Judging the output of a filter - FFT

[atferrari]

A square signal (variously pre-filtered / buffered), enters the same type of filter trying to get the best approach to a sinewave of 3960 Hz. Corrected

For four different topologies, LTSpice gives the attached FFT. My questions:

Which one is the best in terms of purity no matter if that "best" is still a poor one? Could anyone tell briefly what are the main points to know which one is?

How do you judge the height of the harmonics against the height of the fundamental? Or is it the wrong way to see this?

Additional comments appreciated.

[suicidaleggroll]

With a square wave, the problem is harmonics (especially the odd ones).  You should quantify the amplitude of all (or at least the first few) versus the fundamental for each filter.  eg:
Filter 1:
2nd harmonic: -15 dB
3rd harmonic: -10 dB
4th harmonic: -20 dB

Filter 2:
2nd: -16 dB
3rd: -8 dB
4th: -22 dB
etc.

Which one is "best" depends on what you're going to be doing with the signal, and which harmonics are more/less detrimental in your application.

And you do know that none of those are anywhere near 1980 Hz, yes?

[T3sl4co1l]

Step response isn't a good way to test a filter.  Use an AC Analysis.

Step response is for when you want to test the time-domain characteristics: how flat is the step response, overshoot/ringing, rise time, those things.  These are directly at odds with the frequency response (a sharper frequency response necessarily has more ringing in the time domain, while a faster risetime with no overshoot has a very soft frequency response).

Tim