3 ms·
At uni professor was drawing inverse FFT of impulse response in a filter and he only used first few frequencies and then asked "can anybody see what's wrong wit
by dvh 4y ago
At uni professor was drawing inverse FFT of impulse response in a filter and he only used first few frequencies and then asked "can anybody see what's wrong with this chart?". We were all silent...
Turns out if you only use first few frequencies of FFT the filter output will look like it respond slightly before the signal arrives. It's a freaking time machine.
- jononor 4y agoNon-causal filters can be quite useful, and can be realized when the entire input sequence is available. And if one puts in a delay line, then can work without needing access to future timesteps. A bit harder to realize with electronics or mechanics than software though...
- xnoreq 4y agoWhat are you talking about? Do you mean first few frequency bins, and setting the others to zero? That would be a simple linear-phase lowpass filter. And if you do the shifting properly then you will see that it has a positive group delay. There are filters with negative group delay, however. But FFT is irrelevant for that and it's not a time machine either.