rsdio writes: nothing beats a pure sine wave oscillator, except another pure sine wave oscillator at a similar frequency:
This circuit derives a pure sinewave from a crystal-controlled clock source by using a ring counter to remove the highest-amplitude unwanted harmonics, and filtering the result with an 8th-order lowpass, switched-capacitor elliptic filter (MAX7400).
Another (better?) way to approach this may be to feed the clock to an up-down counter set to count up then down repeatedly; then feed the parallel output of the up-down counter into a resistive sine-weighted ladder D/A converter. Hmmm….
@Drone. Excellent idea! The triangle wave generator would only have odd harmonics with amplitudes of 1/(N^2), one over N-squared, and thus it should be easier to remove the harmonics. A triangle wave’s amplitude would be frequency-dependent because less time results in a smaller count, but the SCF already has a fixed ratio between filter corner frequency and input clock, so it would be simple to require a fixed ratio between triangle wave frequency and clock. The counter would just need enough bits to handle the resulting magnitude.