comb filter sound - How can I get rid of the annoying background (carrier) noise?

How to make that sound...
User avatar
KVRist
247 posts since 4 Dec, 2019

Post Mon Apr 26, 2021 12:57 am

This explains a general problem in sound design.
Any helpful contributions to solve this problem would be appreciated!

I'd like to isolate a certain buzzy/screechy sound from its background "carrier" noise:

https://vimeo.com/541739335
Last edited by juno987654321 on Mon Apr 26, 2021 8:58 am, edited 1 time in total.
Nothing can be better than liberty and free time to spend as you choose. Let's give those modern slaveholders the bum's rush!

KVRist
37 posts since 1 Jun, 2019

Post Mon Apr 26, 2021 1:09 am

Turn up the resonance on the comb filter

User avatar
KVRist

Topic Starter

247 posts since 4 Dec, 2019

Post Mon Apr 26, 2021 1:16 am

I have tried this but this leads to a different sound and the background noise stays all the same.
Nothing can be better than liberty and free time to spend as you choose. Let's give those modern slaveholders the bum's rush!

User avatar
9b0
KVRist
344 posts since 23 Sep, 2003 from Hungary

Post Mon Apr 26, 2021 8:03 am

I think, what you want is not something, you'll be able to get this way.

A comb filter is a delay with feedback. The feedback value is called resonance in this case, and the delay time is called cutoff (frequency is the reciprocal of time: you either measure the length of a revolution (time) or how many revolutions you have during a second (frequency)). What you call noise is not noise at all: it's just that you stop moving the cutoff, and then you're getting a constant delay time inside your comb. It's just like having many oscillators, but phase-shifted and with different amplitudes. You won't be able to cancel out the sound source by the oscillators only.

What could achieve an effect you'd like to would be to subtract the output of the comb filter delayed by the pitch of the oscillator from the comb filter itself. This is not a common thing to do, and I'm not sure if this is what you're after, but this somewhat does the trick. I doubt you're able to get this in Serum, or any other non-modular environment.

https://www.dropbox.com/s/mlhf4a87s5t9s ... e.mp4?dl=0
Alpha Forever Modular
Web // Youtube // Facebook //
Instagram

KVRian
1049 posts since 3 Nov, 2010

Post Mon Apr 26, 2021 9:42 am

9b0 wrote:
Mon Apr 26, 2021 8:03 am
I think, what you want is not something, you'll be able to get this way.

A comb filter is a delay with feedback. The feedback value is called resonance in this case, and the delay time is called cutoff (frequency is the reciprocal of time: you either measure the length of a revolution (time) or how many revolutions you have during a second (frequency)). What you call noise is not noise at all: it's just that you stop moving the cutoff, and then you're getting a constant delay time inside your comb. It's just like having many oscillators, but phase-shifted and with different amplitudes. You won't be able to cancel out the sound source by the oscillators only.

What could achieve an effect you'd like to would be to subtract the output of the comb filter delayed by the pitch of the oscillator from the comb filter itself. This is not a common thing to do, and I'm not sure if this is what you're after, but this somewhat does the trick. I doubt you're able to get this in Serum, or any other non-modular environment.

https://www.dropbox.com/s/mlhf4a87s5t9s ... e.mp4?dl=0
Very well said!

I watched the video and was trying to figure out in my head how to explain it because my thoughts on it weren't even clear to myself haha!

Your explanation is perfect even though I'm still not sure if I even fully understand it.

That video demo is AWESOME! AGAIN!
CHOOSX Remakes on my Youtube Channel

User avatar
KVRist

Topic Starter

247 posts since 4 Dec, 2019

Post Mon Apr 26, 2021 9:59 am

Hey, thanks for your reply!
Yes, I know, I've called the sound which I would like to subtract "noise" or "carrier sound" in a rather metaphorical sense. This is because it is something that should only serve as a starting point for the sound I really want to get, but to the sound itself it would just be disturbing if it were to stay there in the end. So I meant "noise" in the sense of something in the background that's rather disturbing.
And by that of course I refer to the saw sound with the comb filter on it as the comb filter does not move. That's right! It's an annoying saw sound across the whole frequency spectrum that I'd like to get rid of so that only the
sound from the MOVEMENTS of the comb filter would remain.

As we could see in the video I actually managed to reduce this sound already which was a first big step and a success already, but the question is how I could reduce it further so as to possibly get rid of it completely. I now need some good ideas for the "engineering". :D

You have again created something that sounds very much like what I'm after and it's no wonder to me any more that with modular synths anything will be possible: in your demonstration you move the cutoff position and it's only then when there's an output signal - whereas if the filter does not move there is no output signal. Great! This is exactly what I'm after.

At some point in time I will almost certainly advance to modular synths as well but as a beginner I would like to try things out with subtractive synths first and, as it were, do first things first. So my aim now is to find out how to do this on a subtractive synth plugin like Serum or Vital.

But then how can this be done in principle? Maybe if you can explain the principles of it there MIGHT still be a way of FURTHER APPROACHING the sound on a subtractive synth with its limited possibilities as well, even though it may be a bit uncommon/awkward... Last time we managed to approach the "aaah sound" in Vital as well and though it was only an approach that was already a good success I think. So maybe there will again be some way to take this at least a little further than I could reach so far - knowing that our tools (common subtractive synth plugins) will always be limited in what they can achieve... :tu:
Nothing can be better than liberty and free time to spend as you choose. Let's give those modern slaveholders the bum's rush!

User avatar
KVRist

Topic Starter

247 posts since 4 Dec, 2019

Post Mon Apr 26, 2021 10:10 am

I understand the comb filter as a kind of combination of many very narrow bandpass filters that take out very narrow bands of frequencies, then output them as amplified and multiply delayed signals again (depending on the amount of feedback you choose). Still it would theoretically and thus also practically be possible to kind of "silence" the incoming dry signal and turn it into a dummy signal, so to speak, and make only the processed signal come out whenever the cutoff position (which in this case determines the center frequency for the "anchor" of the comb filter) moves. That's why I believe it will be possible. We just need the right ideas and a forum is always a good place where people can come together and share ideas. It's just an attempt, though, because I don't want to do this all on my own, you know...
Nothing can be better than liberty and free time to spend as you choose. Let's give those modern slaveholders the bum's rush!

User avatar
9b0
KVRist
344 posts since 23 Sep, 2003 from Hungary

Post Mon Apr 26, 2021 12:19 pm

juno987654321 wrote:
Mon Apr 26, 2021 10:10 am
I understand the comb filter as a kind of combination of many very narrow bandpass filters that take out very narrow bands of frequencies, then output them as amplified and multiply delayed signals again (depending on the amount of feedback you choose).
This is completely wrong, unfortunately. A comb filter is not a filter in the sense a lowpass, or bandpass filter is. A comb filter is a delay effect. If we talk about comb filters, we mostly mean 'feedback comb filters' that looks like this:

Image

This is the diagram of a feedback comb filter. 'z-M' is the delay time, and 'aM' is the feedback value. Comb filters were used initially in reverbs (there's also a thing called feedforward comb filter, which is nothing more than the dry signal mixed to a delayed signal).

If you send a single 1 sample long peak (the digital representation of a DIRAC signal) into a comb filter and check it in a spectrum analyzer, you can see its frequency response. In the example below, you can see a comb filter set to 500Hz (2 milliseconds) with a feedback value of 0.5. The frequency response shows the typical comb peaks starting from 500Hz.

Image
juno987654321 wrote:
Mon Apr 26, 2021 10:10 am
Still it would theoretically and thus also practically be possible to kind of "silence" the incoming dry signal and turn it into a dummy signal, so to speak, and make only the processed signal come out whenever the cutoff position (which in this case determines the center frequency for the "anchor" of the comb filter) moves.
As you can see, this is not possible the way you believe it would work.

My solution was simply based on the idea that all you want to hear is the movement of the comb filter. If we want to catch the movement, we need to calculate the difference of something. In this case, the thing we want to eliminate has the frequency of the oscillator before it. If the comb filter is not being modulated, it also has the fundamental frequency of the oscillator, just with several copies of it phase shifted. So I thought, that by delaying the comb filter by the frequency (as you know already, frequency also means time duration) of the oscillator, and subtracting the result from the comb-filters output I'll be able to obtain the movement of the comb filter. This was the theory, and it worked. For this, you need an extra delay after the comb filter, that can be tuned to the oscillator, and you need to be able to invert the delayed signal and mix it to the comb filter.

EDIT:
if you are interested in how a filter works, I did a video on this topic back in 2019. This explains the basics of filtering:
https://youtu.be/_BVmkifC19M

And this newer one shows some additional things (in this video I'm using ZDF filters):
https://www.youtube.com/watch?v=6lVoa7RNxMc
Alpha Forever Modular
Web // Youtube // Facebook //
Instagram

User avatar
KVRist

Topic Starter

247 posts since 4 Dec, 2019

Post Mon Apr 26, 2021 3:53 pm

9b0 wrote:
Mon Apr 26, 2021 12:19 pm
This is completely wrong, unfortunately. A comb filter is not a filter in the sense a lowpass, or bandpass filter is. A comb filter is a delay effect.
I thought it was a set of bandpass filters because it can at least be emulated like that:
https://www.youtube.com/watch?v=Se_QHZzLhKo

As you can see this guy made a comb filter for Ableton based on bandpass/EQ filtering.
That confused me and I still have to learn a lot about filters. But now I understand that originally a comb filtering effect just results from short delays that lead to this same effect of alternating notches and peaks across the frequency spectrum. (I hope I'm right this way?)

The question to me remains how and if I could get the sound with the means I take as a given for the sake of the challenge. Being able to program how elements should exactly work together with IF - THEN commands would definately be the most efficient way of doing it and modular synths kind of stay very close to the programming language and possibilities. I know, this is the basic problem that limits down the possibilities I have if I stay on such a surface level.
9b0 wrote:
Mon Apr 26, 2021 12:19 pm
My solution was simply based on the idea that all you want to hear is the movement of the comb filter. If we want to catch the movement, we need to calculate the difference of something. In this case, the thing we want to eliminate has the frequency of the oscillator before it. If the comb filter is not being modulated, it also has the fundamental frequency of the oscillator, just with several copies of it phase shifted. So I thought, that by delaying the comb filter by the frequency (as you know already, frequency also means time duration) of the oscillator,
and subtracting the result from the comb-filters output I'll be able to obtain the movement of the comb filter. This was the theory, and it worked. For this, you need an extra delay after the comb filter, that can be tuned to the oscillator, and you need to be able to invert the delayed signal and mix it to the comb filter.
This is still difficult for me to understand it. What should I try?
1. "Delaying the comb filter by the frequency of the osc" - so e. g. if I played C5: 523 HZ,
so a delay of 1/523 sec.? Then I'd better tune the osc to 500 HZ and make a delay of 2
ms for a start?
What do you mean? The first 2 ms after the key was hit and after the first sound
appeared - this should be suppressed and not let through?
Maybe like that?
2. That way you would have subtracted the signal before the filter movement?

But wait: the incoming signal is also sustained. It's not just one burst and then gone. It is there all the time. So, like, every sample would have to be cut off at the first 2ms?
Also: if there is no filter movement for some time there should just be complete silence. So you'd also have to program the output to only be let through if the filter is moving? No?...

Hmmnnn... How does this work in principle? The first step must be to understand how this could work in principle.

I'll check your videos and try to find out something more in the meantime.
I wouldn't be surprised if I got everything wrong again but then you could correct me. :D

Edit:
I've long got THE SOUND I wanted to get! If it was just for the sound: it's easy. You do it in post editing. You sample the thing and then you apply the common noise reduction tools. They come in all kinds of shapes, completely free. But that's not the challenge. The challenge is to do it on a synth without sampling and without any "tricks" - just like with any sound!
Nothing can be better than liberty and free time to spend as you choose. Let's give those modern slaveholders the bum's rush!

User avatar
9b0
KVRist
344 posts since 23 Sep, 2003 from Hungary

Post Mon Apr 26, 2021 4:30 pm

juno987654321 wrote:
Mon Apr 26, 2021 3:53 pm
I thought it was a set of bandpass filters because it can at least be emulated like that:
https://www.youtube.com/watch?v=Se_QHZzLhKo
OMG! That's just simply wrong. :/ This method is kind of OK for formants, but a comb-filter is a delay effect, and cannot be emulated using a bunch of bandpasses. The phase response of a comb is completely different, and it literally delays a signal with milliseconds, not just with one sample.
juno987654321 wrote:
Mon Apr 26, 2021 3:53 pm
That confused me and I still have to learn a lot about filters. But now I understand that originally a comb filtering effect just results from short delays that lead to this same effect of alternating notches and peaks across the frequency spectrum. (I hope I'm right this way?)
No, it's different. It's the same as using a flanger, or chorus. In synthesis, comb filters are mostly used for Karplus-strong style sound design, where you send a short burst of signal into the comb filter, and let it emulate the exponential decay of a string, but it can also sound interesting on sustained notes.

Using TONS OF resonant bandpass filters you can get into the territory of nodal synthesis, but that's again something that's completely different, and is closer to additive synthesis than subtractive.
juno987654321 wrote:
Mon Apr 26, 2021 3:53 pm
This is still difficult for me to understand it. What should I try?
1. "Delaying the comb filter by the frequency of the osc" - so e. g. if I played C5: 523 HZ,
so a delay of 1/523 sec.? Then I'd better tune the osc to 500 HZ and make a delay of 2
ms for a start?
What do you mean? The first 2 ms after the key was hit and after the first sound
appeared - this should be suppressed and not let through?
Maybe like that?
2. That way you would have subtracted the signal before the filter movement?
Well, I really can't explain it more in-depth than like this:
https://www.dropbox.com/s/3gpa3b9jsrmcq ... d.mp4?dl=1
juno987654321 wrote:
Mon Apr 26, 2021 3:53 pm
But wait: the incoming signal is also sustained. It's not just one burst and then gone. It is there all the time. So, like, every sample would have to be cut off at the first 2ms?
Also: if there is no filter movement for some time there should just be complete silence. So you'd also have to program the output to only be let through if the filter is moving? No?...
I'm sorry, but I do not understand these questions. I hope, my video makes clear how I did the effect, but really: this is very far from a common use case. I'm pretty sure, it's not solvable in a simple subtractive synth.
Alpha Forever Modular
Web // Youtube // Facebook //
Instagram

User avatar
KVRist

Topic Starter

247 posts since 4 Dec, 2019

Post Mon Apr 26, 2021 5:25 pm

Thank you so much for these explanations in your video. It will take me some more time to think about everything and try things out.
You make me become more and more interested in your nodal compositions.

The other youtube video really confused me. Someone understood something completely wrong and he built a filter that he mislabled as a comb filter. Sorry about that but I just trusted this for the moment because it was a starting point for me, but now I see that it was just misinformation and I would have found out at some point. Now I already have a better conception. I need to think about all of this and maybe I'll come up with another creative idea and I'd let you know then. (?)

Until then I'd like to state a few intermediate results from all that:
- Maybe this sound could only (if at all) be made in more advanced subtractive synths like Zebra2:
synths that have a more advanced comb filter with dry/wet functions, etc. However, I think they do exist.
- You helped me understand much more about comb filters and also about why exactly modal synths can make
you become freer as an artist if you want to realize your own ideas and really program your own sounds.
Modular synthesis is something I am very open for. It's just that I'm not there yet!
- The sound itself could still be made via the simple subtractive synth methods in conjunction with some post
editing. You may have to apply some
post editing then and you definately get less control over your sounds because for every instance it would
then always be necessary to sample the sounds first and subtract the "noise" from your results afterwards. This
is not something we really want in sound design, though in principle the SOUND itself can still be made with
these simpler methods.
- IF the sound can really be made in Zebra2 (without sampling or post editing) or even in other synths which are
based on subtractive methods:
this to me remains an open question because I'd like to keep an open mind on that question and I'll have to
think about everything. This has been quite a lot of information for today.

In total thank you very much again for those very educational insights into the secrets of sound design.
It means yet another step forward and it has certainly been interesting.

So in your own words: happy noding then and keep up exploring those piano sounds of yours in all its details further!
Nothing can be better than liberty and free time to spend as you choose. Let's give those modern slaveholders the bum's rush!

KVRAF
3224 posts since 2 Jul, 2005

Post Tue May 04, 2021 1:14 pm

Can you play just the part of the sphongle song that has the sound you want? Just that part.
Don't F**K with Mr. Zero.

User avatar
KVRist

Topic Starter

247 posts since 4 Dec, 2019

Post Tue May 04, 2021 2:50 pm

Ah_Dziz wrote:
Tue May 04, 2021 1:14 pm
Can you play just the part of the sphongle song that has the sound you want? Just that part.
Look, I've created some concrete stuff in the meantime and I also (at last) figured out what you meant. You meant that we could use a stutter effect on a saw with bursts of sound, rather than keeping a saw wave sustained and I used an LFO as you recommended, right? This somehow worked in combination with pitch variation (also done with a second LFO) and so I could already approach the sounds I want a little bit better with your method of "impulse saw sounds" as I shall call it.

1. The one sound I wanted in THIS thread here is called "sad screechy sound further post edited on top" because I initially could only get this sound by using noise removal tools and then I sampled the sound as an MP3 file. I removed the noise from the sound but the sound itself is the same one as in the video without the background noise.

2. However, today I already created a patch in Serum that already comes close to creating the patch directly in Serum (without having to use noise removal tools i. e.). This one I called "dinosaur comb" because it sounds a bit like a dinosaur yell and it is based on the comb filter tricks that I learned here on kvraudio.

3. Then there is the other thread with the sounds from sphongle. That was something else again! These are the sounds as you described them from 28 seconds to 45 seconds.
I also created a patch for these sounds today but I'm not yet content with this because I don't know how to set the flanger (or would a phaser be better?) correctly and do the fine tuning and all that right. Note that I'm still a beginner but I'm making some progress already, slowly but surely, and it already sounds a bit like it if you play the notes with the right lenths and pauses on your keyboard (some notes are staccato I think)...
I called this Serum patch "experim comb screech Shpongle Nothing lasts style".

So I'll put all of this in the attachment now as something concrete to have a look at. If we just write one another everything stays way too abstract, so here are the concrete things that I've already made. I think this might work much better for communication. So if you think you could help to improve the patches and if I still haven't understood your suggestions well enough, you could let me know how exactly this could be done even better. I think those comb sounds sound really cool and it's something completely new to me. :wink:
You do not have the required permissions to view the files attached to this post.
Nothing can be better than liberty and free time to spend as you choose. Let's give those modern slaveholders the bum's rush!

Return to “Sound Design”