What are you going to do with the output? Do you need a full class-D amplifier in the end? If so, I second the idea of using an integrated class D amplifier instead of rolling your own.
Is the requirement of using an external clock mandatory? It will certainly make things more difficult. As you saw, many PWM controllers embed their own oscillator (making it a lot easier to generate a sawtooth signal internally). If you're using a full integrated class D amplifier, you may be able to feed it an external clock, but it will most likely be a multiple of the modulation frequency, and not the modulation frequency directly... so you'll probably need a PLL.
What is this last requirement for exactly? Tell us more. Do you just need to control the frequency, or will it have to be fully synchronous? Will the frequency be fixed once and for all, or variable?
If it's a hard requirement, you could always design your own PWM modulator, with a square to sawtooth converter (basically an integrator), but the result will be extremely disappointing unless you spend unreasonable time designing all this. And even so...
Another approach could be to design it all digitally on a CPLD/small FPGA. Either way, if you're going to drive output power transistors, you'll also need to implement dead-time (otherwise your amplifier would have poor efficiency and may even fry.) For this, you could use an additional integrated half-bridge controller.
If I were to implement this digitally for some reason, I would rather implement a sigma-delta modulator, and embed the dead-time controller. It would require an ADC for the input signal if you really need an analog input. Then the input clock would be multiplied with an embedded PLL to feed to modulator.
But again, unless you have a very specific reason to do this, just use an integrated class-D amplifier.