Hi Brian,

thank you for your help.

I have on question left. You say the crossbar is non blocking. Does that
mean it can supply multiply RFNoC Blocks with input data at once at its
full bus_clk speed? Or does it switch between the ports so that some
blocks have to wait until its their turn to get data?

Best regards,

Felix

> On Thu, Sep 19, 2019 at 9:39 AM Felix Greiwe via USRP-users <
> [email protected]> wrote:
>
>> Hello together,
>>
>> I have some questions concerning clock speeds and the corresponding data
>> rates on a USRP x310 (FPGA). As far as I know, there are two different
>> clock speeds on the FPGA, the ce_clk=200MHz, and the bus_clk - I did not
>> find clock speed for this one.
>>
>
> The ce_clk is 214.286 MHz and is usually associated with a 32-bit AXI
> interface.
>
> The bus_clk is 187.5MHz and is usually associated with a 64-bit AXI
> interface.
>
> If you build an FPGA image, you can find these values in
> post_route_timing_summary.rpt that Vivado spits out in your build
> directory.
>
>
>>
>> Is it true, that the ce_clk drives my rfnoc blocks and thus my in- and
>> outgoing data rate of each single block (using sc16 samples) is
>> 200MHz*32
>> Bit/10^9 = 6,4 GBit/s?
>>
>
> It can, and usually does - but just slightly higher as noted above.
>
>
>>
>> I read, that all the RFNoC Blocks are connected to the crossbar which is
>> driven by the bus_clk. First of all: Is this the case?
>> If so, how is the crossbar able to handle the in and output data of each
>> RFNoC Block at once? How many Bytes can it process with each clock?
>>
>> Take for example the flowgraph
>>
>> SignalGenerator ->RFNoC-Gain -> RFNoC-DMAFIFO-> RFNoC-DUC-> RFNoC-Radio
>>
>> which has already four RFNoC Blocks connected to the crossbar, which in
>> my
>> head are 25,6 GBit/s data on the crossbar at once which seems way to
>> much
>> to handle.
>>
>> I think I really miss a point here and would  be grateful for some
>> explanation.
>>
>
> The crossbar doesn't block other ports and is more like a packet switch.
> Since it's a linear flow, the crossbar doesn't have any issue handling
> each
> individual path bandwidth.  Only when 2 packets have to go to the same
> crossbar egress do things become more complicated.
>
> I hope this makes sense.
>
> Brian
>



_______________________________________________
USRP-users mailing list
[email protected]
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Reply via email to