Simple, FM ready, sinewave oscillator with sync.

Hard sync of one oscillator to another has a few things that can be quite fiddly to get right for really nice DSP.

I find the most important thing to get right is the sub-sample location of the hard sync event. If we don't handle this, then we end up forcing the sync events to align to one sample or another, and this causes a lot of aliasing distortion. The true sync event should be at some subsample moment, between one sample and the next. That means on the next sample after a sync event, the follower oscillator should have already be reset to zero and started playing a little part (1-subsample) of the waveform. If you don't take care of that then there's another kind of aliasing added by forcing the sync resets to happen on a sample boundaries. It makes the hard sync sound kind of messier than it should be. Fixing this means you need to know the subsample location where the sync really happens -- that's possible to do very accurately if sync oscillator is a phasor ramp, but involves a lot of guesswork if the sync source is just a pulse.

The other thing is that wherever the waveform currently was, the hardsync event usually means setting the value to zero, or whatever the waveform start value is, and that means a hard jump in the waveform. A hard jump makes a click, and if syncing to another oscillator, this means repeated jumps. Repeated jumps is exactly what a square or saw waveform is made of, and that's part of the buzzy sonority that is added by hard sync. As you probably know, a naive square or saw wave in a digital signal processing world is going to cause aliasing, and the same thing is going to happen with hardsync. So you may want a method to suppress the aliasing caused by the jumps, which usually means some kind of smoothing to spread the jump out in time. The ideal form of this looks like an integrated sinc impulse, but that spreads the energy over a lot of time, so often we prefer a much quicker smoothing.

We look at these issues in the last chapter of the GO book, though in that case we are just working with ramp oscillators. The antialiasing scheme there is a method that is simple, cheap, and *very brief* -- not perfect, but I think not bad at all for what it is, and the brevity of it makes it handle pretty extreme modulation well. It is possible to adapt this patch to synced sines by replacing the ramp-to-curve shapers inside it with ramp-to-sine shapers.

The hard sync still sounds kind of buzzy, but that's true to the nature of it -- it sounds buzzy on analog oscillators too.