|
From: | Patrik Eliardsson |
Subject: | [Discuss-gnuradio] Re: delay through dsp pipeline |
Date: | Thu, 13 Aug 2009 08:38:58 +0200 |
Hi,
The DDC in the dsp_rx pipeline is build with several
filters. Depending on the decimation rate you choose different filter
configurations. See the FAQ regarding DDC
and decimation values.
I guess you want to sample the source from the antenna
synchronous in both usrp2s. The case is the that the dsp_core_rx is reset at
start up of the usrp2. So the receiver will start
to fill their pipeline at different points of time. To get rid of
this problem we have used the 1PPS-signal to reset the DDC in both usrp2 at the
same point of time. Take a look in the dsp_core_rx.v file.
To start receiving the samples synchronous we then just
polled the timestamp and when we received a sample with timestamp = 0 we started
our process.
-Patrik Eliardsson
Från:
address@hidden
[mailto:address@hidden För
Pham, Thanh
Skickat: den 12 augusti 2009 21:45 Till: address@hidden Ämne: [Discuss-gnuradio] delay through dsp pipeline Hello, In several discussions, I recall it was mentioned that the
timestamp reference is at the end of the dsp pipeline. Is the dsp pipeline
consisting just of the DDC or anything else? It was also mentioned that the delay through the dsp pipeline
depending on some factors such as decimation rate. If we have multiple usrps2
using the same decimation rate, center freq, what other factors would contribute
to the difference in arrival of the signal at the end of the dsp pipeline? (if
two usrps2 are receiving from the same antenna and using a 10MHZ gps reference
signal). Is there a way to figure out the delay from the ADC to end of
the dsp pipeline? I also saw a mention about a patch for the
start_rx_stream_at, which tells the usrps2 to start streaming at a specific
time. Can someone elaborate on that some more? Whose time does the usrp2 based
on to do that, and does that ensure two usrp2 start streaming at the same time?
Thanks for any inputs! Thanh |
[Prev in Thread] | Current Thread | [Next in Thread] |