[time-nuts] Capturing NMEA and TICC timestamp data in time-correlated way?

[Forrest Christian (List Account)]

Just to let everyone know what I ended up doing and seems to be working for
me...

I grabbed a raspberry pi.  Hooked up the TICC via USB to it.
I also hardwired my TTL-level NMEA serial output to the serial input on the
raspberry pi's header.

I wrote a tiny script which sets up both serial ports, then uses cat to
pipe them through ts (as suggested by another poster, thank you!), and then
writes them out to a separate file for each serial port, so I have a TICC
and a NMEA file, with each line being timestamped.

The advantage of this is that I can set it on the bench and it will happily
run for some time without worrying about windows doing an update or similar.

This seems to be working for me.   I also wrote a couple of python scripts,
the main one being one which extracts a Channel A and a Channel B timestamp
file suitable for timelab from each of the files.    If I see an oddity I
now have enough data to be able to track the oddity back to a specific
period of time in the NMEA file.  Has already helped quite a bit as I now
don't have to wonder if this was a GPS signal event or something else.

I'll probably spend some time in a couple of days writing another chunk of
python which will extract relevant data from both the TICC and NMEA files
and then try some adjustments of the TICC data based on the NMEA data.

On Tue, Sep 17, 2019 at 8:01 AM Hal Murray <hmurray at megapathdsl.net> wrote:

>
> > I can capture the NMEA data and the TICC data - this is not a problem.
> >  But I'd really like to be able to capture both datasets in some sort of
> > time-correlated way, so I can easily post-process the TICC data using the
> > quantization error data.   I can probably throw something together in
> Python
> > or C to do this, but before I went through the effort, I figured I would
> ask
> > if there is a standard tool I haven't been able to find yet which is in
> > common use.
>
> How accurately do you want to stamp the NMEA data?  If the time on your PC
> is
> good enough, then software will work.  If you can feed a PPS to both the
> TICC
> and your PC, then you can get accurate timings on a second signal to the
> TICC.
>
> With a good PPS, ntpd should hold the time on a PC better than a ms.
>
> You don't actually need good time on the PC, or a good PPS.  All you need
> is a
> signal that is ballpark of once a second that you can feed to both the PC
> and
> TICC.  You can use the PPS capture on the PC to tell you when it arrives
> without using it to control the time.
>
> If you want more accurate timings on the NMEA data, I think you will need
> to
> build something to indicate the start of the NMEA data clump and feed that
> signal to the TICC.  The idea is to turn a long burst of transitions into
> a
> single pulse so you don't swamp the TICC.
>
> [It might just work.  I'm assuming the TICC will be overloaded by the NMEA
> bit
> stream, but it will probably get the first bit and lots of others.  You
> can
> throw away the others.  Has anybody tried something like this?]
>
> You could do it in hardware with a retrigerable one-shot.  This assumes
> that
> the characters come out back-to-back, no extra time between them due to
> software being busy doing something else.  Set the one-shot for a bit
> longer
> than a character time and trigger it from the serial data stream.  That
> will
> give you an output pulse a bit longer than the NMEA burst.  You can feed
> that
> to the TICC.
>
> If your geek hat is on, you will have to subtract off the delay through
> the
> one-shot.
>
> (I'm thinking of one-shots because I was working with a FatPPS recently.
> 74LS123  Thanks John.)
>
> If you prefer software, you can do it with your favorite tiny PIC or AVR
> size
> chip.  That will probably add several cycles of jitter.  I'd have to look
> at
> the data sheets carefully to work out the details.
>
>
>
> --
> These are my opinions.  I hate spam.
>
>
>
>
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to
> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.
>


-- 
- Forrest