Steve Hardy wrote

> 2. Compare module

<snip >
>the high end resolution would be about 1500Hz, which is not as good as
what you >want.  In order to achieve this, your software will need to
increase >the update rate of TMR1 at the appropriate point, and compensate
the >value in CCPR1H/L.  E.g., at the 100Hz point, increase the TMR1
update >rate by a factor of 16, and multiply CCPR1H/L by the same (shift
left 4). >This would improve the resolution at 10KHz to better than 100Hz.

sorry for this late response.

I think I will forget the 505s pulse width, this I can solve with a
monostable multivibrator. It will cost me some more hardware but it will
take a load of the software complexity.

I don't understand why the compare is needed  together with the Timer1 to
generate the wanted signal.

I now use the timer register without the compare and I see some problems
when trying to load the timer registers. When I load the registers, by
sending via I2C another value to the timer registers for example with
0xFF,0x9B I can generate a signale with a frequency of  exact 10.000 hz.
Sometimes when I try to change this value, the generated signal starts
acting strange, sometimes the generated signal disappears for some time
etc.. when I keep transmitting the same values over and over again via I2C
somethimes the signal get stable again.

I suppose it has something to do with the intterupts that I use. Besides
the TMR1 and the I2C I will also be needing TMR0 interrupt and maybe an ADC
interrupt. Is there something I should take care of.

greetings Frank Temmerman