From Steve:
>Try using a hardware divider and generate everything 16x higher
> frequency. That will improve the resolution
OK, brute force way. But not useful to get a fixed frequency resolution.

From Cory:
>You could also try using a 74HC4046 PLL with a programmable divide >by N in the feedback loop. The datasheet from Philips gives a good >example of how to accomplish this.
This is the right way. Build a PLL of your own, with a single or dual-modulus prescaler (dual is better), a LPF and a VCO. Start with the prescaler, here's a datasheet of a 'famous' one:
http://ebus.motorola.com/brdata/PD...5151-2.pdf

Txip.