OK, now we know what you want to measure, but I am still not sure why you want to measure it with a resolution of close to 1 part per million!

Given the fact that you want to measure a PWM signal and the fact that the frequency is 2.5 kHz, this leads me to think that probably the resolution of the PWM is 8 bits (1 in 256). So measuring it with a resolution of, say 1 in 1000 would be quite sufficient.

With a 10MHz clock you can make direct measurements of pulse width with 100ns resolution, which is exactly 1 in 1000 for a 100µs pulse.
Adding some averaging to this, you could easily increase the resolution by 10 or even 100.


Regards,
Rob.