[time-nuts] Re: GPSDO replacement recommendation

[Ole Petter Ronningen]

The trick to getting good adev based on a Mosaic-T GPSDO is to simply
bypass the whole sawtooth problem. Dont compare the PPS from the Mosaic-T
with a PPS derived from your steered oscillator, rather run the Mosaic-T
with the steered 10MHz directly, and use the relevant messages from the
Mosaic-T to run the control loop. No external phase-detector necessary, no
PPS sawtooth to struggle with. It is described in the hardware manual for
the mosaic, appendix D.

ADEV of 2e-13 at 1s on the steered clock (OCXO) has been demonstrated this
way, although with Atomichron real time PPP - but the architecture should
work either way. YMMV.

BR
Ole

On Fri, Apr 19, 2024 at 11:25 AM Zdenek Chaloupka via time-nuts <
time-nuts at lists.febo.com> wrote:

> Hi,
>
> it is very difficult to tame Mosaic-T's sawtooth (clock quantisation
> error) as the correction data for the sawtooth are reported from Mosaic-T
> only after the pulse is out (as oppose to, for example u-blox’s
> ZED-F9T/NEO-F10T, or Furuno’s GT-100). Also note that simply cleaning up
> sawtooth with a PLL is also not that easy, as the sawtooth is not
> uniformly, or normally distributed at times, that means it introduces a
> significant bias which needs to be cleaned up by PLL. Therefore you would
> need a PLL with 5mHz bandwidth or lower to clean up the sawtooth (and bias)
> properly. OCXO’s won’t do, but then a good Rb may be enough to keep PLL
> stable at those low bandwidths.
>
> In general I would go with a good low cost receiver (my recommendation is
> NEO-F10T, which you can buy cheap here:
> https://gnss.store/54-neo-f10t-timing-gnss-modules) and a reasonably good
> PLL to clean up the sawtooth. The sawtooth is reported from ubx receivers
> before the pulse, so you can correct it directly in the PLL, if it supports
> it (supported by Si5383 or AD9545, for example). For some of those PLLs a
> reasonably cheap (300$) evaluation boards exist, so you don’t have to
> design your own PCB.
>
> Cheers
> Zdenek
>
>
> > On 18 Apr 2024, at 15:04, Bob Camp via time-nuts <
> time-nuts at lists.febo.com> wrote:
> >
> > Hi
> >
> > Yup, various folks have indeed done that sort of thing. Some have done
> it on multiple Rb’s.
> >
> > The “filter” in these Rb’s (many of them have that option) is not ideal
> for a GPSDO. It does not handle the “corner cases” very well. Unfortunately
> > 90% of the “smarts” in a good GPSDO involves those corner cases. Yes,
> that makes the term corner case a bit suspect …. still, it’s the term that
> gets commonly used. The “why?” is simply that the feature is primarily
> intended for “quick calibration" against a Cs or something similar.
> >
> > The PRS10 has a bit of a bump in its ADEV curve around 10 seconds. It is
> a result of their fancy mag field management approach. That also (at least
> to me) makes it a less than perfect choice for this sort of thing. A 10
> second gate time on a frequency counter is a pretty common thing …..
> >
> > On the plus side, you can hook up something like a Mosaic-T and have a
> pretty fancy drive into the device. To get the best performance, you would
> need to come up with a way to apply the sawtooth correction data to the pps
> before it goes over to the Rb. That certainly can be done, it just adds
> another layer of excitement to the setup.
> >
> > Fun !!
> >
> > Bob
> >
> >> On Apr 17, 2024, at 6:34 PM, Jürgen Appel via time-nuts <
> time-nuts at lists.febo.com> wrote:
> >>
> >> Hi,
> >>
> >> P,erformance wise I imagine a SRs PRS10 rubidium clock connected with
> its 1pps input to the 1pps output of a ublox zf9t will be hard to beat for
> approx 2. 2kEur with respect to many requirement profiles.
> >> Has anyone here tried that out?
> >>
> >> Cheers,
> >> Jürgen
> >>
> >> On 14 April 2024 15:53:33 CEST, Bob Camp via time-nuts <
> time-nuts at lists.febo.com> wrote:
> >>> Hi
> >>>
> >>> Cost wise, the surplus UCCM Trimble’s (or the Symmetricom version)
> still seem be the one to go with. At the same price, I’d go for the Trimble
> version. Both appear to work pretty well.
> >>>
> >>> If cost is no object, there are lots of crazy choices.  Each of them
> has it's own unique advantages and (usually) unique issues. The SRS FS740
> is one example. It does indeed have its quirks. Not exactly cheap, and
> still far from the most expensive option.
> >>>
> >>> https://www.thinksrs.com/products/fs740.html
> >>>
> >>> At least two list members have one of them.
> >>>
> >>> In this era of multi-band GNSS as well as anti-spoofing / anti-jam
> there are a lot of ways to run up the price :).
> >>>
> >>> How crazy do you want to get?
> >>>
> >>> Bob
> >>>
> >>>> On Apr 14, 2024, at 8:24 AM, Attila Kinali via time-nuts <
> time-nuts at lists.febo.com> wrote:
> >>>>
> >>>> Moin,
> >>>>
> >>>> My house GPSDO (a Trimble UCCM) died recently and I am now looking
> for a
> >>>> replacement. As I have been a bit out of touch with the current state
> >>>> of affairs regarding GPSDOs, I'd like to ask what the current
> recommendation
> >>>> for getting a GPSDO is, these days.
> >>>>
> >>>> Thanks in advance
> >>>>
> >>>>                    Attila Kinali
> >>>>
> >>>> --
> >>>> Science is made up of so many things that appear obvious
> >>>> after they are explained. -- Pardot Kynes
> >>>> _______________________________________________
> >>>> 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
> > _______________________________________________
> > 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