Originally Posted by atferrari

Today I started to experiment with PWM in a 16F877. Got both PWM modules working in minutes.

While period can be finely ajusted with 10-bits precision, the duty cycle seems more complicate due to the two LSB in CCP1CON<5:4>.

I have no specific application in mind but when tried to increase/decrease the duty cycle value I've realized that trying to change it down to the LSB would demand too much code other than INCF / DECF.

For an eventual application requiring a fast adjustment, is it any way to do it or we have to accept adjusting only the value in CCPR1L :?:

You seem to be forgetting the speed a PIC runs at?. If you're clocking the PIC at 20MHz you should be easily able to update all 10 bits in under 2uS. This should be much more than fast enough for any possible application?.

But on the other hand, do you require 10 bit anyway?, and assuming you did? (or thought you did!), updating it far faster than it needs to be is simple.

Because the duty cycle control is transferred to the slave register only at the end of a PWM period, the rate of change of the PWM duty is no faster than once every 51.2usec (@10-bit resolution) even if you found an efficient way to alter CCP1CON<5:4>.

Using a 16-bit shadow register seems TOO MUCH, isn't it?

You'd need one anyway for computation purposes if you wanted to do any thing useful out of the PWM module.