SPL Levels exported from LMS Test.Lab as spectra and third do not match... 1

[Shifas]

Hello everybody,

I am currently doing sound measurements using LMS Test.Lab 6 (very old version - I know). Since product support is outdatted I am hoping somebody has an idea concerning my problem.

My problem is that, when I look at a measured data set in the navigator (I record time, spectra and thirds) the time and spectral data match each other, but the values of the thirds are much higher (at least 6, upto 34 dB!!!), see attachement. All data handling is carried out in LMS Test.Lab upto that point. The differences do not match any usual weighting curve.

http://files.engineering.com/getfil...7bb-90b8-06aeef7b5130&file=LMS_Screenshot.jpg

Has anyone encountered this before? Does anyone know how LMS records the thirds data?

Any help is greatly appreciated. Thanks.

Christian

 

[GregLocock]

That makes no sense to me. I don't, to be honest, understad what you are calling spectra.

Perhaps if you could dump the values out rather than graphs somebody might see what is going on.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[Shifas]

Hi Greg,

I'm not sure, if spectra is the right term (not a native speaker). What I refer to as spectra is the narrow frequency band with a resolution of 1,56 Hz. I can convert this high res. data into the thirds (1/3 octave) band, but I obtain different values than the software is giving me (see below). All data conversion is done using the LMS software itself, but no info is given on what is actually happening - if I do the calculation via excel starting from the time data I obtain the same values as found under for the spectra.

….. ….. "spectra" ….. thirds
Hz ….. dB ….. dB
40 ….. 34,20 ….. 47,14
50 ….. 31,95 ….. 46,79
63 ….. 26,39 ….. 43,54
80 ….. 25,11 ….. 43,32
100 ….. 16,56 ….. 39,94
125 ….. 15,51 ….. 38,74
160 ….. 29,42 ….. 39,16
200 ….. 19,29 ….. 40,53
250 ….. 27,97 ….. 62,76
315 ….. 24,50 ….. 41,14
400 ….. 31,52 ….. 43,90
500 ….. 37,93 ….. 56,16
630 ….. 47,02 ….. 58,81
800 ….. 50,67 ….. 67,63
1000 ….. 40,91 ….. 60,37
1250 ….. 44,68 ….. 54,81
1600 ….. 51,62 ….. 64,49
2000 ….. 44,39 ….. 53,37
2500 ….. 58,53 ….. 65,22
3150 ….. 57,57 ….. 68,68
4000 ….. 55,23 ….. 64,57
5000 ….. 60,36 ….. 70,64
6300 ….. 57,34 ….. 67,47
8000 ….. 59,22 ….. 69,72

 

[GregLocock]

You should have several thousand values for a narrowband spectrum, starting at 0 and then every 1.56 Hz. I think you have just extracted the narrowband result at the centrepoint of each 1/3 octave.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[SomptingGuy]

Try adding (in a log sense) the narrowband values that fall within a 1/3 octave band. E.g. for the 1000Hz 1/3 octave band, all those between 894Hz and 1118Hz.

Other things to consider:

Be aware of window correction factors (assuming your blocks are windowed). It's likely that the 1/3 octave bands are computed using "energy corrected" spectra, whereas the narrowband is possibly "amplitude corrected". There are good reasons for this (not going into them here), but it basically means you may end up looking for a consistent couple (or so) of missing dB when you're comparing your own summed values with the LMS reported 1/3 octave values. Check your signal processing parameters.

- Steve

 

[Shifas]

Hi,

first of all thank you for the replies.

The "energy" vs "amplitude correction" seems to be the issue.
Thanks for the advice