Updated 14 Nov (see bottom of page)
Kevin KB9RLW has posted an interesting series of videos to YouTube recently, exploring the poor SSB performance of his (tr)uSDX transceiver (beginning with this one).
I'm primarily a CW operator and wasn't too concerned about my (tr)uSDX's SSB performance but Kevin's video's piqued my interest and I decided have a look with my SA and an Elecraft Two-Tone Generator feeding a 700 Hz and a 1900 Hz tone into the mic input of the (tr)uSDX.
The purpose of this test is to measure IMD due to non-linearities in the transmitter. In a perfect world, only those two tones would be transmitted.
In our real world, the tones get mixed together and produce new frequencies that also get get transmitted. The amplitude of these frequencies, compared to the two originals, are how IMD is determined.
Typically, 3rd order IMD - those closest to the two originals - are 30 dB weaker than the injected tones. 5th order and 7th order IMD products are even weaker. I was hoping to post screenshots from the analyzer but saved them in Rigol's proprietary format rather than jpgs. But 3rd order IMD products are only 8-10 dB down from the intended sidebands.
In the (tr)uSDX - at least in mine, and quite likely in KB9RLW's and many others - there are numerous IMD products, and they are quite strong. This accounts for the splatter seen in Kevin's videos.
A comment from Manual DL2MAN, one of the (tr)uSDX designers, mentions that the IMD of the radio will improve if a narrower receive bandwidth is selected since receive bandwidth affects transmit bandwidth.
My original IMD tests were performed with the (tr)uSDX set at its maximum bandwidth of 4.0 kHz.
The screenshot below shows the difference in IMD at the rig's maximum and minimum (1.8 kHz) bandwidths:
As can be seen, IMD is indeed improved when the rig is set to minimum bandwidth, however these improvements are mainly 5th order and beyond. Third order IMD is essentially the same but the overall transmitted audio quality is improved by selecting a narrow bandwidth.