frequency weighting vs level weighting

[monotone]

Dear users,
I am looking for a correct way to apply frequency weighting to octave filtered data. I have seen cases of weighting being applied either before or after the octave analysis vi (thus using different weighting vi's, one for octave spectrum or the other for scaled signal waveform), and I always assumed the results would not differ. However, they do, and the wider the bandwith, the bigger differences are seen, particularly in the low frequency region. Could anybody comment on this? Thanks you.

[dsb@NI]

This behavior is described in the SVT User Manual in the Weighting Filters chapter. From the manual:

"When you apply acoustic weighting to a fractional-octave spectrum, a continuous frequency response function defines the attenuation of the weighting filter. When you mathematically weight a spectrum consisting of data from a fractional-octave spectrum, the correction values applied for the weighting are equal to the theoretical values at the center frequency of the fractional-octave band. Applying correction values creates an essentially rectangular filter that does not have a continuous response. The filter applies the same correction to all energy within each fractional-octave band. Applying the correction values to a signal containing a pure tone near one of the fractional-octave filter edges might introduce a measurement error. The error is usually negligible unless an A-weighting filter is used at frequencies below 500 Hz. At frequencies below 500 Hz, the slope of the A-weighting curve is steep. The following illustration shows how the slope of the A-weighting curve can contribute to potential measurement errors at low frequencies."

In the manual there is also a picture of the potential error when applying weighting to the constant proportional bandwidth spectrum.

If you need to comply with a specific sound level standard, you should use the time-domain weighting filters. Just be sure that you allow the filters enough time to settle before trying to make a measurement. This settling could be another reason that you see differences between the time- and frequency-domain weighting levels (especially at low frequency). Applying the weighting to the octave spectrum has the advantage that there is no additional settling time for the weighting filter.

Without more information, it is not immediately apparent which spectrum is more correct. And looking at the results that you provided, it does not appear that the total band power in either spectrum is substantially different as most of the energy is found at higher frequencies.

[monotone]

Thank you.