On Thu, 08 Mar 2007 19:13:30 GMT, "Tony"
wrote:

"Don Pearce" wrote in message
...

1. Generate 1 second of white noise.
2. Use an FFT filter to band-limit it to the band of interest - about
1/3 octave. Normalize to 100%
3. Insert into multitrack
4. In multitrack mode, record with a microphone on another track while
playing the noise
5. Edit the microphone track, using the same FFT filter to exclude
other noise
6. Normalize to 100%
7. Change the right hand scale to dB (right click the scale)
zoom in the vertical and horizontal scale to find the point where the
amplitude crosses -60dB.
8. Subtract 1 second from the displayed time (the length of the
original noise)

Ingenious use of Audition. But I hope you won't mind a few comments from
one who has done, must be hundreds of room RT measurements over the last
25 years.

To get repeatable results you should average, as the results will vary,
particularly at low frequencies. In fact in most rooms the decay at low
frequencies will be very irregular, so you can get quite widely varying
results. You can judge how much averaging you need when you see what the
variation is between measurements - you will need more at LF and less at HF.
You should average over at least 5 mic positions (perhaps 3 at HF), and at
LF you will probably need at least 5 measurements per position. (Repeatable
is important if you are going to modify the room and test again.) But to
reduce the number of measurements greatly, you could use octave bands - that
will be adequate for most people, certainly above 500 Hz where the variation
with frequency should be very smooth.


Yes, but this was really intended to be a quick and dirty test to help
contrast what I know to be a good room with one that appears to have a
2s reverb time.

Even with filtering the received signal, unless you have a quiet room and
hefty speakers you could have a job getting enough dynamic range. But it's
normal to measure over the 30 dB range from 5 dB to 35 dB below the maximum
level, rather than the full 60 dB, and double the decay time to get the
RT60. Starting at the -5 dB point also avoids a couple of causes of
inaccuracy.


My setup allowed me to measure down to about -70-something quite
happily. Again, if I were striving for ultimate accuracy, I might
consider such refinements as starting down the curve some way, but I
was using the mic in the far field and I was confident that there was
no significant step to compensate.

Finally, the filters you use will have their own decay time. At LF this
ring will be long enough to affect the result if your room is dead. So
there is the usual unavoidable time/frequency resolution conflict, and you
mustn't filter too sharply. To check this, at the lowest frequency you're
trying to measure, do a dummy measurement with the output plugged direct to
the input instead of via the speaker and mic, and find the RT of your
measuring system.

I did make sure that the time constant of the bandwidths I was using
was significantly smaller than the kind of decay I wanted to measure
(as an old Marconi Instruments designer, I wasn't going to get caught
out by that one)

Good luck with the measurements and let us know what you find!

Take a look at the vid. In my haste with the demo run, I did miss a
normalization step for the waveform recorded from the mic, but it made
very little difference to the result.

d

--
Pearce Consulting
http://www.pearce.uk.com