[hpsdr] Coherent operation of dual Mercury boards

[Joe Martin K5SO]

All,

I have noticed a slow drift (tens of seconds to produce an observable  
shift)  in the position (angle & phase) of the null when introducing a  
calibrated, constant phase, input to both Mercury inputs, due I think  
to the 122.88MHz oscillators on the Mercury board not being perfectly  
phase locked to the 10 MHz clock input from Excalibur (my guess).

To cure this drift, and achieve long-term coherent operation, I  
recommend using a single 122.88 MHz oscillator of one of the Mercury  
boards to provide the 122.88MHz clock for both Mercury boards via a  
twisted wire pair connection between the two Mercury boards, as we  
initially recommended but later abandoned.

The twisted wire pair connection should connect pin 2 of JP1 (CLK  
SELECT) and ground connection on the "master" Mercury board to pin 2  
of JP1 (CLK SELECT) and a ground connection on the "slave" Mercury  
board.  A convenient ground connection is the middle pin (pin 3) of J1  
(EXT CLK/LVDS) on each board.  The master Mercury board should retain  
its usual jumper across pins 3 and 2 of JP1 ("I" position of CLK  
SELECT) while no jumper should be present across pins 3 and 2 of JP1  
on the slave Mercury board.

Using this scheme I observe an exceedingly stable null condition of  
around -50dB that does not drift with time; indicating that the IQ  
streams from the two Mercury boards are much more coherent than before.

73,  Joe K5SO