PSS Fundamental Frequency Choice

You have the ratios the wrong way around. There needs to be an integer number of periods of each input source in the fundamental (if you think about doing a Fourier analysis, then you'd only have part of the cycles of 768kHz or 5.3kHz in the 40.704M fundamental. The greatest common divisor of the frequencies in this case is 100Hz - so that's what the common fundamental should be.

Note however the performance may not be great with such a high ratio (the 768kHz would be the 7680th harmonic of the fundamental, and higher harmonics of that clock would need to be captured too). 

You might need to look at QPSS, but see how you get on (rather than me explaining QPSS here...)

Andrew

Dear Andrew

Thank you very much for your reply. I sincerely appreciate it. I still have two more questions regarding your reply above.

1. "If you think about doing a Fourier analysis, then you'd only have part of the cycles of 768kHz or 5.3kHz in the 40.704M fundamental"

Could you please elaborate this a little more as I am still a little confused by this concept of "cycles of 768kHz in 40.704M fundamental"?

2. I gave QPSS analysis a read and had a question that may also be linked with this PSS discussion?

How do we determine the number of harmonics required for my system? The chopper clock (Large Signal) is 768kHz and the input sinusoid (Moderate Signal) is 5.3kHz.   

Just for reference, in "Spectre® Circuit Simulator and
Accelerated Parallel Simulator RF Analysis in ADE Explorer User Guide", the sample and hold circuit simulation example for QPSS Analysis has the large signal harmonics set to 15 and the input signal harmonics set to 5. 

When you specify the fundamental in PSS, you are specifying the period over which the analysis will be performed. The requirement is that all signals in the circuit (both inputs and internal signals such as those at the output of dividers) are co-periodic with this fundamental period. In this case, a fundamental of 40.704MHz corresponds to a period of 24.56761ns (approx). Given that the period of 768kHz is 1.3020833us (approx), you're only going to end up with a fraction of a cycle of the 768kHz within the period of the 40.704M. Put simply, the analysis would know nothing about the part of the period it hasn't seen. You need to have an integer number of cycles of all frequencies within the fundamental period to be able to capture a periodic steady state.

If you had a fundamental of 100Hz, then that allows for  53 cycles of the 5.3kHz and 7680 cycles of the 768kHz - so integer numbers of cycles of each (53 is prime so there are no higher common frequencies). If instead of using 5.3kHz you used 768k/145.0 (as an expression), then you could have a fundamental of 768k/145.0. Or if you set the fundamental to 5.3kHz and the clock to 768.5k then the frequencies are much more nicely arranged and would give you a better ratio for simulation purposes.

For QPSS, you need to have enough harmonics of the large to capture the frequencies of interest (it depends on the highest harmonic you are interested in; the large signal harmonics don't particularly affect the accuracy). The moderate signals should have enough to capture the non-linearity that these signals will face - if they are not distorting at the input signal level, then 3 may be enough. The higher you set the moderate signal harmonics too though, the slower the simulation will be - but bear in mind that if you set it too low then you are more likely to have convergence problems (the energy from all the moderate harmonics needs to be conserved, sort of, when you do this kind of analysis)

Andrew

Dear Andrew

     Thank you very much for the excellent and clear explanation. I sincerely appreciate it.

Kenny Lu