Talk:Karplus–Strong string synthesis

I think the basic Karplus-Strong algorithm actually initializes all the taps in the delay lines to random values, though modifications of the algorithm do use the derived impulse responses of real instruments (which is a different thing again from an 'impulse' -- a physical impossibility in a real instrument). The diagram is a bit deceptive, too, since the Karplus-Strong algorithm (and almost all physical modeling string synthesizers) use only low pass filters (typically first order) in the feedback loop. I imagine that the filter pictured is indeed a low-pass filter, but it might also be confused with a bandpass filter. -- Tlotoxl 11:54, 17 Aug 2003 (UTC)

Now that I realize that what I originally called an "impulse" should have been an "excitation", I'd like to point out that the basic algorithm is identical to that of echo, which had been used with excitations other than white/pink/brown noise long before the invention of digital delays, as some recording artists ran a performance through a tape delay line to create echo. When I first experimented with KS, I used Cool Edit's "echo" effect, and it was an easy step to trying to echo other waveforms such as a square wave, a saw wave, or a sine sweep. It appears that some people classify everything before Julius O. Smith III's patent as "KS" and everything afterward as "waveguide"; I just hope the echo effect in audio editors such as Audacity doesn't infringe Stanford's waveguide patents. --Damian Yerrick 04:31, 11 Sep 2004 (UTC)

It's a murky area, I think. The KS algorithm (and all LTI DWGs) are really just comb filters (as are echoes), after all, with relatively short delay lines. Comb filters are widely used in most synthesizers these days - with some synthesizers like Absynth relying heavily on the comb filters for many of its presets - but I can't imagine that these synthesizers have paid (or should have to pay) Julius Smith III/Stanford for DWG patents, even though the end result can be very similar. Implement dual delay lines, though (one for +ive going waves and one for -ive going waves) with some non-linearity (as in a bowed string or a woodwind) and I suspect it would be hard to deny patent infringement, though. --Oarih 09:32, 1 Feb 2005 (UTC)

Dual delay lines is the same thing as 2 taps off a single delay line, which in turn is the same thing as a delay line feeding a 2-tap FIR filter (with a large gap between the taps). If you move the filter out in front of the delay line, then the filter becomes part of the excitation and you get commuted synthesis, as mentioned in the digital waveguide synthesis article. So using this kind of comb with anything but white noise as the excitation might be an infringement, and that's what worries me. --Damian Yerrick 01:34, 28 October 2005 (UTC)

I think this article could probably be merged with digital waveguide synthesis. Both are pretty much the same synthesis method, and furthermore Karplus-Strong is just a simpler version anyways, and as such isn't used anymore because it's easy to make better sounding models with the processing power now available. —Preceding unsigned comment added by 24.203.35.195 (talk • contribs)

I disagree; there's an advantage to keeping the two articles distinct. The Karplus-Strong page provides a particular perspective from the original Karplus-Strong and Jaffe-Smith papers. Merging it with the digital waveguide synthesis article obscures the origins and evolution of the technique.107.217.188.187 (talk) 00:44, 8 December 2013 (UTC)

Older methods such as the Karplus-Strong 1D comb are still used, especially in free software, because they are less likely to be subject to government-granted monopoly problems. --Damian Yerrick (☎) 00:54, 19 December 2006 (UTC)

The section titled "Refinements to the algorithm" mentions a patent indirectly, and it appears that same patent is listed in the References with a link. What is the wiki markup to place that link in the text where the patent can be accessed directly where mentioned? I assume that such a link would still end up appearing in the References automatically. On that note, the formatting of the References section is slightly confusing, because the RefList appears to be part of the 5th reference (in my browser). Is there is a simple formatting change that could make it clear there are two lists in References? BrianWilloughby (talk) 01:44, 10 August 2015 (UTC)