[hpsdr] Frequency offset using Excalibur

[John Ackermann N8UR]

I just discovered something very odd with my Atlas-based HPSDR system 
(Hermes, Mercury, Penny, Excalibur, using W5WC PowerSDR 11/5/12 and 
latest firmware on all boards.)

There seems to be about a 9 Hz frequency offset generated somewhere in 
the system.  I noticed this while monitoring WWV on 10, 15, and 20 MHz 
using the PowerSDR phase scope, tuning the carrier in CWU mode with a 25 
Hz filter to remove modulation.

The minimum phase rotation occurs when I tune about 9 Hz below the 
nominal carrier frequency.  When I change to CWL, the offset reverses 
direction.

I've tried using both GPSDO and lab-quality Rubidium frequency 
references plugged into Excalibur, and get the same result.  I've 
watched while interrupting the external reference, and see about a 12 Hz 
frequency shift between states, so Mercury seems to be locking to the 
external reference.

To verify that it's not some propagation or interference issue, I routed 
the GPSDO output through a bunch of attenuation into the Mercury antenna 
input and see the same result on the very strong 10 MHz signal.  Since 
both the reference and the signal are coming from the same source, there 
should be exactly 0 offset.

In the PowerSDR cal tab, the frequency correction is set to 0.  When I 
run the frequency calibration, the correction factor remains at zero -- 
does this mean that PowerSDR think the frequency is accurate and the 
issue is just in the phase scope?

Either I'm misunderstanding or misconfiguring something for this test, 
or there's some sort of offset being generated in the hardware or 
software.  Anyone have any suggestions?

Thanks,

John