You should look at the DDS synthesizers by Micro Linear.
They are self contained 8 pin devices programmed through the serial
port. These will go up to a few 100khz.

You could square up the signal if you needed a variable frequency
timing signal.

Norman

norman@pcseng.com

> -----Original Message-----
> From: pic microcontroller discussion list
> [mailto:PICLIST@MITVMA.MIT.EDU]On Behalf Of N Steenkamp
> Sent: Tuesday, August 04, 1998 5:43 AM
> To: PICLIST@MITVMA.MIT.EDU
> Subject: Re: PIC as NCO (numerically controlled oscillator)
>
>
> Hi,
>
> > >  *Much* more elegant.  Use DDS (Direct Digital Synthesis).  Your
> > >isosynchronous loop runs at 5kHz (200 instructions at 4MHz crystal
> > >frequency; should allow a fair amount of "thinking").
> > >
> > >  It adds the "frequency" value, expressed in tenths of Hz (i.e.,
> > >1498.5 Hz is the value 14,985 or $3A89) to the two-byte "phase
> > >accumulator", modulo 50,000 and compares the modulus to half that to
> > >determine the output state.  Actually, where all phase increments are
> > >less than half the modulus, it is easier to make the MSB of the phase
> > >accumulator be the output - it will be a little asymmetrical.
> > >
> > >  Even simpler, as you are differentiating this "frequency"
> value anyway
> > >to derive "step" pulses, you may take the compound carry from the
> > >addition (mod 50,000) as the trigger to perform a step.  The whole
> > >procedure becomes a dozen or so instructions as long as the whole loop
> > >is isosynchronous at 200 instructions.
> > >
> > >  Alternately, if you find a "binary" crystal such as
> 4194.304 kHz, the
> > >addition can be performed modulo 65536 which is to say a pure two-byte
> > >add and ignore the remainder, with an isosynchronous loop of
> 160 cycles.
> > >
> > >  Good enough?
> >
> > Great!
> > I tried a couple of examples on paper, and see how it works.
> >
> > I figure (and tell me if I'm making any mistakes)
> > With a 4.096 mhz x-tal, and pa=phase accumulator
> >
> > start:
> > pa=pa+hz*2
> > output '8192' bit (bit 13) of pa
> > repeat the above 4096/times sec. (250 cycles thru loop)
> >
> > This would give me max. freq of 2048 Hz, with resolution of .5 hz.
>
> To me it looks like pa=pa+hz*4.
> You need to accumelate a phase of 16384 for one period of the output and
> at 4096 iterations per second, you need to add 4 to pa each interation if
> you want a 1Hz output.
>
> I would be a bit careful stating a resolution.  On AVERAGE, yes, you would
> achieve it, but in reality the DDS technique
> does exactly what you mentioned as a *kludgy* technique: It varies the
> instantaneous period of the output between two multiples of its "loop
> execution time" so that the average output frequency is within the desired
> of the resolution.  This granularity becomes worse as your output
> frequency increases, for example:
> Suppose you want a 7Hz output.  The period (in loops) of the output
> waveform would be:
>             Low     High
> Period Nr.
> 1           293     293
> 2           292     293
> 3           292     293
> 4           292     293
> 5           293     292
> 6           293     292
> 7           293     292
> 8           293     293
> 9           292     293
> ...
> The first period is thus 250*(293+293)/(4096000/4) = 0.143066.  This is
> 0.14% too long.
> The second period is 250*(293+292)/(4096000/4) = 0.142822.  This is 0.02%
> too short.  The next 6 periods are all the same (0.02% too short), which
> cancels out the 0.14% too long first period.  Then it all repeats.  The
> frequency, based on the first period is 6.9898Hz and on the second period
> it is 7.0017Hz.
>
> Take now an output frequency of 777Hz.  The periods are (In loop
> iterations, once again):
>             Low     High
> Priod Nr.
> 1           3       3
> 2           2       3
> 3           3       2
> 4           3       3
> 5           2       3
> 6           2       3
> 7           3       2
> ...
>
> The first period is 0.0014648, 13.82% too long.
> the second period is 0.0012207, 5.15% too short.
> the frequency, based on the first period is 682.67Hz and the frequency
> based on the second period is 819.2 Hz.
> Taken over 7 periods, the AVERAGE frequency is 774.92Hz.
>
> So, as you move closer to your maximum output frequency, the
> instantaneous frequency can be further and further away from the desired
> frequency.  If you look at your output wavform in the frequency domain,
> then you will see that you get stronger spurious frequency components at
> higher output frequencies.
>
> Niki
>