David Williams wrote:
>
> Hello,
>
> I've got to set a pin every 24 hours plus or minus 10 minutes.  I need the
> timer to work while I'm doing a bunch of other things (I guess this means
> I've got to use interrupts).  Anyway, anyone have any ideas how to set up
> a 24 hour count?  C code is preferred.
>
> Thanks,
> Dave

It remembers me about a device I did long ago, when 60Hz AC was not
available (to synch the unit), and no RTC chip available at that time,
so we needed to track time basically by software, *with* several
interrupts sharing the same microprocessor, it was a Z80.

We had two ideas, the first one was to implement a top priority
interrupt (NMI - Non Maskable Interrupt) each second, so it would be our
3600 counts per hour.  The idea was simple and nice, if some of the
secondary interrupts were not also top priorities... :)

The second idea (that worked), had the processor main routine just
counting Z80 processing cycles.  It just counted certain number or
instructions per second. We accounted how many processing cycles were
used in each one of the 6 possible interrupts. At the end of each one,
the routine just subtracted its particular cycles quantity from the main
timer decrementing counters. It means that the main routine always
decremented its counters to zero at the right time.

There was a tricky thing;  Each interrupt routine should take care of
possible missing "seconds counter" update. This situation only happens
if an interrupt was being handled during the time when the "non
interrupted main clock routine" should reach its counters to zero and
update the seconds counter, but it was not done because the interrupt
routine.

Then we changed this job to the main timer routine. If the decrementing
counters were negative, the routine immediately updated the "seconds
counter" and calculated the misplaced negative value for a new (smaller)
reload, so the next seconds update would be right on time.

Of course, at that time we had no software to take care of how many
machine cycles each routine would take, so we needed to do it manually.
Many pages of paper and lots of handwriting later, we got the right
code.  Ahh, by the way, we had no assembler compiler either, it was
everything done by hand... can you imagine how long it takes to write a
1000 bytes of code?  2708 was the Eprom chip, programmed by hand in a
byte by byte fashion, first to a couple of 2114 (1k sram) and then the
continuous programming cycles... :)

It took 3 weeks to do what we can do today in just 5 minutes or less.

Wagner