>
>    How about using PWM output with a single digital pin and a PWM->analog
> conversion (just an integrator or maybe just a cap?). The PWM
> representation of DTMF might be simple enough to express as a formula
> rather than a monster table (thus some quickly executing code, and possibly
> pretty compact code at that). I guess it would really depend on just how
> pure those sine tones need to be, that would drive how intricate the
> code/table would need to be, how many interrupts each second to yield the
> needed PWM timing resolution etc etc.

You need about 10k samples / second to meet DTMF spec.

Walter Banks
http://www.bytecraft.com