[time-nuts] syncronized clocks

[Pete Stephenson]

On Thu, Sep 17, 2015 at 2:00 AM, Can Altineller <altineller at gmail.com> wrote:
> Hello,
>
> I started putting a test setup together when I suspected the DS3231 RTC's I
> got from dx.com were fake. So I put 2 mcus, each measuring microseconds
> from each 1PPS output with an interrupt, for the unit itself, and another
> interrupt to measure 1PPS time from another unit and display them on a
> nokia LCD.

Out of curiosity, what caused you to think the RTCs were fake?

> I have found out the RTC's differ by 1-2microseconds each second, on of
> them was falling behind 1microseconds each second, relative to other one.

How stable are the oscillators on the microcontrollers? It's certainly
possible that the microcontroller oscillators differ a bit, which may
affect the results.

The specs for the DS3231 says it's accurate to within +/- 2.0ppm. If
my understanding is correct, 1-2 microseconds per second would be
within spec.

> Then I put a i2c gps module, (which does not have 1pps output
> unfortunately) and parse out the time string, and generate a pulse for
> another interrupt.
>
> I found out that time from gps (arriving time of timestamp) varies within
> 10 milli seconds.

That's not a surprise: most non-PPS GPS modules don't make any claims
as to the precision of the serial data stream other than it typically
is emitted after the start of the second described in the string. On
one unit I've seen the start time of each serial data burst differs by
several tens of milliseconds, compared to the PPS signal from a GPSDO.

> So here is the reason I am writing to the list: I am also out to buy a
> rubidium frequency standard, or a trimble gps disciplined clock.
>
> Like:
> http://www.ebay.com/itm/Trimble-GPS-Receiver-GPSDO-10MHz-1PPS-GPS-Disciplined-Clock-Antenna-power-/181810679481?hash=item2a54c2ceb9
> or
> http://www.ebay.com/itm/Trimble-GPS-Receiver-GPSDO-10MHz-1PPS-GPS-Disciplined-Clock-with-rs232-port-/261997391557?hash=item3d0042eec5
>
> I have read somewhere that these newer GPS disciplined clocks are much
> better than an rubidium based atomic standard, like datum modules, or some
> other modules we can get on ebay, since they are based on much more
> advanced atomic clocks that are on board gps satellites, and correct its
> oscillator continously.
>
> Is this true? What would be your recomendation? A GPS disciplined unit, or
> a rubidium standard? What are the differences between them. For right now I
> am interested in making two clocks beat (having the same 1pps output, in
> sync) but later on I might be interested in other measurements. It seems to
> me getting an exact measurement of time is really challenging and
> interesting as a hobby.

I can't comment on that particular receiver, but I've been quite
satisfied with my Trimble Thunderbolt. Unfortunately, the prices for
used Thunderbolts on eBay have increased over the last year or two, to
the point where it may be better to get a new Thunderbolt-E from
Trimble (about $650) with the original factory warranty and get the
antenna separately rather than pay $400 for a used Thunderbolt kit
(including antenna, power supply, etc.). GPS receivers are nice
because they are continuously adjusted to track GPS time and so
inherit GPS' long-term stability.

Used rubidium standards have also gone up in price recently and
there's a wide range in quality. Rubidium lamps also have a finite
(but relatively long) lifetime. GPS receivers typically last
indefinitely modulo electronics failures, though old receivers like
the original Thunderbolt may have issues with GPS week rollovers (this
generally doesn't affect the 1PPS or 10 MHz output, but may affect the
date/time reported over the serial stream).

As a personal choice, I'd prefer the GPS receiver option unless your
devices will be operating in areas where GPS signal will not be
available (e.g. underground, in a room with no view of the sky, etc.).
Your mileage may vary.

That said, before you buy anything, consider your objective: What
precision is necessary? What is your budget?

TCXOs are cheap and widely available as easy-to-integrate modules. 2
ppm drift over a year is only about a minute or so.

If you're within range of the WWVB (North America) or DCF77 (Europe)
longwave time signals, simple low-cost receivers exist and can keep
clocks synced so long as they can receive signal.

As the next step up, timing-grade GPS receivers like the Trimble
Resolution-T (used) or various uBlox modules (new) are available
relatively cheaply (<$50, typically) on eBay or from other vendors. If
all you need is a PPS source that's linked to a global standard, that
might be a good option. They typically interface well with TTL logic
circuits.

Integrated GPSDOs like the Thunderbolt are larger, use several watts
of power, but output both 1PPS and 10 MHz signals. This is useful if
you need a frequency standard in addition to 1PPS. However, they tend
to be intended for telecom use and so require a little bit of tweaking
to work well with other equipment: for example, the Thunderbolt's 1PPS
signal is intended to connect to a device with 50 ohm impedance using
coaxial cable. If the device it's connected to doesn't present a 50
ohm impedance, the 1PPS signal will "ring".

Cheers!
-Pete

-- 
Pete Stephenson