Hello Yeo Jin:

  *   If your maximum sampling rate on Windows is 4MS/s and if you are using 
the MIMO capability of the device, you may want to double-check the USB2/USB3 
issue that Alfredo Gardel mentioned (because (8/channels)MS/s is the 
approximately the maximum rate that you can achieve with USB 2.0).
  *   The Windows GNU Radio distribution comes with a few out-of-tree modules 
pre-installed. You will have to find and install the missing out-of-tree module 
on your Linux system. The modules can be found from http://www.cgran.org/ or 
GitHub. My guess is that you are missing 
gr-radar:https://github.com/kit-cel/gr-radar .
     *   Just want to mention - questions that are related to GNU Radio but not 
related to USRP can be directed to discuss-gnuradio mailing list or the GNU 
Radio IRC channel. They could be better addressed there.
  *   I am not really experienced with RADAR systems, but it seems like that 
the FMCW signals are deterministic (Please correct me if I am wrong). If so, 
you can pre-generate the signals, store them into files (use tmpfs if the drive 
is too slow), and then simply play back the pre-generated files using the "File 
Source".
  *   If your flow graph is not keeping up but the CPU usage is under 100% 
(=not all cores are being fully used), then you can crunch out more throughput 
by parallelizing the block which is acting as the bottleneck. The bottleneck 
can be located by looking for the block (usually a single thread) which is 
occupying the most CPU time during the execution. In my computer, running top 
and pressing H (in upper-case letter!) does the job. If you can somehow 
parallelize that thread into multiple threads (modify the algorithm of the 
block itself, or modify the flow graph), then the maximum throughput may 
improve. But this is not always possible - maybe the block cannot be parallized 
at all!


Regards,

Kyeong Su Shin


________________________________
보낸 사람: Yeo Jin Kuang Alvin (IA) <yjink...@dso.org.sg>
보낸 날짜: 2018년 5월 7일 월요일 오후 12:02:57
받는 사람: Kyeong Su Shin; usrp-users@lists.ettus.com
제목: RE: USRP Underruns "UUUUU"


Hi all,



Thank you for all the help!



I have tested what kyeong Su Shin mentioned, both on windows and Ubuntu. 
However, the maximum sampling rate I can go in Windows is 4MS/s but for Ubuntu 
it can go up to 40MS/s with a only 2-3 U’s to be seen.

The problem for me now is that I want to use the FMCW source generator that can 
be found on my Windows GNU Radio Companion, but I cant find it in Ubuntu, so I 
cant test out my project using Ubuntu.



I’ve tried running https://github.com/scivision/piradar/FMCW_usrp.grc on Ubuntu 
and the maximum sampling rate I can hit is about 8MS/s. What do you mean by 
parallelize the bottlenecks (listed below), is there something that I can do to 
the FMCW_usrp.grc for this? I’ve tried pressing H while running top and nothing 
seems to pop out or show.



Thank you in advance!



---Try using simpler filters, and parallelize the bottlenecks within your flow 
graphs if possible (so as the CPU usage would go up to approx. 100% during the 
execution). You can run top and press H to see which block is taking the most 
CPU time (assuming a typical GNU/Linux distro). ---









From: Kyeong Su Shin [mailto:kss...@postech.ac.kr]
Sent: Friday, 4 May 2018 6:05 PM
To: Yeo Jin Kuang Alvin (IA); usrp-users@lists.ettus.com
Subject: RE: USRP Underruns "UUUUU"



Hello Yeo Jin,



First, find the maximum effective sampling rate that you can achieve with your 
computer. Connect a constant source to a USRP sink (for USRP sink underruns), 
or connect a USRP source to a null sink (for USRP source overruns) and run the 
flow graph. Try different sampling rates to find the maximum achievable 
sampling rate for your hardware (max samp rate that gives no underruns or 
overruns).



If the maximum stable sampling rate that you can achieve is still 4MS/s, then 
you are pretty much out of luck. You can try installing different versions of 
GNU Radio and UHD (older version/newer version, version with AVX support, etc) 
and then try again. If that does not help (very likely), then you WILL have to 
upgrade your computer. In some cases, using a different network card (for 
network-based USRPs) or USB card (for USB-based USRPs) may improve the 
achivable rate (when the bottleneck is the link between the USRP and the PC). 
If that doesn't work either, then you will have to get a faster machine.



If you can achieve a higher sampling rate with the simple flow graphs stated 
above, then the problem is the maximum throughput of your code. You can try 
optimizing your GNU Radio flow graphs and blocks. Try using simpler filters, 
and parallelize the bottlenecks within your flow graphs if possible (so as the 
CPU usage would go up to approx. 100% during the execution). You can run top 
and press H to see which block is taking the most CPU time (assuming a typical 
GNU/Linux distro).



If you do not need to generate your I-Q data in real time, then generate your 
data before the execution of the flow graph and simply play back the 
pre-generated data (with a File Source block). You can also push down some of 
your DSP logics to the FPGA of the USRPs, if you have licences for the needed 
software (and if you are okay with HDL).



Regards,

Kyeong Su Shin



________________________________

보낸 사람: Yeo Jin Kuang Alvin (IA) via USRP-users 
<usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com>> 대신 USRP-users 
<usrp-users-boun...@lists.ettus.com<mailto:usrp-users-boun...@lists.ettus.com>>
보낸 날짜: 2018년 5월 4일 금요일 오후 6:29:09
받는 사람: usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com>
제목: [USRP-users] USRP Underruns "UUUUU"



Hi all,



I am getting underruns and overruns when trying to run the UHD programs, both 
GNU radio and C++. The maximum my computer can handle is 4MHz sampling rate 
before seeing “UUUUU”.

I’ve searched online and most people say, change a new computer into quad core 
etc.



Are there any other ways to solve this problem? I want to hit around 20 MHz, 
but 30 MHz – 40MHz if possible.



Thank you in advance!
_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Reply via email to