Ghostly signals

I have been puzzled and mildly irritated by the 'ghost' traces that appear at plus and minus 100Hz when I receive a very strong signal with WSJT-X.  Torben OZ1TMK wrote to me after we had a JT9-1 QSO to ask if his signal had been OK. He had received requests from a couple of local stations to reduce power because he was 'causing harmonics'. I hadn't noticed anything wrong with Torben's signal but it hadn't been strong enough. It looked to me as if it was the same effect I and a few other JT9-1 users had observed when very strong signals were received, so I decided to investigate.

I had a theory that the +/-100Hz spurious outputs ( +/-120Hz observed in the USA) were caused by ripple modulating the transmitted carrier. I used my general purpose signal generator, otherwise known as the FT-817ND, to transmit a low power carrier (CW key down) into an unscreened dummy load (Elecraft DL-1). I repeated this with the transceiver powered from my bench power supply and then on its internal batteries with the power cable removed. The results in the WSJT-X spectrum display window are shown below.

K3 RX, FT-817 TX on mains supply.

K3 RX, FT-817 on battery power

As can be seen, there are traces at +/-100Hz and at 100Hz intervals on both signals, but the ghosts seem a bit stronger on the signal when the TX is powered from the main supply.

I recalled an issue a few years ago when someone sending CW using their K3 had reports of spurious signals +/- the sidetone frequency. This turned out to be audio modulation of the synthesizer by the sound of the sidetone from the K3 speaker. Elecraft provided a fix in the form of a stiffener for the synthesizer board. My K3 is an old one and does not have this modification. You can see that the synthesizer is affected by physical vibration looking at the trace produced when I rapped on the K3 case.

K3 RX showing the effect of vibration (knock on the case)

My K3 sits on a shelf next door to a heavy linear power supply. Could slight vibration of the mains transformer be modulating the receiver's local oscillator so as to create weak sidebands at +/- double the mains frequency?

To answer that question I repeated the tests using my Elecraft K2 as a receiver, feeding the headphone output at low level into the cheap USB audio dongle I use for computer sound. You can see the result below.

K2 on RX, FT-817 on mains power.

K2 RX, FT-817 on battery power

You can see that the sidebands are much reduced when the signal is produced by a transmitter running on battery power. In fact, some weak +/- 50Hz sidebands are present - possibly the effect of 50Hz AC hum on the un-isolated cable used to connect the K2 headphone output to the sound card.

I'm not sure what to make of all this. It does appear that the 'harmonics' - which are really sidebands - that accompany a strong JT9-1 signal are caused mainly by AC ripple modulating the transmitted carrier, but that hum on the receive side can produce a decodable signal as well. The WSJT-X software is extremely sensitive and can detect these components even if they are 30dB or more below the fundamental carrier.

I would appreciate hearing of other theories or tests carried out to explain this phenomenon. It seems to be that this issue is going to be almost unavoidable when mains-powered equipment is used to generate signals that are decoded by very sensitive software.