[GPS_Standard] Problem with v4, where can I get v3 to try?

[Glenn Elmore]

Hi all,

I'm new to the list but have been struggling to get v4 to work as 
expected. The problem is that with a 10 MHz reference and GPS or even 
with an arbitrary waveform generator providing a 10 MHz reference clock 
and generating a 1 PPS 20% duty cycle square-wave, the frequency count 
wanders around the correct value and thus causes the board to go into 
holdoff. I see counts up to 2000 (decimal) either side of the correct 
0x6800 count. These errors show up in groups and are not constant. Many 
times the count is pretty close. Averaging total counts over very long 
intervals shows that *all edges* are actually being counted as the 
average is essentially correct if taken over tens of minutes. This isn't 
an edge, noise or similar problem. It acts like there is extremely 
severe jitter in the 1 PPS sample clock even though there isn't.

I've looked at the PIC code and do not yet understand how to resolve the 
admonition from Microchip:

    PIC18F2220 data sheet, page 137:
                                                           
    15.3.2       TIMER1/TIMER3 MODE SELECTION
    The timers that are to be used with the capture feature (either
    Timer1 and/or Timer3) must be running in Timermode or Synchronized
    Counter mode. /*In Asynchronous Counter mode, the capture operation
    may not*/
    /*work*/. The timer to be used with each CCP module is selected in
    the T3CON register.

with line 433, where Timer1 (the reference clock) is set up by T1CON as:

     MOVLW b'00000111'                           ;set timer1 and enable
    it Used for CCP1 (Capture)
     MOVWF T1CON

which is asynchronous counter mode.

I'm not at all certain that it anachronism is the problem but I *do* 
experience unstable counts which seems categorically wrong for a 
completely deterministic system.

Has anyone seen this behavior and can someone help with an idea of what 
is going on and/or help by sending me v3 to try to see if it also has 
the problem?  Has anyone verified that v4 does in fact give perfect 
results (constant 0x6800 counts) when supplied with a perfect 10 MHz and 
1PPS?

Glenn
n6gn