Karplus-Strong Synthesis (done the hard way) in Zebra

[Sound Author]

Before I get started, I just want to clarify that I know this can be done rather easily with the comb filters. So, why am I doing it the hard way? The quickest answer: because I can. But I've actually been experimenting with rhythmic delay lines that become musical notes, which is a pretty cool concept I think, so there's a method to my madness.

I've been wondering if there's a way to do this by modulating the delay ratios with KeyFollow, which would save a lot of time fine tuning key values in the mod mapper. I've managed to nail down some tunings intended for use with the "1sec" timing within the delay modules...



Now what I'd really like to figure out (if there is no way to do it with KeyFollow) is a way to successfully program a "pitch vs frequency" curve in the mod mapper (even though I'm not dealing with frequencies at all really) which would look something like this...



^ Only problem with that is the values in the mod mapper don't always play nice with the delay ratios, so high notes are often sharp or flat.

I've attached a zip file containing a preset tuned to a major 9 chord that offers only two octaves between C4 and C6. Anything higher is severely out of tune. You'll notice the key values in the mod mapper.

Any help on this topic is greatly appreciated

[Sound Author]

Also, here's another preset, which is just a simple string model with four octaves (C4-C8) that provides a better example of the curve in the mod mapper.

[Howard]

Interesting. Might be worth using two maps for finer tuning.

[Byrd]

At high frequencies you might still get sharp or flat notes, because most "normal" delays can only delay by an integer amount of samples.

The Bb6 just over the A6 at 1760 Hz, for example, would need a delay of ~23.65045... samples at 44.1 kHz, if I got the numbers right. The Zebra 2 - Delay can probably only delay by 23 or 24 samples?

[Sound Author]

At high frequencies you might still get sharp or flat notes, because most "normal" delays can only delay by an integer amount of samples.

The Bb6 just over the A6 at 1760 Hz, for example, would need a delay of ~23.65045... samples at 44.1 kHz, if I got the numbers right. The Zebra 2 - Delay can probably only delay by 23 or 24 samples?

I didn't know that was the reason high pitches are often out of tune, but yeah, that makes sense. Even if the delay buffer had a higher sampling rate, which would likely be much more CPU intensive, the value table in the mod mapper would need to have a higher resolution for long decimations. But man, if there was a way to somehow translate pitch to frequency, then frequency to milliseconds ala Salamander Anagram's string model in Reaktor ...

[Sound Author]

^ This is when Urs swoops in and says "Yeah, wouldn't it be cool if there was a simple stereo delay tuned to the played note? Oh wait. There is. Use the comb filter."

[Byrd]

Funny, that you posted a Reaktor - Screenshot.
I know of this limitation, because 10 - 15 years ago I tried to build a Karplus-Strong Algorithm in Reaktor and at that time, the Reaktor - Delay couldn't do fractional sample delay, so like you I wondered about detuned high notes, until I found a forum-post on the internet .

[Howard]

^ This is when Urs swoops in and says "Yeah, wouldn't it be cool if there was a simple stereo delay tuned to the played note? Oh wait. There is. Use the comb filter."

He wouldn't say that because he would also have read the very first sentence in the OP.

[Sound Author]

Regardless, I thought it would be funny if he said that. I have a weird sense of humor.

[Sound Author]

Interesting. Might be worth using two maps for finer tuning.

Thanks Howard, that did help! Still only four octaves, but much more in tune now