"True Stereo" Reverbs

[linear.b]

I think we can all agree that what we really need is a fully spatialized quadraphonic reverb module, with a true quadraphonic signal path. Also, it should have a tap-tempo LFO that's not patched to anything and has no inputs or outputs.

[KSS]

A real space will "mix" right and left sounds too.

If you think about that statement for a moment, the problem with this whole line of discussion becomes clear. Stereo is an output format that's supposed to correspond to the signals you receive at your ears. It doesn't actually describe the source of the sound in any useful way, it just reproduces how the sound would "project" onto your 2 ears.

So it doesn't mean ANYTHING to talk about a "stereo input" to an artificial reverb that's supposed to model a space. If we put a real sound source in a real space, it's not a "stereo" source, it's just a source of vibrations in the space. Those vibrations bounce around the space and eventually land on your 2 ears.

Thought experiment: Imagine you have an actual gymnasium or other reverberant chamber.

First, let's imagine a single person singing. How would you get the best and most musical reverberant sound? I would suggest that you put the person in the space and use a stereo mic setup to capture the sound, moving around the person and the mics until you get the combination of direct sound, early reflection, and reverberation that makes you feel all squishy inside. Nice!

Now let's imagine that the singer's not available, but you have a recording of that singer from a close mic. How can we most closely match the live-recorded sound? I would say it makes sense to place a single speaker, playing the recorded signal, in the space, and a stereo mic setup. Likewise, move around the speaker and mics to taste. Pretty similar to the live case. Nice!

Now, let's imagine that you have a live choir. Seems straightforward that you set the choir up in the room, stereo mics, move around to taste. Nice!

Now the hard one: what if you only have a stereo recording of the choir? For the sake of simplicity, let's say they were singing in an acoustically dead space so there's no natural reverb. Now what? Are you going to just place 2 stereo speakers in the room? Where do you place them? There's nothing you can do here that's going to make it anything like the actual choir being in the room, in terms of how the sound interacts with the acoustics of the environment. The stereo speakers are going to phase-cancel each other more or less depending on placement, the best placement would probably have them right on top of each other but that would still have phasing issues in some frequency bands depending on the size of the speakers. Gross!

What if you sum the choir to mono, then play it out of one speaker in the space? Huh, still not like the actual choir being there, but at least it's not a total shitshow of phase cancellation. Mix in the stereo recording as the "dry" sound and you are getting pretty close. Maybe you keep the mics farther back from the source in the room, since you know that you are going to mix the dry sound in stereo later, and just try to capture the room sound. Nice? Definitely not as nice as having the choir there, but better than trying to take an output format and pretend it's a source!

Well said and presented!

If you are going for an unnatural effect, dual mono is the way, but it will sound unnatural for sure.

I appreciate this part of ersatzplanet's post, as room sim is *not* the only reason for reverb! I've added a second tank and channel in my 50SQ 1068 module which otherwise duplicates the 2600's output-pan and mixer-reverb functions. Adding a normal-mono switch to account for the difference.

[KSS]

Stereo modular is this 'season's' new thing. AKA LBD.

[madcaphorse]

I think we can all agree that what we really need is a fully spatialized quadraphonic reverb module, with a true quadraphonic signal path. Also, it should have a tap-tempo LFO that's not patched to anything and has no inputs or outputs.

I think there's a lot of truth in this: when folks talk about wanting a true stereo reverb, what they might be after is true spatialization. Yes, the reverb of a voice placed at on far corner of a sound stage will be much different from one at the other side's near corner, but as has been said much better by others in this thread: dual mono is a poor way to simulate this.

There are lots of tricks to using common reverb algos and mixing tools to emulate this, but to get at a proper simulation of a spaces reverberation for a given input sound, I think you need spatial coordinates - along with either a mono source (or some way of describing a "point" audio source). One somewhat straightforward approach that comes to mind would be a sort of adaptation of the wavetable approach for convolution reverb: simulate stereo impulses at every point on a grid within a given space. Then feed in an input signal and x/y coordinates, use the coordinates to either choose or interpolate between impulses, and process the convolution of the input signal with the resultant stereo impulse.

Does anyone know of any similar implementations in hardware or software? I know there's a ton of work done on sophisticated spatialization, so this is probably pretty naive, but I'm interested in exploring a bit if there's not an existing implementation already.

[cbm]

The early reflections of a sound in a space can be pretty localized, but by the time it really smears out into a reverb tail, it matters less where the sound came from in an L-R panorama, in my opinion.

[revelc]

audio damage
adm23- EOS
https://www.analoguehaven.com/audio-damage/adm23/

Sounds great, sadly has issues that were never fixed.

[Ebotronix]

I use two reverb devices of the same type in dual mono.

[ersatzplanet]

The fact that you think it’s so crazy that so many reverb modules do some kind of summering, and that it defeats the purpose stereo made me think maybe you didn’t quite understand WHY that is (because if you did it wouldn’t be that crazy).

But yeah it’s an interesting discussion, good topic for sure.

I think it all comes down to how much "mixing" of the stereo inputs happen. The stereo outputs are a given to create a true "large space" feel with multiple reflections outputting differently from separate speakers (to recreate a "space" or "room"). The stereo inputs depend on how the designer thinks of the "performer's location". All performers in real spaces will mix together in the room, but they won't mix perfectly and have the same levels in the reflections of the walls etc. If I were to design a stereo reverb, I would design the inputs to have a variable mix control that feeds the two inputs to two sets of reverb generators, which also have a mix control to blend their outputs together. This way you have a unit that can be all the different configurations you may want, dual mono reverbs, dual mono into slightly blended stereo outs, or fully blended stereo reverbs. I would think that this would make dialing in the perfect "space" possible.
This would also allow for some more realistic "location modulation" where a stereo sound could move between the stereo inputs, allowing each channel to be set to different reverbs of a larger virtual room, so a sound can get a specific reverb character added, as it would if you say, sang while walking around in a big room.

[KSS]

If I were to design a stereo reverb, I would design the inputs to have a variable mix control that feeds the two inputs to two sets of reverb generators, which also have a mix control to blend their outputs together. This way you have a unit that can be all the different configurations you may want, dual mono reverbs, dual mono into slightly blended stereo outs, or fully blended stereo reverbs. I would think that this would make dialing in the perfect "space" possible.

Sounds familiar! Have most of that in place already.
1068 of my 2k6 module set is an update of 2600s output section with dual tanks. Option for VC pan and output levels using three faint grey jacks, unplaced in this view. Tested circuit, building it this summer. Will fit in 6U of Euro with shortened middle rails and AdAPTer with 10 jacks upper and lower to replace the MxSws of 50SQ.

Thank you for the underlined! I've been trying to figure out how to get fully to where you just described, and now see the clear path!
Based on your post I'm going to add a D/C pot and lower knob under the orange one. Change the label to Reverb Mix. Add an input arrow from L6,7 to the lower knob.

Then the middle position of the slide switch will fully implement your plan. Was only getting half of it before*.
Left position keeps them fully separate and right position duplicates original ARP2600.

*Within the module. It was possible to use an external mixer into L6,7 and then the orange knob output mixer and slide switch got to where you described. But with the D/C pot and knobset, it doesn't need the external mixer!

[grrrwaaa]

It would be nice to be able to have many mono inserts, each at different "locations". As has been mentioned above, the perceptible differences are mostly in the early reflections; and the diffuse tail can be pretty much shared. So, depending in how the early refections are made, it might just be like a blended mix of write points into the early delay lines/allpasses. It could be something a bit like mid/side or ambisonic encoding on the inputs, feeding the w, x, y weighted signals into three different delays in the early reflection network. Then your parameters per source might be just width & azimuth; you get a complex space in the earlies, and still a no undesirable phase issues in the diffuse tail.

[Zymos]

Stereo modular is this 'season's' new thing. AKA LBD.

I think that’s what everyone was saying “last season” too….

[linear.b]

The excellent info in this thread got me thinking about the nature of human hearing vs "stereo" and reflecting on the impossibility of replicating much of anything "accurately" with two speakers, or five.

If you plug your right ear completely with an earplug and then play a low volume sound on a speaker, say a 1200hz sine wave at 45 db, too quiet for any sound to bone-conduct itself into the right ear, and then close your eyes and have someone move the speaker around in space, you will be able pretty easily to tell them if it's in front of your, or behind, even if it's positioned such that the absolute volume is the same in all cases.

So, what's going on here? Hearing is complex and even though we use the differential between the left and right ear to make huge and broad spatial perceptions, the ear itself is getting sound from all directions, and, critically, the stereocilia inside the organ of corti which actually are the thing thst respond to external audio in a physical way, are not a flat wall but are a three dimensional spiral. So, the orientation of the sound coming into a single ear is to some extent discerned by the firing pattern of different stereocilia that are generally responsive at the same frequency band but oriented in different parts of the ear.

This makes hearing loss sort of insidious, because you may be able to hear a pure tone at 1200hz just fine through an audiologist's test headset, but if that sound was moved to some other place spatially, deficits may exist. My own hearing is notched out pretty good around 14khz, but I can still hear that frequency in both ears -- but, if I make it quiet, and move it around, there's a lot of dead zones that I cannot hear that sound from that direction at.

So, getting back to "true stereo reverb" -- stereo is already such a tremendously stripped down version of "environmental audio data", that I get to "who cares as long as it sounds cool!" My joke about quadraphonics was not entirely a joke, as I have been thinking about setting up 2 mixers and some complex submixers and panners to have a quadraphonic euro setup in my room and be able to pan in both directions and send sound around and doo weird stuff. But, to get close to really simulating what I would call "actual spatial data". I think you would want a room with about 200 speakers in it, at different heights, distances and orientations relative to the listener.

Processing all that audio in real time sounds like one hell of a DSP problem to solve.

Fortunately, stereo effects are "good enough" to mess with my brain and make me go "ohhh. nice warm reverb, I am in warm fuzzy space with happy little beeps, thanks mimeophon!!!!"