Reaktor - schroeder allpas

gentleclockdivider

This is a schroeder allpas taken from a max msp tutorial

For those unfamiliar with max , the tapin~tapeout~object is basically a delay , where tapin~ is equivalent to reaktor dely audio input , and tapeout~ reaktor's delay audio output.

In the max msp implementation you can see that the negative feedback coeeficient(red quare) is mutliplying the tapout ( reaktor delay out ) =feedback comb

THe positive feedforward comb = green

And this is the reaktor implementaion

BUt here is the adsr tutorial , which actually reverses the feedforward and feedback comb

where g=feedback coeef

Which would be like this

SO is the adsr tutorial wrong when it comes to schroeder allpas ??

gentleclockdivider

Hmm , but then this diagram , the negative coefficient '-g' is fter the adder , where the diagram in the first post it's reversed .

Z Gabr

And this is the reaktor implementaion

This is reminiscent of a diffuser delay circuit

gentleclockdivider

Well it's a schroeder allpas in series

colB

iirc, the 'Schroeder Allpass' that everyone uses - the one in your first pic - is a refinement of Schroeder's original (maybe by gardner? can't remember offhand).

EDIT:

Although this is not relevant to your initial post, sorry :)

ADSR tutorials are great, but they are not always 100% correct, or best practice. Go to academia if you want that!

gentleclockdivider

I build some schroeder reverb in reaktor and it's sounds ..hm well interesting

3 allpass in series going into 4 parallel combs

THere are some online papers about combfilters into all pass , which makes it confusing

I thought the original was ap into comb

gentleclockdivider

I am till confused which one is behaving as a schroeder allpas becasue the results are different depending where I put the explicit unit delay .

On the schroeder allpass it doesn't show but since it's digital it has to be somehwere .

First example shows without a manually inserted z-1 , reaktor places it in front of the delay module .

Seconde example , the z-1 is placed before the feedback

input audio =1 unit sample

colB

https://community.native-instruments.com/discussion/comment/142890#Comment_142890

You don't need a unit delay because there is already a delay line in there. Unit delay is only necessary where not having one would require a significant revision to the laws of physics.

This might be problematic in Primary (dunno?), but in Core it just works, assuming the delay has been correctly built with the solid parameter deactivated in the properties pane.

If you need a unit delay in Primary, make the main delay one sample shorter, and stick the unit delay immediately after it, then pretend the output of the unit delay is the output of the main delay line

gentleclockdivider

I am well aware that I don't need to put a unit delay in there .

I was stating that WHEN I do , the result is different becasue of the phase cancelation (feedforward-feedback gain coeff.) occurs depending on where the unit delay is inserted .

I also realised that the schroeder allpas is just a diffuse delay line ( already mentioned earlier in the thread by Z gabr )

It's a bit confusing jumping back and forth between reaktor and pure data .

reaktor diffuse delay = allpass in pure data ( delay line with feedback coeff , and feedforward coeff ) , and ofcourse reaktor diffuse delay is an allpass when delay=0

Now , depending on what the feedforward coeff is multiplying ( see my verry first post about the 2 possible schroeder allpass iplementations ) is equivalent to the negative or positive gain coeff.in primary diffuse delay

colB

https://community.native-instruments.com/discussion/comment/143283#Comment_143283

"I am well aware that I don't need to put a unit delay in there"

why are you using one then?

gentleclockdivider

https://community.native-instruments.com/discussion/comment/143698#Comment_143698

Because then you can decide WHERE to put it in the feedback loop when creating the schroeder allpas .
I have already explained that 4 posts back

There is a difference between the upper ( automaticially inserted by reaktor beefore the delay ) and lower structure (manually inserted after the delay )
Both receive an impulse file , yet the output is different because of the placement of the unit delay , unipolar versus bipolar output
And here's the file

Hereyouare.ens

colB

https://community.native-instruments.com/discussion/comment/143759#Comment_143759

The point I'm making is that in core you absolutely do not need one because there is not implicit one. This is due to there being a delay line.

I'm not sure about Primary, but if Primary does generate an implicit z (which would be quite strange, and should be considered a bug in the context), then the only place to put an explicit one is directly after the delay line.

errorsmith

you need to treat the unit delay as part of the delay. if the delay for the feedback path is different to the delay for the feedforward path it's not an allpass anymore.

colB

I see that pesky z in there. It really shouldn't be necessary, and wouldn't exist in core (you should really be using core for low level DSP like this IMO)… anyway, here is what I would consider to be a more correct version:

In your version, the delay time in the feedforward path is 22, but the delay time in the feedback path is 22+1/SR. In the lower edited version, the feedforward and feedback paths have the same delay time.

For a completely accurate response, You need to adjust the delay time by subtracting 1/SR to account for that unit delay.

Same thing in core. No need for a z^-1 unit delay. Core knows that the delay line means it will just work.

colB

https://community.native-instruments.com/discussion/comment/143884#Comment_143884

hehe, you beat me to it while I was busy trying to promote core

errorsmith

https://community.native-instruments.com/discussion/comment/143967#Comment_143967

yeah, core is the way to go here.