[ntp:questions] backup PPS source

David Lord snews at lordynet.org
Mon Apr 12 11:06:24 UTC 2010

Hal Murray wrote:
> In article <82f3qcFg17U1 at mid.individual.net>,
>  David Lord <snews at lordynet.org> writes:
>> Latest experiment is to attempt to make a backup PPS source and
>> I've now made a start with a watch xtal oscillator and divider
>> that can be synchronised by a PPS source when present (except
>> I've not yet wired up for synchronisation).
> In general, watch crystals aren't very good.  If you want to
> see how good/bad your setup is, just set it up as noselect and
> let it run for a day or week.  Then plot the offset from peerstats.


They may not be as high spec as AT cut oven crystals but
at reasonably constant temperature should be adequate.

Increasing negative 6min increment change turned around 162
minutes ago at -44us and offset -12.8ms. Increment is now at
-2us and offset -15.3ms so offset should start increasing
shortly. It did at last poll with offset now -15.2ms.

That's looking very promising already.

It is set as no select as it is connected to GPS serial DCD
whilst the true PPS from GPS is via parallel port. I'm
logging using 'ntpq -p' from a cron job, and setup for
peer/loop stats and summary last year when I started playing
with radioclocks.

> You can compute the frequency error from the time drift over
> a given time, and the reverse.
>> Any hints/tips welcome, I'm also considering an oscillator using
>> an AT cut xtal but design for that is considerably more complex,
>> however it's then a possibility to convert the HF to be used as
>> system clock.
> The simple way to get a PPS out of a crystal is feed a small
> micro (PIC, AVR, 8051) and do it all in software.

I don't have the tuits for this at the moment. Mid 90's I
was programming COP series but for me they're now too
expensive. I hate PIC, it's back to front for me but I'm
still trying with those and AVR. When I did the COP
project I started with the hardware logic on perfboard
layout that I could understand and get working then
programmed the micro to do the same. COP development kit
is really good with simulator and emulator provided.

With a micro I'd get rid of the trimmer capacitor and have
the calibration automated with appropriate +/- control
of counter. As it is the trim capacitor setting is much too
coarse and switched padding capacitors and varicap for fine
tune look a better method.

> What you probably want is a GPSDO: GPS Disciplined Oscillator.
> The basic idea is that you start with a good oscillator, use
> GPS to keep it running at the right frequency and phase too
> if you want a PPS.  If the GPS fades out, the oscillator will
> coast for a while.  (Magic word is holdover.)  If the crystal

Yes the counter has reset input that just requires a valid
very short PPS and it will set to within 15us and remain
locked to PPS then free run on loss of PPS. I'm checking out
how much or little of a free running drift I can expect so
have the PPS reset disconnected.

> is good enough, it will stay close in time.  How close for how
> long will be in the data sheet.
> Google for HP Z3801A and Trimble Thunderbolt.
> There was an article in the HP Journal on the Z3801A.
> They are used in cell phone towers and 911 offices.
> There are a lot of them on the recycle market.  Check
> EBay.
>> I also keep googling for rubidium sources but not found any
>> within my budget, cheapest I've found (new) starting at ukp 900.
> Try EBay.  Again, there are a lot of them getting recycled
> from telco offices and/or cell phone towers.

I'll have another look. Normally when I search for anything
I get lots of hits from ebay and have to filter them out
but those searches I made for rubidium sources didn't bring
any up.

Spec of the 900+ quid modules is excellent but they were
all "prices start from" and it was obvious I'd need some
of the addons for synchronisation.


More information about the questions mailing list