Csound processes ksmps samples every k-rate cycle. If you wish to have
very small delays lines you need to decrease ksmps. For short delay
lines just leave ksmps at 1, or better still, wrap the code in a UDO
and set the ksmps value locally.

On 18 February 2014 19:03, Forrest Cahoon  wrote:
> Hi, all!
>
> I have been working with what should be a very simple Karplus-Strong string,
> and I'm finding that the frequency I'm varies radically with the ksmps value
> I use.
>
> Here's my csd:
>
> 
>
>
> 
>
> sr=48000
>
> ksmps=1
>
> nchnls=2
>
>
> instr 1
>
> iDur = p3
>
> iAmpl = ampdbfs(p4)
>
> iFreq = p5
>
> iCycle = 1/iFreq
>
> aSignal = noise(iAmpl, 0) * linseg(1, 2*iCycle, 1, 0, 0)
>
> aDelayed = aSignal
>
> aFiltered init 0
>
> aDelayed delay aFiltered + aDelayed, iCycle
>
> aFiltered tone aDelayed, sr/2
>
> aOut = aFiltered * linseg(0, 2*iCycle, 0, 0, 1)
>
> outs aOut, aOut
>
> endin
>
>
> 
>
>
> 
>
> i1 0 20 -10 240
>
> 
>
>
> 
>
>
> and here is the frequency I get, measured with Sonic Visualiser, for various
> values of ksmps:
>
>
> ksmps observed frequency
> 192 122
> 96 162
> 48 193
> 24 214
> 12 226
> 6 233
> 3 236
> 1 239
>
>
> It looks like the delay is a integer multiple of k-cycles? I wasn't
> expecting such radical changes in my output just by changing ksmps. This
> should at least be documented, if not fixed somehow to give more accurate
> delay rates.
>
>
> On a related note, I wasn't able to find the source for the delay opcode --
> I have been grepping for the opcode name, but "delay" turns up zillions of
> lines. How do I find it?
>
>
> Forrest
>
>