Multi instances of reverb vs single bus reverb

In the following video (Session Drummer 2)

http://www.youtube.com/watch?v=-VpzlysXQzE
around 4.05, the reverb is added to diffrerent drums (hi-hat, toms, etc) in the mixer section by dragging the reverb to other drums to create instances of the reverb.

Does this means, each track have it's own reverb inserted ?

Assuming the same setting is desired, isn't it more effective to send all drums sound that requires reverb into a bus, and add just one reverb effect ?

Are there perhaps reverbs that do not create identical sounds repeatedly? If there are, such reverbs would not be recommendable for multi-instance-use like that.
And anyway, I'd be afraid that so many instances would anyway mess the sound somehow.

The point being that there mustn't be any kind of randomness in the reverberation if you use several instances of it, otherwise I believe the sound will be inevitably mashed.

Maybe I should test one day: Play, say, triplets snare/tom/floor tom with both reverb-bus and reverb-per-kitpiece. Somehow I expect to find the bus-reverb-version
more glued together and snappy...don't know...

newbie_hobbyist


I realize this may well be an extremely dumb question, but my intuitive understanding of reverbs on the coding level is almost nonexistent, and this is a question I've found myself wondering about more than a couple times. Namely, I want to know whether a reverb plugin is linear in either of the following senses.

I. Suppose I have two signals S1 and S2, and a reverb R. I can either
(i) route both S1 and S2 through R
or
(ii) create a copy of R and route each signal through a separate copy of R.

Would the resulting signal be the same in both of the cases (i) and (ii)?

II. Suppose I have a signal S, a reverb R and some amount of volume x. I can either
(i) change the volume of S by x and route the resulting signal through R
or
(ii) route S through R and change the resulting signal by x.

Gain vs. volume, as it were. Assuming no clipping takes place at any moment, would the resulting signal be the same in both cases (i) and (ii)?

------------------------------------------

I imagine the answer to both questions would depend on the reverb plugin in use (among other things, whether it does true stereo processing). I would be happy if someone could answer this just in the case of, say, the Sonitus reverb or an impulse-based reverb like Perfect Space. I imagine there is some kind of literature about this subject as well, but quick googling gave me no satisfactory results.

(I apologize if the notation here is unclear or imprecise, I had to write this in a bit of a hurry)

 
1. The results would likely be fundamentally the same if the input did not vary. In other words, if S1 and S2 are mono sources and you routed them together into a stereo reverb, one left and one right, you would get a certain result. If you then seperated them to different tracks and ran each of them thru the -center- of the reverb without panning the initial sources left and right again then of course the end result would depend on whether the reverb algorithm combines the sources (non-discrete) or keeps them isolated (discrete).
 
2. Depends on the reverb. In general you could expect identical results. However with some newer reverbs things like HF damping response can differ depending on the input level (not as common, higher CPU use) but even then the difference is negligible and likely not audible in a mix.
 
Ultimately the question is: is the reverb true discrete or not as you alluded to. If it is truly discrete then the algorithm may process left and right sources independantly and never combine the sources. Likely less common. Most reverbs are probably non-discrete which means that although the input signals remain stereo  (the Dry signal) the process itself first combines the L and R sources internally to a mono source before reverberation occurs and the resulting Wet stereo reverb is mixed back with the Dry. In other words, the reverb algorithm itself actually uses a mono input source and creates a stereo image when it processes the reverb.  I'm sure you already know this, just trying to cover the bases and clarify the discussion for those unfamiliar with the concept.
 
Just my personal experience, I can't cite references for you at this moment although I have beta tested for years with a few different software creators and created reverbs in CSound and Reaktor. Unless someone has a virtual "schematic" or personal code experience it may be hard to get a definitive answer without contacting the company. Or doing some serious testing and analysis.
 
It might be cool to see a switchable reverb that could function discreetly or not. During tracking it could sound almost exactly the same and use half the CPU. During Export (or Mixdown) it could be switched to discreet to improve the quality/imaging. Maybe there is one?

Thanks for the replies!

Seems like my notation was a bit unclear: in case I. I would more likely have two signals (usually stereo), and by "routing them through R" I meant sending them to a (stereo) bus with R on it. Like, say, two string sections on different sides of the simulated hall (with the appropriate pannings fixed before sending them to the bus). Of course, the philosophy in your example is pretty much the same.

Quote: "Ultimately the question is: is the reverb true discrete or not as you alluded to. If it is truly discrete[...]"


Yeah, this is basically my understanding of the term "true stereo processing", a phrase which I've seen used in the context of the Valhalla Room reverb (which I don't have) and the free epicVerb.

I did carry out some preliminary testing by myself and didn't find notable differences, but as you said, there might be differences that are actually difficult to hear, and I'm interested in the theoretical side of this question as well.