It's not a denormal issue.labrat51 wrote:you'd have thought they'd have written a de-normaliser algorithm into it to prevent exactly that (remember digitalfishphones did one to stick on any plugin chain over vst to prevent cpu spikes on a pentium 4? i actually got it to work on cubase sx and kontakt, i was having horrendous trouble when the instrument wasn't playing any notes, spikes of like 35% until i slotted one of these in after. i think it basically cut off anything below a certain threshold)
a denormal routine is done in a number of ways, but one of the most common is to add a very small amount of noise at a very low level. This has the effect of eliminating numbers which fall between the bitdepth limit and true zero . EG : when a delay fades away below the limit for 32 bit audio and then, is it at zero or where is it ? who knows.
Sometimes this can become an large issue - so adding a low level noise definitively eliminates the uncertainty. Unfortunately - low level noise can have implications for the calculations of anything later in the chain, because they think that this is 'signal' and attempt to process it as such. So you have eliminated the nasty 32% CPU denormal spikes, but gain a small 1% for every device that follows. That's my understanding of common denormalisation practise anyway.
sorry, I took the thread off tangent there.