Jacopo Monegato wrote:
> yes, i think so... I need possibly more than 9 bit precision. What i am
> trying to do is digital control of an analog audio vco. I need veeeeery l=
ow
> noise ;) I know that some brands, like Dave smith instruments use 8 bit
> precision for vcf cutoff and even pitch wheel (in older instruments). I w=
ant
> more. I wanted to avoid PWM. In that case I would have used a third order
> butterworth vcvs to have about 0,5-1 mVpp noise, but i wanted to try
> something else. But, at this point, i guess i'll stay on pwm :/

What kind of bandwidth (sample rate) do you need?

If not more than 100 Hz or so, it sounds like you should consider doing
delta-sigma in firmware, which is much cleaner (easier to filter) than PWM.
Linearity is excellent, and resolution is really just a question of the
width of the variables you use in the firmware.

For example, here's some C-ish pseudocode that implements N_CHAN channels o=
f
15-bit DAC. The dac_isr() should run as frequently as possible, at least 10=
x
to 20x the bandwidth that you desire. If you do this in assembler, then you
can squeeze out another bit of resolution, because then you can use the car=
ry
bit out of the addition operation instead of the MSB of the accumulator. Yo=
u
might also want to unroll the loop, which makes mapping channels to output
pins trivial. ("u16" and "u8" denote unsigned 16- and 8-bit integer types.)

struct {
  u16 value; /* sets the output level of this channel, 0 - 0x7FFF */
  u16 acc;   /* internal accumulator (integrator) */
} ch[N_CHAN];

/* Timer interrupt handler */
void dac_isr (void)
{
  u8 i;
  for (i=3D0; i<N_CHAN; ++i) {
    ch[i].acc +=3D ch[i].value;
    if (ch[i].acc & 0x8000) {
      ch[i].acc &=3D ~0x8000;
      /* set the output bit associated with channel i */
    } else {
      /* clear the output bit associated with channel i */
    }
  }
}

-- Dave Tweed
--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.