Impulse Response Research

[chemicalwinter]

Hey all,

So I've gotta do a research project on recording impulse responses in a concert hall. BUT, I need to make it specific within some how. For example, I could test and analyze different transient sources and see what the most effective.

Any one have any ideas of directions I could go with this? Or have any tips/tricks with collecting impulse responses?

Thanks!

[Sound Sorcerer]

Go to:
Impulse Responses

Cheers!

[hpyle]

If you haven't read Angelo Farina's papers on recording impulse responses in concert halls, then start here: Angelo Farina's Publications (search for "impulse")

[David Spearritt]

I have just successfully measured the IR of our local concert hall here in Brisbane. I cannot recommend the following software highly enough.
ARTA home

It is very easy to use and offers three measurement excitation methods, swept sine, periodic noise and MLS. The IR measurement itself is trivial and can be saved directly as a WAV. It runs the signal excitation, recording, averaging all with great competence. No post processing for the IR, and the ETC and decay waterfalls are also derived directly, as is the RT.

Will post the IR here soon. Just get ARTA, a good strong omni source, a laptop and decent soundcard and you are away.

Wouldn't it be great if all of us could measure our local halls and post them here.

[just.sounds]

First thing is to get a pretty omnidirectional loudspeaker. A speaker with dull off axis response will dull more than 75% of the impulse response. And that is the same for mic's.

For microphones dpa4060's are great and cheap. A bit noisy for recording but perfect for sweeping as noise is not a problem and they have no off axis colour. I would suggest 40cm spacing. But that is a taste thing.

For speakers you maybe have to build something or use multiple. Off axis dullness is youre enemy.

[BLP]

Hey chemicalwinter, lots of good advice up above. Definitely read Farina's paper(s). He is, in many ways, the current "wise old owl" of the IR world.

My personal bit of advice is that you need to know exactly what you're after. As I'm sure you know, any change in either source (i.e. speaker) or receiver (i.e. mic) placement will yield different results. If you're trying to gain insight into the basic room characteristics (RT60, bass ratio, envelopment, early reflection strength, etc...), a more "neutral" setup might be advised; source on stage in a location mimicking the location of a solo performer or small ensemble and the mics somewhere in the audience, possible at "the best seat in the house". If the idea is to be more practical, though, you might approach it differently. Work as you would during a session in the hall; place the player/source on stage and then place your mics as you would a main pair.

So, after having spent an inordinate time collecting IRs in the last month, my piece of golden advice is this: IR collection is like recording; you have to know where you want to wind up before you can know where you need to start.



PS What test signals are you considering using? I'm on a bit of a long sine-sweep kick right now, but MLS and Golay codes could also have some advantages, given different situations...

[BLP]

Originally Posted by just.sounds

First thing is to get a pretty omnidirectional loudspeaker. A speaker with dull off axis response will dull more than 75% of the impulse response. And that is the same for mic's.

+1 on this, too! We're currently using an eight-speaker (2x4), 32-driver system with bi-directional and omni-directional speakers (all custom builds). This is a great system to get a full-range test signal, but very dangerous if the system is not put together properly. If the crossover points aren't smooth, your IRs will reflect the peaks and dips in your source frequency response. If you do feel that you must acheive true, full-bandwidth testing and don't have access to a well-tuned multi-driver system, you could try to approximate the system's frequency response and then use it as a "correction curve" to be deconvolved from your IRs...

[chemicalwinter]

I plan on doing research on different impulse transient sources. I will use the same microphone and same speakers and test different sources. I plan to use a clap, a sine sweep, a slap of two blocks of wood, a start pistol, and a balloon pop. I have a Mac so I will be using Logic.

Any suggestions for Mac based software that will allow me to measure the responses or analyze it in any way?

Thanks guys for the help! It's all been very useful so far.

[David Spearritt]

Chemical, you will not be able to get anywhere near the required signal to noise or energy to excite the room from any of the transient sources you list. You will need a swept sine or periodic random noise to do it.

Also, just using DAW software, especially Mac with Logic will make your job very difficult. Borrow yourself a PC, and a copy or ARTA and you will have a chance.

[Audiop]

He may be able to run ARTA on the Mac.

But would it really be that much of trouble just recording to wav and then export it into a PC? ARTA accepts imported wav-files for analysis.


/Peter

[BLP]

Chemicalwinter, the guys are right about your S/N ratio. Sinesweeps will provide you the best S/N ratio, while the others yield quite noisy results. Also, having played around with some of the other sources you're considering, you'll find there are severe frequency anomalies in the sources themselves when not using something more mathematically proper (i.e. sine sweep, MLS, etc.).

As for analysis, MatLab offers the most flexibility, as you can write your own analysis scripts. The software, however, is not something you'll sit down and just use. I would suggest using FuzzMeasure. It's quick and easy and should suite your purposes nicely; it works on Mac, it both generates and records sine sweeps, it performs analysis on the input and there's a 30 day (I think) free trial!