Fwd on behalf of Dan Michaels

Problem and solution here, DM would appreciate any
insights

======================================

I'm writing code to use a 16F873 to time 16 R/C servos:

- 20 mhz xtal

- using timer0 for epoch timing, 2 "polled" overflows w/prescale=256
  and presetting to 244 gives 2*12.5 msec long epochs [works fine].

- using timer 1, in compare mode with CCP1IF interrupts to time the
  servo pulses, internal clock, prescale/1, generate s.w. interrupt
  on match, CCPR1L = 0Ah.

- with this setting, CCPR1H:CCPR1L values from 0-20000 give
  0-4000 usec long pulsewidths.

- I setup for 1st servo pulse, let the interrupt go, and stop timer1.
  Then, back in the main programm, clear TMR1H:TMR1L, load
  new values into CCPR1H:CCPR1L for the next servo, and re-
  start timer1.

The problem comes in as follows:

- if I set CCPR1L != 00h with CCPR1H != 00h, everything is ok.

- but if I set CCPR1L = 00h with CCPR1H != 00h, then the pulse
  widths get crazy.

- eg, CCPR1H:CCPR1L = 1301h --> 973 usec pulse, ok.
      CCPR1H:CCPR1L = 1300h --> tiny pulse, few usec.

I checked the Mchp errata sheets. It seems CCPR1L does
not like a 00h.

===========================
B4 I get a chance to fwd this problem to the list, Dan finds -
===========================

Solution is:

- when in compare mode for timer1, with TMR1 off ....

- if I clear TMR1L and TMR1H registers "before" I set new values into
  CCPR1L and CCPR1H, then I get the funny pulsewidth business.

- but if I clear TMR1x "after" setting new values into CCPR1x, then
  everything works fine. Botta bing.

What gives?

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics