How pink is pink noise ?

When doing audio alignement of screening rooms, you end with level measurements, you know the typical 85dBC, slow,….

But depending on the noise source (Dolby DMU, CP750, Protools, S5 internal pink, minirator,…), you may get strange differences. My friends at MeyerSound did a very good report on pink noise levels for the SMPTE group on acoustics measurements.

I wanted to be able to quickly analyze myself a pink noise, not only level but also frequency response.

So back to the roots :

  • pink noise is a random noise that has constant energy in a frequency log scale, it's called a 1/f noise
  • there is no true pink noise : pink noise is generally white noise filtered at 3dB/oct (but there are other methods)
  • but no 3dB/oct filter does exist !

First, you have to create a mathematically "near perfect" white noise.

And you have to create a filter that comes close to 3dB/oct by combining first order filters (6dB/oct).

Second problem, SMPTE says that noise shoud be rms measured 20dB under max level. Is max level 0dBFs or the rms value of a sine wave at 0dBFs, afaik -3dBFs ? Should the alignement pink noise be at -20dBFs or at -23dBFs ? Dolby does not really answer : the DMU noise is near -21.3dBFs rms !

Two tools are used : 

  • audiogenerator is an old software of mine but now with calibrated fader level in rms values : you can generate and record various signals including pink noise, correlated or uncorrelated.
  • pinkalizer is a new soft to analyze and compare rms, crest factor and frequency responses of wav files, especially pink noises near -20dBFs ! The script is written for Gnu Octave, so you may check it and do your own improvements.

Pink infos can be found here : http://www.firstpr.com.au/dsp/pink-noise/

Some results :

 

 

 

 

 

I compared my measurement to Meyer's ones : very close in rms and crest factors.

We can see that most pink noises announced at -20dB are in fact at -23dBFs rms.

Dolby noises are very smooth in low frequencies (how do they produce the noise ?) but have a general trend of peaking at high frequencies. And Dolby did not decide between -20 or -23. They choose a value in between !