Thanks again Marcus. I really appreciate your help.
I am setting Ref Clock / PPS to external, etc. I get the syncing to work properly around a year ago. Since then, we made various changes and I think one or more change may be causing the issue. Among the changes are: buy N200 units with a new revision (could that be a problem when syncing?), upgrading GNU Radio / UHD, enabling Rs 232 echoing in the GPSDO, etc.
Similar to my last year tests (I entirely unplugged the RS 232 cable), and now it is not detected as an Internal GPSDO. However, I am still having issue syncing the two motherboards.
Attached are two figures for the phase drift between Rx1 vs. Rx2, and Rx1 vs. Rx3. This was generated through splitting the Tx and looping it back to the Rx's. Signal processing was done properly (for dechirping, downsamplng, etc). My application is LFMCW radar, so each 490 samples in the attachment represents one sweep, and sweeping was done continuously.
angle(Rx1./Rx2) looks great. As to angle(Rx1./Rx3), I would expect a linear phase offset between different motherboards. However, as .shown in the attachment, Rx1 vs. Rx3 has weird phase spikes. I think this shows that, for some reason, the Tx is not properly synced with the second motherboard. Any ideas on what might be causing this issue?
I am using Starttech Ethernet Adapter (ST1000SPEX4). The problem I have with this card is that it won't turn AUTONEG off, it is always on. Could this cause the problem?
I tried this on two different Ubuntu machines, with similar results as shown in the attachment.
For the first machine,
UBUNTU 14.04 LTS
linux; GNU C++ version 4.8.2; Boost_105400; UHD_003.008.002-80-ge28d7844
For the second machine,
UBUNTU 14.04 LTS
linux; GNU C++ version 4.8.2; Boost_105400; UHD_003.008.000-46-g5b706d29
I'll highly appreciate any suggestions to solve this problem.