HF Transceiver Frequency Stability

I did some measurements of the phase noise of several amateur (ham) radio HF transceivers, and in the process gained some information about their frequency stability as well. The rigs were:

In each case, I set the rig to generate about 20 watts output in CW mode (except for the Yaesu FT-817, which was set to 1 watt) and fed the signal through appropriate attenuators to put about +3dBm into a TSC 5120A phase noise analyzer. The reference source was a 5 MHz Wenzel Ultra Low Noise oscillator. Each transmission was about 5 minutes long, a compromise to get enough data while not overheating the rig.

The rigs were allowed to warm up for at least one hour before the test; the SDR-1000 had been running for 24 hours.

Remember that these tests were run in transmit mode, and heating from the PA was likely to have a major impact on stability. Performance in receive mode might be quite a bit better, but was not tested here.

You'll notice that the Yaesu FT-817 plot looks like a straight line. Here's that data by itself, scaled to show more detail:

Why is the FT-817 so good? It has a TCXO (temperature compensated crystal oscillator), but so does the IC-746 Pro. One major factor might be significantly lower heating, since the QRP FT-817 was running at 1 watt output while the other rigs were running at 20 watts. Their amplifiers were therefore generating much more heat, and that could affect the stability of the crystal (or in the TS-520's case, the VFO).

Another possibility is that Yaesu just uses a much better oscillator...