[USRP-users] Four channel tx/rx loopback failure

Rob Kossler rkossler at nd.edu
Thu Apr 30 18:35:17 EDT 2015


Update.  This is not a gnuradio issue.  It seems to be device related and I
can duplicate the problem with "benchmark_rate". While I previously thought
it was as four channel vs two channel issue, I now know that I can
demonstrate the issue with just two channels (on the "bad" X310).

Here is the command that demonstrates the issue...
   benchmark_rate --channels="0,1" --tx_rate=5e6 --duration=5

Here is the error message...
UHD Warning:
    x300_dac_ctrl: front-end sync failed. unexpected FIFO depth [0x7]
Error: RuntimeError: x300_dac_ctrl: front-end sync failed. unexpected FIFO
depth [0x7]

Comments:
- If I run the exact same command with this same device but with
'--channels="0"' or '--channels="1"', it runs fine. It even runs fine if I
change the sample rate to 50e6 with only one channel.  So, individually,
each channel runs fine.  It is only when running both simultaneously that
the error occurs.
- If I run the exact same command with this same device but with
'--rx_rate=5e6' instead of '--tx_rate=5e6', it runs fine.  So, it can
handle 2 channel receive streaming just fine.
- If I run the exact same command with a different X310 device, it runs
fine.  Note that I reprogrammed the FPGA on both X310 devices using
"usrp_x3xx_fpga_burner" prior to running the test.  So, both devices should
have the same FPGA.  I also used the same 10G NIC & cable. It is
interesting that the device that is failing is brand new - just received it
in the last couple of days.

Any ideas?

Rob Kossler

PS. as mentioned yesterday, I am using 3.8.3 (UHD_003.008.003-0-g87dfdc3c)


On Wed, Apr 29, 2015 at 6:38 PM, Rob Kossler <rkossler at nd.edu> wrote:

> Hi,
> I am having a failure with the attached GRC flowgraph. It is a four
> channel Tx/Rx loopback using 2 Ettus X310 USRPs.  The flowgraph is simple
> in that the Tx is simply a constant (center freq tone) and the Rx is
> displayed on a QT freq sink.  Externally, the RF signals are looped back
> from TX to RX with attenuation.
>
> There are actually two flowgraphs in this attachment, but one of them is
> disabled.  The one disabled is the two channel version which works fine.
> By "fine", I mean that there is no UHD Warning message reported in the
> reports window, the TX LEDs both come on, and the QT freq sink shows a tone
> on the center of the four RX channels, as expected.
>
> When I run the four channel version (presently enabled in the flowgraph),
> I get a UHD warning (see below and/or in the attached file "tr_test.out"),
> the TX LEDs do not come on, and the displayed spectrum has no tone at the
> center.
>
> There is hardly a difference between the four channel flowgraph and the
> two channel flowgraph.  The four channel version has motherboard and
> channel counts of 2 and 4 while the two channel version has corresponding
> counts of 1 and 2.  Additionally, the device address field has 2 addresses
> rather than 1 in the four channel version.
>
> This failure does not seem to be intermittent.  Every time I have run the
> two channel version, it has been successful and every time I have run the
> four channel version it has been unsuccessful.  Also, this was executed
> after having just re-built both UHD and gnuradio via "./pybombs update".
>
> Here is the pertinent info from the reports window when I run the four
> channel version.  Note that the full output is available in the attachment
> "tr_test.out".
>
> <<< Welcome to GNU Radio Companion 3.7.8git-111-g1425e482 >>>
> ...
> linux; GNU C++ version 4.8.2; Boost_105400; UHD_003.008.003-0-g87dfdc3c
> ...
> Using Volk machine: avx_64_mmx_orc
>
> UHD Warning:
>     x300_dac_ctrl: front-end sync failed. unexpected FIFO depth [0x7]
> thread[thread-per-block[1]: <block gr uhd usrp sink (2)>]: RuntimeError:
> x300_dac_ctrl: front-end sync failed. unexpected FIFO depth [0x7]
>
> >>> Done
>
>
> Finally, on a perhaps related note, it is really not clear to me which
> USRP configuration parameters should be repeated between the UHD source and
> sink.  I am familiar with operating directly with the UHD driver from C++
> where you create a single USRP with various config information such as
> device address, clock ref, timing ref and then create separate RX and TX
> streamers.  When operating from GNU radio with both a UHD source and sink,
> it is not clear if all of that config info should be repeated in both
> source and sink (which is what I did in the attached flowgraph) or if some
> of it does not need to be repeated.  For example, I have both source and
> sink with "sync" option of "unknown_pps".  Perhaps this is only needed in
> one or the other??  Perhaps the clock and timing ref parameters are only
> needed in one or the other??
>
> Thanks.
> Rob Kossler
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ettus.com/pipermail/usrp-users_lists.ettus.com/attachments/20150430/8c8e6814/attachment-0002.html>


More information about the USRP-users mailing list