interesting discussion..
I guess if the customer works in the time domain and insists on a time domain spec, then you have to provide that.
But what I don't see is how a single time domain jitter number i.e. 123 ns could be more useful compared to a frequency domain PLOT that shows the sp ectral density of the jitter.
THe 123 ns could be all random or is could be a single tone at 60 Hz and ju st looking at the number 123 ns tells you nothing about which it is.
The spectral plot of course makes it obvious.
And again you can integrate the spectral density to get a single number.
Your point I think is that the single number result of the integration may not be accurate, because it does not account for the phase relationship am ong the discrete components. I would add also the the result depends on t he frequency range of the integration.
I can see the point the both views are useful.
mark