[time-nuts] Re: General query about disciplining circuitry

Fri Feb 10 09:17:48 UTC 2023

Hi Jim,
If I understand your question, you want a low cost precise method of
measuring 1pps phase relative to an independent oscillator?

If that is the case, then I would propose creating a linear ramp with at
least 10 samples in the linear region of the ramp from the pps and sampling
the ramp with a good ADC such as  ADC3561 (10MHz assumed..).  Least squares
fitting a line through the samples, and calculating the intersection with
the horizontal line preceding the ramp (in sim shown green is PPS in blue
is linear ramp).  This is a cheap single supply version using a pfet
(ignore bss138, it was a bss84) to hold an integrator in reset until the
PPS arrives.  On a dual supply version using a JFET, and a sampling rate
around 50MHz, a PPS arrival time measurement uncertainty of less than 20ps
has been achieved. There is a 3541 version of the TI ADC I referred to
which might be interfacable to a microcontroller with QSPI.  I always use
FPGAs for this so...

Good luck. --mike
[image: image.png]

On Fri, Feb 10, 2023 at 9:09 AM AC0XU (Jim) via time-nuts <
time-nuts at lists.febo.com> wrote:

> Thanks to all for the various comments.
>
> My question was really about how to make accurate phase measurements of an
> oscillator at the 1 PPS time tag output from a GNSS radio. There are some
> methods I can think of, but I am not finding one that has clear advantages
> of accuracy, immunity to all forms of noise on the oscillator output, and
> low complexity and power. I was hoping that someone out there had the
> perfect solution...
>
> Thanks!
> Jim
>
>
>
> At 06:36 AM 2/9/2023, Ed Marciniak wrote:
> >Itâ€™s not going to be budget friendly, but the symmetricom (now Microsemi)
> chip scale atomic clocks might get you there or close at least. Youâ€™d
> probably have to ask someone about what the loop time constants are. If
> they are something like 128 Hz, in theory the onboard quartz would be
> disciplined by the atomic transition as fast as the frequency was
> shiftingâ€¦IF the loop filter once lock is achieved has the same bandwidth as
> when itâ€™s unlocked.
> >
> >Iâ€™m struggling with why youâ€™d need the equivalent of hertz per 100GHz
> stability on that sort of power budget. A moving phase coherent receiver
> suggests whatever youâ€™re trying to transmit or receive should be designed
> with a more forgiving frequency tracking loop. A carrier phase GNSS can
> probably cover most of the other use cases I can think of, and for the ones
> it canâ€™t, youâ€™re talking about things so big a bigger power budget isnâ€™t a
> problem.
> >
> >Get <https://aka.ms/o0ukef>Outlook for iOS
> >
> >----------
> >From: Bob Camp via time-nuts <time-nuts at lists.febo.com>
> >Sent: Friday, February 3, 2023 12:12:17 PM
> >To: AC0XU (Jim) <James.Schatzman at ac0xu.com>
> >Cc: Discussion of precise time and frequency measurement <
> time-nuts at lists.febo.com>; Bob Camp <kb8tq at n1k.org>
> >Subject: [time-nuts] Re: General query about disciplining circuitry
> >
> >Hi
> >
> >With a moving platform, you get into the acceleration sensitivity of your
> oscillator.
> >That could be as high as 2x10^-9 / g. You might find a part down around
> 5x10^-10 / g.
> >Anything much past that gets you into an acceleration compensated device.
> Those
> >typically are a bit expensive and not an eBay sort of item.
> >
> >If the motion is on a typical platform, you will need vibration
> compensation to get to your
> >1x10^-11 range. This is one step up from acceleration compensation. Price
> goes up
> >yet again and surplus availability goes to near zero.
> >
> >One approach for time is to simply use the time pulse out of the GPS
> module. If you donâ€™t
> >care about the PPS edge staying aligned with the (say) 10 MHz output itâ€™s
> a valid solution
> >to the problem. Accuracy wise, itâ€™s going to win every time. Jitter wise
> â€¦ not so much.
> >
> >Bob
> >
> >> On Feb 3, 2023, at 12:37 PM, AC0XU (Jim) <James.Schatzman at ac0xu.com>
> wrote:
> >>
> >> 1W is the notional power budget for the discipliner.
> >>
> >> I am interested in optimizing designs for precision time, precision
> frequency, and precision time and frequency (three cases). I am o.k. with
> the idea of  not adjusting the oscillator phase to perfectly match GNSS but
> to instead record and track the offset without actually tweaking the
> oscillator for time/phase. That might be the best, and maybe that approach
> reduces the distinction between precision time and precision frequency.
> >>
> >> The goal for the loosest case is 1e-11 adev over 0.1 to 1000 sec, but I
> am interested in doing as well as the oscillator will support. In a future
> design I will be using an oscillator with 1e-12 adev @ 1 sec, but the
> current one is good for maybe 1e-11 @ 1 sec at best. The receiver platform
> is moving, and no radio mfr publishes performance specs for this case.
> >>
> >> Thank you for reminding me about the sawtooth. I will need to study the
> sawtooth behavior of the radio. One I am looking at now is the OEM7 series.
> >>
> >> Again, given that there are many possible approaches to precision phase
> measurement, and being pretty sure that I want to do more than implement a
> frequency counter, I am looking for advice.
> >>
> >> Thanks!
> >>
> >>
> >>
> >>
> >>
> >> At 09:31 AM 2/3/2023, Bob  Camp wrote:
> >>> Hi
> >>>
> >>> I would back up a bit first
> >> .
> >>> What sort of oscillator is being disciplined? If the 1W is the total
> power budget, that rules out
> >>> most OCXOÃ¢â‚¬â„¢s. Why ask? The stability of the oscillator does get into
> this pretty quickly.
> >>>
> >>> What is the end goal? Are you after precision time or precision
> frequency? While they are
> >>> connected, the approach to optimize one is not quite the same at the
> approach for the other.
> >>>
> >>> What are the stability goals either for time or frequency?
> >>>
> >>> Is this a mobile / portable application? If so that also impacts a lot
> of this and gets into a whole
> >>> other set of questions.
> >>>
> >>> To at least partly answer the question you asked:
> >>>
> >>> The PPS comes out of the GPS at the same time as the sawtooth
> correction information. In
> >>> most cases, you want to marry those two up immediately. That way the
> correction does apply
> >>> to the data you took. Does this matter in your case? Thus the long
> list of questions above.
> >>>
> >>> Bpb
> >>>
> >>>> On Feb 3, 2023, at 10:16 AM, AC0XU (Jim) via time-nuts <
> time-nuts at lists.febo.com> wrote:
> >>>>
> >>>> I can image all sorts of disciplining circuits for controlling a
> digitally-controlled oscillator against GPS. Presumably, I will measure
> oscillator phase at the 1 PPS times and run those values into a Kalman
> Filter. I think my question is about how best to measure the oscillator
> phase, given the constraint of low electrical power (<1W). There are
> zillions of commercial devices that do this but the mfrs generally do not
> advertize the details. I plan to use either an MCU or an FPGA, depending on
> which is more advantageous.
> >>>>
> >>>> I looked at
> >>>> <<https://hackaday.io/project/6872-gps-disciplined-xcxo/details>
> https://hackaday.io/project/6872-gps-disciplined-xcxo/details>
> https://hackaday.io/project/6872-gps-disciplined-xcxo/details
> >>>> and that seems interesting. I don't have to worry about VCO control
> because my oscillator has a digital control.
> >>>>
> >>>> What about measuring the phase of a large divider PLL locked to the
> reference? With I-Q outputs the phase can be measured with a very slow ADC
> applied to a sample-and-hold triggered by the 1 PPS.
> >>>>
> >>>> Thanks!
> >>>> Jim
> >>>> _______________________________________________
> >>>> time-nuts mailing list -- time-nuts at lists.febo.com
> >>>> To unsubscribe send an email to time-nuts-leave at lists.febo.com
> >>
> >_______________________________________________
> >time-nuts mailing list -- time-nuts at lists.febo.com
> >To unsubscribe send an email to time-nuts-leave at lists.febo.com
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe send an email to time-nuts-leave at lists.febo.com
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 432941 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20230210/b0515e12/attachment.png>