QRP Transceiver 40W Power Amplifier

Updated 201601016
There is no major advance to be made in class AB amplifier design, but improvements in economy are possible. Problems with designing for the HF frequency range are:
  • RF FETs are expensive and often difficult to source
  • The industry is producing transistors for VHF and above…
  • Which means those for lower frequencies are expensive
  • Switching FETs are cheap but have many shortcomings
  • Linearity issues
  • Efficiency issues

FETs designed for switching power supplies have high parasitics and poor thermal characteristics. Nonetheless I embarked on a project to find a usable part, and make a low-cost amplifier using it. Blank PCBs for the project are as shown here.
Aug2016 PCB
The amplifier section is on the left, filter to the right. If anyone wants a low-pass as a separate project, it can be split off by cutting along the dotted line! I am beginning to build with the amplifier as the first block, testing to see which of the power FETs selected perform best.

The most common modern device for HF amplifiers of about 50W is the
Mitsubishi RD70HHF1. They have some disadvantages:
  • High cost >£25 each in small quantities
  • Output impedance too low for an efficient 1:4 output transformer
  • Doubtful linearity at 12V supply

Having discounted the conventional wisdom power devices, I looked through the enormous range of switching FETs available. These are the challenges of designing with cheaper devices:
  • High gate capacitance (except early generation e.g. IRF510), makes a flat 1-30MHz response impossible
  • TO-220 packages have inductive wire bonds
  • Switching FETs hotspot badly and the bias point is unstable…
  • Fairchild Semi agree
  • OnSemi agree
  • Infineon agree
  • Microsemi agree
  • Temperature compensation and other measures are essential

I found these are most suitable in terms of medium gate capacitance, low-ish transconductance and good thermal conductivity:
STP14NF12 (TO-220)
FQP13N10 (TO-220)
STW13N60M2 (TO-247)

The first two are similar, with the ST part having slightly higher V
DSS rating, and the Fairchild part having a track record in citizen band radio output stages. The STW13N60M2 is the lowest gate capacitance available in TO-247, and it's interesting to test a high voltage part against the older generation 100V ones. I spent considerable time looking on manufacturer's websites to find these. Silicon technology is reaching the limits of what is possible, and very few low gate capacitance (<500pF) parts are available.

Even these low capacitance FETs have higher transconductance than RF FETs. So they are more sensitive to bias point when running in linear mode. How much temperature compensation is needed to stabilise them can be estimated from the FQP13N10 data sheet which has a standard graph of V
GS against Id and temperature. It comes out as 0.45V/150C = 3mV/C.

3mV/C is above the temperature coefficient of a single diode but less than two. Having good thermal contact between a plastic FET and diodes is difficult. Hence getting a rapid response of the compensation circuit is difficult. Compensating the bias has a limited effectiveness agains thermal instability. The no-signal bias must fall well into the safe operating area, but dropping the bias too far doesn't reduce dissipation under high signal very much. It has the disadvantage of compromising linearity. So I include bias stabilisation but became aware of its limitations.

Besides the bias, a FET can hotspot and burnout under continuous high signal. So any amplifier like this must be rated in SSB for higher power than in CW modes like PSK31. Having neutralisation (feedback) from drain to gate is actually a form of bias. Having this allows the no-signal bias to be lower, giving the FETs a chance to drop well inside their forward bias safe operating area during speech pauses. The disadvantage is reduced gain and lower efficiency. I decided to use a current limited power supply and intelligently reduce bias using software.

Another concept is using FETs at higher voltage than 12V, because the output impedance at the FET drains is set by the supply voltage. I looked at designing a boost power supply using
TI WebBench. The cost of a 30V, 5A boost converter makes up half the difference to using RD70HHF1 FETs in the first place, let alone the extra constructional complexity.

A solution was found in "250W" boost power supplies from China/eBay. These can put out 150W if heat-sinked, and are voltage adjustable, with current limit for under £3.50! With such a cheap way to get 33V @4A, plans for 50W+ amplifiers were abandoned, and I decided to aim for 30-40W output. The Chinese boost module concept allows a part cost of under £50 for a 40W amplifier (excluding heatsink) and these advantages over of-the shelf HF amplifiers:
  • Operate at the same output power with 10-20V input voltage
  • Improved efficiency output transformer
  • Improved linearity due to higher supply voltage

It's still impossible for me to compete with Chinese amplifiers based on surplus RF transistors. It has to be remembered that the majority of devices sold on eBay do not reach their specified output power. I decided to add a number of extras which push up the cost but make a complete system - RxTx relays; Low-pass-filter; SWR bridge; SWR monitoring; micro controller for automatic band switching and protection.

The circuit diagram is
here and here. I found FQP13N10 FETs the best overall performers. The larger TO247 FETs had some bizarre bias quirks, and are 3x the price of the smaller ones with no benefit in performance. The STP14NF12 has a tendency to self-destruct which I never fully understood but will avoid. Here is the amplifier under test, with the Chinese boost module mounted next to the main board.
This project is complete as a technology demonstrator for low-cost FETs with micro-controller enhancements. Further improvements will take one more PCB revision and include:
  • Use ready wound baluns for the input circuit
  • Use larger, cheaper cylindrical ferrites for output circuit
  • Change 74AC74 to single D-flip-flop gate
  • Use same 3.3V reg for micro supply and bias
  • Add fault protection diodes to gate bias circuit
  • Use 1206 100 ohm resistors in SWR bridge
  • Multi-turn pots for bias adjustment
  • Optimise all SMD pad sizes

Most importantly, design to fit in a practical case, 160 x 100mm. This means having the LEDs on the "short edge" and a PCB mounted BNC socket to fit through an end panel. The Hammond case allows the PCBA to slot in and the FETs to fix onto a heatsink. That case allows room for expansion, to add a battery board or other advancements as described next.

Further thoughts on low-cost HF amplifiers

The real snag with all the amplifiers on this page is their inefficiency. Cheap FETs make this problem worse due to their poor heatsinking. Commercial equipment like 3G base stations have moved to other amplifier topologies to improve efficiency. But schemes like Class-F only improve to 75% at best, with a lot of extra complexity. Doherty amplifiers improve to 80% but are even more complicated, and limited bandwidth. Amateur HF amplifiers primarily have to perform well in SSB mode which is very different to modulation used in short-range mobile infrastructure.

A simpler scheme is modulating the supply voltage. This needs some kind of feed forward dynamic adjustment of the switching power supply. It sounds feasible, especially as the bandwidth of SSB is quite limited at HF. I am looking into this as a plug-in option to the current design. Designing a power supply for 4A , 30V with >1MHz switching is feasible. High switcher frequency is needed to enable a low-loss filtering scheme with a cutoff at several kHz.

Returning to heatsinking, we note the tab insulator will never be so good as fixing the TO-220 directly to a lump of metal. Designing a circuit with the FET drain at ground probably requires a negative supply voltage. As I'm proposing a switching supply anyway, it's an obvious step to use an inverting buck-boost. It turns out the inverting buck-boost is more elegant than a non-inverting!

By definition the drain at ground means the source is floating, with RF on it. A way has to be found to drive the gates to both FETs against the floating source(s). This can be done with a transformer with 3 windings. For the gate bias a photodiode coupler can be used, which gets around the problem of floating with RF on one side.

Another possible future architecture is a class-D (switching) scheme. Unfortunately the technology to switch high power at >50MHz is not quite here yet. Far as I know, the best FETs (even GaN) can't sink enough power or switch fast enough to provide a 20MHz class-D with 90%+ efficiency. Especially in a low cost unit. But as a curious designer I have to keep a close eye on the possibilities in this direction. A scheme based on self-oscillating feedback with constant frequency as used in the latest audio amps might be usable…

No products using any of these concepts are sold commercially. They all play safe and use Mitsubishi FETs running off 12V. A definite opportunity exists for experimentation in this area…

20 Watt Amplifier Iss.C1 (2012)

Previously I designed successive amplifiers to produce 15-20W of RF power from less than 0.5W of drive. Targeted at Soft-rock project builders, though adaptable to boost other QRP or home-brew rigs to the 15-20W level. I added several unique options. All details are in the public domain. Anyone that wishes to build this design, get commercial standard PCBs made, or sell kits, is welcome to do so.

The final batch of PCBs for these boards is now sold out. This is the prototype undergoing tests:

There is also a mini-movie here. It is being driven by an Elecraft KX3, set to 0.3W output. Voltage supply and quiescent current are seen on the power supply. In the foreground is an SWR/Power meter with a dummy load sat on top. The power meter indicates 15W output from a “whistle test”. The next photo shows the completed board mounted on a heat-sink plate:


It captures several years of design thinking, and has features not found on other similar amplifiers:
  • The power MOSFETs have dual footprint, for Mitsubishi RD16HHF1 or power supply FETs
  • Input PI network for matching circuit or attenuator
  • Temperature compensation of bias to help stability
  • Facility for single or multi-band filter
  • Integrated RxTx relay with pull-low or pull-high switching voltage
  • Options for external filter banks

Datasheet parameters to look for in switching FETs are low input capacitance (<350pF), Vds max >40V. Package TO-220 or TO-247. Most MOSFETs optimise RDSon and have high input capacitance. Check websites of ST, Fairchild, NXP, Texas Instruments, On Semiconductor, International Rectifier, Vishay, and others. New types of MOSFETs often come onto the market.

There is a matching multi-band low pass filter
here. The Kits & Parts single band filter is no longer available. I have some replacements.

The bare PCB:
There are several options for external filter banks on the amplifier PCB. An external higher power amplifier can also be used. Look at the circuit diagram and see what is possible. That’s why there are several coaxial termination pads on the board. Anyone that builds this amplifier, then wants to cover more bands with automatic switching, can add an external board. If there is demand I can design a suitable filter bank. My previous design had 4 filter banks, and worked fine from 80m-15m.

If you want better efficiency on the bands <10MHz it is possible to use a BN43-202 twin hole ferrite rather than the type 61 ferrite.

Also there’s an option for a TMP100 temperature sensor, controlled by I2C. MoBo V4.3 users may find it useful to read the output PA temperature instead of the MoBo PA which becomes the driver stage.

Build Data Release (Iss.C)

Gerber-X files (issue C) are found
here. Have a look at them with a free Gerber viewer like PentaLogix SmartDFM. I included top and bottom silkscreen, but to save costs the bottom silkscreen is not essential. Board size is 100 x 60mm. Recommended PCB type is 1.6mm thick FR4, plated through holes, hot air solder level (HASL) finish, 1oz copper. Top and bottom solder masks are required. Any PCB manufacturer who has a clue will accept those files.

circuit diagram is here, the bill of materials (BOM) is here.


Development & Performance Notes - Mitsubishi FETs
At the outset I spent a while surveying available power transistors, before settling on the RD16HHF1 from Mitsubishi semiconductors. These are used by Icom, Yaesu, Kenwood, Elecraft…

The concept was based on the transformer layout by G6ALU which is functionally equal to other transformer layouts, and uses easily available BNxx-202 twin hole balun cores. I made a first attempt with the issue A board, superseded by the issue B which was sold as a kit. It had many features to interface with Softrock RxTx V6.3 and the Ensemble boards with a few component values different.

The power output was near 20W on the lower bands, dropping to 15W on 28MHz. On my PCB the positions for RD16HHF1 connections are indicated on the silkscreen layer if using a commercial standard PCB. The drain and source pins are reversed between the Mitsubishi FETs and switching types.

When using Mitsubishi FETs, performance is equal to other amplifiers, such as
PennyWhistle from HPSDR. The original page for the Pic-A-Star amplifier is still available here. The Mitsubishi FETs are very robust devices. The RD15HVF1 from Mitsubishi is also suitable, and gives slightly better gain at 50MHz.

Development & Performance Notes - STP16NF06L FETs
Switching FETs are much cheaper, but how is their performance compared to the “RF” parts?

A major physical problem with switching FETs is the tab is drain. So tab is “hot” in voltage and RF terms, but has to be heat-sinked. I recommend a relatively thick heat transfer washer without heat transfer compound.

Before fitting the bias network resistors, check the V
GS threshold of your devices. Resistors at top and bottom of the trimpots VR1, VR2 are there to make adjustment easier. So calculate the centre point of VR1, VR2 to be at the point where the device just switches on. Unfortunately switching FETs are designed with a very sharp DC switch-on point, as required in power supplies. This can make the bias unstable. I found the STP16NF06L to be very sensitive to adjust. Despite temperature compensation they will run-away and overheat. Always watch the current drawn by the circuit with a multimeter when adjusting, and observe the variation over several minutes.

An early prototype board undergoing bias adjustment and input network analysis is pictured next:

Cheap PA under adjustment

With the TE Axicom relay, quiescent current is 80mA at 13.6V. It is safe to put 200mA quiescent through each MOSFET when heat-sinked. The input SWR and of course the gain is affected by the bias current. Before running power tests, I looked at the response of the Kits & Parts 20m band low pass filter in-situ as in the picture. The first plot shows SWR and S21 (thru) response with the 20m filter plugged in.

20m filter


The second plot is the input with resistors added to bring up the impedance closer to 50 ohms. It is possible to experiment with the input PI network and achieve better matching, though the input SWR is reasonable and will not be a problem for most driving stages. Certainly the Softrock output stage can drive this without problems. It is worth noting the high SWR on the right of the plot is caused by the low pass filter response, as expected.

The big question is
how good is linearity of these FETs in two tone intermodulation? Running into a dummy load with PSK-125 and listening to the signal on a receiver, while varying the power gives a rough idea. The signal sounds the same from power outputs of a few watts up to 25W at 14.070MHz. There is also no detectable splatter with the receiver.

My test equipment does not run to a spectrum analyser. I only have the SDR-Kits VNWA network analyser, which is brilliant at its primary function, but the spectrum analyser has a minimum resolution of 250Hz. I also have no proper signal generator, let alone the two required for an intermodulation test. The best I can do is generate a PSK-125 signal and sniff the output on the VNWA used in spectrum analyser mode.

Overall gain with STP16NF06L is actually higher than RD16HHF1 on the lower bands. If you only want to operate <10MHz they would be better. But see the compensation notes further down this page.


In a PSK-125 intermodulation test, this is the result from the RD16HHF1 amplifier driven to 16W (post filter) on 21.070MHz. Two tones with no characters transmitted during analyser sweep.
Next the result of the STP16NF06 at same power level. The “shoulders” relative to the main signal are lower than the RD16HHF1 result. The “shoulder” is about -22dB and -25dB respectively, representing intermodulation products from the amplifier. The input power at 250mW is well within the Softrock Ensemble PA linear region.


Driving either type to an output of 25W makes the “shoulders” rise up towards the main signals, to within -10dB. This expected result shows the test is a valid one. Surprisingly the cheaper MOSFET has a better intermodulation performance than the expensive “RF” part. The Mitsubishi RD16 die is probably a standard FET die wire bonded differently to get the tab grounded. There is no magic in its construction.

The ST FETs drop off in gain more than the Mitsubishi's as frequency is increased. At low frequencies they will drive to maximum output (>25W) with <0.2W. But at 10m, you will need 0.8W to get 20W out. It is possible to put a compensation network in which will flatten out the gain considerably. With a bit of simulation and testing, the values are:

  • A 150pF leaded capacitor at the R7 position. There is a grounded via provided in the middle of the R7 position to fit one leg of the capacitor into.
  • A 180nH air spaced inductor at the R1 position. 180nH can be made from 7-8 turns of 22SWG wire wound on a 6mm former.

A picture of the frequency compensation circuit follows.

It is best to start out with 8 turns and adjust the power output on 29MHz. If you find spacing the turns right out increases power, then desolder the coil and remove a turn. Gate capacitance varies widely on these FETs, it is best to adjust your own board individually. Also check the gain on 21MHz is peaked up. With the compensation network, I find the maximum power output to be 25W on 29MHz, with 1.0W input. At 3.5MHz it gives 30W for 0.5W input.

In conclusion, I find the STP16NF06 to be a capable and cheap amplifier device for HF transmitting.

Development & Performance Notes - Fairchild FQP13N06L FETs
An extensive survey of TO220 FETs shows the lowest gate and output capacitance is the Fairchild Semi FQP13N06L. They are cheap, so I tested a pair. With a dummy load, using the Elecraft KX3 as a signal source in FM mode at 0.3W. Quiescent current 700mA combined. Power out was:

7MHz 25W
14MHz 18W
21MHz 14W
29MHz 10W

So the gain drops a lot with frequency which is expected. The gain can be compensated in the same way as for the ST MOSFET. After a lot of tweaking about, I will make the statement that there is nothing to choose between the ST or Fairchild FETS. Bearing in mind the output side is dummy load, the input SWR looks like:

The FQP13N10L is on paper the lowest gate charge TO220 device available. They are often found in the PA stages of 27MHz CBs.

Builder’s Notes
Some of the previous notes are still valid for winding the transformers and setting up. The original builder’s notes file (PDF) is available here. The STP16NF06 or switching FET version only differs in the DC bias (gate) voltage needs to be set lower than the RD16HHF1 version.

The Softrock Ensemble should be arranged to switch the amplifier to transmit by connecting the /PTT line (U4 pin 4) to the J9 pin 1 of the amplifier. So pulling down the base of U3 on the amplifier when the Tx is keyed. Coax input at J11 can be connected straight from the Ensemble RF output. Details of how to terminate coax tails is in builders notes.

Previous 20Watt Amplifier Issue B (2009)

My previous project was an amplifier with integrated filters,
circuit diagram here. Design of HF amplifiers using Mitsubishi RD16HHF1 MOSFETs is old news now. The ones I built in early 2009 are still working reliably. This was the last “issue.B” unit, all of which are now sold !