WA7MLH Set 1

A shot of me and my all home brew shack around 1988 time frame.
This is all HF gear for CW and SSB.
Some of the designs are from W7ZOI and some are my own designs.

My latest 2015 Field Day 80m CW txvr.
IF is 5 crystals at 10 MHz Gaussian to -6dB at 500 Hz BW.
PA is single ended RD6HHF1 LDMOS FET.
RX front end is single ended JFET cascode at about 3mA.
I can hear close to -140dBm.
Audio chain has 4 stages of low noise LPF at about 1kHz BW.

The front end of the 2015 Field Day 80m CW txvr.
A DTC BPF feeding gate 1 of cascoded JFETs and broad band transformer voltage gain
for the gate 2 LO drive.
IF port drives a broad band feed back amp with at about 5 mA collector current.
This rx is a derivative of the S7C receiver in EMRFD.
I opted for this design with a low curent drain agenda and simplified circuitry over
trying to fit in an entire hybrid cascode AGC system.
Field Day operation works very well without IF AGC, or any AGC for that matter.

BFO in the 2015 Field Day 80m CW txvr.
I like a varibale RX BFO, although this could have been simplified with a fixed freq BFO.
The inductors in series with the crystal were needed to pull the osc freq a few hundred Hz
lower than the xtal filter pass band.
The BFO has Po = +6dBm driving a single balanced diode ring product detector.

My latest rig for late 2009 is this 40m Field Day CW txvr .
IF is 5 crystals at 4421 kHz from a box of them I found at SEAPAC hamfest.
These had unusually high Qu, over 150k as I recall.
BW=500 Hz with a Gaussian to -6dB response (optimized for CW pulse response)
A future project is to redesign for 250 Hz BW.
Field Day 2020 had me realize that 500Hz was not selective enough for a contest environment.
RX is a superhet with low side LO.
This txvr features solid state full breakin QSK.
PA is an RD16HHF1 single ended LDMOS FET for 16 watts out, variable Pout from the front panel.

My latest CW filter for my 30m CW txvr .
IF is 4433 kHz (European color burst).
BW=500 Hz with a Gaussian to -6dB response (optimized for CW pulse response).
These glass cased xtals are super high Qu I got from W7ZOI.
Thank you Wes!!

A shot of my 80m/40m SSB/CW txvr.
This is what you hear me on from my Montana shack.
It runs 10 watts out on both bands with a balanced pair of IRF511s at 13.6 volts.
I have a banana jack option on the back panel to run a separate Vdd to the final
drains so that I can use a 24 to 30 volt supply to get around 50 watts out.
The system IF is 6 xtals in cascaded half lattice at 10 MHz using a salvaged
commercial xtal filter (courtesy of W7ZOI).
Rx runs single conversion with high side LO injection at 13.5 MHz or 17 MHz.
VFO is a Hartley with 250 kHz spread on a Jackson drive.
Band switching is done with some low current relays.
AGC is audio derived and works quite well for a general purpose unit like this one.
The antenna tuner is on the right. I restrung my dipoles and don't need the tuner now.

A single ended 15w RD16 broadband amp.

A single ended 15w RD16 broadband amp.

A push pull 30w RD16 broadband amp.

A push pull 30w RD16 broadband amp.

A push pull 30w RD16 broadband amp.

A push pull 30w RD16 broadband amp.

A push pull 30w RD16 broadband amp.

A push pull 30w RD16 broadband amp.

A push pull 50w Si bipolar broadband amp.
This uses VHF transistors out of old police radio carcasses.

Stability for this class C application is maintained via judicious use of resistive
feedback from the series .1uF and 240 ohm resistor from each collector
to respective base.
Class C operation without enough negative feedback will provide really bad instability
problems over variable Pout drive conditions. No negative feedback will provide two levels
of stable operation...low Pout or max Pout.

I think the label inside is left over from a different circuit.

My first pair of RD100HHF1 FETs for a 100w class AB amp.
Idq needs to be on the order of 100mA per stage to get good gain out of them.
Feb 2023: Turns out these FETs work better in a class B bias with Idq=1A.

A UHF FM carcass final now serving as a 50w class C amp for 20m.
LCC impedance matching was used at both input and output ports.

A closer look at the 50w amp.

A push pull pair of 2SC5739 NPNs for 3MHz to 30 MHz at about 15w @ Vcc=13v.
Sadly, this device was discontinued shortly after I had built this unit.

Here is my 40m CW txvr with matching antenna tuner.
It runs single conversion with dual gate MOSFETs for the front end mixer,
IF amps and IF derived AGC. The IF is about 1200Hz BW at 1.98 MHz (courtesy of W7ZOI).
TX final is a single ended IRF511 at maybe 5 or 6 watts max.
This rig begs for an upgrade to an RD16 device.
This rig was intended to be a Field Day rig, which was the reason for the 5W power level.
It does have variable output power. The flash shows the actual speakers that can not
be seen under normal lighting conditions. Some times I use an external amp with this rig.

The interior view of my 3rd generation spectrum analyzer.
This unit has many improvements over the 2nd generation version.
The 2nd LO (fixed at 900 MHz) is contained in a single shielded enclosure for mechanical simplicity.
One box to secure is much better than 5 or 6 separate boxes with .085 semi-rigid tangled all over the place.
This 3rd generation unit also supports 3 different resolution bandwidths: 300kHz, 30 kHz and 3 kHz.
I have 1 Mhz and 100 kHz RBW filters (a set of 4 lumped LC coupled resonators) ready to add to this unit.
This 3rd generation unit also has a 10dB gain increase in the LOG/IF circuitry based on an improvement
in impedance matching between the LOG and IF amps.
An improved front end band pass filter in this 3rd generation fixes a spur problem that plagued the 2nd
generation version. A decent low pass filter was added inside the triple tuned filter after the 1st mixer.
The re-entrant mode of the coupled resonator coaxial filter did not offer much stop band attenuation out
beyond 2 GHz. This deficiency allowed harmonics of the 2nd LO to mix with the harmonic content of the
1st LO getting out the IF port and filter following the 1st mixer to provide a nasty spur at 900 mHz.
This is a very elusive problem to figure out, yet rather easy to solve. The low parasitic inductance of
surface muount chip caps for the added lowpass filter were essential here. Many thanks to W7ZOI for
insights to solving this problem.

Another view in to the 3rd generation spectrum analyzer chassis.
This box will go out to 650 MHz rather clean (in terms of response flatness).
750 MHz is very useable with a few dB of gain abberation.
The unit still has useful response out past 850 MHz.

The basic spectrum analyzer system shown above is an offshoot of what W7ZOI
designed way back in the late 1980s. Wes built the first unit, a 70 MHz box.
I built the 2nd one based on his schematics.

Wes and Terry White, K7TAU, later published the full details in QST, including
the generation of etched circuit boards that became a kit from Kanga (US).
I took to the task of expanding the original design to be a triple
conversion system that uses a Mini Circuits POS2120 VCO, which tunes from about
960 MHz to a little above 2 GHz. A nice unit to start with in that is was a cheap
($22) way to have the LO already done and ready to go. A MAV11 broad band amp follows
the VCO to generate about 17dBm for the TUF5H level 17 mixer (diode ring).

A triple tuned coaxial resonator band pass filter at 1010 MHz (1st IF) and
low pass filter follow the 1st mixer. That 1010 MHz signal (upconverted from
the RF input to the spec an) is mixed against 900 MHz from the 2nd LO, again,
a TUF5H. The 2nd LO is derived from a 100 MHz 5th overtone Butler crystal
oscillator which pumps a Wenzel back to back diode tripler, into a 300 MHz band
pass filter, broadband gain, and then another Wenzel tripler to get 900 MHz.

Further gain and filtering provide 17dBm of LO for the second mixer. The 1010 MHz
1st IF mixes with the 2nd LO down to 110 MHz, in to 4 coupled LC resonators as the
2nd IF. This 110 MHz signal is then mixed against the 3rd LO, another 100 MHz Butler
crystal osc/amp in to a TUF1H mixer down to the last IF of 10 MHz. All mixers are
followed by broad band buffer amps and 6dB PI pads for improved mixer IMD performance.

The last IF drives a selectable bank of band pass filters in order to determine the
resolution bandwidth of the system.

Typical RBW for casual work is 300 kHz.

The last IF is then routed through the LOG amp and IF amp, which provide vertical
signal for the Y axis of a conventional oscilloscope. The sweep gen board sweeps
the VCO (1st LO) and the X input of the scope. Details for the 70 MHz version are
available as pdf downloads from the ARRL web site. More info on the 70 MHz version,
especially upgrades to the LOG/IF circuitry, are posted at

An assortment of built and tested boards built "ugly method" for a 17m SSB/CW
superhet txvr (20 MHz IF) and a 6m CW DSB/CW txvr. In the lower right corner is
a 20MHz xtal osc driving a buffer driving a diode doubler driving a 2 pole band
pass filter to get 40 MHz to mix with a 10.1 MHz Hartley VFO to get a 50.1 MHz
LO source for the 6m DSB txvr. After I get all the bugs worked out I find that the
local surplus store has 40 MHz TTL osc cans for 50 cents each! And they will put
out over 16dBm for 5v@20mA. I am now a fan of osc cans. The catalogs have them for
about $2 new. So the plan now is to go with a can and resistive split the +16dBm
in to a pair of +7dBm LOs--one for the DSB balanced modulator and one for the direct
conversion rx diode ring product detector.

UPDATE Feb 2006--- Yes the TTL osc cans have some merits. However, further measurements
have shown that the phase noise of the cans is really bad! I have since abandoned them,
other than maybe for band edge markers and the like.

A shot of the 20 MHz xtal filters built and tested.

My Cohn topology ladder filter experimentation led me to prefer the idea of making
filters in sets of four xtals each. Then end loading for passband ripple can be
controlled easily. Then cascading a pair of "fours" with a buffer amp and pad between
them for 8 resonators total became my goal.

I found that 5 xtals seems to be the limit for retaining reasonable pass band shape
and ripple control with external end loading. Conversations with Rick Campbell, KK7B,
verified this observation.

More than 5 xtals isolates the inner xtals from the effects of end loading and you get
stuck with pass band shapes and ripple that you have little control over.

These filters are intended to be the heart of a 17m superhet SSB/CW txvr. It turns out
that a 20 MHz IF and a 1.9 MHz LO make for a marginal mixing combination for 18.1 MHz.
Further measurements indicate that the low side LO will probably not work for this band/IF combination.
Mixing spurs/products are not easily removed from the 18.1 MHz preferred signal.

To the left is a 17m low pass filter. I may opt for a 7th order .1dB Chebycheff instead.

On the right is a 2 pole Butterworth band pass filter for 18.1 MHz.
The center coupling cap is an air variable that was used since I could not find a
convenient value out of my minimalist selection at the time I built it.
The 17m rx first mixer and post amp are in the center.

As of early 2005, I have now essentially abandoned the Cohn topology (equal coupling
coefficients between resonators) in favor of real design. With the advent of GPLA
(General Purpose Ladder Analysis from W7ZOI), along with the work of Bill Carver, W7AAZ
and Jacob Makhinson, N6NWP, it is hard to justify not doing it right by designing a
real filter that can actually produce a legitimate Chebychev response (SSB) with 6
or more crystals and well behaved skirts, along with well behaved and controllable
pass band ripple.

If I had the task of getting a filter up and running in an hour, then I would want
to brute force it with the Cohn topology (partially emperical choice of coupling cap values).

However, a few hours of crystal characterization (extracting freq, motional L,
motional C, Qu) and simulation are well worth it to get 10 xtals with .1dB Chebychev
repsonse with only 2.5dB of insertion loss and good control of bandwidth that is essentially
independent of response characteristics (to a first order).
Or a real design for a proper time domain response ladder filter for CW purposes.
XLAD08 (that was included with the 1st edition of EMRFD) defaults to Gaussian to -6dB
with 5 xtals. The Gaussian response has good time domain response performance over Butterworth,
and even more so than Chebycheff.

My all band HF transceiver. This is a superhet rx with 9 MHz IF.
I use a 5-5.5 MHz Hartley VFO reference in an analog one-on-one phase locked loop
for high side LO injection for each band.

The RX front end is a BPF/LNA in to a level 17 diode ring with broad band NPN post amp.
A post LNA BPF is still needed here.

This rig uses IF derived AGC with a pair of cascaded MC1350s.

The TX chain is a diode ring DSB modulator with some gain feeding into the 9MHz
IF xtal filter. Broad band gain stages then feed in to a diode ring transmit mixer,
followed by a double tuned band pass filter driving more broad band feed back amps
in to a balanced pair of IRF511 HEXFETs.

All TX low pass filters are 7th order .1dB Chebycheff.
The low pass filters are relay switched.
The band pass filters are diode switched. This rig is still missing a couple
of crystals and needs some more work on the finals to be complete. Output is maybe
8 watts on 75m at Vcc=13.8v.

My 20 MHz xtal filters for a 17m SSB/CW transceiver.
I opted for 4 xtals, broad band amp and then 4 xtals.
They were surplus xtals and I found that a low side LO works
for 1.9 MHz LO and 20 MHz IF for receive but not for transmit.
I went with the low freq LO so that drift would be easier to manage.
It turned out that a low drift LO at 1.9 MHz was more of a nuisance than
I had thought it would be. This rig needs to be moved to another band or
get a 1/1 anaog PLL ugrade! And I should abandon the cascaded Cohn topology
and do a proper Chebychev response.