Using a spectrum analyser to see if a low frequency TX signal is clean?

[Chris Wilson]

I am not very good with spectrum analysers, so need to ask for advice. I want to look at my signal on 137 kHz (UK LF ham band) from my amp in real time, into my real antenna, not a dummy load. So I stuck a small rubber duck antenna into the input port, with no DC block as some have attenuation on lower frequencies I believe.

 I then transmitted WSPR2 into my real antenna. I see the signal on the SA and played with bandwidth settings. What I am unsure about is how much signal strength do I need to capture? The TX will run 1kW into a dummy load, no problem, but antenna efficiency is of course pitiful on LF given what any normal person can erect. If I hold the rubber duck when TX'ing some sidebands show, well down, but hardly anything if I don't "help" the rubber duck. How do I decide on and acquire enough signal please?

 I have some screen shots of my signal on local LF grabbers, and on the SA. The scope shot is via a "scopematch" where the 2 traces show V / I balance as being near perfect, so the matching into the external loading coil and antenna is excellent. The amp's meter shows near zero reflected power. I *THINK* all looks OK now?

 I have had a recent report of a very dirty signal filling the WSPR waterfall from 2 UK stations. I have run this thing for 18 months with good reports, so something recent has occurred, and I found the sound card level from WSPR to my Kenwood TS-590 exciter had mysteriously changed to maximum. Re-setting it gave a clean sine wave from the pre amp into the main amp. With too much drive into the pre-amp it was a very distorted output. The screen shots show things after I reset the drive level correctly.

I think I have it sorted, but my lack of SA technique causes me a little doubt. Thanks for looking!

The screenshots are at http://www.chriswilson.tv/signals/signals.html   

and at  http://www.chriswilson.tv/signals2/signals2.html

[Paul Price]

You only need to capture enough signal strength to give you something like a 2:1 ratio above the "mud." noise floor.
 Anything more is not needed, just a scaled up sample.