On 2022-02-03 15:35, John Hodgins wrote:
Thanks for the prompt response Marcus. Should the discrepancies
express themselves as a constant offset in cross-correlation or can
the delay measurements also oscillate in a volatile fashion? For
B210-E310 we see a more constant looking offset whereas B210-X310 is
just wild.
Don't forget that the master-clock is quite different between B210 and
X310--I assume that you're sharing a common (10MHz) reference?
Also I remembered this discussion
<https://lists.ettus.com/empathy/thread/2H5WCI2MEWWJKXHXGPMK5UWQXETQ4JRW?hash=L7PPGLOU2FJDCA4ELC2TXJZW75WJQX4C#L7PPGLOU2FJDCA4ELC2TXJZW75WJQX4C>
from sometime ago where rising vs falling edges were used for
different USRPs.
Yes, I think the X310 uses falling-edge, which will make it not-align
with any of the other units that use rising-edge. I think R&D
is revisiting this decision.
Given your answer, can we say that for timed commands it is not a good
idea to mix and match different USRP series?
Absolutely. Realistically, it has never been a design goal to provide
precise alignment among different types of USRPs.
Best,
John
On Thu, Feb 3, 2022 at 11:27 PM Marcus D. Leech
<patchvonbr...@gmail.com> wrote:
On 2022-02-03 15:17, John Hodgins wrote:
> Hi there,
>
> We have a simple set up where we collect iq snippets from 3
usrp's at
> the same rx-times via timed commands. We introduce delays
between the
> usrp's via a reference delay box and try to recover the introduced
> delays via cross-correlation.
>
> Things work great when all the rx's belong to the same USRP family.
> For instance we have been able to recover the correct delay for all
> B210 (with GPSDO/TCXO) or all E310 (with GPS antenna) scenarios.
> However, when we mix different USRP's, the delays computed via
> cross-correlation no longer make sense.
>
> For instance, when we use 1 B210 and 2 E310's the delays between
B210
> and E310's just oscillate wildly but the relative difference
between
> E310 remains relatively stable. We tested 2 B210's with a single
X310
> (with GPSDO/OCXO) as well and we see the same problem as well. The
> interesting thing is though visualizing the spectrum of collected
> snippets from different USRPs show pretty much the spectrum of the
> known applied signal.
>
> Is there some fundamental discrepancy between different USRP
families?
> Or maybe UHD somehow interprets samples differently for different
> USRP's (flipping i and.q, different rx-time, pps implementations,
> some precision issues maybe)?
>
> In any case I would be grateful if anyone can shed some light onto
> this behavior.
>
> Best,
> John
>
> _______________________________________________
> USRP-users mailing list -- usrp-users@lists.ettus.com
> To unsubscribe send an email to usrp-users-le...@lists.ettus.com
The group delays between USRP families will be different due to
different FPGA implementations, the exact "placement" of time-of-day
registers, etc.
It would be very difficult to guarantee phase-coherence across
different
USRP device types, because the implementations are different--not
just
the FPGA bits and
pieces (DUCs, DDCs, etc), but also the analog stages will be
different.
_______________________________________________
USRP-users mailing list -- usrp-users@lists.ettus.com
To unsubscribe send an email to usrp-users-le...@lists.ettus.com
_______________________________________________
USRP-users mailing list -- usrp-users@lists.ettus.com
To unsubscribe send an email to usrp-users-le...@lists.ettus.com
