Sorry to start a new thread. This is an intended reply to previous, but the reply buttons are not functional. (Not a problem listed in FAQs as far as I could find. Perhaps mods can clarify?). Anyway, the original thread, for context:
>> StringTheoryNYC wrote:
>>What is the best way to maintain time-domain accuracy when signals are filtered? IOW, avoid delays in signal within the
>> pass band.
>> I understand that this is possible with the use of FFT filters, but is that the case with Audition's FFT filter implementation?
>> If not, then what kinds of delays can be expected when using FFT filters in near-brick-wall mode?
>SteveG(AudioMasters) wrote:
>Audition has several filter implementations, both IIR and FIR.
>Near-brick-wall isn't a filter mode I've ever come across, so I can't
>tell you directly about that, whatever it is. What I can do though is
>point you at an online reference about digital filters which explains
>approximately how they work, but not in mathematical detail, and
>that's the only way you'll really understand why this is in fact a
>very strange question - you'll have to figure that out for yourself.
Thanks for your reply, Steve. I'm not sure why that would be a 'very strange question' though. 'Brick wall' refers to the sharp transition between pass band and stop band. It's a commonly used term. I prefixed with 'near', since I'd ordinarily moderate the transition to avoid Gibbs phenomenon.
I understand the relative merits of FIR vs IIR re consistent delays, and I would not consider using IIR in this case. The question was intended in context of Audition's filter implementations. It -is- possible to avoid delays by doing an FFT, applying coefficients to the FFT output, then IFFT. But I'm not sure there is more to the approach used in Audition's "FFT Filter". IOW, I was looking for insights on what delays (if any) to expect from the Audition "FFT filter". I hope that is more clear.