[time-nuts] Trimble Thunderbolt 1pps

Chris Albertson albertson.chris at gmail.com
Sat Jan 4 18:37:23 EST 2014


On Sat, Jan 4, 2014 at 1:51 PM, Tom Van Baak <tvb at leapsecond.com> wrote:

> > Pulse quality of single-ended RS232 over unbalanced twisted pair is going
> > to be pretty bad beyond a few feet. If you want to transport the 1pps
> over
> > twisted pair there are a couple of options:
>
> Hi Brian,
>
> I suspect this is true at one level, but what would be helpful to to
> *quantify* it. What is "pretty bad"? What is "few" feet? You are implying
> that 1PPS timing is dependent in cable quality and cable length. I would
> agree. But please provide some numbers, even rough numbers,...



OK, in this use case it is easy to divide "good" for "bad".  Bad means it
does not work.  I could not get interrupts to trigger.  That is "poor"
timing.  If they reliably trigger then for NTP the details past that hardly
matter.

NTP can have very good timing using even not some good jitter on the clock
because it looks at many clock poses over many minutes.  Also on most PCs
the time stamp resolution is one microsecond with about 2uSec accuracy.
The shape of a PPS pulse does not matter, but they DO have to get to the PC
and trigger an interrupt.

If you only have a four wire cable, you are NOT using balanced pairs.  For
me, sending t/l level serial over 60+ feet of install cat-5 cable did not
even result in reliable data transfer.   Bosting the levels to RS-232
worked much better.

Looking at waveforms on a scope using different lentghs of cable, will NOT
be very instructive.  Because in the real world a 100 foot cable is
installed inside a wall and ceiling next to all kinds of real-life things
also found in walls and ceilings, like fluorescent lights, other data
cables, AC mains wire and "whatever".   You'd get real clean signals
looking at wire in a lab.  But in a real cable the result might depend on
the HVAC cycle or if it is day or night.



-- 

Chris Albertson
Redondo Beach, California


More information about the time-nuts mailing list