Forum: GNU Radio USRP2 latency calculations

Announcement (2017-05-07): www.ruby-forum.com is now read-only since I unfortunately do not have the time to support and maintain the forum any more. Please see rubyonrails.org/community and ruby-lang.org/en/community for other Rails- und Ruby-related community platforms.
Pauli Rikula (Guest)
on 2008-11-05 13:02
(Received via mailing list)
I foud this wery usefull:
http://www.gnuradio.org/trac/wiki/UsrpFAQ/Latency

Is there similar calculations about the latency of the USRP2?

If there is not, could someone please do those calculations?

--
- Pauli Rikula
Eric B. (Guest)
on 2008-11-05 15:39
(Received via mailing list)
On Wed, Nov 05, 2008 at 01:01:14PM +0200, Pauli Rikula wrote:
> I foud this wery usefull: http://www.gnuradio.org/trac/wiki/UsrpFAQ/Latency
>
> Is there similar calculations about the latency of the USRP2?
>
> If there is not, could someone please do those calculations?
>
> - Pauli Rikula

I did some preliminary measurements a while ago.  A round trip from
the host to the USRP2 and back was taking about 47us.  It mostly
depends on how much data you've queued for transmission on the
ethernet.  YMMV.

Eric
Firas A. (Guest)
on 2008-11-05 17:01
(Received via mailing list)
Hi,

If USRP2 can respond to PING and reply back, then one should be able to
measure the delay.


Regards,

Firas
Mattias K. (Guest)
on 2008-11-05 17:27
(Received via mailing list)
Hi,

Wouldn't one be able to use for instance "set_freq()" or find() as a
messure? Although there might be some overhead, you will receive an "ok
I have set the freq"/"Ok here I am"- packet from the usrp2. Although one
could think of more elaborate schemes involving a special OP_CODE
(usrp2_eth_packet.h)...

Cheers
//Mattias K.
Eric B. (Guest)
on 2008-11-05 18:24
(Received via mailing list)
On Wed, Nov 05, 2008 at 04:26:04PM +0100, Mattias K. wrote:
>
set_freq would not be a good measure, since it takes a while to
configure the daughterboard.  find() isn't great either, since it
waits 10ms to ensure that it's gathered all responses to it's
broadcast.

set_rx_decim should give a reasonable measurement, as long as you're
not streaming data.

Eric
George N. (Guest)
on 2008-11-05 19:59
(Received via mailing list)
> set_freq would not be a good measure, since it takes a while to
> configure the daughterboard.  find() isn't great either, since it
> waits 10ms to ensure that it's gathered all responses to it's
> broadcast.
>
> set_rx_decim should give a reasonable measurement, as long as you're
> not streaming data.
>
>

BTW, I also have code which can be run which actually measures the
latency on USRP1, rather than just theoretical calculations that you've
found on the wiki:
https://www.cgran.org/wiki/ArchitectureLatencyMeasurements

It uses pings and some modifications to the Linux kernel to measure
latency from host->kernel->USRP, etc.

- George
This topic is locked and can not be replied to.