[time-nuts] ublox NEO-M8T improved by insulated chamber?

Tim Lister listertim at gmail.com
Wed Nov 8 01:47:36 EST 2017

On Tue, Nov 7, 2017 at 8:39 PM, Gary E. Miller <gem at rellim.com> wrote:
> Yo Tom!
> On Tue, 7 Nov 2017 20:16:09 -0800
> "Tom Van Baak" <tvb at LeapSecond.com> wrote:
>> > Which is small compared to the published GPS time resolution
>> > (IS_GPS_200H, page 54) of 90 ns.
>> Correct. GPS performs far better than the original spec. Like the
>> Mars rovers...
> Of course, but then you are on a wing and a prayer, not engineering.
>> As a result we're now all used to ~10 ns level of performance out of
>> GPS, even in a $5 receiver.
> I'm sure we are not talking about the same thing here.  Your talking
> about GPS time?  I'm talking about UTC as output from a GPS, after it
> converted from GPS time.
> I am NOT talking about the 'performance' of GPS. What is performance?
> We talking about frequency stability, or position accuracy, or we
> talking about absolute offset from USNO UTC time?  I'm talking about
> the later. I'm talking about the spec about how close the GPS time is
> to UTC time. Your GPS converts the GPS time to UTC time depending on an
> ephemeris parameter that the GPS owners say is 90 ns (one sigma).
> Sure, you may get better, but when you are looking at subtle error sources
> that is surely one to look at.
> How exactly do you measure offset of your GPS time output to absolute
> UTC time?

That was what Tom's attached plot was showing: the measured difference
from GPS system time distributed through the satellites to UTC realized
at USNO "UTC(USNO)". This is also available (only in arrears) through
the BIPM's Circular T which also give the differences 'UTC-UTC(USNO)'.
With these two sets of offsets and some interpolation (you only get values
every 5 days in the Circular T) you can back-track from GPS time to
"true" UTC but only about 1 month after the observations.

>> Closer to ~1 ns is possible when you dig
>> into the bag-of-timing-tricks like zero-D mode, sawtooth correction,
>> antenna calibration, multi-path mitigating antennae, dual-frequency
>> receivers, external frequency references, post-processing,
>> temperature stabilization of antenna, cables, receiver, etc.
> Yes, of course, but NONE of that fixes the GPS to UTC offset problem.
> It makes the GPS time much better, but does not solve the problem that
> the GPS to UTC offset is only good to 90 ns (one sigma).  Is your GPS
> getting a better offset correction somewhere else?  Otherwis it has NO
> way to compute/calculate/devine that offset.

I've not read the spec, but presumably there must be a way to get better
accuracy and/or the system is being operated and kept to tighter tolerances
than the original written spec otherwise the quoted measured offsets between
GPS time and UTC(USNO), which are quoted in the Circular T to 0.1ns if
I remember right, would be vast overkill.

>> So the
>> industry big boys are getting sub-cm levels of positioning and sub-ns
>> levels of timing. It's all pretty cool. Some time nuts are not far
>> behind.
> I have seen cm level precision myself.  But that is unrelated to the issue
> I bring up.  The positioning depends on stable GPS time, and I agree GPS
> time is much more stable than 90 ns.  I thought the subject was UTC offsets.
>> Note also that relative timing, such as needed by a GPSDO frequency
>> standard is always much better than absolute timing, such as needed
>> by a UTC time standard. This is because many of the unknown offsets
>> (antenna, cable, receiver RF and f/w) magically cancel when used as a
>> GPSDO. This is why some GPSDO can get down to parts in 10^14th
>> frequency stability over a day.
> Yes, I 100% agree, and totally unrelated to my point.  Frequency stability
> is only loosely correlated to absolute time accuracy.  Stable !=
> accurate.
>> There's a slide I remember seeing that shows how GPS timing accuracy
>> has improved since the early days. It's page 9 (attached) of:
> I agree, GPS accuracy is great, but I am NOT talking about GPS timing,
> I am talking about UTC timing accuracy.
> I thought the problem was that the UTC time from the GPS was wandering
> on a diurnal time frame.  The GPS can be perfect to one hundred 9s,
> the GPS position can be perfect to 100 nines, but if the transmitted
> GPS time to UTC time offset is said, by the US Military, to be only 90
> ns (one sigma), then I'd listen to them when it matters.
> Time for us all to actually read the standard and argue what that means.


More information about the time-nuts mailing list