[time-nuts] syncronized clocks

Pete Stephenson pete at heypete.com
Fri Sep 18 04:57:06 EDT 2015


On Fri, Sep 18, 2015 at 3:26 AM, Can Altineller <altineller at gmail.com> wrote:
> Hello All,
>
> Thank you for your illustrious answers. I suspected that the RTC's were
> fake when someone alerted me on a forum, 43oh, when I was trying to make
> them work with energia, and a TI microcontroller. A forum member stated:
>
> "What does have me curious, is how cheap that is--since the DS3231 chip
> itself in single qty's from most retailers is about the price of that whole
> board.  You should do some tests on its accuracy to see if it's a real
> DS3231 or some knockoff."
>
> I had gotten them < 5$ a piece, and even the 1-2 ppm is within range, it
> has other problems, like the battery is not recharging, and it has a diode
> making it impossible to use with 5V (even though the chip itself  is 3.3V)

There's many DS3231 modules on DX. Which one did you order?

As for costs, DX sells a zillion of those modules and can buy the
components cheaply, likely from the local source in China. Comparing
single-unit pricing to bulk-quantity pricing is apples-to-oranges.

That said, the boards I see on DX *are* quite cheap. It's probably
good to be skeptical.

> So from what I understood from your replies is that if a GPS disciplined
> clock is on for a long time with good reception, it can be used to
> calibrate a Rubidium frequency source, thus more accurate. But a Rubidium
> source has less jitter or phase noise, which can be useful in some cases.

Typically GPSDOs discipline a quartz oscillator as quartz usually has
better short-term stability than rubidium or GPS (at least out to a
few minutes). If GPS reception is lost for some reason, the GPSDO will
enter a "holdover" state where the quality of the local oscillator
determines the accuracy of the clock. If you expect holdovers to last
a long time or have very high precision requirements in holdover (such
as for telecom purposes), a rubidium oscillator will be useful.

A ovenized quartz oscillator can be much cheaper and provide quite
good performance -- the oscillator in the Trimble Thunderbolt is
specced to remain within +/- 1 microsecond after two hours in holdover
(within +/- 15 degrees C) and can perform even better with a more
stable environment and some judicious tuning.

<100ns performance has been possible for GPS receivers for a few
decades: the Motorola Oncore UT+ timing receiver ($20 on eBay) can
achieve this with ease. The PPS signal is only emitted on transitions
of the local clock, but the receiver sends data in the serial stream
so a system can correct for this (the "sawtooth correction"). In
general, though, the offset averages out. More modern receivers like
the NavSpark NS-T you mention below are similar, but have a smaller
error due to having a faster clock: "The programmable time pulse is
derived from 81.84MHz clock, it suffers from jitter due to
quantization or granularity of the clock."

> I have the following questions:
>
> a. If I measure the 1PPS from GPS, and take averages, can the average value
> be used for calibration somehow. In other words, if I am able to measure
> variances in a 1PPS pulse, and average them, can I get a more accurate
> reading?

Sure. That's basically what GPSDOs do: the Thunderbolt, for example,
constantly monitors the frequency and phase offset of its ovenized
oscillator and steers the frequency of the oscillator to keep it in
sync with GPS. You can do it yourself with a PLL:
<http://www.jrmiller.demon.co.uk/projects/ministd/frqstd0.htm> or
<http://gpsdo.i2phd.com/>

> b. If we put two rubidium sources, and have them warm up and stabilize, and
> then hook up a 2ch scope and observe the pattern, will there be a phase
> difference between the 10mhz signals? (without the scope doing any syncing)

In theory, the two should be precisely in sync.

In practice, without external disciplining they will slowly drift
apart from each other. With many models you can make fine adjustments
so they drift apart more slowly, but they'll always drift a bit.

> c. If we put two gps disciplined sources after warm up will there be a
> phase difference.

Since they receive external disciplining from GPS, GPS-disciplined
receivers should remain in sync so long as they have signal. There's
likely going to be some short-term jitter but that averages out.

For example, I connected the PPS outputs of a Thunderbolt, Oncore UT+,
Resolution-T, and Garmin GPS 18x LVC to my oscilloscope, with the
scope set to trigger on the Thunderbolt's PPS signal. The pulse of the
UT+ bounced around within a range of about 100 nanoseconds, the
Resolution-T bounced around within ~30 nanoseconds, and the GPS 18x
was within about 300 nanoseconds, all within spec.

> d. I use micro controllers to measure micros and perform my calculations,
> Is there a better way? Can I make my micro controller run with an external
> clock? Since I am not building logic circuits to do the counting, I think
> it is imperative that I have a good cpu clock.

As for the external clock, it depends on the micro. AVR micros like
the attiny85, atmega328, etc. can use external clocks for better
stability.

> http://navspark.mybigcommerce.com/ns-t-precision-timing-mode-gps-receiver/
>
> It is really cheap, and it claims 6 nsec (1-sigma) timing accuracy, but yet
> I would still need a more expensive gear to test it.

Shiny. That's not a GPSDO, in that it only provides an accurate PPS
signal when it has reception (there's no holdover capability), but it
seems pretty cool. I like that it emits both PPS and a
user-configurable frequency output -- that'd be really handy as a
compact, internal frequency reference for ham radios.

Cheers!
-Pete

-- 
Pete Stephenson


More information about the time-nuts mailing list