Room acoustic modeling

Room acoustic modeling
Next generation of reverb plugins?
Needs a pretty face though like REVerence.


Cool! This is something I dreamed about late '80s/early '90s. Planned to implement something like it myself, but realised it would take days to calculate single impulse response (I hadn’t even heard term ‘impulse response’ then) with a PC of those days.

Yes, interesting. Thanks for posting.

Isn’t this kind of software’s used (by architects, etc) to predict how rooms will sound? I wonder how good the predictions are.

Would be fun to play with for a while, but I’d probably get fed up spending time specifying the size and shape of the room and the reflectivity, etc, of the surfaces.

Of course, there’s the Voxengo Impulse Modeler available for $50, but that just works in two dimensions – and I expect that (apart from a few special cases?) you can’t represent the reverb of a 3D space with a 2D model(?). (Perhaps the reverb in a spherical space could be modelled by a (2D) circlular space?) I suppose it could be handy if you wanted to construct reverbs of unusual/unfamiliar spaces. But I don’t think I, for one, could scrape together enough enthusiasm to work with it - though there’s a free trial version (if I ever did want to find out).

Anyway, there are plenty of off-the-shelf IRs available, now. :slight_smile:

would be good if they used the 3D space so you could place sources and destinations visually, then had a click and select surfaces library. Also a library of rooms and objects and things… etc

Yes, that would be essential. There is no way they could reproduce my rooms acoustics without it knowing where my lavalamp is.

Can two different rooms have identical IRs?

If not, is it (in principle) feasible to work backwards from an IR and deduce the rooms size and shape, and the nature of its surfaces? …

… AND pinpoint where Strophoid’s lavalamp is?

Yes they can.

Google just threw this up:


I read the start and end, but skimmed the middle. Essentially, I think it says that, for a restricted type of two-dimensional “room” - convex polygon (all walls of the same type of reflective surface?) - not only is it theoretically possible to deduce the shape of the room from the IR, except for symmetries (eg rooms that are mirror images of each other have the same IR), but also the authors provide an algorithm to determine the shape of the room from the IR. (The IR is measured at the same point as the sound source.)

So, I think that means that, if you created an IR using Voxego’s Impulse Modeler with a lot of restrictions, the algorithm could reconstruct the shape of the room.

But the algorithm can’t say where a two-dimensional lavalamp is in the room, because the method assumes the room’s empty.

Shame it’s not about 3D (a room being a convex polyhedron).

Something like this?

Yes, that’s what I was referring to in two of my posts above. (I didn’t give a link.)

I don’t think it would be possible to deduce the entire 3d geometry of a room from a single mono recording. That’s just too many unknown factors.

I wonder how soon it becomes impossible - eg perhaps if you knew the room was the shape of a shoebox, you could work out its dimensions from the IR. What if you knew it had six faces like a shoebox, but the opposite surfaces weren’t necessarily parallel (and adjacent ones weren’t necessarily at right angles) - I wonder if its size and shape could be deduced from an IR. And so on, through various relaxations of constraints on the shape and the nature of the surfaces.

But … well … do I really care? …

For musical purposes, I like the way convolution reverb can sound stunningly realistic, compared with old-style algorithmic reverbs. But even convolution reverb can sometimes sound like it’s a separate added “layer” of sound, not an integral part of the sound. I’ve read that that’s where the use of multiple IRs can bring back realism - eg using several different IRs, depending on where on a virtual stage the various instruments are.

But thinking again about making personal use of acoustic modelling, it does seem like a lot of trouble to go to, and all for an uncertain benefit - you might spend hours on it and come away with something that doesn’t sound good(?). And, as more and more off-the-shelf IRs come available, I think I, for one, will become less and less tempted to try modelling for myself - and I’m really not much tempted, to start with :mrgreen: .


Just came across this:

“VIRTUAL ROOMs” Impulse Library

… Includes nine free downloadable IRs, created by modelling.

Information about Free IR’s

Here, the Emes website says there’s going to be a CD of such IRs:

though it’s not clear whether that’s recent or old.