FFT windows question

"martin griffith" schrieb im Newsbeitrag news: snipped-for-privacy@4ax.com...

Hello Martin,

If you exactly know the frequency you are looking for, you should set the FFT-time to exactly an integer multiple of this period. Example: F=2.0kHz, T_fft=N*0.5ms, e.g. 10ms In this case the the filter "None" is best.

If you don't have this information, any filter gives a better resolution for the price of a "wider" peak.

I have attached an example to prove that.

Best regards, Helmut

Save this schematic as" fft_test.asc".

Version 4 SHEET 1 880 680 WIRE 96 112 32 112 WIRE 480 112 96 112 WIRE 32 160 32 112 WIRE 480 192 480 112 WIRE 32 304 32 240 WIRE 32 432 32 384 WIRE 480 432 480 272 WIRE 480 432 32 432 WIRE 32 464 32 432 FLAG 32 464 0 FLAG 96 112 out SYMBOL voltage 32 144 R0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR InstName V1 SYMATTR Value SINE(0 0.3m 2k) SYMBOL voltage 32 288 R0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR InstName V2 SYMATTR Value PULSE(0 1 0 4.95u 4.95u 1n 9.9u) SYMBOL res 464 176 R0 SYMATTR InstName R1 SYMATTR Value 1 TEXT -16 32 Left 0 !.tran 10.25m TEXT 272 -8 Left 0 ;FFT 131072 data points TEXT -16 -8 Left 0 !.options plotwinsize=0 TEXT 272 24 Left 0 ;Compare window "none" with "Kaiser".

Thanks Helmut I will look at that in the morning, 5 hours of LTspice is enough for one evening

martin

These days, doing a plain old DFT instead is not that big a deal. And completely eliminates ALL windowing problems.

Many examples at

Can you elaborate? Windowing is needed to prevent edge effects.