so i've been working on this direct conversion 40m transmitter for a few months now, mostly evenings and weekends when i get time. got the VFO stable enough that im happy with it, the crystal filter came out decent, but im having some issues with the final PA stage and wondering if anyone's been down this road before.

im using a IRF510 in the final, running it class C, getting maybe 3-4 watts out right now but the harmonics are pretty rough. like i know i need a proper lowpass filter after the final but even with the filter i built from a design i found in EMRFD the second harmonic is still a bit high. measuring it with my spectrum analyzer and im seeing maybe -35dBc on the second harmonic which i know is not legal for HF.

also the IRF510 gets warm pretty fast. not crazy hot but warm enough that i added a heatsink and now im wondering if i should bias it differently or if thats just how it is with that MOSFET at these power levels. the gate bias is sitting around 3.8v which seemed to work but i just kinda arrived at that by fiddling with the pot honestly.

anyone have experience getting cleaner output from an IRF510 PA or should i be looking at a different device? ive seen builds using the BS170 but thats for QRP stuff right, like under a watt. not sure where to go from here.

the IRF510 warming up is pretty normal honestly, class C biasing on those things is a bit of an art. -35dBc on the second harmonic is marginal yeah, FCC wants -43dBc for transmitters above a certain power level. your LPF design might be the culprit — what cutoff did you use and how many poles? a 5 or 7 pole chebyshev at like 8-9 MHz should be getting you well past that. also check your toroids, if the core is saturating or the winding isnt tight you'll get intermodulation garbage that no filter fixes downstream.

the gate bias thing — 3.8v sounds about right for class C operation but you really want to measure the drain current and work backwards. a lot of the EMRFD builds suggest biasing just below the conduction threshold and letting the RF drive do the work. might be worth spending an afternoon with a dummy load and a power meter just methodically going through the bias pot and watching output vs current. the IRF510 can do 10w no problem if you get it right, i ran one at 5w for years on 40m and it was fine thermally with a decent heatsink.

yeah the BS170 is def more of a QRP thing, i built a pixie clone with one ages ago and it does maybe 500mW on a good day. for your power level the IRF510 is the right call, most of the popular homebrew designs use it. i actually just finished a 20m rig loosely based on the NorCal 40A circuit and had similar harmonic issues early on. turned out one of my toroid windings was a bit sloppy and i was getting coupling i didnt expect. rewound it and things cleaned up noticeably. also double check your filter board layout, any RF getting back in from ground loops or whatever can make your analyzer reading look worse than the actual output. just something to rule out.