Rational Acoustics

Vaughn Pease
November 18th, 2011, 03:06 PM
In an effort to reduce ambient noise in our symphony hall I've been asked to go in and measure the ambient noise levels of different fans, motors, vents, dimmers, etc at the audience listening position. The idea is that if we can identify items in the building that are disturbing the audience we can find a way to reduce the offending noise.

My current plan is to calibrate my mic to 94db with a calibrator and take a screenshot of an RTA/Spectrograph splitscreen for each fan or whatever as one of our engineers turns each one on. This will give me a graphic picture of both the sonic signature of the item under test, as well as a db reading as to how loud it really is at the listening position, great for passing on to the symphony board and the engineers along with a report.

Ideally I'd like to save a Spectrum trace for each one, and subtract it from a measurement of the room in its quietest state, essentially removing the room's ambient signature from the picture. Is there a way to do this? Is there any way to filter the results to ignore the room ambient?


November 18th, 2011, 10:46 PM

As I recall there was a SUBTRACT function in an older version that would allow for that very sort of use. Don't know if that feature is in v7 currently. Let me go look....


November 18th, 2011, 11:04 PM
So it turns you could, not quite as elegantly in v 5 but here's what I found...

A captured spectrum trace can be converted to ASCI text and compared to another. Compare a trace with the DEVICE on and off, and if everything else remains the same you have some empirical evidence that a certain device is the offender.

Here's what one trace would look like. Frequency on the left, relative amplitude on the right:

91.727594 -37.85
97.162795 -39.65
102.920053 -35.69
109.018449 -36.45
115.478198 -38.01
122.320712 -38.11
129.568670 -33.96
137.246096 -36.45
145.378439 -33.58
153.992653 -38.50
163.117291 -41.37
172.782598 -38.57
183.020611 -37.55
193.865264 -41.04
205.352503 -44.88

Another trace might look like:

91.727594 -27.85
97.162795 -29.65
102.920053 -25.69
109.018449 -26.45
115.478198 -28.01
122.320712 -28.11
129.568670 -23.96
137.246096 -26.45
145.378439 -23.58
153.992653 -28.50
163.117291 -31.37
172.782598 -28.57
183.020611 -27.55
193.865264 -31.04
205.352503 -34.88

A visual comparison shows that one trace is 10 dB hotter than the other.
It's not elegant, but it might do what you need.


November 19th, 2011, 11:27 AM
Vaughn, we do a lot of this type of work, for architects, municipalities, reports, etc., and there isn't one absolute best methodology but the reports we produce seem to be accepted with satisfaction. (the much harder part is FIXING these problems!). Usually we are showing these reports to people who don't understand acoustics very well, so you have to be ready to make simple sense out of your findings.
I would think using graphic representations (screen shots of spectrum traces) is a good way to get the point across. And I don't think subtracting the "quietest" room is necessary. It is probably quite telling to show the room in its quiet state and then the a/c, the motors, the dimmers "in action."
You could back that up with the ASCI text and numbers, it'll make them think you have "raw data."
> But the common notation for environmental noise is NC ratings, which current SMAART does not do (yet). There are ways to calculate this from the data you have collected, NC is a curve produced by the static noise levels at frequency bands---easy enough to read up on. You'd need your system to be accurately calibrated for SPL. Showing NC ratings in different scenarios will impress other professionals familiar with this.

Alfredo Prada
November 19th, 2011, 04:38 PM
If you need NC readings:
-calibrate to SPL
-run the spectrum with octave band scale (Flat - no weighting)
-let it settle (I use averaging)
-log into: http://michaelschwob.com/noise-criteria-calculator/

-enter SPL per band data and it calculates the NC and others like NCII with a graph to show!
I attached a sample of smaart 7 and the NC calculator results

I will be a lot faster when smaart adds the NC function