Counting Pulses Method 'Tachometer'

[Ttelmah]

It is perhaps worth pointing out that the count the oscillator approach is always going to be more accurate at these sorts of rates.
At 180Hz incoming frequency, if you are measuring for 1 second, the 'break even' point (where counting the clock will be exactly as accurate potentially as the count the pulses approach, will come when the clock is 180* the frequency of the signal. So 32400Hz. As soon as the clock is faster than this, the potential resolution of the clock approach exceeds that of the count over time approach.
With the 4MHz clock, if the clock was accurate, at 180Hz, the count the oscillator approach could be over 100* as accurate as the count over time approach. However the limiting accuracy is the oscillator. All the code shows you using the internal oscillator. This is only specified to give about a +/-2% accuracy, and is what will be limiting your measurements.
If you want better accuracy, go back to the count the oscillator approach, and switch to using a crystal, rather than the internal oscillator. Otherwise you are just wasting your time....

[carl]

Understood Ttelmah and thanks for your input.

The primary goal is within 1% and if it is achievable without changing the hardware, then that is the main priority. The PCB design is already in use for other encoder applications - so a change in the hardware would be a last resort.

I acknowledge the error rate on the internal crystal, but just wanted to see if 1% was achievable. So far the counting method gives better results, and I can greatly improve this by increasing the PPR Maybe upto 2048. So F = 2048 * 2000/60. F = 68.266KHz.

Of course the drawback is I am always going to loose resolution at lower speeds using the counting method, so I need to determine the minimum speed/tolerance allowed (probably end up being a split tolerance for different min/max speeds). Also the counting method has a long delay.

But I am getting there!

Thanks

[temtronic]

Don't know if you considered this but...
IF you get your 1% solution running on one PIC it may(probably) won't work on the next one. This goes back to using the internal oscillator and 'batches' of PIC will be 'close' but not 100% the same.

I know you've saved a few pennies by not using a crystal and 2 caps but whenever you need accurate and repeatable performance,especially in a commercial product, using a crystal/caps combo will save you a lot of grief,time and money in the long run.

jay

[carl]

Hi Jay,

Its not the cost, but the application is very lower power (WiFi Encoder) that lasts for over a year from battery and ordinarily 'critical' timing is not required for the current applications. So I am trying to see if it is possible to utilise the current hardware (by modifying the firmware) for this new application (tacho + direction) first before considering new hardware.

Thanks
Carl

[Ttelmah]

You have a simple answer. No.

If you start with a clock with a 2% accuracy, you can't get better than this (except by things like 'working short term', and calibrating from a known source).

However if the measurement doesn't need to be made very often, remember you could enable a crystal oscillator, run from this, perform the measurement, then switch it off, and go down to the internal oscillator.

[carl]

Agreed.

Thanks for all your help guys!

[SuperDave]