Windows XP automatically resamples audio at rates that the saoundcard
can't 'inherently' support, rather than report an error. Most
moderncards have an underlying clock based on 48kHz or 96kHz now, so
11025 etc rates are derived from this. However, the resampling is
rarely exact, and for 11025 usually ends up at 11100Hz which is 0.68%
high, or occasionally at 10800Hz. G3PLX reckons they are all
multiples of 300Hz - which looks to be a valid hypothesis, even if
tnot confirmed.
Normally, 12000 (or 48kHz really) based rates are the best as all
modern soundcards use this, but your older machine probably has a
44100Hz soundcard, in which case WinXP is trying to get 12000Hz and
failing to get close enough - the 0.6% you quote seems
non-coincidental!
It is a pity that WSPR does not have the sample rate correction
facility of WSJT - presumably Joe thought that by using 48kHz windows
would just throw up an error if it wasn't possible, but the annoying
windows driver just tries to do the best it can.
Andy
www.g4jnt.com
This email has been scanned for damaging side-effects by the health
and safety police
2009/12/13 Markus Vester <[email protected]>:
> Dear LF,
>
> I have recently installed WSPR 2.0 on one older XP machine and found that
> decodes were erratic, and the frequency readout was about 10 Hz high. The
> reason seems to be that the input samplerate is about 0.6 % low. The output
> samplerate is however correct.
>
> The strange thing is that this samplerate error was not present with the
> former version 1.12. I now have the two versions running simultaneously side
> by side, on the same machine with the same soundcard. WSPR 1.12 is producing
> perfect decodes on the correct frequency, whereas 2.0 reproducibly shows the
> described effects. It looks like the new version somehow accesses the
> soundcard differently.
>
> Has anybody else experienced this?
>
> Best 73,
> Markus, DF6NM
>
|