JT9 and 100 Hz ghosts

Multiple decodes at 100 Hz spacing of K1JT
on 30 m on 28 April 2013, 0101 UTC

From time to time I receive duplicate 'ghost' decodes at 100 Hz intervals on either side of the main signal. Last night I saw the phenomenon on 30 m. You will notice here that I have decoded the message: "TNX 73 GL" four times (press image for better readability):

• -24 dB, 1063 Hz
• -19 dB, 1163 Hz
• -8 dB, 1263 Hz
• -18 dB, 1363 Hz

The actual contact took place at the frequency of the strongest one, 1263 Hz. The station is only moderately strong at -8 dB and at +/- 100 Hz the first sidebands are 10-11 dB down and at -200 Hz the second one is 16 dB down.

In the post "Ghosts on JT9-1" Julian, G4ILO speculated that it had to do with strong signals. His best example showed a very strong example with signal strength as high as 19 dB and with the +/- 100 Hz ghosts 37-40 dB down, much more than in my example. A commenter proposed that it had to do with the mains frequency and that a North American station should have ghosts at +/-120 Hz instead. 

In his second posting, "Ghostly signals", Julian did some tests and found that a mains power supply can indeed result in weak 100 Hz sidebands. Then it was commented on the wsjtgroup that in the US people see these ghosts at a spacing of 120 Hz and that tests suggest that they are generated at the transmitter end.

But my example is from a US station, and the ghosts are still at +/- 100 Hz, so that theory does not seem to be right. My example indicates that it has to do with the receiver and not the transmitter. It comes and goes and the relative levels seem to vary a lot. My guess is that it is 50/100 Hz that enters the audio signal between the receiver and the PC. The software was the latest v0.9, r3195.

And by the way, the US station in my example happens to be Joe Taylor, K1JT, the originator of the JT9 mode and the decoding software.

---

See also "Overmodulated JT65 on HF?