[time-nuts] Re: GPSDO/GNSSDO project: STM32G4 + u-blox ZED-F9T + TDC7200

[Greg Troxel]

John Ackermann N8UR via time-nuts <time-nuts at lists.febo.com> writes:

> I hesitate to post this because I haven't had time to go back and
> refresh thoroughly, but I'm not sure how you would apply RTK
> corrections to the F9T.
>
> I believe that the "T" models can output RTCM data so can serve as an
> RTK base station, but only the "P" models have the built-engine to
> receive corrections via UART or USB and do RTK internally.  So without
> that, I don't see how you can get corrected time directly from the
> module.

A caution that RTCM can transport two different things:
  - pseudorange corrections (with t_0 and rate): typically RTCM2
  - carrier phase reference data: typically RTCM3

The first is what as on the old USCG "differential beacons", 100 or 200
bps on roughly 200-300 kHz.

The second in my view is not really "corrections" even though it leads
to more accurate solutions; that's typically called "RTK network" and
loosely "RTK corrections".

I am not clear on the SBAS data format, but I am pretty sure it's
fundamentally pseudorange corrections (gridded model, but still you get
a range correction and probably a t_0 and range rate correction.

It's true that "P" have the code to accept RTCM3 carrier phase reference
data (along with reference station coordinate) and do RTK, which is
fundamentally a double-differenced carrier phase solution.  But one can
do that externally from raw measurements.  Maybe 5 years ago the Emlid
Reach did this with an M8T and rtklib running under an Intel Edison.
There, you'll probably be able to solve for clock error as well as XYZ,
but I do not know if one can practically get clock error and adjust for
it.

> (I'm assuming that SBAS doesn't count as "RTK" for this purpose.  I've
> never experimented with what that would do to the timing output.)

I've long heard "don't enable SBAS for timing" claims and I've never
understood the full rationale.  In theory, the observed pseudoranges,
after applying the pseudorange corrections, are then used in a
navigation solution.  Everybody believes that one gets better XYZ out of
that.  Time is trickier, as the reference stations have to steer their
clocks as they get measured pseudoranges to diff with calculated
pseudoranges (from satellite positions from the broadcast ephemeris and
their surveyed ARP, with an antenna model).  But I would expect that the
general reduction of errors would lead to a tighter solution in all 4
unknowns.

So the obvious experiment is to take two indentical receivers, one of
F9P, F9T, M8T, run them on a splitter both with SBAS off and compare
each timepulse to a cesium clock or something, to verify that they
behave the same, and then change so that one has SBAS and the other is
still off and look at that.

Can anyone explain a mechanism for the positions getting better but time
either getting worse or not changing?

Greg N1DAM


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20220806/b1408edf/attachment.sig>