Hi Armin,
The DmaFIFO is a FIFO that fits in as follows:  host_tx => DmaFIFO => DUC
=> Radio(Tx).  It's purpose is to provide samples at a steady rate to the
DUC because the host PC will often send samples at an inconsistent rate
(because the PC is multi-tasking). Of course, the PC must provide samples
at the required "average sample rate" or no amount of buffering can help.
When you set skip_dram=1, this buffer is removed and so the PC must provide
samples to the DUC at a steady rate or you will get "U" for underrun (which
is the sample starved condition) or "L" for late (which means that a sample
has been specified to transmit at an exact time but when that sample
arrives at the Tx radio, the time has already past).

Regarding your comment about needing to use the "zero-sample" Tx channels
for Rx, it it not clear to me why you can't use the channel simultaneously
for both Tx and Rx.  That is, the fact that you are sending zero samples
for Tx does not prevent you from using the channel for Rx using the Rx2
antenna port.

It seems from your original post that this is an Ettus bug that has not
been fixed. Potential work arounds are to upgrade to UHD 4.0 or to use the
RFNoC API rather than multi_usrp to implement your gnuradio graph.  This
latter option might not be trivial if you don't have experience using this
API.
Rob

On Wed, Jul 14, 2021 at 9:01 AM Marcus D. Leech <[email protected]>
wrote:

> On 07/14/2021 04:32 AM, Armin Ghani wrote:
>
> Dear Marcus
>
> I didnt get your point about set sample rate to 10Msps with the same
> bandwidth. Would you explain more?
>
> I know that L character at console means starving for samples but what I
> really dont understand is that why it comes up when skip_dram argument sets
> to one with the same sample rate though?
>
> Regards.
>
> I just meant that you could interpolate your signal up to a higher sample
> rate to see if that made the "L" go away.  There were historically problems
>   with lower sample rates on X310 in certain configurations.
>
> I don't know why there's a dependency on skip_dram, since I'm not one of
> the designers.
>
>
> _______________________________________________
> USRP-users mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
>
_______________________________________________
USRP-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to