So reverb is, to some extent, a combination of delay lines filtered different ways with different delay times fed back into eachother.
Reverse reverb, in words, is reverb that gets louder for each note up to a certain point and then stops. But there's no envelope detection; it does that for every note throughout the decay thereof, through glissando, etc. (most famously of course, through the sound of a whammy bar bringing the pitch up to a chord from slightly flat.)
How do you achieve that algorithmically? The only way I can think of involves making the whole reverb for the signal as quickly as possible and passing that into a delay line with many taps each of which is given more weight the further along you are. Is that how it's done?
That would be fairly easy to acheive through convolution (http://www.dspguide.com/ch6/2.htm). Just take a reverb impulse response and reverse it. Of course, that would only work with some DSP systems. What kind are you looking at using?
I suppose ultimately the FV-1 is the only hope I'd have of doing it myself, and I've never even started trying to code one of those. I was just trying to describe it to myself algorithmically really, to work out the "right" way of looking at it.
Some FV-1 related info (http://www.spinsemi.com/forum/viewtopic.php?t=117&highlight=reverse)
What's the role of the all-pass filter in such a situation? I know it's used for changing phase, what does that mean for our purposes?
BTW, take into account that this effect will introduce a high amount of latency to your signal, as the dry signal has to sound at the end of the reverb.
My understanding is that allpass filters are used in most reverbs as a "smearing" and density element. By smearing the signals out in different phase relationships, there will be all kinds of complex phase cancellation, which simulates the dampening that happens in real reflective spaces. But that's for any general reverb, and isn't specifically anything to do with reverse reverb.
I think a better way to describe reverse reverb is that, normally in reverb, density/dampening increase over time, but in reverse reverb, reflection density decreases over time.
If you want to try this in the FV1, it looks like, for example in the program "rom_rev1" available free on the Spin site, you can just change the read-in coefficients for the delay taps to be increasing in volume instead of decreasing. This is super-simple, in fact if you're really interested but don't have the means to test it, if I have time at some point I'll try it myself and report back.
Edit: eh, what the heck. I tried it. Can't say it really sounds very reverse-y to me. I may have misinterpreted Frank's suggestion.
Quote from: Taylor on May 13, 2011, 03:09:58 AM
My understanding is that allpass filters are used in most reverbs as a "smearing" and density element. By smearing the signals out in different phase relationships, there will be all kinds of complex phase cancellation, which simulates the dampening that happens in real reflective spaces. But that's for any general reverb, and isn't specifically anything to do with reverse reverb.
I think a better way to describe reverse reverb is that, normally in reverb, density/dampening increase over time, but in reverse reverb, reflection density decreases over time.
If you want to try this in the FV1, it looks like, for example in the program "rom_rev1" available free on the Spin site, you can just change the read-in coefficients for the delay taps to be increasing in volume instead of decreasing. This is super-simple, in fact if you're really interested but don't have the means to test it, if I have time at some point I'll try it myself and report back.
Edit: eh, what the heck. I tried it. Can't say it really sounds very reverse-y to me. I may have misinterpreted Frank's suggestion.
Did you have a reverb 'going in' or a dry signal though?
What I think he's saying (what I think I was trying to say too) is you have a reverb with no predelay going into a delay line
--¬a-----¬b------¬c------¬d
say, and then you mix .25a+.5b+.75c+1d then the reverb will 'grow'. The speed of the delay line is responsible for the speed of the swell and the input reverb can be anything I guess, something a bit abrupt at the far end like a 'gated' reverb would probably be ideal.
Ah, well, if I understand you correctly*, then that's not what I think most people call reverse reverb. However I have done that kind of thing minus the reverb and it's quite possible.
Indeed, you can have a swelling reverb like that, and no gate is necessary really, but it doesn't sound like when you play backwards music into reverb, record that, then reverse it so the reverb is backwards and the playing is forwards. That studio trick is what I always thought people wanted to emulate with "reverse reverb".
With what you're talking about, the density isn't changing backwards, so it doesn't sound like something is far away and then comes near you. It sounds more like something that's far away, getting louder.
*Looking back at my usual late-night forum posts tells me that my comprehension skills fall off a cliff around 4 in the morning, as it is now, so I can't be sure I do understand you correctly.
Well now I don't know if I'm talking right. The studio trick would require stompbox prescience or the ability to play in time a couple of beats in advance! But could sound really really cool on something like a harmonica, eh? ;)
I guess what I want then is "reverse envelope reverb". Example: (The effect is a shoegaze classic)
http://www.youtube.com/watch?v=oiomcuNlVjk
More recent and much more clear:
http://www.youtube.com/watch?v=i0YwN_NunfA (after the build up)
A 'reverse envelope' reverb is called reverse on many of the rack and now floor units used (digitech hardwire, Marshall reverb, etc.) although the "classic" unit used were the yamaha SPX rack units that called it "early reflections" or something like that.
The classic reverse reverb is on Whole Lotta Love where they flipped the tape over and recorded reverb in reverse.
Obviously you can't make a pedal do this, but you could make it happen on delay repeats, although it would take quite a bit of memory.