ilovefuzz.com

I understand feedback in a delay circuit is the output (the repeats) being fed back to the input.
The delay bit of the Moogerfooger Cluster Flux has positive and negative feedback.
Could someone please explain this to me circuit-wise? What's happening in that part of the circuit?

This is what the manual (http://moogmusic.com/sites/default/files/Clusterflux%20Manual%20Layout%20Web.pdf) says:

In Flange Mode, Positive Feedback emphasizes all the harmonics of
a fundamental frequency equal to the inverse of the Delay time – For
instance 10 msec Delay creates a comb filter with a fundamental frequency
of 100 Hz and emphasis at 200 Hz, 300 Hz, 400 Hz, 500 Hz, etc. Negative
Feedback shifts the frequency response spectrum down by one octave
which causes only odd harmonics to be emphasized. For instance, a 10
msec delay time with negative feedback creates a comb filter with a
fundamental frequency of 50 Hz and harmonic emphasis at 150 Hz, 250 Hz,
350 Hz, 450 Hz, etc

(thank you!)

'Positive' feedback: the input = the input + the output (delayed by t version of input)
'Negative' feedback: the input = the input - the output (delayed by t version of input)

if you think of sinusoids, being delayed then added to themselves, you will see that the same delay time has different effects at different frequencies. For example say a sinusoid with a period of 1 second and a delay time of .5 seconds. If you 'add' the delayed output to the input (positive feedback) you will actually attenuate the wave (adding 2 sinusoids that are totally out of phase). But if you subtract the delayed output from the input you will actually boost the result (essentially the addition of 2 in phase sinusoids). It may blow your mind but most typical frequency filtering IS, viewed froma certain stand point, just delays and summations.

Now keep the delay the same and shoot in a sinusoid with aperiod of .25 seconds. Youll see the result is not the same. IE the system booseted/attenuate different frequencies, differently. THe effect is cyclical over the band in this case which is why you get a comb filter.

maybe google comb filter. there are probably way way better and clearer explanations out there.

[youtube]http://www.youtube.com/watch?v=GJQKjcBC96c[/youtube]

[youtube]http://www.youtube.com/watch?v=LpJTh4tM_2E[/youtube]

even harmonics sounds good on everything. odd harmonics sound bad or dissonant. but all this is irrelevant if it is not tracking the fundamental frequency of the signal coming in. some people like the sound of a sweeping comb filter where all the bands move left to right together. as you can tell from the video, it is possible to get similar sounds to a ring mod. that is not the only sound though. some people have a tighter definition of even harmonics that is strictly octaves. so 100, 200, 400, 800 etc. one way to look at it is the smallest even harmonic divides the larger ones by a whole number. the odd harmonics are perfect fractions. for example

50/150 = 1/3
50/250 = 1/5
50/350 = 1/7

100/200 = 2
100/300 = 3
100/400 = 4

you just need to hear it. all the tech talk is almost worthless.

multi_s wrote:shoot in a sinusoid with aperiod of .25 seconds. Youll see the result is not the same. IE the system booseted/attenuate different frequencies, differently. THe effect is cyclical over the band in this case which is why you get a comb filter.

the affects on a sine are rather confusing to understand comb filters since all you will see is changes in amplitude. you will never see a change in frequency response when looking at pure sinewaves on a scope or FFT. there might be some phasing but thats just a real world side affect that does not relate to the pure math of a phase linear comb filter. the non-linear phase response is probably a big part of the sound of any analog comb filter or EQ.

the easier way to think of harmonics (evn or odd) is

the fundamental frequency is the first harmonic (ie number 1)

even harmonics are frequencies which are even multiples of the first harmonic ie 2,4,6,8

odd harmonics are odd multiple odd the first harmonic ie, 3,5,7,9

it is much more clear from this approach why one is called "even" and the other "odd"

the reason for disonance vs. melody is that the first few even harmonics turn out to be octaves and almost a perfect fifth which are generally pleasing intervals to the ear.

the first few odd harmonics do not have such fortunate characteristics and thus sound a bit less pleasing (subjective of course).

eatyourguitar wrote:

multi_s wrote:shoot in a sinusoid with aperiod of .25 seconds. Youll see the result is not the same. IE the system booseted/attenuate different frequencies, differently. THe effect is cyclical over the band in this case which is why you get a comb filter.

the affects on a sine are rather confusing to understand comb filters since all you will see is changes in amplitude. you will never see a change in frequency response when looking at pure sinewaves on a scope or FFT. there might be some phasing but thats just a real world side affect that does not relate to the pure math of a phase linear comb filter. the non-linear phase response is probably a big part of the sound of any analog comb filter or EQ.

?

multi_s wrote: IE the system booseted/attenuate different frequencies, differently.

?

Frequency response is by definition the magnitude (amplitude?) of the response (output) at different frequencies (input). Your contradicting yourself a bit. You WILL see results on pure sinewaves on a scope. You may also then want to think what the FFT shows and you will see that of course you will see it there too. The FFT is simply an alternative way of viewing the same information. If information was lost in the transform it probably would not be so damn popular.

in the video I posted they use noise to look at a comb filter. all you would see on a fft with a sine wave going through a comb filter is a peak at your sine frequency bouncing up and down. you would not be able to see the frequency response in the way that you see it when looking at noise through a comb filter on an fft.

I was totally wrong. I looked it up. I was thinking that it works just like a notch filter but it's not the same thing at all. wikipedia says that it is just a time delayed signal mixed with the realtime signal done digitally. it is interesting that a notch filter does not add harmonics to a sine wave but a digital comb filter does. I want to download some software and try it. whats better, mathlab or smaart?

[soundcloud]http://soundcloud.com/robpaper/comb-filter[/soundcloud]

the sine wave is at a constant frequency through the entire thing. I am modulating the delay time with a sine LFO. you can hear the harmonics being generated. in my software it has a switch for feedback or decay. this is like positive or negative feedback. the first half of the demo it is on feedback. the second half is set to decay. this is not how I imagined a filter would react to a sine wave. this is some good shit. I need to play with it some more.

EDIT: @ 1:34 the switch is in decay mode.

cool i will check this out when im not an a ffing linux machine with no plugins.

i think that was my original point of contention, that the only thing that creates the filter effect IS THE DELAY. that is all there is nothing else involved. The resonant frequencies are proportional to the length of delay and repeat every 2f. Negative Vs positive will change the "shape" of the comb either with its teeth pointing up (therefore resonance is at even harmonics of the inverse of teh delay period?) or with the teeth pointing down such that teh resonant frequencies are exactly half way between the teeth -> odd harmonics.

If you look up any book on DSP application s to synthesis and music it will definitely have a section on comb filters since their implementation is trivial with a digital delay line.

Also i think it should be feed forward not feed back however the words are sometimes used interchangeably.

I'm guessing you don't like ubuntu. I know lots of hardcore linux users that wont even touch it. if you have a fast machine though its good for flash and firefox.

im at school and the lab uses fedora. i dont mind it really, but since i don't have any privileges on the machine i can't install whatever i would need to enjoy myself. i actually have very little preference for OS. i still run XP at my house.

very interesting, didn't know about comb filters at all;
plus, the relation between miliseconds of delay and the frequencies "comb-filtered" now seems kind of obvious but I can't imagine thinking of that by myself…

I found this flanger circuit by craig anderton (at experimentalistanonymous.com) which has the option for positive and negative feedback:
http://www.experimentalistsanonymous.com/diy/Schematics/Flangers/Anderton%20flanger.pdf

I guess adding or substracting has to do with how you make the op-amp connections… doesn't it?

yes because opamps typically have inverting and non inverting inputs.

Subtracting some signal x(t) from another y(t) has the same as adding the inverted version instead. (: