Dear all,

Busy with transient noise analysis with a fixed time step fixed at a small size like step=0.1ns, i use strobeperiod (say 200 MHz) to reduce the volume of signal data when measuring noise / phase and post processing with Matlab / Octave. Because strobeperiod is sampling signals it should be preceeded by an anti-alias filter such that noise measurements will not be loaded with HF noise folding in band ~[0.0 100MHz]. Such anti-alias filters are not preceding strobeperiod action and one should be careful when interpretating noise results after strobe period.

Does any-one have recommendation vs. the issue? Do we need anti-alias before sampling. How create anti-alias function. Any comments.

Best Regards,
ppg.

 

Hi Ken,

Yes indeed,"Aliasing is innate in strobing", this was expected. Thank you for confirming.

I am simulating a detailed frac-N (AD)PLL models as combination of veriloga blocks (+ some analogs) including non-linear effects and blocks with hidden states. Thus a situation which i believe can only be analyzed using time simulations.

Regards,
ppg

Hi Ken,

Yes, I fully understand "noise added to crossing times" is much more accurate as "noise added to voltages" and i appreciated the detailed reading of the recipe you outlined. Two remarks though:
1. The frac-N (AD)PLL I am simulating with transient noise analysis is already modelled directly in the phase domain or at jitter level (as you recommended).
2. The results from analysis, the power spectral density of the phase deviations, must be accurately plotted for offset frequency ranging from 1 KHz to 100 MHz. Expected in band phase-noise is ~-105 dBc (for 2.4GHz carrier). Simulations run 10-20 ms simtime for smooth and accurate phase noise curves down to 1KHz. Sub-sampling at 200 MHz (from 0.5ns fixed step) via strobeperiod is meant to reduce the amount of data to be enter in 10*log10(psd(.)) function.

Regards,
ppg