My false perception of comb filter, and a question..

[Tepeix]

Hi,

After all this time working with them, i just realize something with comb filter..

Seeing the bode plot or others representation. I was thinking that a comb filter attenuate strongly one or 2 band,
then attenuate less when going to high frequency.
Like this :


But i just realize that it's some illusion from the representation.
If i try to correct the image here's what really happens :

The attenuation band become less larger, but they continue to cancel the sound completely at some frequency ?

But is it really the case in a discrete sample rate system ?
Or is there some point where the comb will not really attenuate in the high ?
Maybe because the cut become to small and have less chance to be the same as an input frequency ?

[adamszabo]

A comb filter does not attenuate anything, it only delays a signal, thus causing frequencies to cancel each other out, thats why you see the comb effect. The only way it could attenuate frequencies is if the delay is interpolated.

[Tepeix]

Yes, sorry for the bad term !

But note, most of the linear filter i know are done with multiple delay adding normally or as feedback,
(is there other's way to make linear one, maybe calculating analog resistance, capacitor ?)
but in case of feedback of 1 sample delay, we combine the cancellation with a slowing down of the signal ?
But even like this, maybe we could consider that a feedback of 1 sample is considered like delays of 1s*9 + 2s*0.81 + 3s*0.729 + 4s*0.6561 etc.

Well maybe with filter using fft there's no cancellation ?