I want to measure a reverberation time in the room. What is a best way to do that? What kind of signal should be used? I know about balloon and pink noise, but are there any better methods?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
6
Well, there is plenty of ways for approaching this problem. Which one is best depends on room itself, your resources and pursued accuracy. I think that best way to start is to look into ISO 3382-2:2008 standard, you will find lot of information. You should remember that post processing of your signals is very important here.
Nevertheless you should consider following methods:
Hand clap, balloon, starter gun. Cheap and most intuitive. You are recording the "impulse" itself from which you are then retrieving the reverberation time and other parameters. Basically almost every other method is relaying on getting the impulse response of your room. After that you filter it in frequency bands of interest and find the slope. Order is not random here. Hand clap doesn't have wide frequency response in low frequencies, balloon is slightly better, and starter gun will give best results.
Intermittent noise. You play the band-pass filtered pink noise (usually in octaves) and register it's level in dB scale. You turn off playback and record the decay. From the slope in dB scale you find the time at which energy curve will cross -60 dB from steady state. This is your reverberation time for this given octave band. It is important to use omnidirectional source and microphone. It is also good if you repeat results for different source-microphone arrangement, i.e. for 2 positions of your source and 6 positions of microphone.
Sweep sine. For that you presumably need a piece of software. You can either write your own (here is how to do it), use a free one (such us REW) or even buy some professional tool (i.e. Dirac, EASERA, etc.). Similarly to above you play signal from speaker, record for few positions and then analyse using favourite application - as simple as that. This is my method of choice.
Maximum Length Sequence. This is white-noise-like signal that correlates nicely with itself and can be used for extraction of impulse response. Personally I do not recommend it as sweep sine seems to work better for me in every case. Although some people like to use it for measurement of speaker cabinets.
-
$\begingroup$ Thank you for detailed answer, I will try to use Sweep sine method for this room.! I started read your second post how to program it myself, it is great!! Do you might have any source code for that? $\endgroup$ Commented Feb 15, 2015 at 18:42
-
$\begingroup$ Good choice. Well, I did my own application some time ago but it is written in MATLAB (yuck!). I think that it would be OK now to make this github repo public. I need to think about that. $\endgroup$– jojeck ♦Commented Feb 15, 2015 at 18:50
-
$\begingroup$ Yes pliis, matlab is great for me!! :-) $\endgroup$ Commented Feb 15, 2015 at 18:55
-
2$\begingroup$ joj, there's a good reason to not recommend MLS in the case where there are any nonlinearities. i tried to spell it out in this decades-old page. a non-linearity in the system will show up as little noisy spikes in the impulse response. and you can average until the cows come home and that spike ain't going away, unless you use different MLS's (with different primitive polynomials) to get the different impulse responses that you ending averaging. then, instead of averaging, i would recommend computing the point-by-point median. $\endgroup$ Commented Feb 15, 2015 at 22:07
-
$\begingroup$ thank you robert bristow-johnson for details!! i will not use MLS $\endgroup$ Commented Feb 15, 2015 at 22:18