Re: Synchronous recording with two USRP B210

[Mark Wagner]

I would add that in my experience it is not enough to make the unknown PPS into a known PPS. For example, if I were to go into the python file that my flowgraph generated and in the __init__() function add something like:

curr_hw_time = self.USRP_A.get_time_last_pps()          #note: curr_hw_time is a 'time_spec_t()' object

self.USRP_A.set_time_next_pps( (curr_hw_time+1.0 ) ) 

self.USRP_B.set_time_next_pps( (curr_hw_time+1.0 ) )

Then I would also have to add something like

self.USRP_A.set_start_time(uhd.time_spec_t(curr_hw_time+2.0 ) )

self.USRP_B.set_start_time(uhd.time_spec_t(curr_hw_time+2.0 ) )

To maintain the relative phase relation of the signals I receive between re-initializations and full reboots of the USRPs. I am not sure why this is the case, but this is the result I have seen while testing

-Mark

 

On Mon, Nov 4, 2019 at 12:39 PM Paul Boven <address@hidden> wrote:

Hi Ali,

Your CDA-2990 (OctoClock?) distributes both a frequency reference signal
(10MHz) and a time reference signal (1PPS). To get both USRP to record
at the same time, you need to first get them to run on the same time
internally, and then tell them both, in advance, at what time to start
streaming.

Unfortunately, that will take a bit of tinkering, and cannot be done
from just regular GRC blocks at the moment.

The first half of the problem, getting them to run at the same time, is
a bit more difficult than it should be, as it involves writing some code
to turn your 'unknown PPS' into a know PPS. Basically, you read out the
PPS status from one of the USRPs, and once you see that it changes (i.e.
a PPS edge has just been detected by it) you read out the time on your
PC, add one second to it, and tell both USRPs that they have to set
themselves to this time value at their next PPS. For more details, see:

https://files.ettus.com/manual/page_sync.html

Once you have this done, you can send a command to each USRP to start
streaming at a particular second. This could be done either from two
separate applications, or just one that talks to both USRPs, using two
UHD source blocks.

I've 'hacked' the required statements into the python code that gets
generated when your GRC file is compiled. That has the huge drawback of
having to re-patch it every time you change the flowgraph. This approach
gets more convenient if you write your flowgraph in Python or C++, and
add the required synchronization commands in there. Most of my
measurements were done with a modified uhd_rx_cfile, where I added the
synchronization logic, and an extra argument for the start time (in unix
epoch seconds).

Regards, Paul Boven.

On 10/29/19 9:04 PM, alireza nazari wrote:
> Hi,
>
> I am trying to use GNURadio to receive signals and record them to files
> from two USRP B210(each has two channels). Both SDRs are connected to an
> CDA-2990 external clock. In gnuRadio I had to put two UHD source blocks
> and use "serial=xxx" to point them to each SDR and connect each output
> channel to 4 file Sink blocks. I also changed the clock and time source
> to "external" and synced to "unknown PPS".  However, a couple of files
> are different in size from the other couple. Apparently if I want to
> start them synchronously, all four channels should be in one source
> block, but how can I do that and point out each 2 channels come from a
> different device? When I try one USRP source block
> with "serial0=31A3C93,serial1=31A3D4B", I face an error bellow:
> /Traceback (most recent call last):
>    File "E:\Alireza\sdr_A20\top_block.py", line 207, in <module>
>      main()
>    File "E:\Alireza\sdr_A20\top_block.py", line 201, in main
>      tb = top_block_cls()
>    File "E:\Alireza\sdr_A20\top_block.py", line 125, in __init__
>      self.uhd_usrp_source_0.set_clock_source('external', 1)
>    File "E:\Program
> Files\PothosSDR\lib\python2.7\site-packages\gnuradio\uhd\uhd_swig.py",
> line 2987, in set_clock_source
>      return _uhd_swig.usrp_source_sptr_set_clock_source(self, source,
> mboard)
> RuntimeError: LookupError: IndexError: multi_usrp::mb_root(1) -
> LookupError: IndexError: multi_usrp::mb_root(1) - path not found
> /
>
> Is there any example available on how I can synchronously start
> recording from different devices? I assume this is a trivial task but I
> have been searching for this for a few days right now and I can't find
> any helpful answer. I am attaching my .grc files.
>
> Thank you,
> Ali

--

Mark Wagner

University of California San Diego

Electrical and Computer Engineering