03-05-2014 10:42 AM
I am synchronizing 3 NI-2920 USRPs. Two are connected via the MIMO cable (what I'll call the "MIMO pair"). An external PPS and 10MHz Ref is connected to one radio in this pair, as well as the third USRP so that all three are synchronized and phase locked. The cable lengths of the PPS and 10MHz are the same to each radio.
Measuring the same event over all three radios, I do observe coherency over my measurement, however I observe some strange behavior in terms of time synchronization. I notice that the event appears at the same sample number between the MIMO pair...but it is delayed relative to the third radio. Once I figure out what this 'delay' is (in terms of samples), it is constant over multiple observations of the event, so long as I don't turn off the radio. So...I'm confident that things are phase locked and the A/D's are firing at the same time....though...I can't explain where this 'delay' comes from.
To be clear...I am confident that my event is reaching all three radios at the same time...so the "delay" seems to be related more to when the A/D's of the MIMO pair "turn on" relative to the third radio. The delay is about 3.25ms +/-0.25ms.
Is this expected behavior? It seems like there might be some kind of delay as the MIMO pair tries to negotiate it's synchronization relative to the external PPS and 10MHz. Don't recall ever reading anything about this.
---
BC
I
03-06-2014 04:50 PM
Hi BC,
Do you have the USRP in the "MIMO pair" that is not connected to the reference frequency source property node set to MIMO?
What results do you get when you disconnect "MIMO pair" and let one of the pair use the external source along with the third unpaired USRP? Basically just have two of the USRPs synchronized to the reference clock. Do you see the same delay then?
03-06-2014 08:19 PM
03-07-2014 01:19 PM
I've basically done the below, but with only 3 radios (receivers). Using the notation below, Rx 2 and 4 are configured to get their timing from the external Ref and PPS, while Rx1 is configured as the "slave", getting the timing info from the MIMO cable. Analyzing the measurement, I'm convinced that everything is phase coherent and that the sample clocks are "in sync". It's just seems like they don't agree as to when t0 is.
03-07-2014 01:23 PM
Noah-
Good idea. I can try to let the "slave" of the MIMO pair be unreferenced/synch'd, and have external sync-ing between the other two radios. I've done this before (before I bought some MIMO cables!) and never had any issues. It's only when I seem to "mix" the two techniques (MIMO cable and external references) do I see the behavior. However...it would be good to try in my current set-up to convince myself it's not operator error!
---
Brandon