On 2022-03-29 12:10, Rob Kossler wrote:
On Tue, Mar 29, 2022 at 10:59 AM Marcus D. Leech
<patchvonbr...@gmail.com> wrote:
On 2022-03-29 09:55, ri28...@mit.edu wrote:
I’m using a UBX-160 daughterboard. My application uses less than 50
MHz of bandwidth at baseband.
In past experimentation I’ve done, changing the RF gain takes on the
order of 1 ms, and I need to adjust for different beam angles an order
of magnitude faster than that.
Ah, in which case, your baseband-based approach makes sense.
So unless your existing codebase is "on the edge" of being unable to
keep-up at your sample rates, then doing a complex multiply in software
would be the way to go.
It could also be done in RFNoC, but you'd have to ramp-up on RFNoC, and
unless there's a strong performance reason for doing it in RFNoC, stick
with the
software side.
Yes, it seems that SW is probably the best way to go. However, if you
did want to do it in rfnoc. The example rfnoc block is a "complex
multiplier" block, so it really is not difficult to do on the FPGA.
The challenge is that you need to build it which requires an expensive
Xilinx Vivado license.
Rob
I don't think it's the complex-mult, per se, that's the issue, but
rather the ability to rapidly change the set of complex-weights across
the beamformer.
So, you have a "schedule" with each item in the schedule being the
complex-weights across the array to effect the desired pattern, and you
need to
"run" that schedule at 100usec timescales. Doing *THAT* in RFNOC
may require more of a deep-dive...
_______________________________________________
USRP-users mailing list -- usrp-users@lists.ettus.com
To unsubscribe send an email to usrp-users-le...@lists.ettus.com
