Frequency Analysis issue: pink noise

[Krakatoa]

Hi Gale,

White noise doesn't have much energy in the low frequency so there will be less perceived overall level change upon introducing a 5k HPF with the white compared to the pink--and this is what one would expect.

Quite simply, a piano keyboard is logarithmic, our hearing is logarithmic, musical harmony is logarithmic, audio is logarithmic, pink noise is logarithmic, and we visualise sound as logarithmic; it's vital for test noises and measurement devices to not blindly considering things in relation to the (absolute) Hz (which is meaningless) but to always put things in context by considering things in relation to the (relativity of the) octave. For example, it would be wrong of me to say that I can typically hear x Hz difference between two tones because that depends upon frequency. It is correct though to say that I can typically hear the difference of x % of an octave.

Best wishes,

Krakatoa

[Gale Andrews]

It doesn't seem you are going to persuade anyone.

Can you show us a graph online showing flat response in pink noise with amplitude against log frequency? Or an image in an audio tool that shows it?

If not, what would your Feature Request be - a graph to show power spectral density against frequency?


Gale

[steve]

I presume that Krakatoa means like this:


source: http://www.ni.com/white-paper/1709/en/

[Krakatoa]

Dear Gale,

It doesn't seem you are going to persuade anyone.

Can you show us a graph online showing flat response in pink noise with amplitude against log frequency? Or an image in an audio tool that shows it?

If not, what would your Feature Request be - a graph to show power spectral density against frequency?


Gale

If I'm not going to persuade anyone then it's a sad day and I'm out. Please accept my apologies for bothering the team with this feedback; I incorrectly assumed it would have been welcomed. By the way, you should easily find the answers to your questions in the thread above.

Very best wishes,

Krakatoa

[steve]

I presume that Krakatoa means like this:

Krakatoa, please say if this is NOT what you mean, and if that is the case, please give some reference / citation so that we know what you do mean.

[Gale Andrews]

Exactly.

Can you show us a graph online showing flat response in pink noise with amplitude against log frequency? Or an image in an audio tool that shows it?

Gale

[schmide]

I like this feature regardless of any miscommunication.

What would be the best solution?

Sum the bins. Basically integrating the graph. (different function)

Skew the graph such that it accentuates the higher frequencies. Basically curving space. (different mapping)

Both?

Although the second is not exactly what Krakatoa was asking for, it would, in my mind, maintain more information while providing the desired function.

I have often thought that better transformations are needed for the frequency analysis graph.

Andrew

[Krakatoa]

Gale, Steve, Offer the user what fraction of an octave resolution they want when in Log mode and then suitably sum the bins to achieve the width selected. The higher the frequency, the more bins will be summed in order to maintain a constant resolution per octave across the spectrum--ie. Logarithmic! Don't do this for Linear mode. Please I request kindly you reread all my posts if there are still questions before asking any more. Thanks and regards.

[Krakatoa]

And I Googled around and this student's report (http://graduateaudioseminar.wordpress.com/2010/03/) shows SMAART live with pink noise flat in 1/12 octave mode. Blue. Hope this helps illustrate. I realised that my excel doc didn't upload is it was too large.

[schmide]

Gale, Steve, Offer the user what fraction of an octave resolution they want when in Log mode and then suitably sum the bins to achieve the width selected. The higher the frequency, the more bins will be summed in order to maintain a constant resolution per octave across the spectrum--ie. Logarithmic! Don't do this for Linear mode. Please I request kindly you reread all my posts if there are still questions before asking any more. Thanks and regards.

I was also thinking more of a set radio buttons to sample, sum or average the samples. It seems to be currently interpolated point sampling the data which may hide some nuances of the spectrum.

Even in linear mode depending on the size of the window and the fft sample data may be skipped.