MIMO vs 10 MHz Reference

[cochenob]

I am attempting to recreate the scenario detailed in the this whitepaper from the Ettus site....

http://www.ettus.com/content/files/kb/mimo_and_sync_with_usrp_updated.pdf

...which uses an external RF source, with a carrier at a small IF relative to the USRP.  The white paper shows that by using the MIMO cable...the measurements between two USRPs can be phase locked and synchronized.  I recreated this with two N210's with WBX's and the LV driver....EXCEPT...instead of the MIMO cable...I used only a common 10 MHz reference.

Upon initiating and fetching packets...I note that within a fetch...the data appears phase and frequency locked, with some constant phase offset.  However, if running continuously...with each new fetch that is displayed...the relative phase between radios appears to change (yet is still constant over that fetch).

My understanding is that the MIMO cable provides not only the 10 MHz reference...but also timing information such that "t0" is the same between radios.  Using only the 10 MHz reference...I understand that can't be assured that the data between radios was in fact taken at the same time.  I anticipated that this "synch offset" would change every time I configured the radio....but should at least be constant when observing multiple fetches over time after a single configuration event.

This is not what I see.  It's as if the constant phase offset changes fetch to fetch.

Any idea what's going on here?  What's different when I use the MIMO cable vs the 10 MHz reference only?

---

Brandon

However...I expected this to result in some kind of phase offset which would remain *constant* between fetches (nothing is changing between fetches). 

[Carl_W]

Hi,

Are you using a PPS signal as well? In the NI-USRP Help document under Programming Reference >> MIMO Synchronization talks about the ways you can do this also.

[cochenob]

Initially, no.  My understanding was that using the 10 MHz would have phase locked the two, but the samples wouldn't be synchronous because I wasn't latching in a time via the PPS.  This should have just resulted in some initial phase offset, which would have been fine for my application.  However...it appeared that instead of a constant phase offset between the two radios...I was seeing a slightly varying phase offset between fetches. 

The code I initially used had two parallel processes to configure/initiate/fetch from two radios.  However...since then...we've combined the configure/initiate/fetch into a single process using the ";" separator and fetching data using the "complex array" polymorphic Fetch VI.  This approach does not appear to have phase 'jumps' between fetches...and results in a constant phase over time.  Not sure yet why the two approaches would have caused what I've seen....but the second approach seems to do the trick.

Still...it would be good just to get a second opinion that by using only the 10 MHz reference and not the PPS that my measurements would still be phase locked...just with some constant phase offset because I'm not synchronous.

[Carl_W]

Honestly using the MIMO cable is the easiest and best way to synchronize. When you are sharing the clocks you will need to match the length of cable for the clock lines which can be tough to do.

[cochenob]

Agreed.  Especially for laboratory use.  I'm trying to package two USRPs in separate underwater canisters, which means that a 10 MHz ref and an external PPS may actually be easier to distribute/link my two housings than modifying/making my own 'underwater' MIMO cable.