[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Measuring Reverberation time

Title: Measuring Reverberation time
Ben Hi,
A quick and easy way of estimating Rt is by ear. Clap your hands and listen for a rough guess (you may also want to look at your watch).
If you're after the numbers, you'd have to use a more copious method.
A basic one would be to run a few dB plots on a Neutric sound level recorder, e.g. Neutrik (provided you can find one). If this is the case, here's what to do:
First thing you'll need is a loud impulse, i.e a very short sound (loud means you have to use ear protection).  For this purpose you may want to fire one of those special acoustic pistols (B&K used to do them). Alternatively, two large pieces of wood or smtng similar (e.g. broomstick) would do. Before you bang! place the neutric analyser and your mic in one end of the room (preferably a pressure-mic, most importantly omnidirectional -Neutric type 3382 usually comes with the plotter and make sure it's a recently calibrated one). Ask a friend to do the banging at the other end of the room (it doesn't matter where the source lies as long as you keep it as far distant from the mic as possible, you want to capture as less of the direct signal you can get away with - nevertheless, you should also avoid placing both source and mic near the walls).
Next: 'activate' the sound source and run the plotter.

[Bare in mind you might have to run a few tests in order to find an approrpiate paper speed on the Neutrik - they do come with a few settings, a sample configuration for a medium sized room I did the other day would be:

Assuming you now have your plot, what you then do is calculate the slope of the decay curve.
In short,
Interpreting the plot you are interested in the decay portion of the graph, which comes shortly after the direct sound has died away. In order to calculate the slope of this - call it a curve (and find the decay rate), try drawing a best fit, find two points and take the ratio of the difference of their coordinates. You can measure the distance between the two points using a small ruler.
Do this and you're done, there's your figure. Note you won't find Rt by looking at the horizontal measure because the plotter motor runs at a specific paper speed, not in real time.

You may repeat the whole process using a noise source.

Another way to do this experiment (Neutrik analysers have gone out of fashion I guess) would be to use a soft sound editor instead (just print out the plot and go through the same calculations (decay rate - best fitting line - two points - gradient). Again the mic needs to be calibrated (and so would your amplifier of course) and the source kept as far from the mic as possible (or indeed vice versa).

You'll need a noise source (white), a sound level meter and octave band filters if you want to measure frequency response. This method is far more complicated and you have to be cautious (I couldn't possibly emphasise any more the importance of using calibrated equipment). If you have this equipment (B&K are the standard) consult the manuals, there you'll find all the information you need to take accurate measurements.
Finally, there are more sophisticated equipment you can use (e.g. LINDOS, melissa boards etc. etc.), but they cost enormous ammounts of money.

Of course, if you know the absorption coefficients you can calculate Rt beforehand (you can look-up the materials in one of those long tables - building standards manuals have them). If the walls are parallel and given the dimensions quoted above you are most likely to get standing waves (the scientific term is eigen tones) which you may also calculate up to a frequency. There's various formulas on the market, which you'll find on the net if you do a keyword search either on reverberation time, or Sabine, or Norris-Eyring, or eigen tones (or eigen modes) and/or other assorted terms.

By the way consult
Kryter, K. D. 1985. The effects of noise on Man. Second Edition. New York: Academic Press.  
on speech intelligibility measures.

Hope this's been helpful.

Sotirios Bakagiannis

P.S. Your graph should look a bit like like this:

Attachment: hand drawn sample plot bmp
Description: application/applefile

Attachment: hand drawn sample plot bmp
Description: Binary data