Sync 2x USRP2s

GNU Radio
1

Hi,

I want to sync two USRP2s, and currently trying to modify the VRT
rx_timed_samples…

usrp2::clock_config_t cc;
cc.ref_source = usrp2::clock_config_t::REF_INT;
cc.pps_polarity = usrp2::clock_config_t::PPS_NEG;
cc.pps_source = usrp2::clock_config_t::PPS_SMA;
cc.provide_ref_to_mimo = false;
u2->config_clock(cc);

// u2->set_time(…); // replaced with…
u2->set_time_at_next_pps(…);

However, the code hangs at the start sampling, and manually applying a
pulse to the PPS SMA does nothing.

What is the correct voltage level and pulse shape required? Manually
toggling anything between 0-1.5v and 0-5v does nothing. I hope 5V isn’t
too much!!!

Ultimately, I want to capture coherent samples with both USRPs, maybe
with the MIMO cable, and a GPS reference + PPS. Would this be just a
case of…
For USRP A (with ref + PPS attached)…
cc.ref_source = usrp2::clock_config_t::REF_EXT;
cc.pps_polarity = usrp2::clock_config_t::PPS_NEG;
cc.pps_source = usrp2::clock_config_t::PPS_SMA;
cc.provide_ref_to_mimo = true;
For USRP B…
cc.ref_source = usrp2::clock_config_t::REF_EXT; MIMO???
cc.pps_polarity = usrp2::clock_config_t::PPS_NEG;
cc.pps_source = usrp2::clock_config_t::PPS_SMA; MIMO???
cc.provide_ref_to_mimo = false;

Kind Regards,
David

2

On 3/19/2010 10:48 AM, David E. wrote:

u2->config_clock(cc);

When do you start_rx_streaming()? I’ve found that it works best to wait
until the next PPS (e.g. sleep(1) if the host PC doesn’t have direct
access to the PPS signal), and then calling start_rx_streaming() with
the desired parameters (i.e. start streaming at current time + some
offset) works.

cc.pps_polarity = usrp2::clock_config_t::PPS_NEG;
cc.pps_source = usrp2::clock_config_t::PPS_SMA; MIMO???
cc.provide_ref_to_mimo = false;

Kind Regards,
David

I haven’t tried the MIMO cable at all, but sync’ing by providing the
same clock/PPS to both USRP2’s works pretty well. You may also want to
review some of the recent discussions on mailing list about the possible
phase offsets you may get, even when all/both USRP2’s are locked to the
same reference, and internal clocks are synchronized. This is due to
the implementation of the PLL on the individual daughterboards, and thus
will be dependent on what you’re using.

–
Douglas G.
Code 5545
U.S. Naval Research Laboratory
Washington, DC 20375
(202) 767-9048
[email protected]

3

Hi Doug,

Thanks for the quick reply.

In rx_timed_samples I basically do start_rx_streaming() right after
set_time_at_next_pps(), so yeah, I’ll try the sleep(1). What do you mean
by “if the host PC doesn’t have direct access to the PPS signal” (read
the serial data from the GPS)?

Yes, I’ve read a bit about the phase offsets stuff, especially the post
from Matt about ambiguities, but at the moment I’m in a learning phase,
and a bit vague about what this means (and also no DSP guru!). I can see
that the RFX boards PLL may start up at an (phase) offset regardless,
but not sure how this would actually affect the sampling/ticks.

Also, as you say, you can split the signals, so why the MIMO cable? I’ll
do this then, its a lot simpler anyway.

Kind Regards,
David

4

You need the sleep(1) to guarantee that the time is latched into the
registers before starting streaming.

Prior to that, the time is unknown. And there seems to be a little bug
in the fpga that makes the streaming hang if this time happens to be in
the past.

When I update the example, I will need to add a note about this sleep(1)

-Josh