Possible to use REW to compare the output of a device vs noise floor of a room?

[walkeryyj]

Hi all,
Sorry if this is a stupid/basic question but I'm looking for a tool to allow me to take measurements of a device (electric motor) and the measurements of a room and then compare both sets of data to generate a spectrograph of JUST the electric motor IE remove the noise of the room to just show the motor. IE use some cleaver math to generate a poor mans anechoic chamber.

Is this something that REW can do?


I have a UMIK-1 and am happy to get a fancier microphone as needed. Recordings are to be done is a makeshift soundbooth but there's still a bit of background noise that I think will be hard to fully eliminate.

thanks so much!

 

[John Mulcahy]

Yes and no. The problem with noise is that it varies, so the measurement you take of the noise alone will not reflect the noise that was present when you measured the motor. You could try and mitigate that by taking a number of measurements of each to narrow the range of uncertainty. REW's RTA will let you capture the spectral content and save it as a measurement, you can then perform subtractions using the trace arithmetic features of the All SPL graph. The result is only likely to be meaningful in regions the motor level significantly exceeds the noise though, you don't really add any useful information by trying to subtract the noise floor and the result is more likely to be misleading. The safest way to present the info would be to overlay the motor measurement on the noise measurement so it is clear where the contribution is due to the motor alone.

 

[walkeryyj]

Thanks John!
that seems to be a reasonable way of showing it off.
Another quick question (again for the uninitiated) presumably if we are running the calibration file with a UMIK-1 the SPL readings will be reasonably accurate? no additional tuning would be required?

 

[John Mulcahy]

Yes, they should be, but if you need to rely professionally on the results it would be wise to check the SPL reading against a calibrated SPL meter or a mic calibrator. There have been a couple of reports of UMIK's being 3 dB off (though I have several and they all read correctly on the calibrator).

 

[BingCrosby]

As a follow up, from an existing impulse response measurement, is it possible for Room EQ Wizard to calculate and display the noise floor?

 

[John Mulcahy]

There is a plot of the noise floor on the distortion graph.

 

[BingCrosby]

Thanks for your reply. Is there a way to export the noise floor to text?

What I'd like to do is compare the SpL of a windowed impulse with the noise floor.

 

[John Mulcahy]

Total SPL or SPL at particular frequencies? The distortion graph shows the fundamental and the noise floor, so the levels at any frequency of interest can be compared there. The noise floor data isn't exported as part of the distortion data, though that could be added in a future release.

 

[BingCrosby]

Thanks for your reply.

I think what I'll do from now on is to do a sweep with the speaker off. That will be my background noise trace on the SPL graph.

Have you considered the option of saving the recorded sweep? Sometimes it is convenient to listen to these. I know there is the option to save the input and load a recorded sweep, but this means different play and record software would be required.

 

[John Mulcahy]

The captured sweep is very large, unfortunately, particularly for longer sweeps, so that would inflict too high a penalty on memory use and file size.

Better to use the RTA for background noise measurement rather than a muted sweep, with no correlation peak to align to a muted sweep may exclude some captured content and the processing of the signal into a measurement alters the spectral content of the noise floor.

 

[BingCrosby]

Thanks John!

Another partially related question (perhaps it should be in another thread). When starting an Leq measurement for long term noise in the logger, I've noticed that the initial values in the Leq trace are always higher than the SPL trace. The noise I'm measuring is highly stationary. Why would this be the case?

 

[John Mulcahy]

'Initial' meaning over what period? Their relative levels will probably be quite dependent on the crest factor of the noise.