Spectral analysis - use pink or white noise?

[Circlotron]

Till just now I always thought it was correct to use pink noise for frequency response testing of speakers, rooms etc. To that end, I have an audio CD I made with pink noise on it and I just shove it in the player and start measuring things. But just tonight (it's 8:46pm Wednesday here as I type) I recorded the pink noise output of my DVD player using CoolEdit and did a spectral analysis and saw to my surprise that it goes downhill at -3db per octave. All these heaps of measurements I have done thinking it would be flat! So I did a spectral response on white noise and it is as flat as you like.

Is pink noise only used when you measure for example 1/3 octave blocks, but white noise if you have a continuous readout?

[SY]

You're close. Pink noise has equal energy per octave. White noise has equal energy per Hertz. For example, a pink spectrum will have the same energy integrated from 100 Hz to 200 Hz as it does from 200 Hz to 400 Hz. White noise will have twice as much energy from 200-400 as it has from 100-200.

[FLZapped]

Maybe this will help:

http://www.embedded.com/2000/0003/0003spectra.htm

-Bruce

[Circlotron]

Still puzzled. Just need a 1 word answer. If I use a spectrum analyser with a continuous line should I use pink or white? Methinks white.

[SY]

It depends on how the bins are set up. If the bins have an equal frequncy spread (i.e., each bin covers the same number of Hertz), use white. If the bins have an equal number of octaves (e.g., they're all third-octave), use pink.

[Pjotr]

A word of caution here!

Circlotron, if you use your test CD with white noise for testing speakers, be careful The energy in the highs is pretty high. If the CD is played too loud you will fry your tweeters.

Since the energy of pink noise is much lower at high frequencies, pink noise is much more preferred for testing speakers.

Cheers