Accuracy Timer2

[GDetienne]

I try to have one interrupt each 1 mSec with the greather accuracy posssible.
I have try with Timer0, without succes. I follow the advice of Mark and I use Timer2.
The test program below give me pulses of 0.99981 mSec ( measured with a counter/frequence meter).
Is it possible to increase this accuracy ?
Test with Pic 16F876A, quartz of 16 Mhz + 2 x 27 pF, PCWH 3.216

[Ttelmah]

Get a more accurate crystal!....
Seriously the final accuracy, is completely determined by the crystal. If you want a more accurate result, change one of the crystal caps into a trimmer, program the chip to give a pwm output at (say) the crystal/16, measure this with your dfm, and adjust the trimmer till the frequency is right. Unfortunatel, remember that if you just try to check the crystal frequency with the dfm, the capacitance of the meter itself will change the raw frequency. Also how recently was you dfm calibrated?. Assuming the dfm is accurate, your current error is 2 parts in 10000, or 200ppm. Normally I'd expect a good quality crystal to do better than this (up to perhaps 20ppm, before you start to need therma control), but I'd also be checking whether your cap values are reasonable. Remember you should include board and pin capacitances in the calculation.

Best Wishes

[GDetienne]

Thanks for your answer. Not easy ...
What the best, Xtal or ceramic resonator ?

Thanks.

[Ttelmah]

Crystal.
Seriously, your figure is below what a reasonable crystal should achieve. If you get the data sheets from various crystal manufacturers, you will find types ranging from perhaps 100ppm down to 20ppm in the 'standard' ranges, then high accuracy types (which will cost more), going down to perhaps 2ppm. Then versions with temperature control, pushing down to perhaps 0.1ppm over short/medium terms.
I would be very suspicious that your capacitances are not correct for the crystal involved (when you include the chips own capacitances). with the right crystal chosen, and right capacitors, and careful board layout, you should be able to improve by at least a factor of ten.
You should also be aware that the internal PIC oscillator, can result in overdriving certain types of crystal (and give very poor frequency accuracy as a result). Look at the MicroChip application notes about crystal selection.
Also consider pre-built crystal oscillator modules instead of a simple crystal. These are available 'pre-trimmed' to much higher accuracies. C-mac for example, do a unit with 20ppm accuracy, that is specified for 15years, over 0 to 70C.

Best Wishes

[GDetienne]

Thanks for ideas. I can start to try with others values of capacitances.

Have a nice week.

[asmallri]

[GDetienne]

I search and let you know when I have found.

Thanks.