I have devised a routine to run on a 20-MHz PIC that gives you EXACT 1 us
increments from 1 us to 65.535 ms. In the application there was some idle
time during which the constants were calculated (admittedly, it took another
2us or so).  The nice thing is that routine exits with one instruction spare
before the time duration is over, so you can use this to perform
output/input instructions as required. (It's the NOP at the end of the routine).

In short, it's a 16-bit UP-counter which terminates when the count 0x0000 is
reached. If you need 2 iterations, you put 0xFFFE in the counter. It will
count to 0xFFFF, and then 0x0000, and then exit.

The list output of the code is given below.
(Because I had to duplicate the code several times I didn't use a label as a
loop reference - so shoot me !)

Regarding adapting it automatically for the clock speed, you are looking for
trouble if you want the PIC to multiply/divide the value in W to adapt to
the clock frequency.

Assuming you know the clock speed at compile time I suggest you write some
MACROs that take a DELAY value in msec, then convert it to a constant
suitable for the delay routine. E.g.,

  00000010          00007 delay1LSB equ 0x10
  00000011          00008 delay1MSB equ 0x11
                    00009
  00000004          00010 CLOCK_FREQUENCY equ d'4' ; for 4MHz
                    00011
                    00012 mydelay MACRO MSEC_DELAY
                    00013   LOCAL NUMBER
                    00014 NUMBER set MSEC_DELAY * d'1000' * CLOCK_FREQUENCY
/ d'20'
                    00015   movlw   HIGH NUMBER
                    00016   movwf   delay1MSB
                    00017   movlw   LOW NUMBER
                    00018   movwf   delay1LSB
                    00019   ENDM
                    00020
0000                00021   org 0x0
                    00022   mydelay d'83'
  0000                  M   LOCAL NUMBER
  000040D8              M NUMBER set d'83'      * d'1000' * CLOCK_FREQUENCY
/ d'20'
0000 0C40               M   movlw   HIGH NUMBER
0001 0031               M   movwf   delay1MSB
0002 0CD8               M   movlw   LOW NUMBER
0003 0030               M   movwf   delay1LSB
                    00023
;+--------------------------------------------+
0004 0270           00024   comf    delay1LSB,f           ;| The actual
constant is modified so that    |
0005 0271           00025   comf    delay1MSB,f           ;| it can be used
in a loop that counts UP.   |
0006 03F0           00026   incfsz  delay1LSB,f           ;| (UP count loops
execute faster)            |
0007 00F1           00027   decf    delay1MSB,f           ;|
|
0008 02B1           00028   incf    delay1MSB,f           ;|
|
                    00029
;+--------------------------------------------+
                    00030 ;
+-----------------------------------------------------------------------+
                    00031 ;  | DELAY delay LOOP
|
                    00032 ;
+-----------------------------------------------------------------------+
                    00033
;+--------------------------------------------+
0009 03F0           00034     incfsz  delay1LSB,f         ;| The LSB is
always incremented. When it     |
000A 00F1           00035     decf    delay1MSB,f         ;| reaches 0x00,
the counter-decrement for    |
000B 03F1           00036     incfsz  delay1MSB,f         ;| the MSB is
skipped so that the MSB is      |
000C 0A09           00037     goto    $-3                 ;| incremented.
When the MSB becomes zero,    |
000D 0000           00038     nop                         ;| the loop exists
|
                    00039
;+--------------------------------------------+
                    00040   end

If you need any more explanation, mail me.



At 07:28 PM 12/17/96 +0000, you wrote:
>Does anyone know any good delay routines for the 16c84 and 16c74?
>I want to be able to enter a value for the clock speed (usually 4MHz, 10MHz
>and 20MHz) and the routine will adapt to the speed.  It must function as a
>subroutine taking the value in the w register as a delay in mS and then
>delay that amount of time.
>
>Any help on this subject would be most appreciated.
>
>Have a happy Christmas
>                        Tim
>
>
Jan van der Watt
[i SAW Elvis. He sAt BETween mE anD BigFOOt on THE uFo.]