Completely homebrew cheap MOSFET amplifier 300W

[mw0uzo]

Got another amplifier project on the go to learn more about them, this time made completely from scratch.It uses 8 x IRF520N mosfets and works up to 20m and will deliver a significant amount of power. Its designed for 250W, but I have had over 400W out of it!!

View of completed module
http://s14.postimage.org/meto9gf8t/SV203353.jpg (http://postimage.org/image/meto9gf8t/)

Drilled heatsink, nicked from an old audio amplifier. And thermal pads.
http://s14.postimage.org/n86a22l9p/SV203358.jpg (http://postimage.org/image/n86a22l9p/)

Underside of amplifier
http://s14.postimage.org/gk9ofh1rh/SV203360.jpg (http://postimage.org/image/gk9ofh1rh/)

Start of development of amplifier controller
http://s14.postimage.org/ba4po6hil/SV203362.jpg (http://postimage.org/image/ba4po6hil/)

The amp has an RX/TX circuit on there so is usable at the moment with a suitable LPF for the band required. Developing a controller with ADCs etc for monitoring SWR and current and TX sequencing. Once tested and working with this smaller amp, it will be used to control the big EB104.

Output transformer is made from ferrites that were clipped to power leads in a dead switch mode audio amp.

It runs off any voltage up to 30V, 24V has been chosen for high power mobile use at 300W. At 13.8V, over 100W of power is available up to 20m.

Total cost very low, £50?

[mw0uzo]

PM me if interested in the circuit diagrams :)

[KA9MOT]

Very cool!

[X-Rated]

With so many transistors, are there issues with phasing them optimally?

[mw0uzo]

With so many transistors, are there issues with phasing them optimally?

Just make sure they're all from the same batch. I have not seen any signs of one device conducting harder than the rest.

For reliable operation up to 30V, use an output transformer secondary of 3 turns. For low voltage e.g. 14v, high power operation use 4 turns then you can get 200w+ out at 13.8V.

The voltage limit is 30V due to use of a 5V regulator (35V max) in the bias circuit.

[mw0uzo]

Gain is in no way flat ... for 5w in, 300w out operation up to 20m I have had to include a switchable first order high-pass filter in the input attenuator. Not perfect, but it works :)

[X-Rated]

Just make sure they're all from the same batch. I have not seen any signs of one device conducting harder than the rest.

For reliable operation up to 30V, use an output transformer secondary of 3 turns. For low voltage e.g. 14v, high power operation use 4 turns then you can get 200w+ out at 13.8V.

The voltage limit is 30V due to use of a 5V regulator (35V max) in the bias circuit.

Beyond matching the transistors from the same batch, I was concerned about matching the phase inputs to the transistors as well. I am not sure what the schematic looks like on this thing, but the phase delays that occur will naturally be more obvious at the higher frequencies.

[NY3V]

Got another amplifier project on the go to learn more about them, this time made completely from scratch.It uses 8 x IRF520N mosfets and works up to 20m and will deliver a significant amount of power. Its designed for 250W, but I have had over 400W out of it!!

View of completed module
http://s14.postimage.org/meto9gf8t/SV203353.jpg (http://postimage.org/image/meto9gf8t/)

Drilled heatsink, nicked from an old audio amplifier. And thermal pads.http://s14.postimage.org/n86a22l9p/SV203358.jpg (http://postimage.org/image/n86a22l9p/)

Underside of amplifier
http://s14.postimage.org/gk9ofh1rh/SV203360.jpg (http://postimage.org/image/gk9ofh1rh/)

Start of development of amplifier controller
http://s14.postimage.org/ba4po6hil/SV203362.jpg (http://postimage.org/image/ba4po6hil/)

The amp has an RX/TX circuit on there so is usable at the moment with a suitable LPF for the band required. Developing a controller with ADCs etc for monitoring SWR and current and TX sequencing. Once tested and working with this smaller amp, it will be used to control the big EB104.

Output transformer is made from ferrites that were clipped to power leads in a dead switch mode audio amp.

It runs off any voltage up to 30V, 24V has been chosen for high power mobile use at 300W. At 13.8V, over 100W of power is available up to 20m.

Total cost very low, £50?

Is it Linear? If so, is it class E?

[mw0uzo]

Beyond matching the transistors from the same batch, I was concerned about matching the phase inputs to the transistors as well. I am not sure what the schematic looks like on this thing, but the phase delays that occur will naturally be more obvious at the higher frequencies.

I don't know. It's not really usable above 18Mhz. I have been trying to increase the frequency response but not got further than that. Any tips would be most welcome!!


Is it Linear? If so, is it class E?

Its class AB, linearity seems to be ok - it transmits SSB nice and clean with no splatter :)

[mw0uzo]

My rough design notes:

http://s15.postimage.org/ryztbyctz/ampbrew2.jpg (http://postimage.org/image/ryztbyctz/)

http://s15.postimage.org/flrkiscjb/ampbrew2_1.jpg (http://postimage.org/image/flrkiscjb/)

http://s15.postimage.org/5cz3cyohj/ampbrew2_2.jpg (http://postimage.org/image/5cz3cyohj/)

Questions or suggestions for improvements welcome :)

[NQ6U]

Remarkably simple device, Dan. Do you think it's a possibility that the phasing issues Jerry mentioned could be what's limiting it to below 18MHz?

[X-Rated]

I am seeing a tuned circuit on the input circuit. What would happen if you took that 220 pF cap out of there? It does not seem like it would be really broadbanded tuned like that.

[X-Rated]

I don't know what the inductance is on the primary winding, but it seems like there is an attenuator on the input with the pi circuit of resistors. Maybe that attenuation could be decreased if the amplitude drops on the lower bands with the 200pF taken out, but it should not roll off at the higher freq bands. If you need the cap in there, it might be best to switch out the cap for the higher bands and use matching caps to tune to the proper band between the inductance of the primary and the reactance of the cap. Ideally, the XL and XC will match and that is impossible between 1.8MHz and 30MHz.

[ab1ga]

I don't know. It's not really usable above 18Mhz. I have been trying to increase the frequency response but not got further than that. Any tips would be most welcome!!


The IRF MOSFETS were intended for use in switching applications like motor control or power supplies, where operating frequencies are low. Also, it was important to keep the channel resistance low when on to minimize waste heat, which limits lifetime and requires expensive heat sinking or a larger package design.

The high-current capability required a fairly wide channel, even for VMOS, TMOS, etc. designs. Put a metal gate across that, and you have a nice capacitor. That capacitance forms an RC circuit with any gate series resistance which limits frequency response. A related parameter, gate charge, provides a useful guide. The bigger it is, the harder it is to move it back and forth to vary channel conductance. The larger the gate charge, the higher the gate current (i.e., drive) required to modulate the channel conductance.

73,

[mw0uzo]

Remarkably simple device, Dan. Do you think it's a possibility that the phasing issues Jerry mentioned could be what's limiting it to below 18MHz?

I don't think so, I would have thought any troublesome phase shifting would have occurred at higher frequency. But that's just a hunch, some hidden inductance or capacitance could cause it. (hmm quite likely!!!)


I don't know what the inductance is on the primary winding, but it seems like there is an attenuator on the input with the pi circuit of resistors. Maybe that attenuation could be decreased if the amplitude drops on the lower bands with the 200pF taken out, but it should not roll off at the higher freq bands. If you need the cap in there, it might be best to switch out the cap for the higher bands and use matching caps to tune to the proper band between the inductance of the primary and the reactance of the cap. Ideally, the XL and XC will match and that is impossible between 1.8MHz and 30MHz.

Yeah the gain of the amplifier is a lot at low frequency, plus the input SWR is all over the place, so attenuator is necessary to reduce gain and provide decent input SWR.

The cap was put in to increase gain at 14Mhz, as it was rolling off quickly. With the arrangement shown, the output power is fairly level across 160-20m, with the switch for 20m.


The IRF MOSFETS were intended for use in switching applications like motor control or power supplies, where operating frequencies are low. Also, it was important to keep the channel resistance low when on to minimize waste heat, which limits lifetime and requires expensive heat sinking or a larger package design.

The high-current capability required a fairly wide channel, even for VMOS, TMOS, etc. designs. Put a metal gate across that, and you have a nice capacitor. That capacitance forms an RC circuit with any gate series resistance which limits frequency response. A related parameter, gate charge, provides a useful guide. The bigger it is, the harder it is to move it back and forth to vary channel conductance. The larger the gate charge, the higher the gate current (i.e., drive) required to modulate the channel conductance.

Yes. Reason for choice is low cost. IRF520N's were the cheapest, lowest input capacitance devices I could find, before getting to the point of headbutting the wall in frustration.

There is clearly a lot of capacitance somewhere in my design as the gain rolls off very early and more drive is required. It is probably what we expect, the mosfets, and tucked away in the bias circuit is a 33k/2n2 RC network which rolls off at 2.2Mhz. I did reduce R significantly, but then encountered some stability problems and lost a mosfet, so it was returned to its previous value.

I also experimented with the gate resistance, removing them gave a good rise in gain. Unfortunately, I also experimented with the output transformer ratio at the same time... and lost some more mosfets, and because of this silliness could not isolate the cause of failure. I now know the failure to have been down to the output transformer ratio and excessive current peaks through the mosfets (I was trying to get as much power as I could at 14V).

[mw0uzo]

(Heh, i'm still surprised I got any power out of it at all, let alone 400W!!! :D)

[NQ6U]

Limitations of the device aside, Dan, the type of experimentation you are doing is truly in the spirit of amateur radio.

[mw0uzo]

Limitations of the device aside, Dan, the type of experimentation you are doing is truly in the spirit of amateur radio.thx Carl :) Make one!

[PA5COR]

Remembering i made one with 20 2SC1969's i had cheap, broadbanded input and output using ferrite bars.

Did pull some serious current on 14.4 volts but did work up to 10 meters then, must have the remnants still laying around in the junk box somewhere...

Good try and for the higher frequencies i would opt for combining sets and combining these outputs again.
That prevents the limitations of the devices you run into now, the IRF 520 should work on 10 meters...

[mw0uzo]

Remembering i made one with 20 2SC1969's i had cheap, broadbanded input and output using ferrite bars.Did pull some serious current on 14.4 volts but did work up to 10 meters then, must have the remnants still laying around in the junk box somewhere...Good try and for the higher frequencies i would opt for combining sets and combining these outputs again.That prevents the limitations of the devices you run into now, the IRF 520 should work on 10 meters...So perhaps a 2(2+2) design with splitter and combiner to avoid multiplying up the gate charge/capacitance?

[X-Rated]

(Heh, i'm still surprised I got any power out of it at all, let alone 400W!!! :D)

Yes. I finally looked at the IRF spec sheet and 400W is a little over spec @ 48W a piece. Also the slew rate is pretty limited to a minus 3 dB at 20MHz. So the transistors appear to limit the frequency range as Dale said.

It is an amazing work though. Good on ya.

[mw0uzo]

Made a lot more progress on the controller since those photos were taken. Got the ADCs working, the 16 bit math working for SWR calculations, pretty power bars, plus some other stuff

Breadboard prototype
http://s12.postimage.org/cqseacryx/SV203370.jpg (http://postimage.org/image/cqseacryx/)
PIC: 16F887
Outputs for attenuator control, lpf control, status leds and a load of pots for simulating ADC input.
(The DDS module is a sideproject to write the code to control it)

Main display with detailed status monitor and menu
http://s12.postimage.org/qlva6kbrt/SV203374.jpg (http://postimage.org/image/qlva6kbrt/)

Main display with power bars and less detail
http://s12.postimage.org/6sj6duye1/SV203376.jpg (http://postimage.org/image/6sj6duye1/)

[mw0uzo]

Some more photos of the build:

Underside in progress
http://s15.postimage.org/m99f9uexz/SV203297.jpg (http://postimage.org/image/m99f9uexz/)

Bias pots and output transformer fitted
http://s15.postimage.org/fk2vtttlz/SV203298.jpg (http://postimage.org/image/fk2vtttlz/)

Heatsink fitted
http://s15.postimage.org/5bfgl4bvr/SV203300.jpg (http://postimage.org/image/5bfgl4bvr/)

Closeup of power distribution arrangement.
http://s15.postimage.org/6t0wwogmf/SV203301.jpg (http://postimage.org/image/6t0wwogmf/)
There were some issues with this. The closeness of the mosfets and rigid arrangement of the power plane did not allow for slight drilling errors, or placement problems when soldering on the ring connectors, The upshot was failure of the plastic insulating bushes when everything was done up tight and heated up. Next time I would allow a small gap between the mosfets and use braided copper for the power distribution.

[NQ6U]

How did you compensate for the stray capacitance created by using that "Manhattan" style of construction?

[mw0uzo]

How did you compensate for the stray capacitance created by using that "Manhattan" style of construction?

I didn't :lol:

[mw0uzo]

Seriously though, the mosfet capacitance would probably be the largest, small 'd'. Even though there's a big area of copper there for the bias circuitry the 'd' is huge, so it can't be significant amount. I don't actually know though....

[X-Rated]

I didn't :lol:

If you are building a design for a single frequency, it would be easy. Too difficult to compensate over a large range of frequencies without a different method of tuning it out for each band.