on the Rx side, I use uhd.tune_request(915000000,10e6)
then I down-sample and display a FFT
the amount of down-sampling is enough that I can see how much frequency offset I get.
I am testing it with a carrier generated from a second B100, using uhd.tune_request(915000000)
even if the reference had an accuracy problem. deriving the clock would produce the same error for all devices connected to it, given the reference signal is within spec
under the scope, the reference signal seems to be a perfect sine with an rms within spec.
currently, for now, I will have to understand more what the tuning does, and why.
since the freq offset I see is constant, I can live with it, for now.