Reaktor - schroeder allpas

gentleclockdivider

THanks for the write up and explanation but I am still not sure .

The first examples you provided is just a feedback-feedforward build around an allpas .

Here , https://dsprelated.com/freebooks/pasp/Nested_Allpass_Filters.html

Otoh , this example of a nested allpas

is exactly like this in core , both allpasses (let's call them master and nested ) have a freq/delay parameter , in your or the dattorro's example only the nested one has a freq param

gentleclockdivider

About the delay taps .

IN reaktor we can't make a tap without having a delay line , 4 taps means 4 independent delay lines whcih may or may not go into a fifth one to control overal feedback .

I could be totally wrong about the following, but in hardware can't we just read from different addres points in the buffer ( taps ) from 1 delay ?

colB

https://community.native-instruments.com/discussion/comment/145800#Comment_145800

IN reaktor we can't make a tap without having a delay line , 4 taps means 4 independent delay lines whcih may or may not go into a fifth one to control overal feedback .

Maybe in Primary that is a restriction, but core has existed for quite some time. It's trivial to implement.

This includes the logic for fake zeroing the buffer. With that removed it can be much tighter code, but that was more work, this approach was just a cut and two pastes.

(you could probably remove DNC modules for taps that are not in the loop too)

.

gentleclockdivider

And another example of a nested allpass ,just like the one in my previous post

http://gdsp.hf.ntnu.no/lessons/6/33/

colB

https://community.native-instruments.com/discussion/comment/145799#Comment_145799

Hmmm, I think I might know roughly what's going on, but I'm no expert, so ideally, someone with more filter knowledge can chime in.

(take with pinch of salt ;)) normally a 1 pole all-pass would be built just using unit delays (those z^-1). With the coefficient (k in the diagram you posted) set to 0.5, the 90degree shift frequency would be at SR/4 or SR/2, not sure which.. :( ). You would change the frequency by changing the coefficient, biasing it more towards 0 or towards SR.

(you can verify this for yourself by setting up an all-pass with unit delays, and variable k, then subtracting its output from the direct sound, for a high pass with variable cutoff, or adding it to the inverted direct sound for a low pass... varying k will change the cutoff frequency)

With these Schroeder style diffuser all-passes, it seems that using longer delays is similar to reducing the effective sample rate... so even if the coefficient stays at 0.5, you can reduce the cutoff frequency in steps by increasing the delay. But the longer the delay is, the less like a normal phase shifting all pass filter it becomes... the individual delays become more and more granular as they tend towards becoming recognisable as individual delays rather than a homogenous diffusion of sound... all pass comb filter maybe?...

anyway, this what I _think_ is going on, so don't go quoting me :)...

Fortunately, you can still use them to build a nice reverb without 100% clarity on this point!

ANDREW221231

https://community.native-instruments.com/discussion/comment/145827#Comment_145827

all pass comb filter maybe?

i believe that's correct

all this talk of allpasses is really making me want to go hunt down something crazy i did with them one time

was building a drum synth and one of the sounds was insanely annoying. so ended up feeding it through some kind of cursed allpass effect i rigged up to turn the rinky dink rimshot percussion sound into terrible and very convincing screams of bloody murder

think it was some combination of a series of heavily modulated allpasses subtracted from the original signal?

anyway it completely turned the sound inside out in both pitch and timbre. was surprised to see such an effect was possible