[time-nuts] U-blox teaser

Tom Van Baak tvb at LeapSecond.com
Sat Feb 27 22:15:42 UTC 2021 

Dana,

 > During my Arecibo Observatory days we used NIST's TMAS service to keep
 > our H-maser-based station clock synced with UTC.

And before that, the observatory used Tom Clark's Oncore & SHOWTIME and 
later Rick Hambly's CNS clock & Tac32Plus, yes?

Rick continues to develop the CNS clock, having switched from the 
Motorola and iLotus receivers to u-blox T receivers. His papers are on 
cnssys.com or gpstime.com. Check out a recent one like:

"High-accuracy Time and Frequency in VLBI "
https://www.cnssys.com/files/TOW/High-accuracy_Time_and_Frequency_in_VLBI_2019_sem.pdf

Bonus: lots of graphs and photos, masers, receivers, etc.

Main publication page: https://www.cnssys.com/publications.php

Check out the performance he's getting. This is with a 6T. That's not 
dual-frequency or multi-constellation. Just plain old L1 GPS. It's way 
better than +/- 20 ns. So I'm really confused by what you're saying 
below. Did Arecibo get rid of the CNS clocks?

/tvb


On 2/27/2021 8:18 AM, Dana Whitlow wrote:
> Thanks, Bob.
>
> It seems to me that, depending on the positions of sats visible to one's GPS
> antenna and the spatial distribution of free electron density in the
> ionosphere,
> the ionospheric contribution to position errors could sometimes largely
> cancel.
> But that observation may (or may not) reflect strongly on one's ability to
> get
> accurate absolute time from GPS on "average" days.
>
> During my Arecibo Observatory days we used NIST's TMAS service to keep
> our H-maser-based station clock synced with UTC.  Our user community
> (mainly VLBI and pulsar timing people) seemed pretty satisfied with +/-
> 100ns
> accuracy, so I tried to do better by keeping things well within +/- 50 ns
> during
> my reign.  IIRC, NIST was claiming that TMAS could produce results mostly
> within about +/- 20 ns.
>
> To be honest I'm baffled by how time transfer much better than that could
> be achieved in practice.
>
> Regarding Q3, yes I'm aware that *some* GPS receivers do the estimation of
> ionospheric delay.  What I was asking was:  Do any of the relatively
> inexpensive
> receivers to which we time-nuts have access do this?  Here I'm speaking of
> those being sold for no more than a few hundred USD.
>
> Dana
>
>
>
> On Sat, Feb 27, 2021 at 9:08 AM Bob kb8tq <kb8tq at n1k.org> wrote:
>
>> Hi
>>
>>
>>
>>> On Feb 27, 2021, at 9:41 AM, Dana Whitlow <k8yumdoober at gmail.com> wrote:
>>>
>>> I've long understood that ionospheric delays and variations thereof lead
>> to
>>> *position*
>>> uncertainties in GPS navigation receivers, to the tune of perhaps 10m
>>> (2DRMS IIRC).,
>>> and that these are said to constitute the single largest GPS error
>> source.
>>> Q1: Would this not imply timing errors of comparable magnitude (10's of
>>> nsec)
>>>        for a single band GPS?
>> Once all the signals “hit” the antenna, the delays are mostly common mode.
>> Instead of showing up as a position error, they show up as an error in the
>> time estimate. Since time is one of the things you estimate in the
>> solution
>> (along with X,Y, and Z) it get’s it’s own independent solution.
>>> Q2: Why have I not seen this issue raised in connection with the present
>>> discussion
>>>       about achievable absolute timing accuracy?
>> GPS time transfer is often done to the sub-ns level. There are a number of
>> papers on this.
>>
>>> Q3: Do any of the "modern" timing GPS receivers available to civilians do
>>> dual-band
>>>       reception in a way that includes estimation of (and correction for)
>>> said delays and
>>>       their variations?  I know that Garmin, for one, is now selling L1/L5
>>> handheld GPS
>>>       receivers (GPSMAP66sr and GPSMAP65s), but I've seen no indication
>>> that these
>>>       units make any attempt at doing such corrections.
>> Yes, some receivers do an estimate of ionospheric delay based on the
>> variation of that delay with frequency. This does not help with
>> tropospheric
>> delay or all of the various “common mode” issues we have been talking
>> about.
>> It is also unclear how the “unknown” timing variation between the bands
>> due to the antenna impacts these solutions…..
>>
>> Bob
>>
>>> Dana
>>>
>>>
>>>
>>> On Sat, Feb 27, 2021 at 7:43 AM Bob kb8tq <kb8tq at n1k.org> wrote:
>>>
>>>> Hi
>>>>
>>>> The same 20 or so ns delay in a saw would also apply to the
>>>> saw (or tight filter) in some timing antennas. It also would apply
>>>> to the saw(s) in some modules. Even if the tolerance is “only”
>>>> a couple ns on each of them, you *could* have 3 or more in the
>>>> chain.
>>>>
>>>> Lots of numbers to crunch to get to 5 ns “absolute”. One could go
>>>> grab a GPS simulator and start poking. First step would be to find
>>>> a simulator that is spec’d for a < 5 ns tolerance on the PPS into
>>>> GPS out. I do believe that rules out the eBay marvels that some
>>>> of us have lying around …..
>>>>
>>>> Simpler answer would be a quick “clock trip” with your car full
>>>> of 5071’s …… hour drive over to NIST and then back home.
>>>> That sounds practical for most of us :) :)
>>>>
>>>> Bob
>>>>
>>>>> On Feb 26, 2021, at 9:29 PM, John Ackermann N8UR <jra at febo.com> wrote:
>>>>>
>>>>> A while ago I tried doing a decidedly non-anechoic measurement with a
>>>> VNA exciter going to a 1500 MHz ground plane and the receiver connected
>> to
>>>> the antenna (with a known delay cable) and I got a similar result, but
>>>> there was enough noise that I didn't think I could nail it down to
>> within
>>>> 10 ns.
>>>>> I've also measured GPS antenna splitters and they tend to have 20-ish
>> ns
>>>> delays, mainly due to the SAW filters.  I did surgery on an HP splitter
>> to
>>>> remove the filters so it could be used for L1 and L2 and that dropped
>> the
>>>> delay down to only 1 or 2 ns.
>>>>> So there's definitely lots of stuff to calibrate if you want to get
>>>> accurate time transfer.
>>>>> John
>>>>> ----
>>>>>
>>>>> On 2/26/21 8:02 PM, Michael Wouters wrote:
>>>>>> Typical L1 antenna delays range from 20 to 70 ns.
>>>>>> I know of only one antenna for which a delay is given by the vendor
>> and
>>>> the
>>>>>> technique used was just to measure the electronic delay ie by
>> injecting
>>>> a
>>>>>> signal into the circuit. To do it properly, you need a setup in a
>>>> microwave
>>>>>> anechoic chamber with transmitting antenna etc. The practical
>> difference
>>>>>> may be small though, 1 or 2 ns ( sample of one antenna!).
>>>>>> Cheers
>>>>>> Michael
>>>>>> On Sat, 27 Feb 2021 at 11:42 am, John Ackermann N8UR <jra at febo.com>
>>>> wrote:
>>>>>>> They're claiming "even better than" 5 ns for relative time, which
>> given
>>>>>>> the 4 ns jitter seems at least sort-of reasonable.  But until someone
>>>>>>> shows me otherwise, I'm still thinking that getting better than 25 ns
>>>>>>> absolute accuracy is a pretty good day's work.
>>>>>>>
>>>>>>> John
>>>>>>> ----
>>>>>>>
>>>>>>> On 2/26/21 5:26 PM, Bob kb8tq wrote:
>>>>>>>> Hi
>>>>>>>>
>>>>>>>> I can’t think of many antennas (multi band or single band) that
>> claim
>>>> to
>>>>>>> know their
>>>>>>>> delay to < 5 ns. Simply having a *differential* delay spec of < 5 ns
>>>> is
>>>>>>> quite good.
>>>>>>>> Same thing with delay ripple, you see specs out to around 15 ns on a
>>>> lot
>>>>>>> of antennas.
>>>>>>>> None of this is getting you to the actual total delay of the
>> antenna.
>>>>>>> It’s a pretty good
>>>>>>>> bet that number is a bit larger than either of these.
>>>>>>>>
>>>>>>>> Some of the ripple probably comes out in the standard modeling. I’m
>>>> not
>>>>>>> sure that
>>>>>>>> the differential delay is taken out that way. Total delay, not taken
>>>> out
>>>>>>> in any obvious
>>>>>>>> fashion ( at least that I can see). If the F9 has a built in antenna
>>>>>>> database, that’s not
>>>>>>>> mentioned in the doc’s. Any benefit from the corrections would have
>> to
>>>>>>> be part of
>>>>>>>> post processing.
>>>>>>>>
>>>>>>>> No, that’s not the same as talking about the F9 it’s self doing X
>> ns,
>>>>>>> but it would be part
>>>>>>>> of any practical system trying to get close to 5 ns absolute
>> accuracy.
>>>>>>>> 5 ns *relative* accuracy between two F9’s? I probably could buy that
>>>> if
>>>>>>> the appropriate
>>>>>>>> one sigma / on a clear day / with the wind in the right direction
>> sort
>>>>>>> of qualifiers are
>>>>>>>> attached.
>>>>>>>>
>>>>>>>> Bob
>>>>>>>>
>>>>>>>>> On Feb 26, 2021, at 4:27 PM, John Ackermann N8UR <jra at febo.com>
>>>> wrote:
>>>>>>>>> It's interesting that they talk about the F9 receivers offering 5
>> ns
>>>>>>> absolute time accuracy.  Does anyone know of tests confirming that,
>> and
>>>>>>> what sort of care was required in the setup to get there?
>>>>>>>>> John
>>>>>>>>> ----
>>>>>>>>>
>>>>>>>>> On 2/26/21 9:34 AM, Robert LaJeunesse wrote:
>>>>>>>>>> FWIW. No detailed content, and a rather quick read. "Five key
>> trends
>>>>>>> in GPS".
>>>>>>>>>> https://www.u-blox.com/en/blogs/insights/five-key-trends-gps
>>>>>>>>>> _______________________________________________
>>>>>>>>>> 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.
>>>>>>>>> _______________________________________________
>>>>>>>>> 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.
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> 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.
>>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> 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.
>>>>>>>
>>>>>> _______________________________________________
>>>>>> 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.
>>>>> _______________________________________________
>>>>> 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.
>>>>
>>>> _______________________________________________
>>>> 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.
>>>>
>>> _______________________________________________
>>> 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.
>>
>> _______________________________________________
>> 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.
>>
> _______________________________________________
> 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.

More information about the Time-nuts_lists.febo.com mailing list