Reverb Subculture

[acreil]

Mostly I'm just curious to see how they work; I suspect they may use unique (though not necessarily worthwhile) topologies. I'd probably build and enhance the algorithms in PD. I like the Ibanez a lot, but as for the Quadraverb I'm more interested in finding what causes that awful metallic ringing . I do have other reverbs I'd rather analyze, but these allow enough control over density that it's actually possible.

[seancostello]

Originally Posted by acreil

Mostly I'm just curious to see how they work; I suspect they may use unique (though not necessarily worthwhile) topologies. I'd probably build and enhance the algorithms in PD. I like the Ibanez a lot, but as for the Quadraverb I'm more interested in finding what causes that awful metallic ringing . I do have other reverbs I'd rather analyze, but these allow enough control over density that it's actually possible.

I'd be interested in seeing the topologies, delay settings, that sort of thing. Although if there is nasty ringing, the delay settings are probably far from ideal. Or the allpass coefficients are turned up too high.

From my emails with Keith Barr, I think that the Quadraverb algorithms would be parallel combs with embedded allpasses. Similar to the Quantec algorithm, but less combs, more allpasses. Output taps taken from within the comb - not necessarily from the end of the comb. Barr avoided taking output taps from within an allpass delay, to avoid the metallic sound that can result from that.

[acreil]

Originally Posted by seancostello

From my emails with Keith Barr, I think that the Quadraverb algorithms would be parallel combs with embedded allpasses. Similar to the Quantec algorithm, but less combs, more allpasses. Output taps taken from within the comb - not necessarily from the end of the comb. Barr avoided taking output taps from within an allpass delay, to avoid the metallic sound that can result from that.

That's partly why I'm interested. I've also somehow developed an allergy to the Quadraverb's gross reverb, so I kind of want to revisit it to see if it's the algorithm's fault or my own clueless misuse at 19-ish. It's probably both.

Autechre were pretty big Quadraverb offenders too.

YouTube - Autechre - Yeesland

There's your "metallic allpass coloration". You'd think they were messing it up on purpose, but I have the exact same disaster on some of my old recordings.

[zmix]

Originally Posted by acreil

Quick hypothetical question: if I were to analyze the Alesis Quadraverb and Ibanez SDR-1000 algorithms in moderate but not exacting detail, would that be of interest here or just a total waste of time? This would include topology, delay lengths, parameter mappings (i.e. at least as much detail as the Dattorro algorithms) but not an exhaustive examination of loop filters or coefficients at specific decay times, etc.

The Ibanez SDR 1000 is identical to the SONY MUR-201 and these share quite a lot of DNA with the famed SONY DRE2000.

I have several SDRs and a DRE2000 and these are capable of some incredibly well articulated reverb sounds.. unlike anything else I've heard (and I've heard / owned then all!)

The basic reverb structure in the MUR/SDR is a true stereo in / out and is based on the same "tapped ring" FDN topology as the Lexicon / Ensoniq / etc... but are constant density, like the AMS and EMT digital reverbs.

[acreil]

That's why I wanted to try it. I'd noticed previously that an impulse into one channel will sort of pan back and forth between the outputs, like a Lexicon algorithm. But as you say it is constant density: with a long decay time and high "size" setting (diffusion), I can put in an impulse, then while it's decaying turn the size down and the tail will have the same low density as it would if the size parameter had been low to start with, indicating that the allpasses come after the main "tank" structure. This makes me think it won't be that difficult to analyze.

It seems like a unique and otherwise unknown topology, not that it's necessarily super useful today. I'd be interested to know how it differs from the DRE2000, also. I've heard (from you, probably) that it's the same basic algorithm, but is it significantly cut down or enhanced?

Obviously this isn't urgent, but I'll do it if I'm ever in the mood to do something that takes much longer than I expect.

Also, I didn't know AMS was constant density, I'm actually a little surprised. Any more info on that?

[zmix]

The SDR-1000 / MUR-201 differs from the DRE2000 in several important ways.
First of all he DRE2000 is a mono in / stereo out device. It has 4 reverb algos and 2 delay and 2 Echo (with feedback) algos.

The DRE2000 reverb algos have no size or diffusion parameter, just HP, LP (4 choices of each), High and Low Rt multiplication (fixed frequency), decay time (0-9.9s) and two dry delays feeding the L and R outputs (except in the "A reverb" algo where the first delay is sent to the outputs centered and dry and the second feeds the input to the reverb and is called "sub reverb")

The 4 reverb algos vary in density (A is the loosest, D is very very fine grained..)

[zmix]

See this post for more info:
Reverb Subculture

[acreil]

Well it seems different enough that I shouldn't worry about emulating both at once, but it's nice to see what the algorithm descended from.

[chrisj]

I'm finding I'm able to get more soundstage depth by doing a thing sort of like companding- applying a saturation effect, doing the reverb and then trying to apply the inverse of the saturation effect again. On an actual sound, it goes thin and nasty, but does drop back in the sound picture. Applied to the reverb output it seems interesting

mp3 versions and other links in this post- http://www.gearslutz.com/board/6368824-post85.html

[theBackwardsman]

Originally Posted by zmix

Below I have provided a screen shot showing the output each of the 4 reverb algorhythms of my 1980s era SONY DRE-2000, fed by a single sample impulse and with the decay adjusted to 0.1s (minimum is 0.0s but that produces no output..). Note how diffuse algos B, C and D are..!

EDIT: The file lengths pictured below are exactly 233ms from beginning to end.

You wouldnt mind sharing the impulses?

[Froombosch]

Originally Posted by acreil

Also, I didn't know AMS was constant density, I'm actually a little surprised. Any more info on that?

Maybe Warp69 should chime in

[FETHead]

This thread is way better than any technical reverb paper that I've ever read. You guys rock!
In anyway - there is at least one Sony reverb algorithm that is at least partly laid out - and it's not in a reverb box - but in the Sony D7 delay - here is one of the pathes called Zoomverb:

Pre-Delay 1: CH1 41.44 CH2 52.25 100%
Pre-Delay 2: CH1 75.52 CH2 80.50 72.95%
Early Refl 1: CH1 24.02 CH2 27.67 50%
Early Refl 2: CH1 29.31 CH2 33.69 -50%
Early Refl 3: CH1 37.33 CH2 40.06 50%
Cross PreD : CH1 94.75 CH2 99.46 40.96%
Tap 1: CH1 0.208 CH2 3.021 97.67%
Tap 2: CH1 91.19 CH2 104.19 -73.24%
Tap 3: CH1 201.25 CH2 195.77 48.82%
Cross Tap 1: CH1 31.29 CH2 41.48 90.33%
Cross Tap 2: CH1 131.31 CH2 122.96 -47.36%
Cross Tap 3: CH1 229.85 CH2 241.19 29.44%
Fdbk Time : CH1 292.17 CH2 309.92 25%
Cross Fdbk Time: CH1 317.85 CH2 342.50 10.84%
Fdbk Treble Shelf 6.3kHz -3dB

Overall EQ: LowShelf 250Hz -2.7dB ; Peak 1.25kHz +4.2dB BW1.414

No time modulation is supported by the D7 btw.

I could take an IR sometime if somebody is interisted. By the above - it's easy to realise that it is a true stereo verb - what makes it unique to me - is all the "Cross" delays as well as the feedback delay...

On a completely random note: does anybody have an example (or paper)
of the Gerzon vector allpass stucture/matrix?