[time-nuts] PN on internal/external references and Cross correlation

[John Miles]

> In a PN measurement system with a dual channel cross correlation is there a
> simple rule of thumb for how low should be the PN of the (external)
> references compared to the DUT? (even with a 20 dB noise floor
> improvement with 10000 cross correlations)

If you have two references, then it's OK for them to be somewhat noisier than the DUT.  Their contribution will average out of the cross spectrum over time just like the rest of the instrument noise.  There is no penalty in accuracy, only in measurement time.

Given a choice, you're better off using references with good close-in noise performance rather than good broadband performance, since it takes longer for those FFT segments to converge.  You can achieve 20 dB of noise floor improvement at offsets >10 kHz within a few minutes, but a 20-dB improvement at 1 Hz might require running overnight or even longer.

> In the case of the references
> have an equal noise contribution compared to the DUT will the results suffer
> from loss of accuracy?

With two references it's not a problem.  If you have only one reference source -- or if your measurement setup doesn't do cross correlation at all -- then the reference needs to be at least 10 dB quieter than the expected DUT performance to keep its contribution below 0.5 dB.  

-- john, KE5FX
Miles Design LLC