23cm 150+ W affordable Power Amplifer

When I needed a compact and small "driver PA" to power my large 23CM PA for the upcoming 4.8-meter dish for EME activities, I searched for a suitable solution.

The ICOM IC-9700 I use for EME operations can only output 10W on 23cm.Given that my 23CM PA requires significantly more power, and considering the 3 dB loss from the radio to the PA module mounted on the dish, the output power would end up being only around 5W at the input of the PA.

By introducing this small driver PA, the power could be boosted to over 150W if needed, well above the requirement for full output on the large PA, which only needs a maximum of 30W.

Various possibilities and modules are available, and after discussions with Mark 9H1BN, he shared information about an affordable module utilizing a surplus LDMOS device (MRFE6S9160) that he uses.

(This PA can of course also be used as a "final" PA for 150W+ on 23cm!)

This version of the PA design was crafted by Yuan, BG3MDO, and Feng, BG0AUB, drawing inspiration from the original work done by Henning, DF9IC.

(Please see the addendum by Dani YO5LD also)

Design by BG3MDO/BG0AUB.

Original design by DF9IC.

Another design using same LDMOS from IK3GHY.

500W PA with four LDMOS, design by F5JWF, and his tips and tricks

Datasheet for the LDMOS MRFE6S9160.

I can deliver blank PCB's for the amplifier for €5,- each (or you can have them produced yourself from the data available).

I can also deliver milled copper heat spreaders as shown below with "mirror polished surface treatment" for €44,- each. Find my contact details on my CV/Contact page.

First, I ordered LDMOS devices from EBay from this seller. I initially ordered 5 pieces, out of these only ONE device worked (and only marginal)! One of the remaining had very low gain, and the rest (three pieces) would not draw any drain current at all. It

seems from this (and from other peoples experience) that buying these from EBay is a bit like a lottery, I would NOT recommend to do that, unless you want to experiment and take some chances :)

Instead I got 5 devices from "UTSource", I was recommended this company by others: MRFE6S9160HSR3.

I swapped two of the "EBay" devices, and immediately saw 160+W out on the two test modules I had built!

The total cost depends on what you have in "your drawer", if you want to make your own copper heat spreader etc. In case you buy everything from new, the list will look something like this, depending on where you source components from:

LDMOS, from UTSource, € 19,-
Copper heat spreader from JLCPCB, US$ 24,- (select material "copper" and surface treatment "Mirror polished")
PCB from JLCPCB (at 30 pcs), US$ 1,- (a little more expensive at 5 pcs)
Components from Mouser, €15,-
Capacitors and SMA connectors from RF Microwave, €36,-
A suitable heatsink
A suitable power supply, capable of 28V/15Amp
Various hardware, box, wire etc.

Add to that shipping (Mouser is free shipping above a certain amount), if you do a group buy (increasing quantities), most of the items drop somewhat in price.

Design files:

STEP file for copper heat spreader, STP file
Info for threaded holes in copper heat spreader, PNG file
Drilling template for heatsink, can be 3D printed, STL file
Gerber files, package that can be uploaded to JLCPCB directly (select 0.8mm FR4, ENIG surface), ZIP file
3D printable file for heatsink drill guide, STL file
Drill guide for heatsink, PNG file

Design

The design is pretty straightforward. The only thing missing is temperature dependent bias. The design does not originally have this. Testing will show how much this is needed in my case, I plan only on running the PA at low power, so it might not be an issue. More on this when testing is done.

As Yuan/Feng has published the Gerber files for the 0.8mm thick FR4 PCB, I ordered a small batch of the PCBs from JLCPCB.

I designed a copper heat spreader for the board and had that one produced as well. I added two threaded holes at each end of the copper spreader so a SMA female connector could be mounted at each end.

I also designed a drilling template (orange 3D printed part on picture below) that makes it easy to mark/drill the holes needed to attach the copper heat spreader on the heatsink (file for this available above).

Below are some pictures of the parts, more will be added once I receive some LDMOS devices.

I have both empty blank PCB's and heat spreaders available if you need, look at the top of this page.

Addendum to design

Dani, YO5LD, reported that he had previously built the modules and he did see some high temperatures from the 100 pF coupling capacitor on the output (C6 on the schematic below).

He later on changed the 100 pF capacitor to two capacitors of 47 pF in parallel (0603, 250V) and did see a good improvement on the temperature.
I have not yet tested this if it is a problem or not in my modules, just putting the information here in case anyone else stumbles over the problem!

Schematic/BOM

The schematic and BOM information are both shown on the Github page of BG3MDO Yuan and BG0AUD Feng. I have included both below just for refence.

It is a good idea to mount a small Zener diode across C5, I use a 3.6V Zener diode (normal range for the bias voltage is around 3.0 to 3.3V for 1.3 Amp idle current @ 28V according to datasheet.

PLEASE note that C16 is shown as 1 pF in schematic, in BOM list and original article it is 2.2 pF (this is also what I use)

I ordered parts for the PA module from Mouser and from RF Microwave.

The BOM list for mouser can be downloaded from here (approximately €12,-)

Apart from the parts from Mouser and RF Microwave, you will need a small trim capacitor (3 pF), I ended up using a cheap foil trimmer for this on one module, and a 0.7 to 5 pF Sky trimmer on the other. Both worked fine, please notice that the adjustment is quite "sharp". By careful adjustment, I could get an almost perfect match on the input.

The Sky trimmers can be purchased from HF Berg in Germany. You need to send Michael an email about what you need, you can catch him on: info@hf-berg.de.

The trimmers are part number M3983, mention that to Michael when ordering.

Besides the components from Mouser, I ordered the two SMA connectors and the ATC capacitors from RF Microwave:

From RF Microwave (approximately €36,-):

1 pcs 100B-100P-J500

1 pcs 100B-2P2-B500

6 pcs 100B-3P3-B500

2 pcs SMA-73-04 (remove the PTFE sleeve and shorten the center pin)

You will also need two pieces of enameled copper wire, 1.5mm in diameter for the two inductors that supplies the drain with power.

Assembly

Assembly of the boards are very easy. There is not much to it, and everything is excellent described on the Github page of BG3MDO Yuan and BG0AUB Feng.

The PCB board is delivered as one single PCB, you need to split it in two. I used a sharp knife from both sides of the board, did a few cuts with the knife and could break the board in two. Do a slight sanding of the edges.

The two SMA female connectors are fastened using two M2.5 screws (4 to 5 mm long) at each connector.

I ended up placing the capacitors differently than on the following pictures. You can see some of the ATC caps are "standing on their side", this turned out to be a bad idea (lower output). I placed them "as normal" instead. Also I needed to move the 3 top ATC caps on the drain approximately 5 to 7 mm to the left to increase output. You will need to experiment a bit with this.

Below the next couple of pictures is a picture of the boards once they were optimized.

You can also see that I swapped the input trimmers with another type, and mounted it differently than indicated on the PCB. This gave (by far) the best results.

Soldering of the LDMOS followed the same method as I used for the large PA modules, I shot a video of the process.

Bringing up the board

On the Github page of BG3MDO Yuan and BG0AUB Feng there is a description of how to bring up the boards the first time.

DC check and idle current:

Make sure the bias trimmer is at minimum, you need the lowest possible voltage on the gate to start with.
Mount a suitable 50 Ohm dummyload at the output connector.
Terminate the input in 50 Ohm.
Connect a current limited power supply (28V at max 3 Amp) to the board.
Connect +12V (approximately) to the input of V1 (the 78M05 regulator) to turn on bias to the LDMOS.
Adjust the bias trimmer (slowly and careful) so that you see around 1.3A quiescent current (Idq) flowing into the drain.
Remove the +12V bias voltage from V1.

Adjustment of input match:

Re-apply the +12V to the bias input (V1). Confirm that the idle current is around 1.3A.
Using the 3 pF trimmer on the input, adjust to best possible return loss (SWR) on the input. You should be able to achieve a SWR better than 1:1.5.
Remove the +12V bias supply again (V1).

Testing with RF:

Adjust the current limit of the 28V power supply to 10A (if possible).
Apply +12V bias again (check that idle current is around 1.3A).
Apply very low RF to the input, observe the drain current and the RF power output (watt meter).
Slowly increase the RF input, at around 2W input, current consumption should be around 8 to 9A and power out should be 60W or more.
Remove RF and +12V bias.

You can try and move the two groups of 3xATC caps (3.3 pF) 1 mm left and right to optimize the output power (move one group at a time). Retest several times while optimizing the position of the 2x3 capacitors.

Once you go above around 90W output, you will need to increase the current capability of the 28V power supply from the 10A. At full saturated power (around 170/180W), current consumption will be around 14A (or more).

Gain will be around 15 dB.

Test of module #001

I did a quick test of the first board assembled (module #001), I got a max of 176W at 4W input and 12.4A (efficiency 51%).

This is using the LDMOS from UTSource, the ones from EBay I was unlucky with a few, so its a bit of lottery!

For my tests I used a 20 cm long piece cut from a Wakefield 125410 heatsink(1 meter long). Thermal resistance for this 20 cm long piece is 0.45 K/W in natural convection (1.2 K/W for 3 inch).

For the test, I used my 450W attenuator with a connected mW meter. The attenuator was before use, calibrated on my VNA.

Current measurements was done using a 1 mOhm resistor in series with the 28VDC supply, readout using a DMM6500 multimeter (1 mV/Amp).

On this I tuned for minimum indicated SWR on the built-in meter of my ICOM IC-9700. I connected my VNA after I was done with testing, just to see how the return loss was measured "correctly". It seems to be very nice, -31.2 dB

Test of module #002

Module #002 was tested, I got a max of 180W at 4W input and 11.7A (efficiency 55%).

This is also using the LDMOS from UTSource.

As I need to use a module as driver for a larger 23cm PA, I tested the power/gain at different Vdd voltages (this was done before optimizing the position of the 6 capacitors at the output/drain!)

One thing that still needs to be tested, is the quality of the output signal, this has not yet been done.

Using module #001 as driver at 12VDC

I did a test of module #001 running at only 12VDC input. In my 23cm PA for EME, I will be using one of these modules as driver as the IC-9700 will only be able to supply around 10W. To overcome cable losses etc, I will mount one of the modules as a "driver" in the PA box located on the dish.

You can see some final testing of a module used as driver running 14V supply voltage on my 4.8M EME dish project.

Mechanics/housing

I needed a small aluminium "housing" for the PA modules so I designed a box and cover for it. The files for both parts can be downloaded below (I might be able to deliver a small number of these if needed).

The housing uses two SMA connectors for RG-402 cable (.141 size).

Connectors used: 2 pcs SMA for RG-402

Two pieces of RG-402 cable (each around 40mm long).

60mm fan (many different ones can be used): 60mm fan

Guard for the 60mm fan: fan guard.

Feedthrough capacitor for Vdd: feedthrough capacitor.

Feedthrough capacitor for bias: feedthrough capacitor.

(other possibilities exists for the feedthrough capacitors).

18 pcs M2.5 x 5mm screws for cover and SMA connectors.

8 pcs M4 x 10mm screws for fan and guard.

8 pcs M4 locknuts for fan and guard.

10 pcs M3 x 10mm screws (plus washers) for attaching aluminium box to heatsink.

3D file for the box, CNC milled aluminium, 6061: STEP file.

File with info for threads: PNG file.

3D file for the cover, CNC milled aluminium, 6061: STEP file.

On the pictures below there is a temperature sensor installed, this is a DS18B20 based 1wire sensor (connects to my REPAM module)

Please also note that the shown heatsink is NOT big enough for full power from the PA module! It is only mounted on the small heatsink as the module is going into my larger 23cm PA as a driver (running only at 12VDC supply)!

Carsten Groen's (OZ9AAR) personal place