Here is the trick (but only ksmps=1)

<CsoundSynthesizer>
<CsInstruments>

sr     = 44100
ksmps  = 1
nchnls = 1
0dbfs  = 1

instr 1
aout rand 0.5
out  aout
endin

</CsInstruments>
<CsScore>
t0 [44100*60]
i1 0 12345       ; duration is 12345 samples
e
</CsScore>
</CsoundSynthesizer>


csound -d test.csd
...

sndfile-info $SFDIR/test.wav |grep ^Frames
Frames      : 12345

The grid for the score time is ksmps. The result of the previous
example with ksmps=64 is 12352 because

  64 * 192 = 12288 < 12345
  64 * 193 = 12352 = 12345 + 7

therefore the maximum error is (ksmps - 1) samples.

tito

On Sat, Jun 23, 2012 at 10:44:14AM -0400, Ralph Bluecoat wrote:
> Hello again,
> 
> I'm currently playing around with some low-level algorithms whose notes and
> durations are based on sample rate and control rate. Is there a way in
> Csound that I could write to an audio file and specify amount of samples I
> want to write to disk?
> 
> Thanks,
> 
> Paul