when should one perform fft based filtering?
in [DSP]
|
From: Steve Pope on 27 Jul 2010 18:37 robert bristow-johnson <rbj(a)audioimagination.com> wrote: >there are multiple products that implement this convolutional reverb >(McDSP, Waves, Digidesign, i think). i am *sure* that fast >convolution is used for at least the dense reverb tail. i do not know >what the throughput delay is, but i know that they're real-time or can >be real-time. Sure, because if you produce the first 100 msec or so of reverb by some sort of direct convolution, your transform-based tail has that much extra time to catch up. So it can spend that on its internal latency. Or put another way, a typical mixed FIR/IIR reverb would have a delay line in series with it Schroeder (IIR) reverb anyway. The really interesting question, totally unrelated to the above, is whether you can produce a suitable (for your market) reverb purely with linear filtering. It's pretty easy to demonstrate, for example, that a plate reverb is nonlinear. Steve
From: Les Cargill on 27 Jul 2010 22:31 Steve Pope wrote: > robert bristow-johnson<rbj(a)audioimagination.com> wrote: > >> there are multiple products that implement this convolutional reverb >> (McDSP, Waves, Digidesign, i think). i am *sure* that fast >> convolution is used for at least the dense reverb tail. i do not know >> what the throughput delay is, but i know that they're real-time or can >> be real-time. > > Sure, because if you produce the first 100 msec or so > of reverb by some sort of direct convolution, your transform-based > tail has that much extra time to catch up. So it can spend that > on its internal latency. > DAWs with plugins can use latency compensation. They cheat. SIR is a convolution engine which publishes a latency spec. > Or put another way, a typical mixed FIR/IIR reverb would have > a delay line in series with it Schroeder (IIR) reverb anyway. > > The really interesting question, totally unrelated to the above, is > whether you can produce a suitable (for your market) reverb purely with > linear filtering. It's pretty easy to demonstrate, for example, > that a plate reverb is nonlinear. > > > Steve Purely convolution based plate reverb sims are pretty compelling and widely used. -- Les Cargill
From: Steve Pope on 28 Jul 2010 00:34 Les Cargill <lcargill99(a)yahoo.com> wrote: >Steve Pope wrote: >> The really interesting question, totally unrelated to the above, is >> whether you can produce a suitable (for your market) reverb purely with >> linear filtering. It's pretty easy to demonstrate, for example, >> that a plate reverb is nonlinear. >Purely convolution based plate reverb sims are pretty compelling >and widely used. Thanks for this information. I've been away from this market segment for so long, this conversation is one of my first glimpses at even remotely recent thinking on the subject. Steve
From: robert bristow-johnson on 28 Jul 2010 00:48 On Jul 27, 6:37 pm, spop...(a)speedymail.org (Steve Pope) wrote: > > The really interesting question, totally unrelated to the above, is > whether you can produce a suitable (for your market) reverb purely with > linear filtering. well, the Schroeder kinda reverb (without time-varying tap movement) is an LTI system. and if the taps vary slowly, it's still linear. so i think the answer is "yes". i think all of these digital reverbs are pretty linear, if you ignore quantization effects on the words. > It's pretty easy to demonstrate, for example, > that a plate reverb is nonlinear. well, if you (over)drive the damn thing into non-linear response, sure. just like any device. but for very small signals, i would expect a plate reverb to be describable as an LTI system. i know a plate 'verb can sorta "resonate" on some frequencies, depending on its dimensions, but does a plate reverb create new frequency components? if so, is that not just due to the transducers going nonlinear? r b-j
From: Steve Pope on 28 Jul 2010 01:34
robert bristow-johnson <rbj(a)audioimagination.com> wrote: >i would >expect a plate reverb to be describable as an LTI system. >i know a plate 'verb can sorta "resonate" on some frequencies, >depending on its dimensions, but does a plate reverb create new >frequency components? if so, is that not just due to the transducers >going nonlinear? If different frequencies travel at different velocities through the plate medium, is that not a nonlinearity (at least in terms of behavior of the medium)? Steve |