Reverse IR to counter room acoustics

[dagosto]

The following quote came from the comments of this story:

http://www.gearwire.com/sound-proof-window.html

Originally Posted by Ben

Dan I saw your impulse response video's and I figured something. Maybe it has already been done or maybe it sounds stupid, but just to clarify: You can capture a room response, and make the entire mix sounding as if it were playing in the IR'ed room by inserting the captured IR on the master, right? Would it be possible to reverse-IR a mix as well? For example, if you are mixing your tracks in a room with some acoustic problems, too much early refelctions. You could capture the room IR and subtract that IR from a dry audio loop, so it misses character. That way you could play the loop in the same room, and the missing character is added by the room again, so the signal sounds very "dry" in the room. off course, just for listening purposes. When bouncing you should disable the subtracted IR. Or is this not possible?

[dagosto]

Ben, I love the idea. I think it would actually be easier to try and do something like this. Here's how I would do it.

-Take a post-fader send off the master bus to a new bus with the IR plug-in on it. The plug would have the control room response in it.

-Reverse the polarity of the IR bus and have both busses either going to a third bus ore ot of your interface

This should actually remove a lot of the frequencies that are being hyped in the room. You could also play the two busses in phase and listen on headphones to see what the room is doing.

3 caveats though:

1. What a room does is very complex and I don't think this is something you should necesarily mix through. It might be good to reference though, similar to how you might reference your mix in mono.

2. The placement of the mic(s) when you take the IR is very important. If your room is pretty even sounding everywhere you will want to put the mic(s) further away from your speakers. Most rooms are not even throughout which means that the logical spot would be in the monitors sweet spot. The problem with this is that there is mostly direct sound there so when you use this method you will mostly just get a drop in bass response. that kind of makes this a bad method. For this reason You may want to only do this for situations where you will be listening from further back in the room (something I always check) with an IR from a similar spot.

3. Latency caused by IR plugs needs to be addressed if there is no auto compensation in the DAW.

I'll have to try this out myself to see if it is actually useful but I do like the idea.

[Ben]

Originally Posted by dagosto

Ben, I love the idea. I think it would actually be easier to try and do something like this. Here's how I would do it.

-Take a post-fader send off the master bus to a new bus with the IR plug-in on it. The plug would have the control room response in it.

-Reverse the polarity of the IR bus and have both busses either going to a third bus ore ot of your interface

This should actually remove a lot of the frequencies that are being hyped in the room. You could also play the two busses in phase and listen on headphones to see what the room is doing.

3 caveats though:

1. What a room does is very complex and I don't think this is something you should necesarily mix through. It might be good to reference though, similar to how you might reference your mix in mono.

2. The placement of the mic(s) when you take the IR is very important. If your room is pretty even sounding everywhere you will want to put the mic(s) further away from your speakers. Most rooms are not even throughout which means that the logical spot would be in the monitors sweet spot. The problem with this is that there is mostly direct sound there so when you use this method you will mostly just get a drop in bass response. that kind of makes this a bad method. For this reason You may want to only do this for situations where you will be listening from further back in the room (something I always check) with an IR from a similar spot.

3. Latency caused by IR plugs needs to be addressed if there is no auto compensation in the DAW.

I'll have to try this out myself to see if it is actually useful but I do like the idea.

That's exactly what I mean! Wow, great idea! Unfortunately I am without mic for the next 2 weeks otherwise I would be cupturing IR as we speak But even if it results in absolutely nothing usefull, I think the idea could have some other use to it? Like.. if you want to record vocals for example. You have a great mic, a good preamp etc but the one thing you cannot afford (money/space) is a acoustically decent vocal room. You could capture the room and subtract it from your vocal recordings made in that room (using your inversion idea). OK maybe it's not perfect but it could help you get rid of a lot of unwanted room verb or refections.

About your second caveat: I think that is the biggest issue to get the most realistic IR. Small rooms (vocal stuff) could be easier, but a more complex room should be handled more precisely as you said. Still I think it is somehow possible, if you for example only take one spot in measure. But in the end it's all maths and can be calculated, I guess? I would very much love to see you doing a little testing with this, if you feel like it!

[abarnett]

Originally Posted by Ben

But in the end it's all maths and can be calculated, I guess?

Although computers are a useful tool for calculating acoustics, wave behaviour in 3 dimensions is one of the most complex PHYSICS problems out there. I recently wrote a paper about this, and I was surprised that things like altiverb are possible.

What you are getting with IR is an approximation. There are so many variables. Air density is a variable. The advantage of the computer is that it can calculate many mundane tasks quickly. However, large calculations such as 3D wave behaviour would still take years for an exact answer. That is where the algorithm comes in. The algorithm in general is a detailed list of instructions on what to repeat and what to do when you get to the nth member. However, when you are dealing with a ton of data, a good algorithm is going to have to have some sort of approximation to get things done.

This is probably more than you were asking for, but I'm just trying to make it clear that computers are still far from the point of a perfect map of wave behaviour. Your actual wav files are also an approximation, so while I think you have a great idea, I would think it more useful to use the reverse IR just to get an idea of what frequencies are removed