[srslte-users] Why setting sampling rate multiplied by 4?

[Tom Tsou]

On Mon, Apr 10, 2017 at 7:09 AM, Ismael Gomez
<ismael.gomez at softwareradiosystems.com> wrote:
> Yes I think it should work well setting the master clock rate to the
> sampling rate. I think at some time we found better results in terms of USRP
> latency when operating at low sampling rates. But that might be fixed
> already. Have you tried setting it to the sampling rate?

I just want add that there are two considerations here focusing on the
B200/B210.

1. The long term stable UHD release series - 3.9.x - supports master
clock rates down to 5 MHz, which means that the lowest LTE sample
rates (1.92 and 3.84 MHz) require oversampling. The regular 3.10
release series does not have this limitation and can handle all
relevant LTE sample rates directly.

2. Regarding latency, oversampling will generally result in shorter
latencies due to faster clocking with both the FPGA and RFIC. That
said, oversampling group delay is affected by the sample rate
converters, so the behavior is not a simple one-to-one relationship.
For latency critical applications, developers will need to either
perform timing measurements and/or testing for the required
configuration(s).

  -TT