> From: Florian Bloemer <F.BLOEMER@LINK-R.DE>
>
> Hi All!
>
> Is it possible to count and calculate a frequency of an external source
> (about 30kHz) using the TMR0 of a Pic16c84 running at 4 MHz. If it is
> possible how can this be realized on the software side.
>

Herr Blvmer:

It certainly is posible.  This is the way I would go about it:

Use TMR0 in external counter mode i.e. apply the frequency to be
measured to RA4/T0CKI.  Assign the prescaler to TMR0 (1:256) so that
more than 256 cycles of the frequency can be measured.  Use instruction
counting techniques to determine how many PIC cycles occur in the time
that it takes for TMR0 to overflow from 255 to 0.  You can capture the
counter overflow event by using the 16C84's interrupt facility.

Initialisation code:

        bcf     intcon,TMR0IE   ; Disable counter interrupts
        clrf    lowbyte         ; Clear time counter (24-bit)
        clrf    medbyte
        clrf    highbyte
        clrf    tmr0            ; Clear counter + prescaler
        bcf     intcon,TMR0IF   ; Clear previous interrupt flag
        bsf     intcon,TMR0IE   ; Enable counter interrupts


Code to increment a 24-bit counter at a constant rate is given here.
Each increment of the 24-bit counter takes 9 PIC cycles (9us at 4MHz).

x1      bcf     intcon,GIE      ; Disable interrupts
        incfsz  lowbyte
        goto    x2
        incfsz  medbyte
        goto    x3
        incf    highbyte
x3      bsf     intcon,GIE      ; Reenable interrupts
        goto    x1
x2      goto    x3

Note that interrupts are disabled around the counter increment so that
the increment appears as an 'atomic' operation.  This works in spite
of the Microchip caveat regarding clearing of the GIE flag.  The interrupt
routine does most of the real work:  when TMR0 overflows, it is known
that 256 times prescaler cycles of input have occurred.  Also, the
24-bit counter is captured at this time so it is known how many PIC
cycles have occurred in the same time.  A calculation can be made to
determine the average input frequency over the sampling period.

The calculation is

  f = 256 P / (0.000009 C)

where   f is frequency (Hz)
        P = prescaler (2, 4, 8, ... 256) as selected
        C = 24-bit count.

The max frequency you could count would be about 50MHz.  At this frequency,
the maximum count would be about 145 hence an accuracy of better than 1%.
At 30KHz, the potential accuracy would exceed your xtal's accuracy.

For further information, read Microchip's data sheet cover-to-cover.

Regards,
SJH
Canberra, Australia