Noise Data - Real Time Octave Analysis vs. FFT-synthesis
Noise Data - Real Time Octave Analysis vs. FFT-synthesis
(OP)
I work for a global automotive company with several NVH labs worldwide. Although many of these labs are dedicated to testing specific products, the need often arises to take on testing from a sister lab. Lab to lab correlation is required at our company (calibration stuff). However, exact testing methodology is not required.
Recently, differences between noise data collection became a rather contentious point.
Our lab takes most noise data via Real Time Octave Analysis algorithms(1/3 octave, A-weighted, etc.) Whereas our sister lab takes FFT noise data and then synthesizes it. (1/3 octave, A-weighted, etc., etc.)
Since both labs had performed similar experiments with differnt results, a more thorough correlation study was requested. We recently performed a correlation study involving this issue using both methods and obtained some troubling results. (different values for both octave bands as well as overall values) Both of our labs have swapped set-up information and equipment lists. It is understood that many set-up parameters could effect the measured results.
The BIG question? For taking noise data, which method is better? Which method is more precise? Which method is more accurate? Which method is right? Some feel (myself) that it's better to go right to Real Time Octave Analysis algorithms while others feel it's better to perform base FFT and then synthesize the data. Some feel that the answers should be identical. Still others feel that errors and loses could occur from doing the FFT - synthesis method. Can any one shed some light on the problem at hand?
Thanks
Recently, differences between noise data collection became a rather contentious point.
Our lab takes most noise data via Real Time Octave Analysis algorithms(1/3 octave, A-weighted, etc.) Whereas our sister lab takes FFT noise data and then synthesizes it. (1/3 octave, A-weighted, etc., etc.)
Since both labs had performed similar experiments with differnt results, a more thorough correlation study was requested. We recently performed a correlation study involving this issue using both methods and obtained some troubling results. (different values for both octave bands as well as overall values) Both of our labs have swapped set-up information and equipment lists. It is understood that many set-up parameters could effect the measured results.
The BIG question? For taking noise data, which method is better? Which method is more precise? Which method is more accurate? Which method is right? Some feel (myself) that it's better to go right to Real Time Octave Analysis algorithms while others feel it's better to perform base FFT and then synthesize the data. Some feel that the answers should be identical. Still others feel that errors and loses could occur from doing the FFT - synthesis method. Can any one shed some light on the problem at hand?
Thanks





RE: Noise Data - Real Time Octave Analysis vs. FFT-synthesis
Obviously, you should also consider the format of the relevant standards for your industry.
The FFT and the 1/3 octave formats should agree in terms of overall level.
There are several calculation pitfalls with each format, however.
If you send me some measured time history data in ASCII text format, then I will perform each calculation for you. Then you can compare your FFT and 1/3 octave results with my results.
Tom Irvine
Email: tomirvine@aol.com
http://www.vibrationdata.com
RE: Noise Data - Real Time Octave Analysis vs. FFT-synthesis
I would have to be very confident of my test methodology to reduce the total data taken during a test to a single number.
Cheers
Greg Locock