Robin Whittle wrote,

> Peter Kearton  wrote about what I think 
> might be called the "1/f Pink Noise" problem":
> 
>     How to create pink noise such that there is the same noise 
>     energy for each octave?
>

As I intended it my question was actually different to  Robin's 
interpretation. For psychoacoustical purposes I need to generate 
samples of 'white' bandpassed noise with various bandwidths, with a 
constant centre frequency, and constant noise power density across 
frequencies (both above and below the centre frequency, a 
constant 'average' density should be enough). Generating broadband 
white noise with 'rand', then filtering with 'butterbp' there is 
amplification of the overall output signal, increasing with 
bandwidth. So I have been multiplying the 'out' signal for each 
bandwidth by a fraction that brings the 'max output amp' (as 
indicated by CSound execution) for each bandwith to the same value. 
Thinking this would be enough I was frustrated once I analysed the 
generated samples with a frequency analyser (WAVE SE on PC). It seems 
with the method described above I obtain decreasing average power 
density with increasing bandwidth. Not having much knowledge about 
DSP my problem was whether this could be an artifact of the 
frequency analysis, and that I really do have the power distribution 
I need.  If no, then the additional question was can anyone help 
with a more appropriate solution ?!,

                        Pete Kearton, lost in DSP.