[USRP-users] E310 - Maximal rate for rx_samples_to_file example

Claudio Cicconetti ccicconetti at mbigroup.it
Tue Mar 22 06:21:10 EDT 2016


Dear Marcus,
Can you elaborate a bit more on your comment?

With iperf in my setup (E310 connected to host via a GbE Ethernet
switch) I can reach 450 Mbit/s with UDP and > 600 Mbit/s with TCP.

The same _cannot_ be achieved with sshfs due to the protocol/encryption
processing overhead, which becomes a limiting factor with sampling rate
> 2 Msamples/s (i.e. about 64 Mbit/s over the wire and one core at 100%
utilization).

I suggested sshfs as an example of remote dumping samples with i) low
sample rates, ii) unlimited storage capacity (well, limited by the
host), and iii) almost zero configuration effort (provided that the host
is a Linux machine you likely already have an ssh server up and running).

I suppose other less cpu-consuming options (nfs, samba) could be
explored as an alternative to sshfs to achieve higher sample rates.

Having said this, I understand raw baseband dumping is not a target
application for an E310, so the discussion is somewhat academic at this
point.

Best regards,
Claudio

On 03/21/2016 10:31 PM, Marcus Müller via USRP-users wrote:
> I'm afraid I can't recommend going through network to increase
> performance; the network hardware really isn't up to that in my experience.
> 
> Best regards,
> Marcus
> 
> On 03/21/2016 05:48 PM, Claudio Cicconetti via USRP-users wrote:
>> Dear Roy,
>> Another option is to write to a remote server, provided you have enough
>> network capacity between the E310 and the remote server.
>>
>> This also solves any issue you might have with the limited size of the SD.
>>
>> For instance, I use sshfs on the E310, that is quite straightforward:
>> you just need a Linux box with the ssh server configured.
>>
>> Best regards,
>> Claudio
>>
>> On 03/21/2016 05:36 PM, Kyle A Logue wrote:
>>> You should definitely be able to maintain 1MSPS. We have a gang of the UHS-1 and UHS-3 cards and are happy with the performance. Remember that the spec on the outside of the box (300 MB/s!) is only for the first 50 megabytes or whatever cache they have onboard. Look up sustained write speeds for a better estimate.
>>>
>>>
>>> I'd recommend writing to /var/volatile/tmp or /run since you won't run into a bottleneck of write speed writing to the ram-disk, just file size.
>>>
>>>
>>> Kyle Logue · The Aerospace Corporation · ETG-DCID · Senior MTS · 310.336.3038 · kyle.logue at aero.org<mailto:kyle.logue at aero.org>
>>>
>>>
>>> ________________________________
>>> From: רועי <royk at eng.gov.il>
>>> Sent: Monday, March 21, 2016 09:33
>>> To: Kyle A Logue; Claudio Cicconetti; James Humphries; Ettus Mail List
>>> Subject: ‏‏RE: [USRP-users] E310 - Maximal rate for rx_samples_to_file example
>>>
>>> Hi Claudio, Kyle and Trip,
>>>
>>> Thanks a lot for the help.
>>> I tried running with the --null option as Claudio suggested, and there were no overflows. I guess this means that the problem is indeed with the writing.
>>>
>>> I had a similar problem with a B210 board in the past, and what worked for me is setting "num_recv_frames=512" in the "args" part. However, when I try doing the same with the E310, I get the following Error: "AssertionError: zf_peek32(ctrl_base+ARBITER_RB_ADDR_SPACE) > 0 in virtual void e300_fifo_mb::release()".
>>>
>>> If changing "num_recv_frames" can't do the trick this time, is it safe to assume that upgrading the micro sd card to "UHS Speed Class" 3 (which works at 30MB/s according to http://www.pcper.com/reviews/Editorial/SD-Card-Speed-Classes-Grades-Modes-and-File-Systems-Explained) or newer will enable writing to a file at a rate of 1Msps continuously?
>>>
>>> Best regards,
>>> Roy
>>> ________________________________
>>> מ: ‏‏USRP-users ‏[usrp-users-bounces at lists.ettus.com]‏ בשם Kyle A Logue via USRP-users ‏[usrp-users at lists.ettus.com]‏
>>> ‏‏נשלח: יום שני 21 מרץ 2016 18:22
>>> ‏‏אל: Claudio Cicconetti; James Humphries; Ettus Mail List
>>> ‏‏נושא: Re: [USRP-users] E310 - Maximal rate for rx_samples_to_file example
>>>
>>>
>>> Roy,
>>>
>>>
>>> I have successfully used the following higher sample rates on an E310:
>>>
>>>   *   6 MSPS in 2 second intervals to Metafile on PNY Turbo Performance 64GB UHS-3 SD Card
>>>   *   16 MSPS in 170ms intervals to /var/volatile/tmp every 2 seconds, then moved to SD Card
>>>
>>> Trying to get data off via Ethernet is a fool's errand, so don't bother. I haven't tested the newest Sandisk Extreme Plus SD Cards, but they may be faster.
>>>
>>>
>>> Kyle Logue · The Aerospace Corporation · ETG-DCID · Senior MTS · 310.336.3038 · kyle.logue at aero.org<mailto:kyle.logue at aero.org>
>>>
>>> ________________________________
>>> From: USRP-users <usrp-users-bounces at lists.ettus.com> on behalf of James Humphries via USRP-users <usrp-users at lists.ettus.com>
>>> Sent: Monday, March 21, 2016 08:09
>>> To: Claudio Cicconetti
>>> Cc: USRP-users at lists.ettus.com
>>> Subject: Re: [USRP-users] E310 - Maximal rate for rx_samples_to_file example
>>>
>>> Hi Roy,
>>>
>>> In addition to Claudio's suggestion, keep in mind that a 1MHz sampling rate is already approaching the max write speed of the micro-sd card. It may not be able to sustain those write speeds for very long.
>>>
>>> At 1 MS/s,
>>>
>>> 1 MS/s * 4 bytes/float * 2 floats/sample = 8 MB/s.
>>>
>>> This doesn't consider that the OS is also loaded from the SD card and may be consuming some time reading/writing to the SD card as well.
>>>
>>> -Trip
>>>
>>> On Mon, Mar 21, 2016 at 9:10 AM, Claudio Cicconetti via USRP-users <usrp-users at lists.ettus.com<mailto:usrp-users at lists.ettus.com>> wrote:
>>> Dear Roy,
>>> Did you try the --null option? It will do everything except storing the
>>> samples, just in case you suspect the issue may have something to do
>>> with the actual writing part.
>>>
>>> Best regards,
>>> Claudio
>>>
>>> On 03/21/2016 11:37 AM, רועי via USRP-users wrote:
>>>> Hi,
>>>>
>>>> What is the maximal sampling rate possible for sampling to a file without getting overflows (E310)?
>>>>
>>>> I ran the rx_samples_to_file example on the E310 with the default sampling rate (1MHz), and after about a minute I started getting overflows ("O"s).
>>>>
>>>> Additional info:
>>>>
>>>>
>>>> 1.       The E310 image is "fido-rfnoc-test" (http://files.ettus.com/e3xx_images/alpha/fido-rfnoc-test/)
>>>>
>>>> 2.       The FPGA image is "uhd-images_003.010.rfnoc-282-gab4e9eaa" (http://files.ettus.com/binaries/images/)
>>>>
>>>> 3.       The "rx_samples_to_file" example was cross-compiled from "rfnoc-devel" branch (https://github.com/EttusResearch/uhd/tree/rfnoc-devel), following the instructions in http://files.ettus.com/manual/page_usrp_e3x0.html#e3x0_sdk.
>>>>
>>>> 4.       The micro SD card is UHS speed class 1 (10MB/s).
>>>>
>>>> Thanks,
>>>> Roy
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> USRP-users mailing list
>>>> USRP-users at lists.ettus.com<mailto:USRP-users at lists.ettus.com>
>>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
>>>>
>>>
>>> _______________________________________________
>>> USRP-users mailing list
>>> USRP-users at lists.ettus.com<mailto:USRP-users at lists.ettus.com>
>>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
>>>
>>>
>>
>> _______________________________________________
>> USRP-users mailing list
>> USRP-users at lists.ettus.com
>> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
> 
> 
> _______________________________________________
> USRP-users mailing list
> USRP-users at lists.ettus.com
> http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
> 





More information about the USRP-users mailing list