Maybe a starting point would be to look at actual samples, and try to deconstruct them with sonogram producing software. I suspect that using low frequency 1/F noise to sweep a filter would be part of it. Paul perry Melbourne Australia