On 30/05/2023 17:44, Brian Padalino wrote:
On Tue, May 30, 2023 at 5:32 PM Mena Ghebranious <m...@chaosinc.com>
wrote:
I apologize, I think I must be missing something. This is the
filter (Xilinx IP) I see implemented in the N320 master code:
https://github.com/EttusResearch/uhd/blob/master/fpga/usrp3/top/n3xx/dboards/rh/n3xx.v#L3431
This HBF seems to be external to the DUC, but I doubt it's killing a
bunch of samples from coming out.
We are planning on running various sample rates running from 1 to
possibly 61.44Msps - for our current experiment we are using
20.480Msps
This is the most likely culprit. Currently you interpolate by 12 if
you use a 245.76 Msps master clock rate using the DUC (HBF (x2) -> HBF
(x2) -> CIC (x3) = 12).
What you can do is interpolate to 245.76 Msps on the host for TX and
you've got full control down to the sample.
See what happens when you interpolate on the host instead of using the
FPGA to do it for you. My guess is that it will be closer to what you
want.
Brian
The HBF is used (at least historically) when the effective
decimation/interpolation ratio is odd.
Also, a lot of the "logic" (at least, again, historically) of which bits
and pieces of the DUC/DDC chain were "in/out" of the
"circuit" was decided by the host side, and it would just set some
bits in the FPGA registers to "make it so". But my knowledge
has drifted quite a bit since RFNoC was introduced.
_______________________________________________
USRP-users mailing list -- usrp-users@lists.ettus.com
To unsubscribe send an email to usrp-users-le...@lists.ettus.com