You can generate accurate 8970Hz frequencies from a single PIC DDS provided you clock it with a decent source to start with.   The surplus market is awash with suitable oscillators as a reference, and the exact starting frequency does not even matter provided it it within the range for the PIC, so 5 - 20MHz will do.

 

See http://www.g4jnt.com/PIC_DDS.pdf  for a DDS based on 16F268 using an external R-2R ladder.   8 bit DDS resolution (spurii typically -45dBc) and 32 bit DDS resolution, with an internally generated clock in the 100kHz region, so output frequency setting to 23uHz  and a stability whatever the input allows.   A off-the-shelf typical TCXO with 1ppm stability is good for 9mHz  8970Hz

 

It will offer you FSK so instead of on-off keying you can implement DFCW, WSPR or whatever.

 

Andy

www.g4jnt.com

 

 

On 27 April 2012 12:01, wolf_dl4yhf <[email protected]> wrote:

Hello Peter, Stefan and all,

From the curved traces visible in Peter's high-resolution spectrogram I'd say this thermal drift is normal for a soundcard.
You may be able to shorten the initial time (after a "cold start" in the true sense) by setting the tolerance range (in ppm of the soundcard's "calibrated" sampling) to a much larger value. Markus DF6NM pointed this out a few weeks ago.
>From the "Components" menu, select "Sampling Rate Detector". There's a field labelled 'Max deviation from initial sample rate'. Be generous if your soundcard oscillator drifts fast and 'far', for example enter 10000 ppm in that field.

All the best,
  Wolf (@ DF0WD this evening, hoping for some crossband activity 500 kHz / 35XX kHz).


Am 27.04.2012 11:22, schrieb PA1SDB, Peter:

Hello Stefan, VLF,

I think you are very lucky with a PC that generates such a stable signal.

 

>Here i used a more than 10 year old notebook. Also if people like to use the digimode

>termninal or also the frequency generator, not beeing limited to frequencies given by the

>steps of a divider, SpecLab is much more comfortable. Also on the receive side i

>remember to have received DF6NM's first tests in a 47 uHz window, with a stable trace. :-)
>

>73, Stefan

I tried two notebooks now. That one where I was talking about yesterday is also a 10 year old Acer. Nice to see here (grab below) how good that "Sampling rate detector" works. In the first few hours I forgot to un-set the "measure only" option.... The signal drifts all over the graph. Then it stabillizes when I did the adjustment in the cfg.

Unfortunately that 2th PC that I test here, a "modern" 2 year old Acer Extensa 5235 notebook generates also a terrible trace at 47uHz.

 

 

One more PC to go :-)

 

73's , Peter - PA1SDB