still don't have solution!!!

[rmteo]

I thought so too. Anyway, I tried it and indeed you can get a 2MHz PWM with 3 different (4 if you include 0%) duty cycles with an 8MHz clock. Checked it out in hardware and on an oscope. I can post a picture of the scope but you said you can't see it.

With a 4x PLL, you can get a 8MHz PWM. Of course you can just use CLKO and get these frequencies, albeit without any duty cycle resolution.

[skimask]

I thought so too. Anyway, I tried it and indeed you can get a 2MHz PWM with 3 different (4 if you include 0%) duty cycles with an 8MHz clock. Checked it out in hardware and on an oscope. I can post a picture of the scope but you said you can't see it.

I'm on a different PC now (the server that doesn't let me see pictures is at work ), and of course, not near my hardware.
No crazy glitches or anything? (rough edges, ok, but nothing else wierd?)...
Which PIC are you using anyways?

[rmteo]

I am using a PIC18F4620 (a PIC16 with CCP/ECCP should work too) just because I have it handy. Here is the scope trace:

Horizontal is 200nS/div and Vertical is 2V/div. Upper trace is the PWM (50% duty cycle) and lower trace is CLKO (Fosc/4).

[skimask]

Horizontal is 200nS/div and Vertical is 2V/div. Upper trace is the PWM (50% duty cycle) and lower trace is CLKO (Fosc/4).

Well then, that settles it...
Either the datasheet has to be a bit more concise, or I have to read a little bit LESS into it.