Impulse Recording for Convolution Reverb

[dcharrison]

I have a project coming up where I will need to simulate the acoustics of a large limestone cavern. I'd like to use it as an excuse to try my hand at impulse recording, but I have never done this before. I understand the theory--use a recording of a short, loud burst of broad spectrum audio and its resulting reverberations as the basis for processing another audio file to give it those same reverb characteristics. It's kind of the audio equivalent of ray tracing.

Unfortunately, that's about all I know.

I have seen numerous software packages for convolution reverb, but I don't know which ones allow the importation of wav based impulse files. Are there open source tools like this available?

I was planning on using my little FR2-LE as the recorder and an AT 825 stereo mic, but how do I generate the impulse? I have heard of using everything from Japanese pop guns, starter pistols and signal generators.

Suerly someone who has done this before can mentor me. Help me out and if I pull it off, i'll send you an impulse file from a limestone cavern in the Texas hill country.

[Chaellus]

you can use a starter pistiol..is the fastest way record setup and shot let the recorder capture everything from the start of the reverberation to the end of the tail....when you got it recorded then import the wav file in a program and truncate the rest of the material before and after the start and tail...also make sure that the file isnt distorted...then mark it as what ever it is ex. limestoneimpulse.wav or etc then with a convo plug it will ask you to load a wav file then it will read it and then it will use its charcterists on what ever track you highlight..very simple.....you can also take impulse from diffrent angels to get diffrent sides of the cave which you could use for surround if you get 5 impulses of the same place from diffrent angles

[pkautzsch]

A good freeware convolution reverb is SIR: News

Just load the impulse response wav file into the plugin. Be sure to remove the impulse itself, since this is what the "dry" signal in conventional reverbs is.

[David Spearritt]

A pistol shot may not give you enough energy and S/N ratio depending on the size of your cave. Swept sine or MLS will give better results. Angelo Farina has written extensively on this, for example:
http://www.acoustics.net/objects/pdf...e_farina03.pdf

and there are some other recommendations here.
Acoustics.Net.
http://world.std.com/~griesngr/dgaes92b.pdf

[yrplace]

Altiverb is the best convolution program imo, but with any such program you can burn a CD of the impulse sound and use a portable player and speaker as the source and whatever mics/ pre/ DAW combo you have available to record the space. Or you can playback the impulse in say Pro Tools while recording the space on another set of tracks.

I've done this using a laptop and an M-box which worked fine..... Mark

[dcharrison]

Once I have my impulse response trimmed and ready, are there any rack mount efx boxes that will allow me to load my own impulse file? It might be better if my final installation didn't involve a computer.

Probably too much to ask.

[just.sounds]

I would use a swept sine. Try to get an omnidirectional speaker for this purpose.

The Sony DRS-777 and Yamaha SREV1 are convolution reverbs but i don't know if they can load a sample.
And the new quantec yardstick

For me an A-B worked better then a X/Y or MS because the reverb sounds somewhat mono. an A-B between 30 and 40 sounded best to my ears. If you go the altiverb route keep in mind that they include an impulse generator and a deconvolver program. But it will only work in altiverb. Fuzzmeasure pro is also capable of doing this. The key to good impulse responses is the off-axis coloration the source and the microphones have to have as little as possible.

[David Spearritt]

Here is another deconvolver to remove the excitation signal from the measured response.
Voxengo

[dcharrison]

Hardware Convolution
The Yamaha SREV does indeed allow you to create your own samples. They even include the software, but Zorks! It's a little pricey to bury into the wall at an installation. I'm guessing it must take some significant DSP power to compute all those reflections in real time. Can any of the plug-in impulse verbs run a stereo verb in near realtime? I might be able to run this on an embedded processor with open-sourced software.

Bullets & Banging
It's one thing to carry a pistol down into a cave, backpacking a PA in might be a little tricky. I have a target pistol that will take a .22 long shell. I'm guessing that a .22 long makes a bang somewhere near 140 dB--the threshold of pain area. As I understand it, the gunshot may not distribute its energy evenly across the spectrum, but getting a swept wave close to that loudness would take several Mackie SRM 450s at 127 dB, each. (the closest thing I have to a PA)

Sweeping Sines
Wouldn't the swept sine wave have to complete its cycle before the return of the first reflection in order to distinguish it from the rest of the sample? Even if the first reflection were 50 feet away, that only gives about .05 seconds to sweep from 20 to 20K. What if the ceiling is only 20'? That's down near.01 seconds! Is that enough time? Geez, a full cycle at 50 Hz is .02 seconds. I can't wrap my head around this.

Post Processed EQ
If I used the pistol and then processed the signal after the fact, in theory, I could correct for any nonlinearity in the impulse and it's resulting response. Is that reasonable?

Thanks in advance for all your good guidance!

[just.sounds]

Yes the DSP is quite intense as it has to multiply each sample with all the samples in the impulse response. The swept sine technique is not intended to use real time. The idea is to touch all frequencies with that sine wave and later it goes trough quite eleborate FFT processing. It is also known as mlssa if i'm correct. This process will calculate the sweep (wich can be minutes) back to the wanted impulse. The longer you sweep the better S/N you get. But beware every tick or click during the recording of the impulse results in a whine in the impulse. ((sine=tick)-->(tick=sine)) That is what happens in the so called deconvolution. When you use a starter pistol equalisation is needed. Try to go to a resonance and reverb free place and record the starter pistol a few times and you can see on a FFT spectrum analyser what has to be done. Don't forget in a reverberant place that you are a quite efficient absorber yourself.

Here are some ideas to use boomboxes or small monitors.
The Altiverb™ IR Pre-Processor

[andy_simpson]

Originally Posted by dcharrison

...
Bullets & Banging
It's one thing to carry a pistol down into a cave, backpacking a PA in might be a little tricky. I have a target pistol that will take a .22 long shell. I'm guessing that a .22 long makes a bang somewhere near 140 dB--the threshold of pain area. As I understand it, the gunshot may not distribute its energy evenly across the spectrum, but getting a swept wave close to that loudness would take several Mackie SRM 450s at 127 dB, each. (the closest thing I have to a PA)

If you will be shooting a gun in a confined space, it would be a good idea to take hearing protection. Not least because it is hard to shoot a gun remotely!

According to the literature I have to hand, you can expect 152-157 dB (linear) SPL at the ear firing a .22.

Regarding the Mackie speakers, I carry a pair of those to location recordings quite often.

Last time I took them to an orchestral date I carried the pair (one in each hand) up a flight of stairs and for about 500 metres - I was in a hurry - my heart was beating like a hammer after that! Not recommended unless you are conditioned better than me.

If you have an original 'italian' (rcf drivers) pair they can crank out an orchestra at 2-3 metres better than most studio monitors I've ever heard (better with a sub).

Andy

[mohthom]

If you use Logic Studio, Impulse Response Designer does what you want while taking away the stress. Point, click, done.

MohThoM