[USRP-users] Frequency offset between two USRPs B210 synchronized with use of Octoclock-G
perper at o2.pl
Mon Oct 9 03:36:26 EDT 2017
W dniu 06.10.2017 o 23:22, Marcus D. Leech via USRP-users pisze:
> On 10/06/2017 03:47 PM, Piotr Krysik via USRP-users wrote:
>> W dniu 06.10.2017 o 21:33, Piotr Krysik via USRP-users pisze:
>>> Hello everyone,
>>> I synchronized two USRPs B210 with use of Octoclock-G and the attached
>>> gnuradio-companion flowgraph.
>>> I've connected signal generator generating sinusoid to both devices (it
>>> was on frequency 694.5MHz while USRPs were tuned to 694.3MHz, signal
>>> power was about -40dBm). What I would expect is to see constant phase
>>> offset between signals received by both USRPs (perhaps changing
>>> with temperature etc.). However what I see is constant changes of phase
>>> that corresponds to frequency offset between two devices. Moreover this
>>> frequency offset changes significantly from one execution of a
>>> to another (look at the image):
>>> relative phase of two B210's
>> Missing image of phase offsets in the attachment.
>> Best Regards,
>> Piotr Krysik
> Try disabling automatic DC-offset and I/Q correction. I've run into
> this before--since the DC offset and I/Q correction logic is somewhat
> stochastic, you would
> expect precise phase-measurements between two devices to wander
> around a bit with respect to one another.
Regarding your previous e-mail - I don't expect phase offset to be
constant from run to run. The issue is a frequency offset between the
devices that shouldn't be there. The offset is proportional to slope of
the lines presented on the image. The fact that it changes from run to
run is a secondary concern - I only shown this effect because it might
be important to find a solution.
Disabling I/Q imbalance and DC offset correction (with
set_auto_dc_offset(False) and set_auto_iq_balance (False) calls to UHD
source objects) didn't solve the problem. Only change is that I
additionally started to observe effects of these impairments. Anyway -
thank you for your ideas, I didn't try to disable automatic corrections
on B210 before.
One thing that helped was replacing one of the USRP B210 in the pair
with another one. So it seems like a flaw in particular devices that
causes the problem. For four devices that I checked I observed this
problem on 2. Which is quite a large percentage, but it's hard to make
conclusions with such small statistical sample ;) (if I just didn't have
luck or if is it some production problem).
Later this week I will check more devices. Anyway, I will try to get
replacement of every B210 with the flaw that I have with use of NI's
More information about the USRP-users