 13).
C1—Glass piston trimmer, 1-5 pF
L1—Master oscillator coil, 450 nH,
T-37-17 powdered-iron toroidal core,
17 turns. Adjust turns to pull the
crystal to the proper frequency.
Compressing the winding or
expanding to fill the core greatly
affects the inductance.
(See Reference 14.)
L2—Offset oscillator coil, 500 nH, same
as L1 except 18 turns.
L3, L4—570 nH, T-37-6 powdered-iron
toroidal core, 14 turns.
VC1—Varactor diode, nominal 6.8 pF.
Motorola type MV2101. NTE type 610
(see Reference 11).
Y1—Master-oscillator crystal, third
overtone, 43.1 MHz, type CS-1, ICM
#471360-43.1 (see Reference 12).
Y2—Offset-oscillator crystal, same as
Y1 except 34.2835 MHz; ICM #471360-
34.2835 (see Reference 12).

Nov/Dec 1999 45
bipolar amplifier follows the receive
mixer. It overcomes filter losses at the
input to the IF board and provides
enough signal at the first IF amplifier
to ensure high IF sensitivity.
Transmit Mixer
The balanced MOSFET transmit
mixer circuit is shown in Fig 8. The
mixer is followed by several buffers and
careful filtering. There is no impedance
matching at the mixer input, since high
mixer gain is not required here. The
voltage level on the σ9T input line is
sufficient for the mixer gates. Keeping
the level from the IF board high on
this line minimizes the effect of any
possible carrier leakage into the cir-
Fig 5—PTO mixer schematic diagram. For general notes on schematics, refer to the caption for Fig 3.
BPF1, BPF2—Band-pass filter,
bandwidth 2 MHz, center 30.685 MHz.
The coils are each 520 nH, T-37-10
powdered-iron toroidal core, 13 turns. proper inductance before assembly
Formulas predict 14.4 turns, but coils
46 QEX
cuit. The total residual hum, noise and
carrier on the transmitted SSB signal
is more than 65 dB down.
CW keying is accomplished in the
buffer stages. The keying waveform is
adjustable using two trimpots in the
simple timing circuit shown in Fig 9.
The make and break trimpots act inde-
pendently. The bias trimpot is needed
because of the variation in individual
MOSFET cutoff voltages. A fixed bias
level high enough to accommodate any
MOSFET would result in an unwanted
lag between the key closure and the
start of the transmitted element. The
bias is set just high enough to achieve
full cutoff. The timing trimpots are set
for 2- to 3-ms make and break times.
on these cores often have more
inductance than indicated by the
manufacturer’s data. Adjust turns for L1—FT-37-43 ferrite toroidal core, 12
To ensure the absence of key clicks, the
adjustment is made while monitoring
with a receiver.
Tunable Noise Channel
The noncrunching noise blanker has
been described previously in QEX.6 It
consists of several parts. The tunable
noise channel is situated here on the RF
board. The noise amplifier, pulse detec-
tor, signal channel and blanker gate are
located on the IF board, described in
Part 2. The circuit for the tunable noise
channel is shown in Fig 10. The main
40-MHz filter and the 40-MHz TR reed
relays are also shown on this diagram.
The MOSFET noise preamplifier,
with its high input impedance, taps
and again in-circuit with a sweep
generator.
turns, tap 3 turns from low end.

Fig 6—LO-amplifier schematic diagram. For general notes on the schematics, refer
to the caption for Fig 3.
BPF3—Same as BPF1 (see Fig 5).
L1—2 µH, T-37-10 powdered-iron
toroidal core, 28 turns.
Fig 7 (below)—Receive-mixer schematic
diagram. For general notes on the
schematics, refer to the caption for Fig
3. The JFETs comprise a matched pair.
Resonance in the mixer drain circuit is
with about 80 pF. The tuning capacitor
may be made up of convenient
components; for example, a 62-pF
silver-mica capacitor in parallel with a 9
to 35-pF ceramic trimmer. The 64:1
transformer in the drain circuit
presents a load of 1600 Ω to each drain,
which is optimum for best IMD
performance (see Reference 4).
LPF1—Low-pass filter, cutoff frequency
13 MHz. Inductors, 1 µH, T-37-6
powdered-iron toroidal core, 18 turns.
T1—Mixer input transformer, FT-37-61
ferrite toroidal core, 8 trifilar turns.
T2—Mixer output transformer, T-50-6
powdered-iron toroidal core. Primary,
4 µH, 32 turns, wound over full length
T1—FT-37-61 ferrite toroidal core, 8 of core, center-tapped. Secondary,
bifilar turns. 4 turns, close-wound at the center of
the primary winding.
T3—FT-37-43 ferrite toroidal core,
6 bifilar turns.
Nov/Dec 1999 47
Fig 8—Transmit-mixer schematic diagram. For general notes on the schematics, refer to the caption for Fig 3.
BPF1-BPF2—Band-pass filter, piston trimmer capacitor; cover a 1-inch L1—FT-37-61 ferrite toroidal core,
bandwidth 2 MHz, center 39.5 MHz. piece of #16 bare wire with teflon- 11 turns, tap 2 turns from ground end.
Inductors, 1 µH, T-37-6 powdered-iron sleeving, then wrap #24 bare wire T1—FT-37-43 ferrite toroidal core,
toroidal core, 18 turns, center-tapped. completely over a 1/ 2-inch range of the 6 trifilar turns.
C1, C2—Nominally 0.5 pF; adjust for teflon. Adjust by sliding the teflon, with T2—FT-37-61 ferrite toroidal core,
desired passband with sweep wrapping, partially off the larger wire. 6 trifilar turns.
generator. Homebrew fractional-pF

Fig 9—Keying-circuit schematic

diagram. For general notes on the
schematics, refer to the caption for Fig
3. The bias trimpot sets the cutoff bias.
When the key is closed, the transistor is
turned off and the Break trimpot has no
effect; the Make trimpot determines the
rise time. When the keyline is open, the
Make trimpot has no effect and the
transistor is turned on; the Break
trimpot determines the decay time. The
voltage-follower op amp isolates the
timing circuit from external influences.
C1—Low-loss timing capacitor, 0.47 µF.
Panasonic P-series polypropylene,
#ECQ-P1H474GZ. Digi-Key #P3474
(see Reference 8).

48 QEX
noise off the main 40-MHz line with no serves mainly as a buffer. The noise ning oscillator tuned by a varactor
significant loading of the main receiver mixer is a simple MOSFET, which is diode with a front-panel control. The
circuits. With no input matching, the adequate for this task. The LO injec- total range is about 2 MHz, providing
preamplifier has only moderate gain; it tion power is provided by a free-run- considerable over-range. The 1 MHz

Fig 10—Tunable noise-channel schematic diagram. For general notes on the schematics, refer to the caption for Fig 3. This
diagram includes the main 40-MHz band-pass filter and the TR relays. Reed relays are used to allow high-speed, break-in CW
(QSK) operation.
The βN control line is provided by the Logic board. It switches from –15 to +15 V when the blanker is turned on. The control
line is used here in three different ways. Most conveniently, for the noise-channel preamp and the noise-channel post-mixer
amp, it controls the MOSFETs by means of gate 2. For the noise mixer, a diode is added; this simulates a µ-type control line
and controls the MOSFET using gate 1. The βN line is also used to directly power the noise channel’s local oscillator. Later
designs avoid µ-type control lines and use only β-type control lines. This simplifies the logic board circuit; only a diode is
needed for a β-type control line to simulate an µ-type line. The βX line switches from –15 V in receive mode to +15 V in
transmit. Thus it powers one or the other of the TR reed relays.
BPF1—Main band-pass filter, center FL1—Crystal filter, 25-kHz bandwidth, L1—Noise local-oscillator coil, 1.8 µH,
39.5 MHz, bandwidth 1.4 MHz. four-pole. This filter was salvaged T-50-6 powdered-iron toroidal core,
Inductors, 800 nH, T-37-6 powdered- from an irreparable CX7; it has an 21 turns, tapped 5 turns from the
iron toroidal core, 15 turns. impedance of 3200 Ω. ground end.
BPF2—Noise band-pass filter, center K1, K2—Miniature reed relay, SPST, NO, T1—FT-37-43 ferrite toroidal core.
39.5 MHz, bandwidth 2 MHz. Inductors, 12 V dc coil, 1450 Ω, 8 mA; Gordos Primary, 11 turns; secondary, 2 turns
1.2 µH, T-37-6 powdered-iron toroidal #0490-1478DZ; Hosfelt #45-191 (see wound on low end of primary.
core, 19 turns, center-tapped. Reference 11); Jones #3806-RL (see VC1—Varactor diode, nominal 6.8 pF.
C1, C2—Same as C1 and C2 in Fig 8. Reference 15). Motorola type MV2101; NTE type 610
(see Reference 11).

Nov/Dec 1999 49
desired range is conveniently obtained
over the upper 180° swing of the panel
knob. Drift problems were anticipated
with this free-running circuit, but in
seven years of operating, this has not
been a problem. The diode typically
covers a range of 4 to 14 pF when re-
verse biased from 15 to 0.1 V. Less
range is actually required here, and the
panel control is wired accordingly. This
prevents conduction in the varactor di-
ode, which might otherwise result from
the RF voltage in the tank circuit. The
mixer output at 9 MHz is filtered by a
two-pole crystal filter and amplified by
a MOSFET, with the output going to
the noise amplifier on the IF board.
Construction
The RF board is shown in Fig 1. The
general method of construction was
described in Part 1, where the need for
careful shielding and lead filtering was
discussed. Part 2 gave further construc-
tion details, most of which also apply to
the RF board. The power and control
leads are filtered as described in Part 2,
except for the π-section filters in the
power and control lines feeding the cir-
cuits operating above 30 MHz. Those
instead consist of two 1-nF bypass ca-
pacitors and a 100-µH RFC. The board’s
underside is shown in Fig 11.
Some of the circuits are built on
perfboards: some with no solder pads,
some with pads and some with a
ground plane on the underside. These
were all poor choices for RF circuits.
Plain copper board and true dead-bug
construction—as used in the noise
channel and later in the IF board (Part
2)—is much better. In addition, the LO
amplifier should be in a separate com-
partment. All these circuits are sched-
uled for rebuilding.
Test and Alignment
The BFO trimpots in the IFS circuit
are adjusted in the transmit mode for
8816.5 kHz in USB and 8813.5 kHz in
LSB. This assumes selection of first-
rate, prime-condition SSB crystal
filters with the correct passband of
8814-8816 kHz. The USB and LSB
BFO-buffer-level trimpots are ad-
justed to obtain –16 dBm at the IF port
of the BFO mixer. The trimpot at the
BFO section output (terminal ϕ9A
leading to the AF board) is adjusted for
100 mV P-P output. The BFO voltage
on this line is kept low to minimize
leakage into the IF strip and else-
where. A BFO amplifier on the AF
board provides the proper LO injection
levels for the product detector and
balanced modulator.
50 QEX
Fig 11—Bottom view of the RF board. Effective filters are installed at each terminal
and coax cables are soldered directly to the double-sided circuit board; see the
discussion in Part 2. To minimize connector troubles, the board is hard-wired to
the radio; a 12-inch-long bundle of wires and cables allows the board to be easily
lifted and serviced.
The master oscillator is adjusted for
43.1 MHz. The offset oscillator is ad-
justed for a range of 34.2835 MHz
(for 0 Hz offset) to 34.2825 MHz (for
1000-Hz offset). The PTO trimpots are
adjusted to provide –16 dBm at the IF
port of the PTO mixer. Injection to
both the BFO and PTO mixers at the
LO ports is about +4 dBm. Although
the doubly balanced mixers (DBMs)
are +7 dBm devices, the lower LO
power results in only about 1 dB more
conversion loss. This lower LO power
is in accordance with the manufac-
turer’s suggestions for situations
where dynamic range is not a factor.7
Lessened LO power reduces harmonic
mixing and decreases stray LO power
in the system.
For convenience, RF probes that
measure LO injection level at the main
mixers are permanently built into the
circuits. This enhances the repeat-
ability of measurements, difficult to
achieve with external RF probes and
their questionable grounding leads.
The proper dc voltages at the receive-
and transmit-mixer LO test points are
1.1 and 1.4 V, respectively (on a 10 MΩ
meter). The two trimpots at the LO-
amplifier output compensate for the
different impedances of the two mix-
ers and for circuit reactance; one
trimpot is kept at maximum. The GAIN
ADJUST trimpot in the LO amplifier is
set as required to obtain the specified
injection levels.
The RF-board receiving circuits
operate at a gain of 15 dB, overall,
between terminal σ40 (the rear-panel
40-MHz jack) and terminal σ9R, which
leads to the IF board. The mixer drain-
circuit tuning capacitor is peaked on a
weak signal. The input to the transmit
mixer at terminal σ9T is adjusted for
200 mV P-P using trimpots on the IF
board. The transmit mixer’s balance
trimpot is adjusted to minimize LO
energy at the output. The trimpot at
the output of the transmit-mixer sec-
tion is set to obtain –7 dBm at the
40-MHz jack on the rear panel of the
radio. Other alignment specifications
are included above in the discussions
of the individual circuits.
Summary
This article gives a complete de-
scription of the RF board in a high-
performance homebrew transceiver.
The board establishes the 40 MHz to
9 MHz transitions.
References
1M. Mandelkern, K5AM, “A High-Perfor-
mance Homebrew Transceiver: Part 1,”
QEX, Mar/Apr 1999, pp 16-24.
2M. Mandelkern, K5AM, “A High-Perfor-
mance Homebrew Transceiver: Part 2,”
QEX, Sep/Oct 1999, pp 3-8.
3A long-term temperature-compensation
program was carried out over a six-month
period while work continued on the next
board in the radio. A garage workshop,
quite cold on winter mornings, provided
daily temperature cycling.
4E. Oxner, KB6QJ, “Designing FET Bal-
anced Mixers for High Dynamic Range,”
Application Note LPD-14, pages 9-114 to
9-145 in the Low Power Discretes Data
Book, 1989, Siliconix Incorporated, 2201
Laurelwood Rd, Santa Clara, CA 95054-
1516; tel 408-988-8000, fax 408-727-
5414; www.siliconix.com.
 A High-Performance
Homebrew Transceiver:
Part 4
Here is a complete description of the AF board.
D
esigning and constructing the
AF board of a radio is a rela-
tively easy task. Op-amp
circuits are perfectly predictable,
special features can be added with little
effort; signal levels are high and easily
measured with simple equipment. The
AF board in this radio contains the BFO
amplifier, product detector, noise limit-
er, speech amplifier, balanced modula-
tor and CW sidetone oscillator.
Part 1 gave a general description of
the K5AM homebrew transceiver,
built for serious DX work and contest
operating. 1 Parts 2 and 3 described the
IF and RF boards. 2, 3 This article gives
1Notes appear on page 56.
5259 Singer Rd
Las Cruces, NM 88005
k5am@roadrunner.com
By Mark Mandelkern, K5AM
circuit details for the AF board. A
general description of this board was
given in Part 1; the board is shown here
in Fig 1. The block diagram in Fig 2
shows the arrangement of the various
stages.
Features
Some of the special features of the
AF board are:
• High-pass transmit audio filter to
eliminate hum.
• 60-Hz receive audio filter to
eliminate hum.
• Electronic attenuators in all circuits
with front-panel controls to avoid
routing audio signals to the panel.
• Class-A audio output stage with only
0.3% total harmonic distortion.
• Noise limiter for reduction of
atmospheric static.
Circuit Description:
BFO and Carrier Amplifiers
Fig 3 gives the schematic for the
stages used to provide LO injection
power for the product detector and
balanced modulator and that provide
a carrier for CW operation. Special
effort was made to prevent carrier
leakage into the IF board during SSB
operation. Both a switched MOSFET
and a diode switch are used to gate the
carrier. The total residual hum, noise
and carrier on the transmitted SSB
signal are more than 65 dB down.
Receiver Circuits
Fig 4 is the schematic of the stages
used to process received audio signals.
Signals arrive from the IF board at
terminal σ9R1 and are fed directly to
the product detector. Measuring LO
Jan/Feb 2000 47
injection at a DBM (doubly balanced
mixer) is difficult with simple
equipment. The diodes in the DBM
cause distorted scope patterns at the
LO port that are not easy to interpret
as to power level. The 47-Ω resistor at
the product detector’s LO port helps
isolate the port and allows meaningful
readings at the test point indicated.
The 2-V(P-P) level specified at the test
point is estimated to correspond to a
+4 dBm LO injection level. This is
appropriate, although the device is
rated at +7 dBm (see the discussion in
Part 3).
A special low-noise op amp was
selected for the first AF amplifier. Care
was taken to install this low-noise
amplifier immediately adjacent to the
product detector to avoid hum pick-up
on a connecting cable. The gain of the
low-noise op amp is adjusted for 300 mV
(P-P) at terminal σPD. The AM and FM
detector circuits are adjusted similarly
and the three JFET gates are used to
select the appropriate detector.
Noise Limiter
As hams painfully know, noise
blankers do not blank static caused by
atmospheric thunderstorms or corona
noise produced by high-tension power
lines. In the eternal struggle to copy DX
signals (no matter how weak) under
noisy conditions (no matter how severe)
hams have always been eager to try any
conceivable noise-reduction method.
In times past, when AM was the
primary mode for voice work on the ham
bands, receivers had noise limiters.
One common type was the shunt
limiter, a diode clipper applied to the
detected AF signal. In some radios, the
clipping level was adjustable. Other
radios had an automatic noise limiter
that used the detected carrier—a dc
signal sometimes also used to drive the
S-meter—to establish an appropriate
bias level for the clipping diodes.
Modern commercial receivers have
no noise limiters. Many hams, espec-
ially 160-meter DX hounds, have used
diodes on the headphone line to clip the
noise. This can be very effective. A
disadvantage is that the degree of clip-
ping depends on the AF voltage level at
the headphone jack, the headphone
impedance and the AF gain (AFG)
setting selected by the operator. The
main disadvantage is that to obtain a
fair amount of clipping, the receiver AF
output level must be much higher than
normal; this is likely to cause distortion
in the receiver AF circuits, degrading
signal intelligibility.
These problems are avoided by
putting the noise limiter inside the
radio, ahead of the AF GAIN control.
Clipping at low AF levels is a simple
matter. With no AM carrier to set the
bias level, we cannot expect an auto-
matic noise limiter, but we do want an
adjustable noise limiter. We need a
panel control to set the clipping level
according to conditions. In Part 1,
Fig 3, the NOISE LIMITER control is just
below the meters, third knob from the
left. The limiter is quite effective,
especially with atmospheric static for
weak low-band DX signals.
“Limiter” is a good term. It reminds
us that the circuit cannot eliminate
the noise, as a good blanker can under
certain conditions, but that it only
limits the noise to a level set by the
operator. This level is usually the peak
level of the desired signal, so that the
noise and the signal are evenly
matched in a fair contest to register in
the operator’s brain.
Receiver sensitivity measurements
are based on the minimum discernible
signal (MDS) specification. This means
that the (S + N)/N is approximately
3 dB. Experienced DX operators know,
however, that this ratio is no “mini-
mum signal”—it is “arm-chair copy,” or
in other words, “loud.” DX operators
can copy CW signals that are well
below the MDS. Thus, the limiter can
turn barely detectable signals into
reasonably good copy.
Don’t expect the noise limiter to
eliminate thunderstorms! Its effective-
ness is only apparent on a narro
range of weak DX signals, those within
a few decibels of the ambient noise
level. When signals fade below that
level, nothing will help; when signals
build up a few decibels, a noise limiter
won’t be needed. Still, most new
countries added to an oper-ator’s 160-
meter DXCC list involve signals within
that narrow range.
The noise limiter is normally used
with the AGC on. This is the safest
operating method (see Part 1, pp 21-
22). Receiver gain is enough that
ambient antenna noise on any band
activates the AGC system. This
insures that the AF level into the noise
limiter is essentially constant. The
resulting headphone volume is thus
also constant. Signals do not “jump out
the noise,” as is often said about
receivers reported to be “quiet.” Such
receivers merely have inadequate

Fig 2 (see right)—AF board block

diagram. Signals from the IF board
arrive at terminal σ9R1. After detection,
amplification and filtering, the audio
signal travels from output terminal σAF
to the audio-output module on the rear
panel. For transmitting SSB, there are
dual speech inputs: a MIKE jack on the
front panel and a line-level jack on the
rear panel. The DSB output is at terminal
σ9 DSB leading to the IF board, the SSB
Fig 1—Top view of the AF board in the K5AM homebrew transceiver. At right, from filters and RF speech clipper.
the top, are the balanced modulator, the BFO amplifier, the carrier amplifier for CW Potentiometers labeled in all capital
and the product detector. The two-stage transmit high-pass AF filter is included in letters are front-panel controls; others
the balanced-modulator compartment, eliminating any possibility of hum pick-up are circuit-board trimpots for internal
on a connecting cable. Similarly, the first receiver AF amplifier, a special low-noise adjustment. An explanation of the
op amp, is included in the product-detector compartment. All other AF circuits are terminal designators is given in Part 2,
in the large compartment at the upper left; the lower left compartment is for the Table 1. The control lines are provided
AM/FM circuits. (See Note 8.) by the logic board.

48 Jan/Feb 2000
Jan/Feb 2000 49
Fig 3—BFO and carrier-amplifier schematic diagram. Except as noted, each resistor is a 1/4-W, carbon-film type. All trimpots are
one-turn miniature types, such as Bourns type 3386; Digi-Key #3386F-nnn. The unmarked coupling and bypass capacitors are
all 10 nF disc ceramic types. Also, each control and power terminal has a bypass capacitor that is not shown. Electrolytic
capacitors have 25-V ratings and values given in microfarads; those less than 100 µF are tantalum types. Capacitors labeled
“s.m.” are silver micas, with values given in picofarads. The values of RF chokes (RFC) are in microhenries. Part-supplier
contact information appears in Notes 9, 10, 11 and 12.
The MOSFETs are all small-signal, VHF dual-gate types. Type 3N140 is used here, but any similar type may be substituted.
Replacement-type NTE 221 is also available. The JFETs are J310s. Except where otherwise indicated, the diodes are small-
signal silicon types, such as 1N4148; the bipolar transistors are type 2N2222A. Except as noted, the op amps are sections of
LM1458Ns. Unless shown otherwise, each op amp is powered by the +15 V and –15 V rails. Not shown is the filtering at each op
amp power terminal. This varies with the application. The minimum is a 100-nF, monolithic ceramic bypass. More sensitive and
low-level circuits also use a 100-Ω isolation resistor and a 10-47 µF tantalum electrolytic at the terminal.
Potentiometers labeled in all-capital letters are front-panel controls; others are circuit-board trimpots for internal adjustment.
The control lines are provided by the logic board. The signal levels indicated at various test points are scope readings.
L1—FT-37-61 ferrite toroidal core, 14 turns of #26 AWG enameled wire.

AGC systems; they are not necessarily
“quiet” with regard to sensitivity or
noise figure. Listening for weak
signals in the noise on such receivers
requires one to set the gain too high,
creating an aural hazard when a
strong signal suddenly appears. The
flat AGC characteristic in this receiver
allows the noise-limiter clipping level
to be easily set. If “noise-jumping”
signals are desired at other times, the
IF GAIN control may be used to raise the
AGC threshold (see Part 1, pp 21-22).
The noise-limiter circuit is shown in
Fig 5. Op amp U1 raises the receiver AF
level from 3 V to 14 V (P-P), while op
amp U2 reduces it to the former level
after the clipper. When clipping is used,
however, the AFG control will need to
be increased. With both controls on the
Fig 4—(see right) Receiver AF circuit
schematic diagram. For general notes
on the schematics, refer to the caption
for Fig 3. Capacitors labeled simply “1”
are 1-µF monolithic ceramics (see Note
10). The DBM used as a product
detector may be obtained in small
quantities directly from the
manufacturer (see Note 12). Signal
levels indicated on the diagram are
given with the AF GAIN set for 2 W
output with an 8-Ω resistive load at the
speaker jack.
C—Polypropylene capacitor, 2.7 nF, 2%
tolerance. Panasonic #ECQ-P1H272GZ,
Digi-Key #P3272. Where 2C is indicated,
use two of these in parallel; this
ensures the closest match.
C1—Polypropylene capacitor, 1 nF.
Panasonic #ECQ-P1H102GZ, Digi-Key
#P3102.
C2—Polypropylene capacitor, 470 nF.
Panasonic #ECQ-P1H474GZ, Digi-Key
#P3474.
C3-C4—Polypropylene capacitor, 220
nF. Panasonic #ECQ-P1H224GZ, Digi-
Key #P3224.
R—Metal-film resistor, 1 MΩ, 1%
tolerance. Digi-Key #1.00MXBK. Where
R/2 is indicated, use two of these in
parallel; this ensures the closest match.
R1—AFG potentiometer, 5 kΩ. This has
been replaced five times in the last
eight years. The current resident, a 2-W
wirewound type found at a hamfest,
seems headed for a long life.
U1-U2—Electronic attenuator, Motorola
MC3340P; discontinued. Available as
replacement component NTE #829,
Mouser #526-NTE829 (Mouser). A
datasheet is available on the Web:
http://www.mot-sps.com/books/dl128/
pdf/mc3340rev6.pdf. Omitted from the
schematic is the filtering at the +15-V
supply terminal: a 100-Ω isolation
resistor, a 100-nF monolithic ceramic
capacitor and a 47-µF tantalum
electrolytic capacitor at pin 8.

50 Jan/Feb 2000
Jan/Feb 2000 51
front panel, this is convenient; the
clipping level is normally set according
to conditions and not changed until the
thunderstorm is over.
The clipping level is set by the NOISE
LIMITER panel control. Voltage follower
U3 is required to provide a low-impe-
dance negative bias source for the
negative-clipping diode. Inverter U4
tracks U3 and provides reverse bias to
the positive-clipping diode of equal
magnitude but opposite polarity. This
insures that the positive and negative
halves of the signal waveform are
clipped equally. When the panel control
is fully clockwise, the diodes have no
reverse bias; each clips to 0.7 V. Thus
the 14-V (P-P) level allows a maximum
clipping depth of 20 dB. With the
control fully counterclockwise, the
diodes are completely reverse-biased,
and the noise limiter is effectively off.
When noise limiting is needed for
CW, the control is usually set near
maximum; distortion on CW signals is
of no consequence. For SSB, the full
setting causes noticeable distortion,
but signals are very intelligible. For
less distortion, retard the control
somewhat. With 6 dB of clipping, the
distortion is barely noticeable. At
those times when signals can be heard
through the static without the limiter,
use of the limiter makes listening far
more pleasant and reduces fatigue.
Also, the limiter can be left at 3-6 dB
at all times; it won’t noticeably effect
the signals, but will reduce the noise
from key clicks and splatter.
When a DX signal is too weak to hear
during a static crash, the noise limiter
still serves a valuable function and
may enable an otherwise impossible
contact. It saves the operator’s ears!
This is both a long-term safety feature
and a short-term operating aid. An
operator’s hearing ability is depressed
after some minutes of listening to loud
noise. With the AFG high enough to
hear a weak station between static
crashes, one’s hearing threshold may
be degraded within ten minutes. On
the other hand, if the gain is low
enough so that static crashes don’t
cause desensitization, a weak station
won’t be heard. The solution to this
dilemma is an adjustable noise limiter.
The noise-limiter schematic dia-
gram is separated from the main
diagram so that it might be built and
installed in other receivers. Details
and pin-outs are given. If the receiver
AF level at the point of insertion is
other than that specified the gain of
the input and output amplifiers may be
changed easily.
52 Jan/Feb 2000
This noise limiter has been very amplifier to compensate for the loss in
helpful in DX work. One can only the filter. A simple 20-dB pad in the
wonder: “Why don’t all receivers have radio at the loop’s return jack restores
noise limiters?” the AF signal to the original level.
The external amplifier is shown in
DSP Loop for External Filter Fig 6. It has controls to set the proper
The DSP loop was added only a few levels both to and from the filter.
years ago to accommodate a Timewave These need be set only once: The filter
Model DSP-9+ filter. The loop is can be switched in or out at any time
positioned before the AF GAIN control. with no readjustment of receiver,
This is done so that the AFG knob on amplifier or filter controls. This proce-
the radio can be used independently of dure eliminates the need for any modi-
the filter. This avoids the usual cum- fications to the external DSP filter and
bersome arrangement wherein the allows adjustment for a variety of
AFG on the radio must be continually different external units. A switch on
readjusted for the proper input level the rear panel of the radio cuts the loop
and the AFG control on the external out of the circuit. This avoids the need
filter used to control headphone for a patch cable when the external
volume. This is the main reason for filter is not connected. When
adding the DSP loop. There are also connected, the filter is switched in and
other problems: out by its own controls.
• There is a 6-dB loss in the external
filter between the line input and line Electronic Attenuators
output jacks. To avoid routing audio signals to
• Signal levels are low and hum can be front-panel controls and possible hum
introduced. pickup, electronic attenuators are used
• External devices installed on the for the AFGAIN function and all other
speaker line can suffer from annoy- front-panel audio level controls. The
ing RFI during transmissions. attenuators [MC3340Ps —Ed.] provide
The solution is to raise the AF level a 60-dB range, with a pin 2 control
20 dB for the loop and use an external voltage range of 3-6 V. The maximum
Fig 5—Noise-limiter schematic diagram. For general notes on the schematics, refer
to the caption for Fig 3. The 1-kΩ resistor at the output of U1 is needed to prevent
excessive current drain on U1 when the diodes conduct. The gain-setting resistors
for U4, while they might be of any (equal) value to set the gain at –1, should be of
no lower value than shown. The circuit of U4 establishes a virtual ground at the
inverting input; a low-value input resistor would unduly load the potentiometer
circuit, resulting in less available reverse bias. At the full counterclockwise
position of the control, a reverse bias of 9 V is presented to the diodes. This is
enough to cut off the diodes, since the peak AF voltage is 7 V in either direction.
U—Quad op amp, LM348N. Each of the four op amps is shown separately. A fifth
symbol shows the power connections, but all is contained in a single, 14-pin DIP
package. Note that while type LM324N is used in other parts of the radio, it is a
single-supply type vulnerable to crossover distortion. It is useful for dc-control
circuits because it has a wider output voltage swing, but it should not be used for
AF. The dual LM1458N and quad LM348N types used on the AF board are standard
low-distortion 741 types.
voltage gain of the attenuators is four. the other hand, too little capacity will hum, whether from internal circuits or
The circuits are configured for 300-mV allow clicks to be heard in the head- included on received signals. The
(P-P) input, with output varying from phones. The divider for the blocking component values specified result in
1.2 to 1200 mV (P-P). This same setup voltage also affects the timing. The a calculated notch frequency of
is used for the MIKE GAIN, SIDETONE circuit shown here mutes the receiver 58.9 Hz. The component manufac-
LEVEL and CW OFFSET SPOT LEVEL smoothly and quietly. Stations can be turers’ tolerances result in a worst-
circuits. (One attenuator for each cir- heard breaking in while the radio is case error of 1.7 Hz. This possible error
cuit.) Each of these circuits requires transmitting CW at 50 wpm. is one reason for using an adjustable-
only a single, unshielded dc lead to the Q filter: to ensure that the notch
front panel. Simplicity of wiring and 60-Hz Notch Filter includes 60 Hz. The trimpot is
avoidance of hum was well worth the The next stage is an adjustable-Q adjusted to obtain a 3-dB band-reject
extra trouble of installing the ICs. 60-Hz notch filter.4 It eliminates any range of about 55-65 Hz. Measured
I like the normal operating positions
of all controls to be between 10 and 11
o’clock, with 40-dB of gain reduction at
the full counterclockwise position and
20-dB increase when fully clockwise.
Satisfying this particular demand was
more difficult than expected—the
characteristics of the devices varied
from sample to sample. Each circuit
requires two trimpots to set the desired
curve, and the adjustments interact
greatly. Alignment was a bit tedious,
but the attenuator circuits have func-
tioned perfectly and have required no
further attention.

Muting Gate
After the attenuator is the JFET
muting gate, used for TR switching and
squelching. The deceptively innocent-
looking capacitor, C2, in the source
circuit deserves attention. Since all
other circuits in the radio are quite fast,
this capacitor determines receiver
recovery time. Some delay is needed, or
the radio will squeal in protest for a few
milliseconds at the end of each trans-
mission, or after each dit or dah when
operating CW QSK. Too much capacity
here will cause excessive delay and
destroy the radio’s break-in ability. On
Fig 6—External-DSP-loop-amplifier schematic diagram. For general notes on the
schematics, refer to the caption for Fig 3. The amplifier may be housed in a small
aluminum box. Defense against RFI from the transmitting antenna is not critical,
since the amplifier and filter are positioned before the receiver muting gate.
The amplifier is powered by the same 13.6-V dc power supply used for the numerous
operating-bench auxiliary gadgets. The amplifier is fitted with a 6-foot shielded
power cable terminated in a two-pin Jones plug. The 13.6-V dc fused distribution
boxes on the operating bench contain two-pin Jones sockets. These connectors are
solid, reliable and easy to use. Since the shack has them only for this purpose, this
system is safer than one that uses other common connectors, such as phono plugs.
P1—Jones plug, two-pin, Cinch #P-302-CCT, Digi-Key #CJ102P. Mating socket, Cinch
#S-302-AB, Digi-Key #CJ302S.

Fig 7—Sidetone-oscillator schematic diagram. For general notes on the schematics, refer to the caption for Fig 3.
C1-C3—Polypropylene capacitor, 47 nF. U1—MC3340P. See caption for Fig 4.
Panasonic #ECQ-P1H473GZ, Digi-Key
#P3473.

Jan/Feb 2000 53
notch depth is 50 dB.
The notch filter circuit is a modern
version of the traditional “twin-T”
network. The network alone has a Q of
only 0.3, but with feedback derived from
the upper voltage-follower op amp, the Q
can be adjusted from 0.3 to more than 50.
The trimpot determines the amount of
feedback. The lower voltage follower is
needed to drive the network from a low-
impedance source so the circuit is un-
affected by the resistance of the trimpot.
Notch depth depends on component
match (see Note 4).
The last stage is a summing op amp
that combines the signals from the
detectors, sidetone oscillator and CW-
offset spot mixer on the RF board. The
gain of the summing-amplifier stage is
adjusted to provide the correct audio
signal level at terminal σAF, which
leads to the AF output module on the
rear panel.
Sidetone Oscillator
The sidetone oscillator is shown in
Fig 7. Adjustments are provided for
pitch, feedback and output level. The
feedback control was originally in-
tended to allow adjustment for the
minimum amount needed to sustain
oscillation, thus to obtain the purest-
possible sine wave. However, a pure
sine wave may produce fatigue over
long operating periods. The feedback
control may be used to introduce some
Fig 8—AF-output-module schematic diagram. For general notes on the schematics, refer to the caption for Fig 3. The 1-Ω
resistor in the +18-V supply line to the output stage is a shunt for current metering during tests. A DMM set to the millivolt
range will read out in milliamps directly. These 1-Ω shunts are also included at many other points in the radio where a current
measurement is required. The complementary-output transistors are available at Hosfelt.
C—Polypropylene capacitor, 1 nF, 2%
tolerance. Panasonic #ECQ-P1H102GZ,
Digi-Key #P3102. For 2C use two of
these in parallel; this insures the
closest match.
54 Jan/Feb 2000
harmonic distortion, which may result
in a more pleasant sound. Recall that
a good violin produces many nice-
sounding harmonics. When in either
TUNE or PULSE TUNE mode, the side-
tone oscillator is disabled.
AF Output Module
The AF-output module is mounted on
the transceiver rear panel, with the
output transistors using the panel as a
heat sink. The circuit is shown in Fig 8.
Two stages of low-pass filtering are
used, each with a 3-dB cut-off frequency
of 3000 Hz. The measured composite
response is –3 dB at 2500 Hz and –6 dB
at 3000 Hz.
Each filter stage is a unity-gain,
maximally-flat Butterworth filter with:
2
Q= (Eq 1)
2 plugs that impede control-knob access.
in a voltage-follower configuration.5
This configuration is the simplest to
apply with respect to component
selection. It also involves the simplest
formulas. One first chooses the cut-off
frequency, f (in Hz), and the capacity,
C (in farads). Then the resistance R (in
ohms) is simply:
1 levels are 30 mV (P-P) at the front-
R= (Eq 2)
2 2 π fC
The 10-kΩ resistor at the non-
inverting input does not affect the
R—Metal-film resistor, 37.5 kΩ, 1%
tolerance. Digi-Key #37.5KXBK.
response; it serves merely to protect
the op amp (see Note 5, p 126).
The output stage operates class A.
The biasing is chosen for 50 mA idling
current. Total harmonic distortion
(including hum and noise) of the AF-
output module measures 0.3% at
normal levels and 1.5% at the full 2-W
output. In a test involving the entire
receiver, an RF signal on the 20-meter
band produced total harmonic distor-
tion (including hum and noise) of 0.6%
at normal levels.
Transmit Circuits
Fig 9 is the schematic of the stages
used for transmitting. The speech
amplifier has two alternative inputs.
The front-panel MIKE jack is rarely
used. I don’t like a clutter of cables on
the operating bench or protruding
I always use the rear panel SPEECH
jack.6 This jack provides a high-level,
600-Ω line input and accepts speech
signals from the station’s speech-
distribution system and the digital
voice keyer.7 Running this input at a
high-level minimizes hum pick-up—
often a problem when digital voice
recorders are used. The nominal input
panel MIKE jack and 300 mV (P-P) at the
rear-panel SPEECH jack. The inverting
op-amp summing stage combines the
two inputs with no interaction.
RFC1—RF choke, 10 µH. The choke
must have a dc resistance of 1 Ω or
less.
 The JFET microphone amplifier is
located directly at the front panel on a
very small circuit board attached to the
jack. This minimizes hum pick-up. A
single shielded lead runs from this
board to the AF board. The microphone
amplifier will accept high-impedance
mikes; for low-impedance mikes, an
appropriate load resistor must be
provided externally. For my Heil HC-5
mikes, I install a 2.2-kΩ resistor in the
mike plug. This arrangement provides
the greatest flexibility (see Note 7). The
RC network at the mike jack filters RF.
Silver-mica and polypropylene capa-
citors are used for their low-noise
characteristics.
After the summing amplifier, there is
an electronic attenuator for MIKE GAIN
Fig 9—Transmitter AF-circuit schematic diagram. For general notes on the schematics, refer to the caption for Fig 3. The
signal levels after the mike-gain attenuator are given for the normal control-knob position of 11 o’clock (see Part 1, p 21 for
information about microphone calibration). The DBM used as a balanced modulator may be obtained in small quantities
directly from Mini Circuits.
C, C1—Polypropylene capacitor, 10 nF,
2% tolerance. Panasonic #ECQ-
P1H103GZ, Digi-Key #P3103.
control. It is adjusted as described
above in the “Receiver Circuits” section.
The electronic attenuator avoids the
need to route speech signals to the front
panel, again minimizing hum pick-up.
Capacitor C2 at pin 6 of the attenuator
is for high-frequency roll-off; it may be
altered as desired. The value given for
C2 results in a 3-dB roll-off at 5 kHz.
After the attenuator is a JFET gate.
This, along with switched stages on the
IF board, eliminates any possibility of
modulation during CW or TUNE opera-
tion. The next op amp drives the high-
pass filter.
High-Pass Filter
Two stages of high-pass filtering are
used, each have a 3-dB cut-off
C2—Polypropylene capacitor, 5.6 nF.
Panasonic #ECQ-P1H562GZ, Digi-Key
#P3562.
frequency of 300 Hz. The measured
composite response is –3 dB at 380 Hz
and –50 dB at 60 Hz. Care was taken to
install the high-pass filter immediately
adjacent to the balanced modulator,
thus eliminating any possibility of hum
pick-up on a connecting cable.
Each stage is a unity-gain, maxi-
mally-flat Butterworth filter with
2 (Eq 3)
Q=
2
in a voltage-follower configuration (see
Note 5). This configuration is the
simplest to apply in regard to compo-
nent selection. It also involves the
simplest formulas. One first chooses
the cut-off frequency f (in Hz) and the
capacity C (in farads). Then the
R—Metal-film resistor, 75 kΩ, 1%
tolerance. Digi-Key #75.0KXBK. For R/2
use two of these in parallel; this insures
the closest match.
U1—MC3340P. See caption for Fig 4.
Jan/Feb 2000 55
resistance R (in ohms) is simply:
1
R= (Eq 4)
2π f C
With a fixed value of C, the cut-off
frequency may be changed (within lim-
its) by simply using the inverse relation
between f and R. For example, with the
value shown for C, a cut-off frequency
of f = 200 Hz can be obtained with
R = 112.5 kΩ. The 10-kΩ resistor at the
non-inverting input does not affect the
response; it simply protects the op amp
as previously described. The 10-kΩ
filtering resistor at the +7-V bias
supply point also does not affect the
response; the 47-µF capacitor provides
a virtual AF ground for the network.

Construction
The AF board is built as shown in
Fig 1. The general method of construc-
tion was described in Part 1, where the
need for careful shielding and lead
filtering was discussed. The power and
control leads are filtered as described
in Part 2. Most of the purely-AF cir-
cuits are hand-wired on perf boards.
Plain copper board and truly “ugly”
construction is used for all circuits
containing 9-MHz signals. The board’s
underside is shown in Fig 10.
Test and Alignment
Normal operating levels at various
points of the circuit are indicated on the
schematic diagrams. The associated
trimpots are adjusted to obtain the
specified oscilloscope readings. There
are quite a few trimpots in the radio.
When building a single unit, it is often
easier to install a trimpot than to
individually select a component. This
minor expense is justified by saving of
hours of work in selecting and changing
components. This is especially true
when a radio is designed and built
stage-by-stage. Often, a section is built
without full knowledge of what lies
ahead. Desired levels are apt to change
in the light of final testing. None of the
trimpots has required readjustment
since the radio was built. Exceptions
occur when modifications are
introduced; then the trimpots become
especially useful.
Fig 10—Bottom view of the AF board. Effective filters are installed at each terminal and
coax cables are soldered directly to the double-sided circuit board (see the discussion in
Part 2). To minimize connector troubles, the board is hard-wired to the radio; a 12-inch-
long bundle of wires and cables allows the board to be easily lifted and serviced.
remainder of the radio, flexibility and
operating convenience are prime design
factors.
Notes
1M. Mandelkern, K5AM, “A High-Perfor-
mance Homebrew Transceiver: Part 1,”
QEX, Mar/Apr 1999, pp 16-24.
2M. Mandelkern, K5AM, “A High-Perfor-
mance Homebrew Transceiver: Part 2,”
QEX, Sep/Oct 1999, pp 3-8.
3M. Mandelkern, K5AM, “A High-Perfor-
mance Homebrew Transceiver: Part 3,”
QEX, Nov/Dec 1999, pp 41-51.
4R. C. Dobkin, “High Q Notch Filter,” Linear
Brief #LB-5, in Linear Applications Hand-
book , National Semiconductor Corp,
Santa Clara, California, 1980.
5W. Jung, IC Op-Amp Cookbook (Indianapo-
lis: Howard W. Sams and Co, 1974),
pp 331-333.
6Similarly, I don’t use the front-panel HEAD-
PHONE jack, but only the rear-panel
SPEAKER jack. On the operating bench is
a 40-year-old speaker/headphone switch
box with a headphone-level control for
equalization. This provides instant switch-
ing without fussing with the headphone
plug or readjusting the AFG.
7M. Mandelkern, K5AM, “The AMSAFID: An
Automatic Microphone Switcher Amplifier
Filter Integrator Distributor,” QST, Nov
1995, pp 47-49.
8Sharp-eyed readers may notice that the AM/
FM compartment is empty. This only proves
that this shack follows true ham-radio tradi-
tion: Nothing is ever really finished.
9Digi-Key Corporation, 701 Brooks Ave S,
PO Box 677, Thief River Falls, MN 56701-
0677; tel 800-344-4539 (800-DIGI-KEY),
fax 218-681-3380; http://www.digikey
.com/.
10Hosfelt Electronics, 2700 Sunset Blvd,
Steubenville, OH 43952; tel 800-524-
6464, fax 800-524-5414; E-mail hosfelt
@clover.net; http://www.hosfelt.com/.
11Mouser Electronics, 2401 Hwy 287 N,
Mansfield, TX 76063, tel 800-346-6873,
fax 817-483-0931; E-mail sales@mouser
.com; http://www.mouser.com/.
12 Mini Circuits Labs, PO Box 350166, Brook-
lyn, NY 11235-0003; tel 800-654-7949,
718-934-4500, fax 718-332-4661; http:
//www.minicircuits.com/.
See Part 5 in March/April 2000 QEX

Summary
This article gives a complete descrip-
tion of the AF board in a high-perfor-
mance homebrew transceiver. Double-
balanced mixers are used for both the
product detector and the balanced
modulator. Special filters are included
to avoid hum both in reception and in
the transmitted signal. As in the

56 Jan/Feb 2000