[ntp:questions] Very large jitter and offsets on GPS ref clock after upgrade to "p5"

unruh unruh at invalid.ca
Tue Jan 10 19:52:48 UTC 2012


On 2012-01-10, David J Taylor <david-taylor at blueyonder.co.uk.invalid> wrote:
> "unruh" <unruh at invalid.ca> wrote in message 
> news:yb%Oq.28020$d52.7176 at newsfe22.iad...
> []
>>> I was expecting that, for timing purposes, the four pairs would be used
>>> as:
>>>
>>> - TXD/ground
>>> - RXD/ground
>>> - DCD/ground
>>> - unused
>>>
>>> i.e. that every relevant RS-232 line was carried with a ground on the
>>> twisted pair.  This would be better than some random "pushing of pins"!
>>
>> The problem of course is that all of those grounds are common grounds at
>> the DB9 interface ( and probably at the gps as well) which means that
>> the ground current would be carried on all three ground wires, giving a
>> common mode signal on each pair.
>
> But I think that would be less of a problem as:
>
> - the signals are carried on individual twisted pairs
> - ground is now 1/3 of the resistance it was before
>
> If possible, I would not be earthing the GPS other than via the cable (and 
> use the unused pair to carry +5V to the GPS).
>
> I would, of course, be interested in any actual waveforms, which have been 
> unfortunately absent in this discussion so far.  We have no idea whether 
> the problem is related to the cable length or not, and if so, why.

Agreed waveforms would be nice. 
I guess the problem could be common mode interference. Eg, if the RX
line carries a 1 bit (-12V) enough of that -12 volts could spill over
onto the PPS line. to lower the pulse height that the serial card would
not recognize it as a pulse. 

But as you say, a waveform on the CDC line would be nice. Not only of
the start of the pulse itself, but also of the full one second to see if
there is crosstalk from the other lines. It should not really matter
since the nmea sentences usually come between the PPS pulses, not
overlapping, but the oncore uses its own system and its output may
overlap.

Better might be to carry the PPS on a completely separate line from the
serial signals to reduce crosstalk.

>
> Cheers,
> David 
>



More information about the questions mailing list