In a packet-switched system, because a whole packet must be buffered before being passed through the crossbar, your delay is going to be governed by your packet size and your sample rate. If you are decimating down to a lower sample rate, you will incur larger delays.
To a first approximation, you can say the delay of a block will be (packet size / sample rate), plus a bit of overhead. If I understand the Sunderlin paper correctly, that overhead is what he's measuring. I am able to get loopback operation in an X310 down to about 26us latency with several blocks in the loop through careful optimization of packet size and by keeping the full 200Msps sample rate (even though my application only requires 30kHz of bandwidth). If you can run your custom block at full bus speed, that's the best way to ensure low latency. If you require narrow bandwidths, consider a cascade of FIR filters instead of DDC-[block]-DUC. Nick On Fri, Sep 28, 2018 at 2:44 AM Sebastian Bräuer via USRP-users < usrp-users@lists.ettus.com> wrote: > Hi, > > I am currently working on a custom block that implements > Listen-Before-Talk directly on the X310 and I have question regarding > the latency that I can expect in between RFnoC blocks. This article [1] > claims that each block delays the signal by about 2us which would put > the turn-around time for my block to about 10us (RxRadio -> DDC -> > custom block -> DUC -> TxRadio) but I am measuring turn-around times of > roughly 150us, far higher than expected. My custom block does not buffer > anything besides some flops. > > Has anyone an idea what might be the cause of this? I already tried > reducing the spp of the radio with very limited success. > > [1] https://pubs.gnuradio.org/index.php/grcon/article/view/37 > > > _______________________________________________ > USRP-users mailing list > USRP-users@lists.ettus.com > http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com >
_______________________________________________ USRP-users mailing list USRP-users@lists.ettus.com http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com