[sdiy] Prophet VS a phase accumulator design

Scott Nordlund gsn10 at hotmail.com
Mon Feb 7 01:38:03 CET 2011


> Keith Barr said once indirectly anything above 15Khz are
> waste in reverberation.

I have some theories about this... First, of course an 
actual acoustic space doesn't have much energy at 15+ 
kHz. But also once you increase the bandwidth of a "good"
reverb algorithm, it can start to sound crappy. What was
once a smooth decay now sounds grainy; I think the limited
bandwidth serves to smear impulses into something more 
continuous. And this also means that the brickwall anti
alias filters (typically high order elliptic or Chebyshev
filters, sometimes at a quite low frequency like 7 kHz)
play a part too. I generally think of such things as 
sounding "unmusical" due to lots of ringing, phase 
smearing, etc. but on reverb it may contribute something
positive to the sound. Often reverbs may increase 
echo density by chaining allpass comb filters (average
delay time maybe 6 ms); this spreads out an impulse in
time to multiply echo density at the cost of some
coloration (despite flat frequency response, all
frequencies don't arrive at the same time). I'm wondering
if chained biquad allpass filters might be used as a sort
of "micro diffusion" for similar effect, independent of 
brickwall filtering. It's known also that plate and 
spring reverbs have very high dispersion, resulting in 
a characteristic chirp-like response. This can be 
be emulated by embedding high order allpass filters in
a reverb structure. But aside from the group delay 
changing the modal decay characteristic, I haven't 
found this to hugely impact the sound past the initial
part of the reverb tail (assuming the reverb topology
increases echo density over time). So maybe if they were
chained at the input rather than distributed throughout
the  structure, they could have a more dramatic and 
controllable effect...

Haven't tried it yet, but it seems plausible.

> I thought you just said you had no idea about the Fairlight. ;)
> The DDS1 aren't completely solved yet, nor DPX either.

Well, if it's got a microprocessor per board, that says 
a lot.
 		 	   		  


More information about the Synth-diy mailing list