Hello Rob, Thank you very much for your advice. The benchmark test is working on all 4 channels for 1 MSamp/sec, which is sufficient for the current application (though sometimes there are some S and U at startup). The ram drive did not solve the problem, however if I set a settling time for 1 (sec) for the txrx_loopback it started working for the 4 channels. In this case the RX starts only 1 sec after the TX. So it kind of seems like there is a problem at startup and after that everything is fine. I don't know if that makes sense though.
I tried to make the settling time thing happen in gnuradio with issue_stream_cmd after init/start with a python snippet, but oddly there is a very short time at startup when both TX and RX are active, and the command seems to take effect only after that. Maybe I'm missing something here. So now I'm trying to merge and customize the txrx_loopback with the tx_samples from file examples hoping that it will work. Also I checked the CPU load with htop and it is only 15% on all 4 cores during the loopback. Any further advice is still welcome! Balint Rob Kossler <rkoss...@nd.edu> ezt írta (időpont: 2020. szept. 3., Cs, 19:55): > Hi Bálint, > You may want to step backwards once more and run the "benchmark_rate" > utility with 4 channels with tx & rx rates at 1MHz. This will eliminate > the file I/O considerations (both for tx and rx). If benchmark_rate is > successful, see how high you can increase the rate. The n4200 is not a very > powerful CPU, so you might be significantly limited. If you are using the > 1Gb interface, then the max sample rate you can achieve (assuming your CPU > is not limiting you) for 4 channels is ~6.25 MS/s because of the 1Gb limit > (4 chan x 6.25MS/s x 4 bytes/sample x 8 bits/byte = 800Mb/s). > > Once you are satisfied with the USRP streaming rates, you can try putting > your tx and rx files in a ram drive in order to improve the file I/O so > that it does not impact overall performance appreciably. Then retry > txrx_loopback_to_file. > Rob > > > On Thu, Sep 3, 2020 at 3:23 AM Bálint Horváth via USRP-users < > usrp-users@lists.ettus.com> wrote: > >> Hello, >> >> I have a USRP N310 and in my application I want to do a 4 channel >> loopback. Transmit sources are files and the received samples are >> processed. However, I ran into stability issues receiving a lot of Late >> packet errors (LLLLL). To trace the problem I executed the basic UHD >> example txrx_loopback_to_file on 4 channels with 1Msamp/sec sample rate. >> There every time I got ERROR_CODE_LATE_COMMAND. When I executed it for 3 >> channels I got it working, though on some occasions the same error would >> occur. With netstat I see that in case I receive the L-s, the UDP packet >> error is increasing. >> >> After reading through the manual and threads I tried several things but >> could not resolve the issue. >> - I set the net.core.rmem_max and wmem_max as suggested >> - I set the MTU on both host and N310 to 8000 (also tried 1500) >> - I put the CPU to performance mode with cpupower >> >> My system config is: >> - Ubuntu 20.04, UHD version 3.15, (gnuradio installed with apt) >> - quad core intel n4200, 8GB RAM >> - N310 is loaded with default HG firmware (sfp0 is 1GBit) >> >> Do you have any tips on how to resolve this? >> >> Thanks! >> >> Balint >> _______________________________________________ >> USRP-users mailing list >> USRP-users@lists.ettus.com >> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >> >
_______________________________________________ USRP-users mailing list USRP-users@lists.ettus.com http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
