Different looking response charts when 1/6 octave bar chart and full res

[Peter Zupančič]

This might not be exactly a problem with the application, more a cry for help in understanding the measurements.

I am trying to measure my speakers' frequency response as precisely as possible.

I make my measurements using white noise and pink noise, and 1/3 Octave chart and full resolution.

Why do measurements look so much different when using one or the other display method? When using 1/6 Octave and pink white noise, the response looks flat, and when using white noise and full res, the response also looks flat. However, pink noise and full res produce falling response, and white+1/6octave a rising response. Which measurements is "correct" and why?

[Julian Bunn]

Hi Peter,

I'm not sure I can do full justice to this question, as it's quite far-ranging.

The definition of white noise is that it is noise that occurs at all frequencies with equal probability. So, if you measure white noise in Full Res mode in AudioTool, the spectrum should be flat - the noise doesn't prefer any particular frequency.

Pink Noise is like white noise except the probability that the noise occurs at a given frequency f goes as 1/f. This means that higher frequencies are less likely to appear. If you measure pink noise in Full Res mode in AudioTool, then the level of the spectrum should fall as the frequency increases.

The octave modes (as opposed to Full Res) are sums of energy across frequency bands. For example, in 1/1 octave mode, the 8000Hz bin shows a level that comes from summing up all the energy between the 4000Hz bin below, and the 16000Hz bin above. So the higher the bin, the greater the range of frequencies it contains. If you look at pink noise in 1/1 octave mode, then the level of the spectrum should be flat - the 1/f fall is compensated by summing more frequencies as you move upwards in frequency. 

White noise in 1/1 octave mode should show a spectrum that rises with frequency.

I think the reason we use pink noise to make acoustic measurements is that our ear hears pink noise as "flat" - it doesn't sound accentuated at any particular frequency, as opposed to white noise, which sounds "hissy".

But others more experienced than I am may have more to add.

Julian

--
You received this message because you are subscribed to the Google Groups "AudioTool Discussion Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to audiotool-discussion-group+unsub...@googlegroups.com.
To post to this group, send email to audiotool-discussion-group@googlegroups.com.
Visit this group at https://groups.google.com/group/audiotool-discussion-group.
For more options, visit https://groups.google.com/d/optout.

[colan...@gmail.com]

Hi, I've also been confused by this difference between the octave and full res modes.  I think what's confusing is that, in full res mode with logarithmic frequency, the "area" under the curve doesn't reflect total power (while it does in the octave modes).  The opposite is true with linear frequency -- in this case, it is "area" under the curve of the octave modes that is misleading.  (I say "area" in quotes because what I really mean is the horizontal integral of 10 to the power of the vertical axis, since it is in units of dB.  But I digress.)

Perhaps a useful option would be to scale the displayed levels by frequency, say about 1 kHz, such that the area under the curves reflect total power.  This means, in full res logarithmic-freq mode, scaling e.g. 2 kHz +3 dB, 500 Hz -3 dB, etc.  And in octave linear-freq mode, scaling e.g. 2 kHz -3 dB, 500 Hz +3 dB, etc.  So, while in logarithmic-freq mode, with this option enabled, pink noise would create a "flat" graph in both octave *and* full res modes, rather than just octave mode.

This would make full res mode much more useful to me when equalizing pink noise.  As-is it's hard for me to visualize the -3 dB/octave slope.  Working in 1/6 octave mode makes equalization easier (since pink noise appears "flat"), but makes it harder to spot resonances and nulls.

To post to this group, send email to audiotool-dis...@googlegroups.com.