sk de en
Main page > Hobby Projects and Research > Sound Modelling - Karplus-Strong Algorithm

Sound Modelling - Karplus-Strong Algorithm

© Dipl.-Ing. Radovan Konečný - radkon, 2011

About the project

After past attempts with the Karplus-Strong algorithm, I decided to remake the project, and try to create the sound I have in mind for a long time. So far, I have not been able to make this sound with other algorithms.

This is a classic Karplus-Strong algorithm excited with filtered and suitably modified noise. The resulting sound is, however, composed of a certain amount (e.g. 32) of such generators that are suitably detuned and have the suitable volume. In addition, they are alternately distributed between the right and left channels, creating a pleasant stereo effect.

From the point of view of frequency spectrum, this is a signal where each harmonic component of the tone has a certain sideband. Similar sidebands are created in instruments where there are more strings that are very little detuned (e.g. piano), or in the case of several musicians whose played tones are not exactly tuned (e.g. string ensemble). Of course, these detunings are small (units of cents).

Since the classic Karplus-Strong algorithm does not allow fine-tuning, the individual Karplus-Strong generators are additionally detuned by resampling. Without resampling, they would produce the same frequency. In the same way, the generators were "tuned", because at higher tones where the delay line is short, the correct pitch cannot be easily obtained.

Due to the complexity of the calculation, the sound cannot be created in real time.

Sound samples

(small audio defects are due to mp3 imperfection)

- 4 tones - 1 generator

- 4 tones - 16 generators, detune spacing of 0.8 cent

- 4 tones - 32 generators, detune spacing of 0.8 cent

- 4 tones - 64 generators, detune spacing of 0.6 cent

- music - 32 generators, detune spacing of 0.8 cent

There are only slight differences in the "robustness" of the sound, but the difference between 1 and 16 generators is significant.

Use in practice

This way all the tones have been created and a sound set is created for my Sampler.

Copyright © 2009-2018 Dipl.-Ing. Radovan Konečný - radkon. All rights reserved.