I’m back on wifi, so thank you! @sezare56
1 Like
Messing around with quasi Karplus Strong. Using a bugbrand weevil to ping thru machines with comb filters on them. Tweak and repeat. The unstable nature of the weevil makes for some really nice textures, which cant be done with a white noise source.
The video is a couple of years old, and takes a while to get going, I didnt edit it.
9 Likes
It’s also fun to use live input with a really fast, short envelope in the amp page gating it as your impulse for K-S.
Yep, that’s what I was doing in the video. Live weevil signal into Short plucky amp envelopes.
Repeat across 6 or 7 tracks, reverb on the master. Sort of a poly KS patch… kinda
Keep meaning to revisit the idea and flesh it out into a proper composition.
1 Like
What are differences between comb filters and KS?
@microtribe. I made something similar with a guitar, wanting to emulate a sitar. 6 x 2 fixed comb filters iirc, half mix.
1 Like
As I understand it, KS is a sound running through a filtered delay, where delay time matches musical pitch.
A comb filter is a filter with a delayed version of itself feedback through again, this creates the comb notches.
here’s a video i did with some octatrack experimentation
it’s mostly me making use of resampling and FX. i don’t even go past 16 step sequences. i go in with the intent of designing “one-shot” psychedelic sounds.
9 Likes
So cool!!
interesting. will watch it later 
1 Like
Your saving routine looks slick. Could you explain a bit about it?
How do you get your samples into that folder you have already made?
Yeah, for the most part K-S is a feedback comb filter with a lowpass filter in the feedback loop after the delay and the delay, more or less. Pretty similar. I don’t know where the lowpass filter is in the OT’s comb filter, if it’s in the feedback loop after the delay then you actually CAN do full K-S synthesis with the comb filter. I also don’t know off the top of my head if it’s a feedback or feed forward comb filter.
The main advantage of building a K-S algorithm from scratch the way I described is that it exposes everything to direct control, but that’s also the main disadvantage. They’re both useful.
Since phaser, flanger and chorus are all actually allpass filters with slightly different topologies it doesn’t make sense to worry too much about it. Karplus-Strong is basically an application of filtered-feedback comb filtering. A phaser is technically the dry signal passed through an allpass filter and then blended with the dry signal (but an allpass filter is essentially a comb filter with only the delayed signal so no comb filtering happens, so…). If you want to be really pedantic about it, reverb is usually just a bunch of allpass filters (and filtered-feedback comb filtering is a simple mathematical model of a wave moving in one dimension in an enclosed space so you could argue that Karplus-Strong synthesis is a type of reverb). They’re all variations on the same basic concept, and the only thing that really matters for musical applications is that they all sound a bit different so it’s nice to try out all of the different variants you can.
I’m not an engineer, that’s all broad-strokes rambling, but that’s also kind of the point - for our purposes the differences aren’t really clear cut and it just comes down to trying different things and seeing how they sound.
1 Like
I think it’s a setting on a per project basis. When you make a project, you can tell it to save samples in that folder or to save samples in the regular AUDIO folder.
As for how I discovered that sort of workflow, you can go here to my topic where I worked out the kinks and had a couple revelations thanks to other users here.
1 Like
My favorite teacher in community college (who now works as a composer and sound designer for a video game company) once told my class that all effects are amplitude and/or delay. I still think this concept today and your post seems to fit right in.
3 Likes
Mostly.
Except convolution, that’s multiplying samples. Probably some other exceptions but I’m not thinking of them. Filters are kind of a type of delay.
Ironically I learned about this stuff from one of the professors at college who I found a bit annoying but was undeniably good at what he did, and USED to moonlight as a sound designer for video games (he did Cruisin USA back in the 90s apparently). So kind of the opposite of yours.
2 Likes
There are two settings, you can choose if OT saves samples into audio pool or project folder.
Project Menu -> Personalize -> Save Samples to
Nice video @GirTheRobot !
You mention in the video that you like to play random samples through a already existing track with fx, p-locks and stuff. That’s such a great source for happy accidents. 
Are you filming on a green desk or how do you get that effect btw?
1 Like
Karplus-Strong uses the same tools but has a different conceptual framework, since it was articulated as plucked string or drum modelling. Basically they use noise exciters to model picks and strikes.
Comb filters were made to process any sound, but use very short delays to create constructive/destructive resonances which can be repurposed for K-S synthesis.
Isn’t multiplication ring modulation? I thought convolution was more like the integral of the product of the two functions after one is reversed and shifted. I suspect that’s how the plate and spring reverbs work on the OT but I might be wrong.
1 Like
Dude it was my first foray into a green screen! It’s actually literally an old green shirt I dont wear anymore. You can honestly do it with any color though, doesnt have to be green. You just tell the video software “take all of this color and replace it with this image” that’s why you can see little artifacts because it wasnt lit properly and also the “green” LEDs on the octa end up just being the background.
9 Likes
Lovely doggo! <3
2 Likes
From a mathematical point of view the convolution of two signals in the time domain is a multiplication in the frequency domain.
So - of course - there is much more involved then simply multiplying two signals sample by sample.
- calculate the spectra of both signals at a specific sample location (fourier transformation)
- multiplying the spectra
- transform the result back into the time domain
And of course these calculations needs to be done over and over again, because the spectra can change with each sample position.
2 Likes
Laplace is turning in his grave, after all the hard work he did to capture transient behavior in discrete systems and Fourier gets all the credit. But regardless of whether you are using a Fourier or Laplace transform function, you are using calculus. When you get right down to it, calculus is being used so that you can go back to using add and multiply but you can’t omit the hard part.
I think it’s somewhat analogous to the way that we intuitively use semitones, going up and down semitones like they were linear. But they are based on a exponential conversion that is non intuitive to most people when explained in math terms. Simple to understand, but more difficult to articulate.
2 Likes