Improving Fourier (FFT) Accuracy in PSpice

In PSpice, you can often improve the accuracy of your Fourier (spectral) analysis by modifying your transient analysis setup.

Here are some guidelines for getting more accurate results:

_Final Time_

where Tss = Time to reach steady state.

The output must reach steady-state before you can even begin to look at the data. Otherwise, the signal is not periodic, and it will introduce noise in the Fourier transform. You can set the No-Print Delay (from the Analysis menu, choose Setup, then click Transient) to Tss to avoid having to restrict the data when using the FFT feature of Probe. Although the simulation still begins at Time=0, No-Print Delay effectively truncates the leading portion of the data file, making it smaller without affecting the accuracy of the FFT.

N = Integer number of cycles to be transformed (Frequency domain resolution = F0/N)

It is usually best to use 5 or more cycles; the magnitude of the frequency components between harmonics (which should be zero) gives you an idea of what the effective noise-floor of the transform is.

F0 = Fundamental frequency

In multi-tone analyses, F0 is the highest frequency of which all input frequencies are integer multiples, i.e., F0 = Fi / Ni for all input frequencies Fi, where Ni are arbitrary integers. If you are using the Fourier Analysis function of PSpice, then you should specify the fundamental frequency as F0/N to get the frequency resolution discussed above.

_Step Ceiling (and Print Step)_

Nfh = Highest harmonic component with a "significant" amplitude.

The Nfh component must be well resolved, at about 25-100 points per cycle (Ntres)."Significant" indicates relative amplitude compared to components to be measured.For example, consider a 1kHz sine wave into a highly nonlinear system. To study the first 5 harmonics, set the Step Ceiling (and Print Step) to 1/(5*1kHz*25)=8us. However, if the accuracy required is 1mV, and one of the neglected higher harmonics is greater than 1mV, then the higher harmonic component can "fold back" (a kind of aliasing) and disrupt the accuracy of the harmonics being measured.

25 This range was derived empirically, but it is interesting to note that if one period of a sine wave is represented by 25 evenly spaced points with linear interpolation, then the maximum possible error is approximately 0.8%,and with 100 points, the maximum error is approximately 0.05%.

_Simulation Accuracy_

Simulation accuracy is limited to about 3-3.5 digits (60-70dBc). The accuracy can be improved by tightening tolerances and taking the above considerations to extremes, but not by very much. This may not be considered sufficient for some applications, but keep in mind that even the best simulation models are probably not accurate enough to provide better results.

Do share your insights which one from the above helped !!.

Happy Learning.