hackrf: gr-osmocom sink tx buffering

Sebastian Böhm Sebastian.Boehm at b-tu.de
Thu Dec 10 08:06:53 UTC 2020


Hi Eric,

thanks for your reply!
I understand what you mean by the performance aspect, but 'immediately' 
in my case did not mean that no latency should occur. And I think my 
problem does not necessarily belong to the size of the buffer.

However, I'm surprised that this buffer is only flushed to the Hackrf 
hardware when it is completely filled, whereas it is actually streamed 
continuously with 4M samples per second via USB.  This is a major 
limitation.

Here is an example of what I mean:
If I create a simple communication protocol in grc where only a response 
is made by sending a small acknowledgement packet for a request, but 
otherwise waiting, the osmocom sink will never send this packet because 
the buffer is not yet full. This inevitably leads to misbehavior in the 
communication flow.

I could adjust the size of the buffer so that it is already completely 
filled with the smallest of my packets, but his is not a very elegant 
solution.

However, I'll try it!

Sebastian

Am 09.12.2020 um 16:39 schrieb Eric Wild:
> Hi Sebastian,
> since the hackrf is half-duplex and has no timestamp support (only 
> continuous streaming) the buffer size was chosen at the time to 
> provide a reasonable tradeoff between latency and performance for 
> higher sample rates - I suppose you could just decrease it if that is 
> the problem, but there is still no way to just send something 
> immediately, there will always be a considerable latency between the 
> host and the rf output.
>
> -Eric
>
> Am 09.12.2020 um 13:23 schrieb Sebastian Böhm:
>>
>> Does anyone have any idea what could be the reason for this behavior?
>>
>> Am 01.12.2020 um 16:28 schrieb Sebastian Böhm:
>>> Hi,
>>>
>>> i try to transmit 2.48 ghz radio packets (beacons) at a fixed time 
>>> interval (e.g. every 1 s) via hackrf using the osmocom sink in 
>>> gnuradio. Without exception, the beacons are all received by my 
>>> sniffer "golden device" hardware, but burst.
>>>
>>> It seems that the packets are only transferred via USB to the hackrf 
>>> device hardware when the buffer with the fixed BUF_LEN 262144 
>>> (32*16*512) is completely filled. For my setup this are e.g. 18 
>>> beacon frames, which are transmitted as burst. When I look at the 
>>> USB interface, I can see this burst of data every 18 seconds in a 
>>> full USB request block (URB) of  size 262144 byte. All other URBs 
>>> (also of size 262144 - every approx. 200us) transferred via USB are 
>>> filled with zeros only. So this burst behavior should be caused by 
>>> gnuradio.
>>>
>>> How can I achieve that the complex data of the osmocom sink's input 
>>> is immediately (with the next URB) transferred to the hackrf 
>>> hardware? Am I missing any options of the GRC module?
>>>
>>> Thank You in advance!
>>>
>>> Regards,
>>> Sebastian
>>> -- 
>>> __________________________________________________________
>>>
>>> Sebastian Boehm
>>> Brandenburg University of Technology Cottbus - Senftenberg
>>> Computer Networks Communication Systems Group
>>>
>>> E-Mail: sebastian.boehm at b-tu.de <mailto:Sebastian.Boehm at b-tu.de>
>>>
>>> P.O. Box: 10 13 44
>>> D-03013 Cottbus
>>> GERMANY
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osmocom.org/pipermail/osmocom-sdr/attachments/20201210/4809f1fb/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5485 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.osmocom.org/pipermail/osmocom-sdr/attachments/20201210/4809f1fb/attachment-0001.bin>


More information about the osmocom-sdr mailing list