LF: Sound cards/Frequency stability]
Mon, 01 Feb 1999 18:17:13 -0500
> >Which perhaps raises another matter ..... How good is the "timebase"
> >in each of the PC based spectrum displays? Each must rely on some
> >internal oscillator or bus data for reference, and it is probably an
> >oscillator with other than a tight specification. I have used GRAM
> >(uses a SB card) for some audio checks, and it seems to be agreeably
> >"correct" from a general point of view, but I do not know the actual
> >tolerance. No doubt there are readers who can fill in the gaps on this
> In some cases absolutely appalling !!
> The 11025Hz setting is usually the best in terms of accuracy, but I have
> seen variations up to 20 Hz eror (500 ppm). Separate Sound cards,
> especially the true 'Soundblaster' ones are usually good, ie. within the
> limits of an indifferent crystal oscillator say 20ppm. Those built onto PC
> motherboards are often much worse as they presumably derive their clocks
> from whatever oscillator happens to be around.
> At 8kHz sampling my Toshiba laptop actually samples at around 8064 Hz - I
> only noticed this when merging two .WAV files generated on different
> machines to give a demonstration on PSK31 in QRM at the HF Convention last
> October. A good musician could hear this level of sampling error by ear.
> In fact Peter G3PLX, in producing the software for PSK31 for the
> Soundblaster card, has had to include the facility for users to set their
> actual sampling frequency in the software, as measured in a setting up
> session, otherwise unacceptable frequency errors can creep in.
> Think I'll keep to a separate DSP card, at least the oscillator on that can
> be locked to a reference either in hardware or software.
> If anybody on LF has a Motorola DSP56002EVM - I have produced 8 and 1 baud
> versions of PSK31 (PSK08 and PSK01 respectively). LF is the only medium
> stable enough to support this very efficient mode in these lower speed
> varients, and it would be nice to have someone to talk to. If there is
> sufficient interest then G3PLX will do slower versions for the Soundblaster
> but there is a chicken/egg situation here - until there is interest he has
> other higher priority things to do.
> The PSK31 mode has really taken off on HF where it seems to be the most
> popular data mode in use at the moment - and no one is using it on LF !!
> Also as a thought, has anyone considered very narrow multi level FSK for LF
> - say 1 Hz shift with 32 different frequencies for eg. the RTTY alphabet.
> This could even be read by eye from a spectrogram display and be a lot
> quicker than trying to read dots and dashes off the screen. It shouldn't be
> too difficult to generate this in simple hardware - a varicap pulled,
> divided down, crystal oscillator driven by a D/A converter from the printer
> port would probably do the trick.
> Andy G4JNT / G8IMR