Re: [gpsd-users] Ublox EVK-M8T

[Denny Page]

> On Sep 20, 2017, at 13:25, Gary E. Miller <address@hidden> wrote:
> 
> 
>>> A pulse has two edges.  PPS uses the eading one, and ignores the
>>> trailing edge.  
>> 
>> Okay, my bad. I misinterpreted what the “reject” meant.
> 
> Very common misunderstanding.  I'm just changed 'reject' to 'ignore' in
> git head, but Im not sure that is really better.  Any recomendation for
> how we could improve the message?
> 
>> So, any speculation you can offer as to what the problem is?
> 
> Uh, which problem?  I thought you just said the PPS is fine?

Hey Gary,

I’m embarrassed to confess that no longer own a scope. Last time I did was 
around 30 years ago. I understand that they’ve changed a bit. :)

Sorry I wasn’t more clear previously. The problem was that the PPS was actually 
incorrect, off by 100ms. This caused chrony to reject the PPS, either via 
kernel pps or via socket from gpsd.

The core issues is that in the default configuration on the EVK-M8T, the RS-232 
signal is inverted on the PPS line. In other words, DCD is held high until TOS, 
at which time it is pulled down for the pulse width. The default duty cycle is 
10%, so this results in the a falling edge at TOS, followed by the rising edge 
100ms later. As far as the kernel is concerned, it’s a 900ms pulse. :)

The M8T has a “polarity” value for the time pulse, which defaults to “rising”, 
but the effect on the DCD signal out of the EVK appears to be the opposite. 
It’s unclear whether this is particular to the M8T, or if there is an inverter 
on the EVK’s RS-232 output. I’ll need a scope to know for sure. :)

Thanks,
Denny