refclock 28 gone wacky on me

[Mike]

On 06/09/2016 10:04 PM, Gary E. Miller wrote:
> Yo Mike!
>
> Back on list since I think I see a real problem here.
>
> On Thu, 9 Jun 2016 21:41:32 -0400
> Mike <bellyacres at gmail.com> wrote:
>
>> On 06/09/2016 09:26 PM, Gary E. Miller wrote:
>>> Yo Mike!
>>>
>>> On Thu, 9 Jun 2016 21:22:29 -0400
>>> Mike <bellyacres at gmail.com> wrote:
>>>   
>>>
> Yes, one GPGGA and one GPRMC each second, but they are both for the
> same fix, one fix per second, always ending in 0.940.
>
> That is bad.
>
> gpsd assumes that the fix data for each second starts to arrive just
> after the beginning of the second.  But yours starts so close to the end
> of the second that data will still be spilling into the next second, before
> anything is computed.
>
> Some of your lines are 78 chars long.  At 9600 each char takes 0.001
> seconds.  78 chars is 0.078 seconds.  So even if the GPGGA is started
> at exactly 0.940 into the second, the sentence can not be received, and
> checksum checked, before the next second starts.  0.940 + 0.078 is 1.018
> seconds.
>
> gpsd may then pair the PPS with the wrong second.  Which is the sypmtom
> you see.
>
> Groan...  I'm gonna have to sleep on this one...
>
> RGDS
> GARY
Okay so I can understand that explanation.  Thank you for the detail.  
What I fail to understand is why this just seems to have appeared out of 
thin air.  It's not like I just hooked this up yesterday.  I have toyed 
with this thing for probably three years off and on, never had anything 
like the offset I'm seeing.  Of course there is no saved data from times 
when it was running well.

Mike