Please list True Stereo Reverbs

[toothnclaw]

But the summing changes what's being fed to the tail (and how)*, so the reflections this generates are different from what you'd hear if there was no summing.

And in real world there's no strict separation between ERs and tail, so that's really complex.

Or, more likely, I don't understand this topic quite well.


* Probably something like this:

- ValhallaRoom can have independent processing for left and right channels, with controllable mixing time for energy moving between the left and right. I'm not sure how useful that is, but it's in there.
- ValhallaVintageVerb has fixed left/right mixing, where the speed depends on the algorithm used, the reverb size, and the decay time.

[Warp69]

Any digital reverb engine is an approximation of a real phenomenon and is therefor not true - but we can advance the technology to improve the emulation.
Some have stated numerous times in this thread, that "real" is not always the goal in the design process.

A Lexicon based allpass loop (Lexion, ValhallaVintageVerb, etc.) will eventually mix the left and right channels even thou it has stereo inputs.

[toothnclaw]

A Lexicon based allpass loop (Lexion, ValhallaVintageVerb, etc.) will eventually mix the left and right channels even thou it has stereo inputs.

I'm not saying it has to be "real". It can't be! Please note the word "eventually".

the fact that mixing a stereo microphone signal to mono will create weird artifacts. If these artifacts are injected into a reverb, it adds coloration that can't be taken away. If the mixing happens inside the reverb network in some manner, the coloration doesn't happen.

I'm not taking sides here, or arguing, just trying to understand it better. Still far from it. But Sean said that in contrast to feeding the late reverb with a mono mix, the 4 signal sources can be fed into the reverb network at different points. Which, obviously, is very different and more complex.

And that's where

but we can advance the technology to improve the emulation.

comes in.

[valhallasound]

Any digital reverb engine is an approximation of a real phenomenon and is therefor not true - but we can advance the technology to improve the emulation.
Some have stated numerous times in this thread, that "real" is not always the goal in the design process.

A Lexicon based allpass loop (Lexion, ValhallaVintageVerb, etc.) will eventually mix the left and right channels even thou it has stereo inputs.

Most "true stereo" reverbs will eventually mix left and right channels. However, the PHASES of the left and right signals will have been totally scrambled by the time that the signals are mixed. Moreover, there will be a huge number of different mixes, i.e. a signal that has been filtered through N allpasses will be mixed with a delayed signal that has been filtered by N different allpasses, while the next mixing will have different allpasses and different delays between left and right. This produces something where the mixes are constantly changing, and happen at the OUTPUTS of the reverb, as opposed to before the INPUT of the reverb.

I just realized that the only mixing of left and right inputs internally in a stereo allpass loop reverb happens with energy that is "hanging" around in the allpasses. This information will be totally scrambled in phase, and constantly changing and rescrambling. Otherwise, all the left/right mixing happens in the reverb outputs, and your typical allpass loop has a LOT of outputs.

In an FDN reverb, the left and right inputs MAY mix in the FDN matrix (eventually they WILL mix in most FDN reverbs). The left and right signals will be summed, which will cause phase cancellations. However, there are a few things that complicate this:

- The left and right signals usually have a bunch of different delays, so you are summing delayed signals that have far different phase cancellations than simply summing the left and right inputs.

- The reverb matrix doesn't just sum. It differences. In other words, with LeftDelay1 and RightDelay1 (to hugely simplify things), there will be two outputs: Left+Right, and Left-Right. These generate complementary notches in the frequency spectrum, so that the total frequency response in the FDN network remains flat. Each of the individual FDN matrix outputs will have different frequency responses, but they add up to an "allpass" response.

The coloration I have talked about with mixing a stereo microphone signal down to mono before feeding the reverb has nothing in particular to do with the reverb itself. Instead, it is a consequence of the stereo microphone signal itself, and the delays and phases between the two microphones being used. The signal itself sounds bad mixed to mono, regardless of what happens after the mixing.

I figured this out during the development of ValhallaShimmer. I was using "Sprout and the Bean" by Joanna Newsom as a test signal, and it sounded really bad with my mono input reverb. It turns out it sounds REALLY BAD just mixing the two channels down to mono, without any reverb. However, a stereo input reverb sounded great with "Sprout and the Bean."

Sean Costello

[Warp69]

That still doesn't change anything - As I wrote earlier : The preprocessing by the ER engine before the mixing could contain multiple taps and/or multiple allpass filters and/or something else. You could scramble the sources as much as you want before the mixing stage.
An ER engine is not necessary equal to a 4 tap delay line (it could, if you decide so) - it could also contain 80 taps, 16 allpass filters with modulation in finely tuned structures for each channel before the mixing stage.

Im unsure what we are discussing - NONE of current algorithms are remotely close to true behaviour anyway, no matter how much processing is done before a mixing point. The proposed algorithm produce a L->L + R->L, L->R + R->R output just like allpass loops, fdns and similar structures - I thought we already established that.

If you want to argue - it's not what happens in a real environment, then that's correct - BUT, none of them are remotely close (or closer than the proposed algorithm) to real behaviour anyway.

[valhallasound]

That still doesn't change anything - As I wrote earlier : The preprocessing by the ER engine before the mixing could contain multiple taps and/or multiple allpass filters and/or something else. You could scramble the sources as much as you want before the mixing stage.
An ER engine is not necessary equal to a 4 tap delay line (it could, if you decide so) - it could also contain 80 taps, 16 allpass filters with modulation in finely tuned structures for each channel before the mixing stage.

If you sum the ER outs, you will get a signal with a SINGLE comb filter response. 80 taps and 16 modulated allpasses will generate a complicated comb filter response, but the frequency response will always be just that one response, circulating through the reverb.

Meanwhile, if you had 2 or more inputs to the late reverb, you could avoid the comb filtering from summing the ER outputs. Or, if you had a pure mono input to the late reverb (i.e. the late reverb is just fed a sum of the left and right inputs to the entire reverb), you could avoid the comb filtering that happens from summing signals with different delays and allpasses, and just take a chance that most input signals will sound good that way.

Im unsure what we are discussing - NONE of current algorithms are remotely close to true behaviour anyway, no matter how much processing is done before a mixing point. The proposed algorithm produce a L->L + R->L, L->R + R->R output just like allpass loops, fdns and similar structures - I thought we already established that.

I'm not saying that it doesn't produce those outputs. I'm just saying that better structures exist, and that there will be coloration in it versus other techniques that are described as "true stereo." What possible advantages are there for summing the inputs in your hypothetical structure, versus having a late reverb with two or more inputs?

Sean Costello

[Warp69]

Remember that the Lexicon based reverb loop is a comb filter by it self and should receive as much care and tuning as other structures - some of the complexity could be moved outside of the main loop. The Lexicon algorithms already have moved some of the complexity outside (diffusion and FIR(Shape/Spread)) - I could easily use the same injection point to the loop without anyone noticing a perceived difference. Lexicon have already done it - one injection point, but two separate diffusion paths that get summed. Others do it too as mentioned. You would always need careful tuning, but the same is true no matter which approach you choose.

I do not favor one approach over another - I don't even think that the current delay loop structures is the way forward.

[Andrew Souter]

play nice guys!

this is starting to remind me of:

[valhallasound]

play nice guys!

this is starting to remind me of:

"He that is without reverb theory geekery in this thread, let him first cast a stone."

- Hal Chamberlin, Musical Applications Of Microprocessors, p. 247

[Andrew Souter]

"He that is without reverb theory geekery in this thread, let him first cast a stone."

- Hal Chamberlin, Musical Applications Of Microprocessors, p. 247

I think Hal may have been paraphrasing...

[valhallasound]

"He that is without reverb theory geekery in this thread, let him first cast a stone."

- Hal Chamberlin, Musical Applications Of Microprocessors, p. 247

I think Hal may have been paraphrasing...

That clip is a bit rough to watch. Mainly because I expect Willem Dafoe to start talking in his Bobby Peru voice. *shudder*

[Andrew Souter]

That clip is a bit rough to watch. Mainly because I expect Willem Dafoe to start talking in his Bobby Peru voice. *shudder*

Awesome score though!

Despite being predominantly real instruments, it's one of those things you just simply are required to include as an influence if you work in electronic music. It's kinda like Bladerunner in this regard...

I think I took an ethno-musicology class at USC b/c of this score...

[Michael Carnes]

I thought all the Exponential Audio reverbs were true stereo. Maybe Michael Carnes can weigh in. They sound fantastic. I just wish they weren't ILok.

Yes, they're true stereo. The nature of the stereo output reflects the nature of the stereo input in terms of the way the signal propagates through the system.

[antithesist]

Just when I thought this thread had died the utube death, Somebody Special arrives.

Not that the other "Revererati" aren't special... or, uh... spacial?

Where Casey?

OK, back to hooking up two quadraverbs in "true stereo."

Don't say it's not!

I can't hear you (ears covered).

Oh, "Passion," as well as some ethnomusicology, are highly recommended.

Better yet is to listen and play.

(Book learnin' can be fun, though.)

[Cues: Smithsonian Folkways "Mbuti Pygmies of the Ituri Rainforest."]

It might be just about time for the other half of my Chipotle Barbacoa burrito, too.

Carry on (in mono, stereo, 7.1 or whatever you got).

[Sparky77]

There must be others. 2cAudio's reverbs?

Aether and Breeze each have a true stereo modes. Pretty sure B2 does as well.

B2 is a duel stereo reverb I believe.