This is probably easy for those that might have done it. Maybe?
Got an RF output at 7Mhz c400W using one of those broadband dual SSPA LDMOS things, NXP sent me some nice samples even though explained I was qrp, anyway pretty sure there might be some mixing product output at 3.5Mz and 14Mhz as well as higher (no surprises really) before necessary filtering. Exciter sending either single carrier or 2 tone stuff (I have something was from Far Circuits that can send a matrix of up to 8 different tones in pairs). So just to see how effective different filtering options were as I originally though the extra signals/ 'problem' was due to a cheap audio amp 65V SMPS being used but it wasn't as I borrowed an expensive supply and saw the same issues.
I was messing about with use of various filters between the exciter c1Watt and different 1st stage transformers into the amp the later seemed to make a huge difference as do coax stubs (not terribly practical in a PA even if just a coil of RG174) but pretty sure even though well down the mixing product signals are just detectable at the output. Definitely well down on the primary and pretty sure they comply (c-50dB) but trying to work out exactly how far 'down' they are and any other mixing products as working out RF budgets. Having been qrp for years you don't really want a readable signal either.
I first noticed I could detect 'something' on a PHSNA (see this blog, https://groups.io/g/PHSNA or Google for it, another ARCI/GQRP homebrew thing) so decided I had to to see what it was and where.
Not having a £££ HP spectrum analyser to hand I decided to use an RSP1 (original version), with the latest spectrum analyser code, and sampling the output using, originally the head from a homebrew LP100A watt meter but found the signals seem better (and I can muck about with it without worrying too much about the commercial head) with a simple T68-2 transformer (about 30-40 turns of 18-22SWG c90% of the core is filled when turns spaced out) with the primary the center of a length of RG213 all on some PCB in a dicast box. This feed goes to a nice 5 chip resistor dummy load suspended in an olive oil can. All in all it looks just like one half of your typical vswr meter. I used my old 500Khz NoV current and phase meter on the feed in and out more for curiosity to see if I had attenuated any of the output signal before and after (no surprise no noticable effect). The max signal off the transformer based on a 400W RF carrier is about 5mV give or take, I might remake it to lower this but need to be sure the harmonics are still readable (if there). The harmonic signals are several magnitudes lower in the uV range. You can still 'see' those signals on the RSP much clearer as the noise floor is much lower than the PHSNA (I noted the background noise floor was almost 10dB higher when up in Cambridge). The exact transformer ratio (or the maths calculations anyway) wasn't that important just as long as the voltage produced is well below what the RSP front end can reasonably handle. It didn't look like enough but I didn't want to attenuate ALL the signals so using some sample filters I had been thinking for a use for (just like those TOKO cans) I created a band stop that attenuated it further at 7Mhz, I wasn't worried by how exactly much as long as well well below the needs of the RSP1 front end and the signal peak was well down. As the RSP and software combination was not really calibrated (I guess you could), I used a separate signal generator (a USB board with an Si5351 thing I think G0BBL gave me a few years back) to select the exact same frequency and match the amplitude on the SDR with an adjustable output (simple pot really) and then using a cheap chinese AD8137 RF detector (temporarily borrowed from the QO100 project!) and DVM (the AD amplifies the signal back up again by about 55dB) so you can roughly calculate the RF power levels. I hope that is simple enough explanation? The original voltages are down at the uV level I might rewind that transformer to increase it bearing in mind that the fundamental increases too.
I am sure lots have done something like this themselves which investigating HF amps and harmonics. I might be interesting to see any comments? Especially anyone that does this professionally.
I assume I might be able to work out some of the third order intercepts etc if I can crack this.
Not sure I'd ever use the finished PA as intended but this is an interesting albeit perhaps academic exercise all the same.