[time-nuts] advice sought on basic gpsdo

[Jim Harman]

Hi Bob,

I have built several GPSDOs based on Lars's design; comments are below

On Fri, May 8, 2020 at 7:07 PM Robert Melville <bobmelville1 at gmail.com>
wrote:

> Fellow time nutters:
>
> Any input or advice on the following project would be appreciated.
>
> I want to make a small, portable, batter-operated clock that will
> spend most of its time docked into a charging bay with access
> to a signal from a GPS receiver -- either 1 PPs or 10.00000000 MHz.
> The device will have a voltage-controlled TXCO that will be disciplined
> by the GPS input unless the unit is traveling.
>

What accuracy do you want from this clock, and how long will it be
un-docked? In GPSDO parlance, this is called "holdover mode" and would be
specified as X usec max drift over Y hours.


>
> I have seen several designs for disciplined oscillators using a uP,
> such as an Arduino or a PIC. The main decision seems to be between
> 1 PPs or 10 MHz as the input. 1 PPs might be preferable because
> not all cheap GPS receivers seem to provide 10.000000 MHz out.
> Mostly, I want accurate time but a frequency reference with
> decent phase noise would be useful.
>
> 1. I am tending towards a published design from Lars Walenius, in part
> because
> it uses a familiar uP (an Arduino) but am open to other designs.
>

As you have probably learned from Lars' design, it compares the local 10
MHz oscillator to a 1 pps pulse from the GPS.

>
> 2. What advice to people have on glitch-free switching when
> docking/undocking the unit from the GPS???
> This seems to be something like the de-bouncing problem for a
> push-button.
>

Lars' design will switch smoothly from disciplined from holdover mode.
Basically if the PPS pulses stop, the default is to maintain the most
recent DAC setting. It also maintains 5-minute averages of the DAC value
over the past 12 hours and 3 hour averages over the past 18 days, so you
could easily have it switch to an average value if you want. You would have
to be careful to ignore glitches on the 1 pps signal caused by docking and
un-docking, but this is not difficult. Depending on how long the unit has
been un-docked, when the PPS is restored there may be a significant offset
between the expected pps time and the actual time. You will need to decide
how to deal with this depending on your application - "jam-sync" it to the
correct time, or adjust it gradually to make up for the gained or lost time.

>
> 3. Has anyone used the Arduino time library withOUT the Dallas RTC chip --
> i.e., some other source of time such as the from the locked oscillator?
>

Yes, Lars' design maintains precise time by driving one of the Arduino
counters from 5 MHz derived from your 10 MHz oscillator. It does not depend
on the accuracy of the Arduino's clock. You can use this along with the
Time library functions to display date and time. The basic time resolution
is 10 msec, driven by a timer interrupt.

You may also want to use an NMEA library to decode the time reported by the
GPS on its serial port. I have used the TinyGPS++ library with good
results. Also you might want to consider using a 32u4 based Arduino
processor because it has a hardware serial port that can be dedicated to
listening to the GPS. Most recently I have been using the Adafruit Itsy
Bitsy 32u4.

>
> 4. Can anyone share experience with conditioning the power going into the
> TXCO -- to what extent can digital noise from the uP or counters
> contaminate the phase noise of the TCXO? Does a separate isolating buffer
> help for the "osc out" port? I have had good success in the past with
> so-called "active bypassing" to deliver very clean power to an oscillator.
>
> Others may wish to comment on the phase noise questions, but I think the
power supply sensitivity will depend a lot on the oscillator you choose.
High stability oscillators generally have built-in regulators that reduce
their sensitivity to the power supply. Likewise, you will often see a spec
on the effect of output loading on oscillator frequency.

Thanks to all for your attention to this message -- I am glad to look at any
> and all possible designs.
> Happy time-keeping!
> _______________________________________________
> 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.
>


-- 

--Jim Harman