On 2022-05-04 08:43, Martin wrote:
Hi,
I have data corruption problems with high speed dual channel capturing
on USRP E320 over 10 gbit ethernet.
Medium rate, dual channel, everything seems fine.
High speed, single channel, everything also seems fine.
High speed, dual channel capturing. Problems. High amount of random
high values in datastream even with low gain and low signal level)
I do not seem to get U or O (Underrun, Overrun) errors
I was under the impression that with E320 you should be able to
capture dual channel 56 Msps signals. If that is not feasable then we
have a serious problem.
I also do not see any error messages.
We tested on two different E320ś on two different host-PC's. Same result.
Do you also have these problems?
Do you have tips on how to solve or check what the issue is?
Thanks in advance.
Please see below for details
When captured with uhd_rx_cfile or our own capture scripts.
The output data shows random very high, usualle even full scale values
even if the input signal is small and gain low.
With int 16 output values are between 32768 and -32767. With complex
float from -1.0 to 1.0
The capture is done from a host-pc using the E320 as remote capture
device over 10 Gbit ethernet.
The USRP E320 is running UHD 4.0.0.0
root@ni-e320-321D09C:~# uhd_config_info --version
UHD 4.0.0.0-0-g90ce6062
The host pc is also running uhd_4.0.0.0:
nldudok1@noisegate:~$ uhd_config_info --version
UHD 4.0.0.0-133-g7ec04886
dual channel high speed test with complex int16 (complex short)
wrong values in both channels. Whatever the gain. values go from
-32767 to 32768
uhd_rx_cfile -a master_clock_rate=56e6 -c 0,1 -g 44.0 -r 56e6 -f
105.0e6 -N 56e6 -s tmp_cshort_56M_g44.raw
uhd_rx_cfile -a master_clock_rate=56e6 -c 0,1 -g 0.0 -r 56e6 -f
105.0e6 -N 56e6 -s tmp_cshort_56M_g0.raw
Same result with complex float32 output, values fo from -1.0 to +1.0
whaterver the gain.
uhd_rx_cfile -a master_clock_rate=56e6 -c 0,1 -g 44.0 -r 56e6 -f
105.0e6 -N 56e6 tmp_cfloat_56M_g44.raw
uhd_rx_cfile -a master_clock_rate=56e6 -c 0,1 -g 0.0 -r 56e6 -f
105.0e6 -N 56e6 tmp_cfloat_56M_g0.raw
Same test at medium datarate. Values are OK. complex float or complex
short output. Note
uhd_rx_cfile -a master_clock_rate=16e6 -c 0,1 -g 44.0 -r 16e6 -f
105.0e6 -N 16e6 -s tmp_cshort_16M.raw
uhd_rx_cfile -a master_clock_rate=16e6 -c 0,1 -g 44.0 -r 16e6 -f
105.0e6 -N 16e6 tmp_cfloat_16M.raw
Single channel test at high rate also seems OK with eather complex
int16 or complex float32 output values.
Note that with complex float output the same writespeed to the ssd is
needed as with dual channel and complex int16 output. So it does not
seem to be a slow disk issue. Would also not be very likely since we
are storing only 1 second of data which should fit in linux disk write
buffer.
uhd_rx_cfile -a master_clock_rate=56e6 -g 44.0 -r 56e6 -f 105.0e6 -N
56e6 -s tmp_cshort_56M_g44_single.raw
uhd_rx_cfile -a master_clock_rate=56e6 -g 44.0 -r 56e6 -f 105.0e6 -N
56e6 tmp_cfloat_56M_g44_single.raw
You can look at complex float output data with gr-plot_iq and gr_plot_fft
complex short data can be converted to complex float with our own
little conversion script.
cshort_to_cfloat.py -i inputfilename.raw -o outputfilename.raw
(I can send the conversion script if you need it. It is just a file
source, interleaved short to complex block and file sink)
Example:
uhd_rx_cfile -a master_clock_rate=56e6 -c 0,1 -g 0.0 -r 56e6 -f
105.0e6 -N 56e6 -s tmp_cshort_56M_g0.raw
#first channel
cshort_to_cfloat.py -i tmp_cshort_56M_g0.0.raw -o
tmp_cshort_56M_g0_converted_to_cfloat.0.raw
#second channel
cshort_to_cfloat.py -i tmp_cshort_56M_g0.1.raw -o
tmp_cshort_56M_g0_converted_to_cfloat.1.raw
#show data
gr-plot_iq -R 56e6 tmp_cshort_56M_g0_converted_to_cfloat.0.raw
gr-plot_iq -R 56e6 tmp_cshort_56M_g0_converted_to_cfloat.1.raw
Best regards,
Martin Dudok van Heel
_______________________________________________
Martin:
I've forwarded your note to Ettus R&D.
But in the meantime, are you running the latest UHD?
What happens if you reduce the sample rate "a little" -- like down to
52Msps?
I know that on other platforms that use the AD9361 (as the E320 does)
that the maximum supported rate for dual-channel is 32Msps, but that may
just be because of the
type of clocking+data interface used on those platforms--the AD9361
supports a couple of different interface regimes...
_______________________________________________
USRP-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
