Forums

1/3 Octave Analysis

Started by Alisson Vieira May 27, 2014
On Wednesday, June 4, 2014 1:49:47 PM UTC+2, Bob Masta wrote:
> On Wed, 4 Jun 2014 00:46:40 -0700 (PDT), Alisson Vieira > > <alissonvieira01@gmail.com> wrote: > > > > <snip> > > > > >Thanks Bob. Very nice from you. See. Sorry about this, but when I said that I've used a microphone I have actually used a sound level meter 2238 Mediator from B&K.In this case what would change? thank you > > > > Now I'm really confused. From what I read about the 2238, > > it already comes with 1/3 octave analysis software. And > > coming from B&K you should certainly be able to trust the > > calibration. > > > > So, if you don't use any of your own software, just the B&K > > analysis, do you get reasonable results? > > > > If so, then my first guess would be that when you are using > > raw data from the B&K to do your own analysis, that data > > might not include any calibration. It might be raw volts or > > A/D counts or something. > > > > Best regards, > > > > > > Bob Masta > > > > DAQARTA v7.50 > > Data AcQuisition And Real-Time Analysis > > www.daqarta.com > > Scope, Spectrum, Spectrogram, Sound Level Meter > > Frequency Counter, Pitch Track, Pitch-to-MIDI > > FREE Signal Generator, DaqMusiq generator > > Science with your sound card!
Let me try to be clearer. I am measuring the noise of an electronic steering lock device. The noise only appears when I press a button from outside the anechoic chamber. When I do this the device will lock or unlock (it will move a pin upwards or downwards).This process takes about 1 second. So, I have this device 2238 mediator that uses serial filters. Probably, you know everything about this but I will put here what the manual says about the meter just to try and explain my situation. Serial Filters - the sound level meter measures in one band at a time. The meter will usually be fitted with a single filter circuit, which will be electronically switched to measure the different bands. With most modern meters you can set the filters to scan through from 31.5Hz to 16kHz automatically. The advantage of serial octave band filters is that the meter has only one filter circuit, keeping the cost and power consumption down. So, I can't use the device's frequency analysis software to get the 1/3 octave bands because I would have to run the device at the precise moment that it changes the frequency band. In order to solve this problem I've connected the sound level meter electrical outputs to the data logger to get the SPL variations in real time and then used the program in Matlab that uses parallel filters to obtain the 1/3 Octave Bands. When I connect the sound level meter to the data logger I can obtain two results the voltage variation and the exactly SPL in dB, which the sound level meter is showing on the display. My question is: in order to filter this signal should I use the raw voltage or should I transform these values in db to pressure as I did before?
On Wed, 4 Jun 2014 05:34:29 -0700 (PDT), Alisson Vieira
<alissonvieira01@gmail.com> wrote:

>On Wednesday, June 4, 2014 1:49:47 PM UTC+2, Bob Masta wrote: >> On Wed, 4 Jun 2014 00:46:40 -0700 (PDT), Alisson Vieira >>=20 >> <alissonvieira01@gmail.com> wrote: >>=20 >>=20 >>=20 >> <snip> >>=20 >>=20 >>=20 >> >Thanks Bob. Very nice from you. See. Sorry about this, but when I said t= >hat I've used a microphone I have actually used a sound level meter 2238 Me= >diator from B&K.In this case what would change? thank you >>=20 >>=20 >>=20 >> Now I'm really confused. From what I read about the 2238, >>=20 >> it already comes with 1/3 octave analysis software. And >>=20 >> coming from B&K you should certainly be able to trust the >>=20 >> calibration. >>=20 >>=20 >>=20 >> So, if you don't use any of your own software, just the B&K >>=20 >> analysis, do you get reasonable results? >>=20 >>=20 >>=20 >> If so, then my first guess would be that when you are using >>=20 >> raw data from the B&K to do your own analysis, that data >>=20 >> might not include any calibration. It might be raw volts or >>=20 >> A/D counts or something. >>=20 >>=20 >>=20 >> Best regards, >>=20 >>=20 >>=20 >>=20 >>=20 >> Bob Masta >>=20 >> =20 >>=20 >> DAQARTA v7.50 >>=20 >> Data AcQuisition And Real-Time Analysis >>=20 >> www.daqarta.com >>=20 >> Scope, Spectrum, Spectrogram, Sound Level Meter >>=20 >> Frequency Counter, Pitch Track, Pitch-to-MIDI=20 >>=20 >> FREE Signal Generator, DaqMusiq generator =20 >>=20 >> Science with your sound card! > >Let me try to be clearer. I am measuring the noise of an electronic steerin= >g lock device. The noise only appears when I press a button from outside th= >e anechoic chamber. When I do this the device will lock or unlock (it will = >move a pin upwards or downwards).This process takes about 1 second. So, I h= >ave this device 2238 mediator that uses serial filters. Probably, you know = >everything about this but I will put here what the manual says about the me= >ter just to try and explain my situation. >Serial Filters - the sound level meter measures in one band at a time. The = >meter will usually be fitted with a single filter circuit, which will be el= >ectronically switched to measure the different bands. With most modern mete= >rs you can set the filters to scan through from 31.5Hz to 16kHz automatical= >ly. The advantage of serial octave band filters is that the meter has only = >one filter circuit, keeping the cost and power consumption down. >So, I can't use the device's frequency analysis software to get the 1/3 oct= >ave bands because I would have to run the device at the precise moment that= > it changes the frequency band. In order to solve this problem I've connect= >ed the sound level meter electrical outputs to the data logger to get the S= >PL variations in real time and then used the program in Matlab that uses pa= >rallel filters to obtain the 1/3 Octave Bands. When I connect the sound lev= >el meter to the data logger I can obtain two results the voltage variation = >and the exactly SPL in dB, which the sound level meter is showing on the di= >splay. My question is: in order to filter this signal should I use the raw = >voltage or should I transform these values in db to pressure as I did befor= >e?
If I understand correctly, you are not getting SPL from the data logger, but only from the sound level meter; the logger is only giving voltage with an unknown calibration factor. You can use the SPL reading to calibrate the logger voltage. Take the FFT magnitude spectrum of the voltage (not 1/3 octaves). Apply the same weighting curve that the SPL meter uses. (You mentioned previously that the SPL reading was A-weighted.) Now take the RMS sum of all the spectral lines, or at least those under the curve. That should be equivalent to the metered SPL. You can now find a gain factor to apply to the spectrum data to give this same SPL. You can double-check this by using a different weighting curve on the meter, which you likewise apply to your FFT spectrum as above, and you should get matching results. You can set the meter to show a single 1/3 octave SPL, then verify that your own 1/3 octave scheme gives the same SPL when applied to this single octave. Best regards, Bob Masta DAQARTA v7.50 Data AcQuisition And Real-Time Analysis www.daqarta.com Scope, Spectrum, Spectrogram, Sound Level Meter Frequency Counter, Pitch Track, Pitch-to-MIDI FREE Signal Generator, DaqMusiq generator Science with your sound card!
use a speaker and an oscillator and create a single frequency tone that you can measure directly with your SLM.

Run the data for the single tone through your analysis and see what you get.

Repeat with two tones

Repeat with N tones

Repeat with white noise

Soon you will see the pattern to point to whatever is the problem.

Mark

On Thursday, June 5, 2014 9:14:01 AM UTC-3, Bob Masta wrote:
> On Wed, 4 Jun 2014 05:34:29 -0700 (PDT), Alisson Vieira > > <alissonvieira01@gmail.com> wrote: > > > > >On Wednesday, June 4, 2014 1:49:47 PM UTC+2, Bob Masta wrote: > > >> On Wed, 4 Jun 2014 00:46:40 -0700 (PDT), Alisson Vieira > > >>=20 > > >> <alissonvieira01@gmail.com> wrote: > > >>=20 > > >>=20 > > >>=20 > > >> <snip> > > >>=20 > > >>=20 > > >>=20 > > >> >Thanks Bob. Very nice from you. See. Sorry about this, but when I said t= > > >hat I've used a microphone I have actually used a sound level meter 2238 Me= > > >diator from B&K.In this case what would change? thank you > > >>=20 > > >>=20 > > >>=20 > > >> Now I'm really confused. From what I read about the 2238, > > >>=20 > > >> it already comes with 1/3 octave analysis software. And > > >>=20 > > >> coming from B&K you should certainly be able to trust the > > >>=20 > > >> calibration. > > >>=20 > > >>=20 > > >>=20 > > >> So, if you don't use any of your own software, just the B&K > > >>=20 > > >> analysis, do you get reasonable results? > > >>=20 > > >>=20 > > >>=20 > > >> If so, then my first guess would be that when you are using > > >>=20 > > >> raw data from the B&K to do your own analysis, that data > > >>=20 > > >> might not include any calibration. It might be raw volts or > > >>=20 > > >> A/D counts or something. > > >>=20 > > >>=20 > > >>=20 > > >> Best regards, > > >>=20 > > >>=20 > > >>=20 > > >>=20 > > >>=20 > > >> Bob Masta > > >>=20 > > >> =20 > > >>=20 > > >> DAQARTA v7.50 > > >>=20 > > >> Data AcQuisition And Real-Time Analysis > > >>=20 > > >> www.daqarta.com > > >>=20 > > >> Scope, Spectrum, Spectrogram, Sound Level Meter > > >>=20 > > >> Frequency Counter, Pitch Track, Pitch-to-MIDI=20 > > >>=20 > > >> FREE Signal Generator, DaqMusiq generator =20 > > >>=20 > > >> Science with your sound card! > > > > > >Let me try to be clearer. I am measuring the noise of an electronic steerin= > > >g lock device. The noise only appears when I press a button from outside th= > > >e anechoic chamber. When I do this the device will lock or unlock (it will = > > >move a pin upwards or downwards).This process takes about 1 second. So, I h= > > >ave this device 2238 mediator that uses serial filters. Probably, you know = > > >everything about this but I will put here what the manual says about the me= > > >ter just to try and explain my situation. > > >Serial Filters - the sound level meter measures in one band at a time. The = > > >meter will usually be fitted with a single filter circuit, which will be el= > > >ectronically switched to measure the different bands. With most modern mete= > > >rs you can set the filters to scan through from 31.5Hz to 16kHz automatical= > > >ly. The advantage of serial octave band filters is that the meter has only = > > >one filter circuit, keeping the cost and power consumption down. > > >So, I can't use the device's frequency analysis software to get the 1/3 oct= > > >ave bands because I would have to run the device at the precise moment that= > > > it changes the frequency band. In order to solve this problem I've connect= > > >ed the sound level meter electrical outputs to the data logger to get the S= > > >PL variations in real time and then used the program in Matlab that uses pa= > > >rallel filters to obtain the 1/3 Octave Bands. When I connect the sound lev= > > >el meter to the data logger I can obtain two results the voltage variation = > > >and the exactly SPL in dB, which the sound level meter is showing on the di= > > >splay. My question is: in order to filter this signal should I use the raw = > > >voltage or should I transform these values in db to pressure as I did befor= > > >e? > > > > If I understand correctly, you are not getting SPL from the > > data logger, but only from the sound level meter; the logger > > is only giving voltage with an unknown calibration factor. > > You can use the SPL reading to calibrate the logger voltage. > > > > > > Take the FFT magnitude spectrum of the voltage (not 1/3 > > octaves). Apply the same weighting curve that the SPL meter > > uses. (You mentioned previously that the SPL reading was > > A-weighted.) Now take the RMS sum of all the spectral > > lines, or at least those under the curve. That should be > > equivalent to the metered SPL. > > > > You can now find a gain factor to apply to the spectrum data > > to give this same SPL. You can double-check this by using a > > different weighting curve on the meter, which you likewise > > apply to your FFT spectrum as above, and you should get > > matching results. You can set the meter to show a single > > 1/3 octave SPL, then verify that your own 1/3 octave scheme > > gives the same SPL when applied to this single octave. > > > > Best regards, > > > > > > > > > > > > Bob Masta > > > > DAQARTA v7.50 > > Data AcQuisition And Real-Time Analysis > > www.daqarta.com > > Scope, Spectrum, Spectrogram, Sound Level Meter > > Frequency Counter, Pitch Track, Pitch-to-MIDI > > FREE Signal Generator, DaqMusiq generator > > Science with your sound card!
Thanks again Bob. I will try this out and see what I get.
On Thursday, June 5, 2014 11:14:36 AM UTC-3, mako...@yahoo.com wrote:
> use a speaker and an oscillator and create a single frequency tone that you can measure directly with your SLM. > > > > Run the data for the single tone through your analysis and see what you get. > > > > Repeat with two tones > > > > Repeat with N tones > > > > Repeat with white noise > > > > Soon you will see the pattern to point to whatever is the problem. > > > > Mark
Thanks Mark for replying. See, I've tested something similar. The issue is that by doing what you're telling me I would test mainly the filters and I don't thinks the problem is in the filters, once I have got good results for white noise, pink noise and sine waves at different frequencies. The problem is probably in how I've been handling or acquiring the data.