Feedback Alters Freq Balance

[Perfect Human Interface]

This would have been correct... I think.

Feedback 3.PNG (45.71 KiB) Viewed 13588 times

[nix]

If you want to make an echo delay,
loop the output of the delay back to the in,
punch 11025 samples delay, and put a knob that goes up to .9
in the link that feeds back from the output to in.
ATM- u r not creating feedback.
Delays do not have feedback built in, u have to route it.
Cheers

[Perfect Human Interface]

If you want to make an echo delay...

No actually, I know how to do that. I was just testing the theory that feedback loops are delayed by 1 sample per cycle (and that this was the nature of the high-freq attenuation) by recreating the effect to some degree using delay modules. I guess at this point I'm just confusing people.


By the way, here's another question that may be important... If you (I mean you, not me; I've done it) create a feedback loop without decreasing the signal, does it spit out a ridiculous digital buzzing noise? This is a stupid question but I've been experiencing an extremely troublesome bug for a while now involving this very noise happening in certain schematics every time I add or remove a module or link (ie something that would cause the schematic to recompile). I'd really like to get to the bottom of what's causing it.

[KG_is_back]

The original model had feedback topology (delayed output goes to input), this uses feedforward topology (delayed input into input), they are totally different. Feedback is one-pole filter whether feedforward is all-zero filter, which basically FIR filter (or a very simple convolver). It is however often used in oversampling as a lower quality CPU-friendly alternative (I think is also in the oversampling toolkit).

one pole filter has an impulse response of exponential decay. If you what to recreate it using delay modules each delay has to be x % quieter than previous. like 1, 0.5 , 0.25 , 0.125 , 0.0625,...

[Perfect Human Interface]

So is this how all filter q/resonance works, by adjusting feedback level? If I put a 6dB lowpass filter into a feedback loop will that be the same thing?

[KG_is_back]

So is this how all filter q/resonance works, by adjusting feedback level?

Eh, yes and no... digital filters are all based on adding feedbacks and feedforwards to the signal. the number of feedbacks and feedforwards are usually equal and are called filter order. The coefficients for them are computed via Z-transform, which is very complicated mathematical method about the filters. To make the long story short, it lets you define the frequency (and phase) curve based on set of points. First order filter is usually defined by amplitude of 3 points in the spectra (DC=0hz cutoff frequency and nyquist). Second order filter has another parameter called Q(uality) or Rezonance or Slope. All of them define the steepness of the curve. Yes, empirically higher the steepness higher the feedback, but without Z-transform and Fourier transform, you have very little idea on what is actually going on.

[trogluddite]

From a more "musical" point of view, the "multi-feedback delay" schematic is a phaser or flanger effect - but without the LFO that normally makes them "swoosh" around the frequency spectrum.
The bizarre frequency changes that KG alludes to are what give those effects their signature "psychedelic" sound. Very useful to make a sound more "synthetic" sounding and draw some attention - but not so good for unobtrusively EQ'ing your lead vocals to sit in the mix!

[Perfect Human Interface]

From a more "musical" point of view, the "multi-feedback delay" schematic is a phaser or flanger effect - but without the LFO that normally makes them "swoosh" around the frequency spectrum.

Right, of course. Which is why I was initially baffled by the "single pole lowpass" and the drawn out mathematical explanation. To my more simplistic musical understanding, combining a delayed signal with the original yields phase cancellation, and feedback enhances that effect (though I realize as well that it will eventually lead to resonant sounds, which I can understand that webpage was endeavoring to explain mathematically). It just seemed needlessly complicated to me.

Yes, empirically higher the steepness higher the feedback, but without Z-transform and Fourier transform, you have very little idea on what is actually going on.

Can you predict what will happen if I put a lowpass filter (module) into a feedback loop? I was just experimenting with that and for the most part it seemed to just boost the signal, whatever was left of it below the filter cutoff. But then at one point when I was toying with the resonance and feedback level I got a rather shrill resonant sound and then the output went NaN, which is always nasty. So is that just a bad idea in general? I mean I imagine if it's a long delay feedback (i.e. "echo delay") it wouldn't be a problem, but it was something I was planning on trying out in my current project.

[KG_is_back]

Yes, empirically higher the steepness higher the feedback, but without Z-transform and Fourier transform, you have very little idea on what is actually going on.

Can you predict what will happen if I put a lowpass filter (module) into a feedback loop?[/quote]

No, not me, but someone who actually knows the math can . But it reminds me my older project of waveguide synthesis. Feedback loop (several miliseconds long) with a filter (usually low pass) within simulates the way a string oscillates. Lower frequencies are damped only a little which makes them decay slowly, but higher frequencies decay faster (due to low pass filter) and also are slightly detuned because they propagate through the string slower (the phase shift of the LPF).
Here is the topic from SM forum. I'm not sure if the links are still valid - it was from that period when site was broken and we couldn't upload files.

Anyway, the result of putting a lowpass filter into feedback will be sort of comb filter with slight detuned ringing probably. However make sure that the filter response nowhere exceeds 0dB (never boosts - only cuts). Otherwise you'll end up with self-sustaining loop which results into NaN or inf or other unfortunate numbers.

[Perfect Human Interface]

Hi all, quick question:
Since we've established that feedback loops inherently have at least 1 sample delay, will a feedback loop that contains a delay module always have 1 extra sample of delay added or will it only be 1 sample of delay if there's no delay modules present? I just need to know so I can compensate for it if need be and I couldn't readily think of any way to test this.

Thanks.