Accuracy of GPS 1 PPS signal under invalid fix conditions...

[Steve Olney]

Hi All,

I am playing around with a GPS unit (GPS 16HVS) with the idea to use the 1 PPS output to time stamp data.

I am trying to make it as simple as possible and one of the critical aspects is what happens to the 1 PPS accuracy when the unit loses a valid fix.

Googling reveals all manner of detail about nanosecond accuracy - but assume the unit has a valid fix.  There will be times when this is not true and so this has to be catered for.

I can configure the unit to continue outputting the 1 PPS signal during periods of invalid fix.

My question is, how stable is the 1 PPS when a valid fix is lost?

As far as my understanding goes the internal GPS unit clock is slowly synchronised in phase to GPS time within +/- 1us over time under valid fix conditions (I gather from the need to wait at least 15 minutes after fix for the 1 PPS to be within +/- 1 us that the time constant is of this order).  There is some jitter on this - but that is not a problem as I only need accuracies of +/- 1 ms.

I am guessing that when a valid fix is lost the accuracy reverts back to the internal TCXO(?) with a drift rate equivalent to the time it takes to be phase-locked when a valid fix is present. On the other hand the offset may be stored and used to hold-over the 1 PPS rate during invalid fixes - in which case it is the short-term accuracy of the TCXO that prevails (say < 0.1 ppm).

Some worse-case *guesses* (no hold-over) for a 15 minute duration loss of valid fix (pretty rare for a fixed position and a reasonable view of the sky)...

1. Internal clock crystal bog-standard (+/- 100 ppm): ~ 90 ms error worse-case.
2. Internal clock crystal TCXO (+/- 5 ppm): ~ 4.5 ms error worse-case.

This link (https://blog.dan.drown.org/gps-pps-drift-when-it-has-no-signal/) states that a unit with a TCXO only drifts by 2 ms over 6 days with no signal.  This seems way too good (~4 ppb !!!).

Does anyone have any *solid* data for this ?

Don't bother with guesses - I have a pile of those already of my own :-)

Cheers

Steve Olney

HawkRAO

[Paul Oxley]

Steve

You will need to research your specific GPS-DO to determine the basic stability of the unlocked oscillator. It likely is perhaps 10-12 or higher Parts which is typical for a Oven Controlled Crystal Oscillator  (OCXO) . If it is a Temperature Compensated Crystal Oscillator  (TCXO) , it would be less stable. In either case, it is important to warm up the unit to gain temperature stability. A non compensated crystal oscillator is about 10-6 stability.

Paul

--
--
You received this message because you are subscribed to the Google
Groups "Society of Amateur Radio Astronomers" group.
To post to this group, send email to sara...@googlegroups.com
To unsubscribe from this group, send email to
sara-list-...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/sara-list?hl=en
---
You received this message because you are subscribed to the Google Groups "Society of Amateur Radio Astronomers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Steve]

Hi Paul,

On 11/03/2019 12:01 pm, Paul Oxley wrote:
> You will need to research your specific GPS-DO to determine the basic
> stability of the unlocked oscillator.

Yep - that was the question.  I was asking if anyone has that
information - or can point me in the right direction to arrive at that
information.

I have been unable to find it.

Cheers

Steve

[Marcus D. Leech]

You might look at the uBlox application note here:

https://www.u-blox.com/sites/.../Timing_AppNote_%28GPS.G6-X-11007%29.pdf

That talks about their 6 series, but would be similar (better) for their newer modules.

Basically, all GPS modules that I know of that produce 1PPS are doing so from their master oscillator that is steered by GPS data.  So when that data goes away, they are in "holdover" state, which is generally dominated by the accuracy of the VCTCXO, which will continue to receive the last control voltage it had before the satellites went away.  Some units try to do some modelling and prediction "stuff" during holdover.  But worst-case, you get
whatever the TCXO gives you...

[Jeff Kruth]

Information on all manner of frequency and time issue is available on the group run by my friend John Ackermann, N8UR, "time-nuts", google them, then you will not only find a wealth of info in the archives, but true experts in the field you can ask questions of.

Jeff Kruth

WA3ZKR

In a message dated 3/10/2019 9:19:56 PM Eastern Standard Time, hawkr...@gmail.com writes:

Hi Paul,

On 11/03/2019 12:01 pm, Paul Oxley wrote:
> You will need to research your specific GPS-DO to determine the basic
> stability of the unlocked oscillator.


Yep - that was the question.  I was asking if anyone has that
information - or can point me in the right direction to arrive at that
information.

I have been unable to find it.

Cheers

Steve

--
--
You received this message because you are subscribed to the Google
Groups "Society of Amateur Radio Astronomers" group.
To post to this group, send email to sara...@googlegroups.com
To unsubscribe from this group, send email to
sara-list-...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/sara-list?hl=en
---
You received this message because you are subscribed to the Google Groups "Society of Amateur Radio Astronomers" group.

To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+unsub...@googlegroups.com.

[Jeff Kruth]

Here is some info for the SARA folks if they are not aware, this is a great group of people, world wide, world class expertise.

Jeff Kruth

Send time-nuts mailing list submissions to

    time...@lists.febo.com

To subscribe or unsubscribe via the World Wide Web, visit

    http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com

or, via email, send a message with subject or body 'help' to

    time-nut...@lists.febo.com

You can reach the person managing the list at

    time-nu...@lists.febo.com

When replying, please edit your Subject line so it is more specific

than "Re: Contents of time-nuts digest..."

To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+...@googlegroups.com.

[Steve]

Hi Marcus,

On 11/03/2019 12:33 pm, Marcus D. Leech wrote:
> But worst-case, you get
> whatever the TCXO gives you...

..which is the question I am asking.

Does anyone have any *specific* information about the behaviour of the
GPS 16HVS 1 PPS under no-fix conditions.

All the general stuff I already know.

Cheers

Steve

[Marcus D. Leech]

On 03/10/2019 10:10 PM, Steve wrote:
> Hi Marcus,

>
>
> ..which is the question I am asking.
>
> Does anyone have any *specific* information about the behaviour of the
> GPS 16HVS 1 PPS under no-fix conditions.
>
> All the general stuff I already know.
>
> Cheers
>
> Steve

I missed the subtlety of the specific model number. I got nuthin...

[Steve]

Mmmm - better say "*Most* of the general stuff I already know".

Steve

[Steve]

Hi Jeff,

On 11/03/2019 12:41 pm, 'Jeff Kruth' via Society of Amateur Radio Astronomers wrote:

Information on all manner of frequency and time issue is available on the group run by my friend John Ackermann, N8UR, "time-nuts", google them, then you will not only find a wealth of info in the archives, but true experts in the field you can ask questions of.

Jeff Kruth

WA3ZKR

I've already tried to find an answer to the 'GPS 16HVS 1 PPS behaviour under invalid fix' question there - but gave up after manually scanning through the archives (lost the will to live after 30 minutes ;-)).

Most of the stuff I see there is talking about ns issues - not a lot about a fairly basic question in the vein I am asking.

Nonetheless - I have subscribed and once approved will try my luck asking.

Thanks.

Cheers

Steve

[James Brown]

One of the ublox GPS modules, I don’t remember which, has a feature that when you pull one of its inputs down it returns the GPS of that event down to 10’s of nanoseconds.

--
--
You received this message because you are subscribed to the Google
Groups "Society of Amateur Radio Astronomers" group.
To post to this group, send email to sara...@googlegroups.com
To unsubscribe from this group, send email to
sara-list-...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/sara-list?hl=en
---

[Marko Cebokli]

If you have two units, get them running, compare the 1ppses on an 2-chan oscilloscope, then diconnect the antenna from one, and observe the drift.

 

Marko Cebokli

[Paul Oxley]

Marko

I like your proposed test. However, seems to me that the observation would be needed over a very large number of seconds to see the drift of the TCXO that is not locked. The expected stability estimate would be greater than 1 second in 10^10 seconds.

Paul

[Marko Cebokli]

Hello Paul,

 

you can easily measure 1us on a 'scope. 1E-10 will amount to 1E-6s after 1E4s, somewhat less then three hours.

Of course, you would want to repeat the measurement several times under different temperatures, supply voltages, etc.

 

But I doubt very much that an unlocked unit would keep at 1E-10.

 

Marko Cebokli

[Paul Oxley]

Marko

Yes you can observe 1 uS on a scope. However, the scope's time base stability would need to be considered. In addition, the 1 pulse per second signal would not provide a 1 uS signal. One might count the 1PPS signals from one locked and one unlocked systems and compare the counts over a long enough period to determine the desired accuracy.  The counts would need to be at least 1e6 seconds (277.7 Hours) in duration to determine 1 uS difference between the two systems.

The difficulty with measuring oscillator stability is that if you want very high stability, observations need to be made over very long periods of time.

Stressing the environment (power, temperature, etc.) would be good to disqualify a system if it fails under stress. However, it still could not be stated that a given parts in 10^X was being met.

Paul

[Steve]

Hi Marko,

On 12/03/2019 1:57 am, Marko Cebokli wrote:

If you have two units, get them running, compare the 1ppses on an 2-chan oscilloscope, then diconnect the antenna from one, and observe the drift.

 

Good idea.  I do have more than one (or will have shortly) and I have a CRO with a 100 ns/div hor. timebase - so a drift between channels of 1 ms should be easy to see.  It is a dual channel CRO - not dual trace.   However, the information on the 'chop' setting states it can be used down to 0.2 ms/div.   I'll see how it goes...

The responses on 'time-nuts' so far overshoot my question intent (except KB8TQ's).  But I think from responses all round that I will not find what I am looking for.

So your suggestion to try and measure it is a good one.

Thanks.

Cheers

Steve

[Marko Cebokli]

Hello Paul & Steve,

 

I'm afraid a DSO is needed here - or maybe an analog storage scope.

You only get a scan each second, and you need a high X speed to see 1us. This will be extremely dim on a non-storage CRT.

On the positive side, any cheap chinese DSO will be OK.

 

One thing you must take care is triggering - you must use a single channel to trigger both traces. You must NOT use "vertical mode" as the trigger source - that will always make the pulses look simultaneous, and the time difference information will be lost.

 

I does not matter which single channel is used as a trigger source, so you can always switch it to the pulse which arrives first, to see the rising flank of both.

 

As far as I know, the GPS receivers do not shift their clock oscillators to the correct frequency. They let them run free, and output the 1pps at the closest flank of the free running oscillator. With a (hypothetical) 10MHz clock, you get a +-50ns quantization error on the 1pps timing - so when observing on a scope, you will probably see 100ns jumps instead/in addition to a steady drift (100ns assuming 10MHz clock - if I remember right, GPS units use something around 14MHz or multiples thereof)

 

Marko Cebokli

--

[Paul Oxley]

Marko

The GPS-DO that I have is a refurbished unit from a Cell site. It has a 10 MHz Oven Crystal Oscillator that provides the hold over function. This oscillator is locked to multiple satellites that are in view. I think that there is some comparison of the multiple satellite signals to make doppler corrections, etc.

The 10 MHz is available on two outputs (One square wave and one sine wave). There is also a 1 PPS signal available as an output. 

The software monitors all of the functions and reports when everything is locked (and valid). It also identifies the satellites that are being used and how strong the signal are from each. A sky map is produced that shows the satellite positions and relative strength.

Since it uses multiple satellites, I have never observed any dropouts of the locking signals. It likely could occur if your location is overhung with foliage blocking all of the satellites that are above the horizon. 

I use a cone antenna that has a very good pattern for the satellite reception. It is placed outdoors with a good view of the sky.

Paul

[Marko Cebokli]

Steve,

 

you can avoid chopping by using the "early" pulse as external trigger and scope in single channel mode.

 

Marko Cebokli

 

 

On Tuesday, March 12, 2019 07:28:38 AM Steve wrote:

--

[Steve]

Hi Marko,

On 12/03/2019 8:18 am, Marko Cebokli wrote:
> you can avoid chopping by using the "early" pulse as external trigger
> and scope in single channel mode.

But probably still have the trace visibility problem with my CRO as you
have mentioned.

If the GPS satellite signal info doesn't condition the internal GPS
clock in some fashion how is the 1 PPS signal produced ?

I thought the GPS unit clock was steered such that the 1 PPS output
comes from some division of that clock.

In any case, it may be all not a problem.  I have run the GPS unit on
the roof for 15 hours now and there have been no invalid fixes.

Have to run now as I am going out.   Back later today.

Cheers

Steve

[Paul Oxley]

Steve

The 1PPS could be created by counting down the satellite signal. However, the more likely case is to lock the local oscillator to the satellite signal and then count it down for the 1 PPS signal. Designer's choice. 

If the direct satellite count down were used, it would need to switch during any loss of signal.

Paul

Cheers

Steve

--
--
You received this message because you are subscribed to the Google
Groups "Society of Amateur Radio Astronomers" group.
To post to this group, send email to sara...@googlegroups.com
To unsubscribe from this group, send email to
sara-list-...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/sara-list?hl=en
---
You received this message because you are subscribed to the Google Groups "Society of Amateur Radio Astronomers" group.

To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+unsub...@googlegroups.com.

[Marko Cebokli]

I believe the only thing that gets locked in these single chip GPS devices are internal, software defined DPLLs, all of this clocked from a free running TCXO.

The software then decides each second, which flank of the free running clock is closest to the true second, and clocks the 1PPS from it.

The scope test should show this very clearly.

I heard that some chips will tell the quantization offset of each pulse in the NMEA output, but never saw it myself. Should be described in the chip's datasheet.

 

Marko Cebokli

To unsubscribe from this group and stop receiving emails from it, send an email to sara-list+...@googlegroups.com.