A
Homebrew Solid State HF/6m Amplifier
Project
by John DePrimo, K1JD
Background
Following tibia stress fracture and torn meniscus injuries late September of 2018, it quickly sunk
in that I was facing months of very limited physical activity. Seemed to be a very good time to
consider building something. For several years I’d been visiting the W6PQL amplifier site and
considered taking on an HF amplifier project using the now widespread laterally-diffused metal-
oxide semiconductor (LDMOS) technology. If not now, when? The idea of a 1 KW QSK amp
suitable for daily rag-chewing took shape. The Alpha 9500 is for heavier DX lifting when needed.
There is a lot of information about LDMOS technology on the Web, but for
this article I’ll mention but one aspect – heat dissipation. For comparison
consider an 8877 triode that has heat fins and uses forced air to cool it. The
anode fins offer a fair amount of surface area from which to remove hundreds
of watts of heat dissipation. In contrast, a LDMOS device (this amp uses a
BLF188XR, nominally 1.6” x 0.6”) is a relatively tiny device from which a similar
amount of heat must be removed. This is accomplished by using a thick
copper plate “heat spreader” to which the LDMOS device is flow soldered to
avoid any gaps, and attached to a heavy aluminum finned heatsink. The surfaces are machined
to be very flat, such that the two pieces of metal seem to stick together! This approach allows
the LDMOS device to work fine with small DC fans blowing air through the heat sink cooling
fins.
Visiting Jim Klitzing’s W6PQL website, you’ll find an impressive array of amplifiers from HF to
microwave. I can’t begin to estimate the number of hours Jim has invested in these products
but it’s got to be in the thousands. Jim offers a wide variety of module kits from which to build
your own amplifier, including a very capable control board. His 1KW HF version looked good
except for relay T/R switching - I wanted solid state, silent T/R QSK switching. It became clear
that my task would have 4 major parts to it:
1. The W6PQL 1 KW LDMOS amplifier, built and tested “as is” with Jim’s modules
integrated together
2. Design, implementation and testing a solid state 1KW QSK switch
3. Electrical integration of the QSK switch with the basic amplifier
4. The physical layout including internal metal work, obtaining a suitable enclosure and
front/rear panel designs.
This project spanned the 10-month period from October 2018 to August 2019.
1
Part 1 – The W6PQL amplifier
Part one went well, and quickly too. I had found a
suitable enclosure from Buckeye Shapeform in OH
(Jim W8ZR’s recommendation), and used its internal
dimensions as a guide by marking off a piece of
plywood – a true breadboard. The thought was that
by doing this I could transfer cable harnesses from the
breadboard to the final implementation. Seemed like
a good idea at the time, but there were too many
placement adjustments to realize much advantage.
The two major parts of the PQL amplifier were
ordered in early November: the RF deck and the Low
Pass Filter board (LPF). I ordered them as kits, and
found the provided assembly information to be
adequate.
By mid-November, the RF deck kit was completed and
testing had commenced. The top photo shows the RF
deck kit complete and mounted on the heat
spreader/heatsink, weighing in at almost 10 pounds.
In the middle photo, the test setup shown is for just
the RF Deck. I was able to confirm 1KW out into a Bird
dummy load. The small power supply on the left is
providing 13vdc bias to the LDMOS and running the 3
cooling fans (total of 93 CFM).
On November 24th 2018, about 3 weeks after ordering
the RF Deck and LPF, the LPF board has been attached
and also a Binary Coded Decimal (BCD) band decoder
board by Unified Microsystems interfaced to the LPF
relays and ACC jack in the K3. With this arrangement,
the LPF selection tracks the transceiver band
automatically.
Just a few days later, I designed and ran a partial
system test. My objectives were to make sure these
basic amplifier components were working properly, to measure the amplifier’s linearity and to
get a sense for Pin vs Pout vs Id (drain current).
2
Excerpts from a Power Point file I made to document these test results follow, including a block
diagram and sample data collected. Note how much power is generated from so little drive.
Partial Amplifier System Test
11/26/2018
. 12v
Band Select & 12v
BCD .
Decoder 12v
ACC:
3 Fans - 93cfm
BCD Out
0 – 1000 watts
. 12v K3
RF Drive Out
LDMOS
Pallet
12v bias
34” RG-142 LPF* Bird Bird
RF Out Wattmeter DL
12vdc
0.01 ohm
Supply 1% DVM ID amperes = voltage x 100
IDQ set to 2 amps
51.6 vdc
HP 50v/57a PS * LPF transfer function/band checked with
ESP120 Rigol DSA815 on 11/26/2018. Confirmed to be
close to W6PQL published responses
230 vac
K1JD 1
A surplus HP commercial supply rated for 50vdc @57 amps continuous duty 24/7 was
purchased. It’s an overkill for an amplifier rated at 50v/35 amps with a limited duty cycle, but
cheap enough. I bought 2 on eBay for $100.
A DVM was used to measure Id and Idq, drain current and drain quiescent (bias) current
indirectly through a 0.01 ohm 1% resistor .
Noted on the block diagram slide is that Jim W8ZR and I used his Rigol DSA815 sweeping
spectrum analyzer to measure the transfer function of the LPF’s on each band for comparison
to W6PQL’s published responses. We found them to be very close.
The length of RG-142 between the LDMOS deck and LPF is noted on the diagram as 34”. In his
provided documentation, Jim W6PQL explains that the LDMOS push-pull amplifier output has
very low even harmonics but higher level odd harmonics. By adjusting the length of the coax,
one can minimize the effects of power from these harmonics reflecting back from the LPF and
3
in doing so maximize power output out of the LPF. This cut-and-try process is time consuming
but worth the effort.
This sample data from
Power Out vs Input and Current per 160m and 80m testing
Band displays good linearity and
1000
160m
1000 40
high gain >300. In the final
1000
80m
1000 40
900
830
35 amplifier, I use a 13dB
900
820
34.9 35
attenuator in line with the
800 800 31.6
30.9
30 30
28.8
700 700 27.9
input to the RF deck, so a
26.5
640
25 25
600 600 600
500
400
20
mistaken 100w out from
500
400
20
the transceiver hits the
15 15
300 300
10 10
200
100
5
amplifier with only 5 watts.
200
100
5
0
1
Input Power (watts)
Drive Power, watts
2
Input Power (watts)
3
0 0
0.5
Drive Power, watts
1 1.5
0
Pout (Watts) ID (Amps)
Pout (Watts) ID (Amps)
11/26/2018
K1JD 2
The final effort in Part 1
was integrating the control
board that manages timing
for bias switching and the
T/R relays. Several dozen
QSOs were made using this
less than elegant
breadboard setup. The
relays were noisy enough
to further motivate the
development and
integration of silent diode-
based QSK, so onward to
Part 2.
Part 2 Design, Implementation and Testing a Solid State 1KW QSK
Switch
I had no intention of designing a QSK circuit from scratch. As it turned out, I used design
elements from the Ameritron QSK-5 and Elecraft KPA500. The final design is essentially that of
the KPA500 scaled for 1KW+ power output and with modified timing circuitry. Both QSK-5 and
KPA500 use switching diodes with low voltage (5-6 or 12vdc) to control forward bias, and high
voltage (450vdc) for the blocking bias.
After completing the prototype design schematic, I
ordered the parts and decided to build it using perf
board construction. Initial measurements of the QSK
switching logic from Mid-December 2018 to early
January 2019 looked promising.
The O’Scope photograph was taken while running 100w from the K3 through the switch. The
amplifier input/output connections were replaced by a wire (i.e., a unity gain amplifier) for this
initial testing. The RF waveform is contained safely within the T/R diode switching envelope –
good news!
The same timing control signals for T/R switching are also useful for the LDMOS bias switching,
so the decision was made to include the bias switch on the same board as the QSK switch. A
13vdc power supply was added plus additional circuitry to control its switching for the LDMOS
device’s bias. At this point in time, the +450vdc supply was cobbled together from parts
scavenged from my old QSK-5 and the two lower voltage supplies were small switchers.
5
Preliminary testing for RFI from the switchers didn’t reveal significant issues (spoiler alert – I
didn’t do a complete band scan in the preliminary testing!).
Encouraged by the satisfactory
switching waveforms, the prototype
QSK device/bias switch was interfaced
with the RF components from Part 1 on
January 11, 2019. By this time, multi-
segment LEDs , a SWR bridge and a few
other components to be used in the
final implementation were added.
Power was applied to everything, the
power out of the K3 was turned down
to zero and I sent a single dit.
Everything had been going so well, I
could not have anticipated that the
single dit at 0 watts drive would have
destroyed the LDMOS device, but it did. The old Doris Troy hit “Just one Look” from 1963, the
year I got my Novice license, kept playing in my head as “Just one Dit”. The culprit turned out to
be large switching transients (some >15vdc) on the LDMOS bias waveform that created a gate-
source overvoltage condition. These devices are rugged enough, but they do not tolerate this
kind of voltage spike. Since the device is flow soldered onto the copper heat spreader, I sent the
RF deck back to W6PQL for repair. His repair costs are very reasonable but added to the cost of
a new LDMOS device comes to about $250/failure! Avoid them if at all possible.
Now attention had to be directed at
understanding the source of the large
spike (yellow trace) on the switched bias
and investigate ways to mitigate it. It was
clear that the spike was somehow caused
by the high voltage (450vdc) blocking bias
switching within the QSK switch. About 6
weeks on and off were devoted to
mitigation of the spike that were
somewhat successful, but not wanting to
destroy another LDMOS FET meant taking
those efforts further. The decision was
made to transition from the prototype
QSK perf board to a proper printed circuit board (PCB) with a ground plane flooded on the
bottom side. Many larger components could be changed to high voltage SMDs thereby reducing
stray inductances.
6
Jim W8ZR has been using a program called CIRCAD by OmniGlyph to design PCBs for several
decades and he loaned me the design software. The software package doesn’t come with a
primer to get you started nor is the user
interface at all intuitive, so I had to impose on
Jim to guide me. In all, about 25% of the time
to complete the amplifier was devoted to
learning and using OmniGlyph. One begins by
drawing the schematic diagram with the tool.
The tool then registers all of the connections
and does its own internal error checks. Once
the errors are cleared up, you then change to
another part of the software package and
create the PCB layout. By the beginning of
April, I was ready to upload the PCB design
(Gerber Files) to Advanced Circuits at
https.4pcb.com located in Aurora, Colorado.
When the boards arrived on April 22 2019, I
was very impressed by the quality.
Populating the boards and through subsequent
testing, I was relieved to find that the switched
bias waveform was much cleaner (blue trace)
with just a tiny overshoot on the leading edge;
in fact, some of the components on the board
could be eliminated or modified. The voltage
spike problem on the switched LDMOS bias
output – that took out a LDMOS FET - has been
completely resolved by transitioning to the PCB
implementation. Time well spent!
Before integrating with the LDMOS amplifier
however, I hooked up the QSK “box” (board
plus power supplies were installed in an
aluminum Bud Box) to my Alpha 9500 to verify
that the QSK PCB board could handle the
power. The low voltage switched bias was not
used of course, and the 9500’s “amplifier key
in” was shorted for the test. The QSK box was keyed by the K3 and performed perfectly running
1200w out. Several QSOs were made at this power level and the QSK was very smooth.
Unfortunately, this is when I noticed some big frequency dependent RFI problems (I had not
checked every part of the bands) with the switcher supplies, so I tore apart the original QSK
aluminum bud box and built another using only linear supplies. The final configuration is shown
in the next part of the article.
7
Part 3 Electrical Integration of the QSK Switch with the Basic
Amplifier
This part covers the electrical interfaces between the W6PQL amplifier components and the
QSK box, revising the layout within the QSK box, and fabricating cable harnesses.
Preparing the QSK box for
integration meant having to modify
it to reduce/eliminate RFI either by
using only linear DC supplies or
finding quiet switchers. The former
was selected as the best approach.
The box was also configured to
come apart easier than its
predecessor by partitioning the RF
and power supplies on opposite
sides and using Molex connectors
to connect wiring harnesses
together. The QSK board is
mounted high on the right side and
the 3 linear supplies are mounted
lower on the left side. When the
box is pushed together and closed, the ground plane on the bottom of the QSK board
effectively separates them. Testing confirmed that here was no more RFI of any kind. This is the
final QSK Box configuration that has been integrated with the amplifier.
Marrying the electronics in the QSK box to the W6PQL amplifier components is the main point
of Part 3. Here are some of the decisions that were made:
- The PQL input and output T/R relays, instead of being controlled by the PQL Control
Board in a “semi-QSK” fashion when the control board is keyed, are latched closed by a
front panel pushbutton switch selecting “operate” and released in “Bypass”. In Operate
mode, the QSK box amp key line (not the control board amp key) is keyed and
accomplishes the T/R function electronically and silently.
- The PQL control board controls the heatsink fans through a ground return. They are
nominally off until the heatsink reaches about 110 degrees F and the ground return then
turns on the fans. I wanted the fans to run continuously at low fan speed and kick into
high speed when needed. A 15 ohm power resistor was wired in series with the fan
return lines. A simple MOSFET circuit when turned on though the PQL control board fan
return shorts out the 15 ohm resistor, running the fans at high speed and then returns
to slow speed after the heart sink cools down. Perfect.
- The PQL control board responds to various signals such as high SWR, high temperature
and other faults within the amplifier by disabling the “amp key” line within the control
8
board; but wait, the amplifier key line does not run through the control board at all, it
only keys the QSK switch! Fortunately, every fault detected by the control board is
represented by a common signal going low to ground. I had connected the various
sensors distributed throughout the amplifier to the control board, so I needed a way to
use the protect circuits’ signal to ground. Instead of
disrupting the control board key line, what if the fault
signal going low opened the T/R relays and dropped the
amplifier into “bypass”? This was indeed the answer, and
seems to work flawlessly. There were an extra set of relay
contacts on the low power input T/ R relay, so I routed the
“amplifier key in” signal through them as well. As soon as a
fault is detected, the key in line is also disabled as the
relays open. The photo shows the 2 MOSFETs for fan speed and fault handling installed
on a perf board above the W6PQL control board in the front partition of the amplifier.
Part 4 Physical Layout, Enclosure and Panel Design
Early on, the expected contents of the amplifier enclosure were studied and enclosure product
DII-87-4-13 from Buckeye Shapeform in OH was chosen, its 17”/15.8” width x 13”/12” depth x
8.7”/8.4” height external/Internal measurements seemed adequate. I thought that my choice
would leave a generous amount of open space inside the cabinet, but that was not so: Good
thing I didn’t try to go smaller. The cabinet arrived on 29 January 2019, around the same time
the repaired RF deck arrived from W6PQL.
Along the way, I’d take some of the components from Parts 1-3
and fit them inside to get a better sense for what might be an
optimal layout. One could not count on the lightweight
aluminum bottom shell delivered with the cabinet to support
the weight of the amplifier components, so I began reinforcing
the bottom with angle aluminum and bar stock. This photo was
taken on February 15th and shows the repaired RF deck/heat
spreader again mounted on the heat sink along with the fans
resting vertically against it and
the blank back panel in place. The
aluminum bud box on the left
would house the QSK board (the
perf board at this time) and the
switcher power supplies that
were later replaced. With that
major portion of the cabinet
occupied, the LPF seemed best installed vertically against the back of the right side. The fans
were mounted using channel and angle aluminum stock. The way the layout was evolving,
maybe the back volume of the cabinet could be dedicated to the RF components, a vertical
9
partition installed in front of the fans with outside air coming from
a rectangular hole in the bottom of the cabinet, and the resulting
space in front for control-like components and front panel parts.
This approach was not a dead end, fortunately, and proved well
suited to housing the rest of the amplifier. In this photo, the
amplifier is essentially complete. The QSK box shown was replaced
with the newer version using linear supplies. The partition behind
the front panel has the control board, MOSFET custom circuitry
and BCD boards attached to it on the front panel side. All of the
front panel switches and component are ready to install. This
brings us to the front and rear panel designs. The company “Front
Panel Express” has been around for a long time and I had
previously used them for a tiny custom panel that fit into a Pelican
micro-case for one of my
SOTA radios. Their design
software isn’t particularly
difficult to use, and so
both front and rear panels
were designed and
submitted for them to
manufacture. The layout
on the computer screen
was the information Front Panel Express used to make the black panel. Since all the front panel
component harnesses had been previously made, everything just needed to be placed in the
proper mounting holes.
I had decided early on to not display information to a high degree of resolution, choosing
instead to, at a glance, make sure the amplifier is operating within a “safe operating envelope”.
The first red LED segment in each of the 10 segment displays, left to right, indicate 50vdc, 35
amps, 1KW forward power and 100w reflected power.
The back panel was a little easier since it didn’t have to
look “pretty”, just allow for the needed connections and
for the heatsink exhaust air.
The amplifier has been essentially completed and in daily
use for several weeks now. It’s a joy to use with its silky
smooth QSK, power delivery at low drive levels, and auto-
band switching with the K3.
73,
John K1JD
10
