Hi Walter,

That is very useful, and (I think) it explains an issue I have been seeing
myself since beginning programming the 4x2 ADCs.

Many thanks for this information. Very much appreciated!

Cheers,
Colm

On Thu 10 Apr 2025, 5:02 PM Walter Klahold, <[email protected]> wrote:

> Hi Sebastian,
>
>
>
> I wanted to close the loop on this one, since I started out with the same
> questions and it didn't seem like you received an answer. I think you more
> or less have the correct understanding of how each of the clocks are
> derived, but to reiterate:
>
>
>
> -The ADC clock rate (i.e. the sample rate), is derived from the reference
> clock supplied by the LMX2594 to the RF data converter, the frequency of
> which is what you enter into the "Reference Clock" field of the RFDC Casper
> block. Assuming you have "Enable Tile PLLs" checked in the RFDC block, the
> PLL internal to the RF data converter is responsible for this.
>
> -The user clock rate is derived from the PL reference clock supplied by
> the LMK04828 to the FPGA fabric, the frequency of which is what you enter
> into the "RFPLL PL Clock Rate" field in the RFSoC4x2 XSG block. The PLL
> associated with the FPGA's MMCM is responsible for this.
>
> -The output clock is used for clock forwarding from one ADC tile to
> another, although I believe this is currently not implemented for the
> Casper RFDC block
>
>
>
> What clock rates one can obtain for each of these are constrained by the
> dividers and VCOs internal to each of the corresponding PLLs. For the RF
> data converter, these can be found here:
>
>
>
> https://docs.amd.com/r/en-US/pg269-rf-data-converter/PLL-Parameters
>
>
>
> Whereas for the MMCM, the constraints can be found here:
>
>
>
> https://docs.amd.com/r/en-US/ug572-ultrascale-clocking/MMCM-Attributes
>
>
>
> There is also the further constraint that your user clock must match the
> required AXI4 stream clock, which is equal to:
>
>
>
> User clock = AXI4 stream clock = Sample rate / Decimation / Number of AXI4
> stream samples per user clock
>
>
>
> The Casper RFDC block will automatically compute this for you when you
> enter something into the "Sampling Rate" field. It will also automatically
> compute the ADC reference clock, but interestingly the PLL constraints it
> assumes aren't quite the same as what's specified in the ADC documentation
> above, and ultimately the reference clock it gives you will result in a
> sample rate that doesn't quite match up with the one you specify in the
> Sampling Rate field and is incompatible with the required AXI4 stream clock
> shown (have a close look at calc_refclk.m in the casper-astro/mlib_devel
> repo). Similarly, the jasper backend will mis-configure the MMCM if you do
> not specify a user clock rate that is compatible with the constraints of
> the corresponding PLL (see clk_factors.py in
> casper-astro/mlib_devel/jasper_library, it will only give the closest
> frequency to what you specify).
>
>
>
> For this reason, I do not rely on the RFDC block to guide me to the
> correct clock frequencies. In practice what I do is I start with PL
> reference and ADC reference clocks supplied by the LMK04828 and LMX2594
> (e.g. 122.88 and 491.52 MHz in the default Casper config files), and use
> the above PLL constraints to work forwards and calculate all possible
> sample rates I can use given those clock rates and my required decimation
> and AXI4 stream sample number. If desired, I can send along a handy Python
> script that does this.
>
>
>
> -Walter
>
>
>
> --
> You received this message because you are subscribed to the Google Groups "
> [email protected]" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To view this discussion visit
> https://groups.google.com/a/lists.berkeley.edu/d/msgid/casper/PH7PR14MB5368930D7E91BE73519BEF6EACB72%40PH7PR14MB5368.namprd14.prod.outlook.com
> <https://groups.google.com/a/lists.berkeley.edu/d/msgid/casper/PH7PR14MB5368930D7E91BE73519BEF6EACB72%40PH7PR14MB5368.namprd14.prod.outlook.com?utm_medium=email&utm_source=footer>
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"[email protected]" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/a/lists.berkeley.edu/d/msgid/casper/CAEx9wh9hhz_4%2Bm-yw7gQz9dkgvVjbJFFNq3tF-HcgFxWowzhVw%40mail.gmail.com.

Reply via email to