[time-nuts] U-blox F9T performance (was: U-blox teaser)

[Thomas Abbott]

Hello Time Nuts, first post from a long-time lurker.

In late 2019 I evaluated the Ublox F9T for a project. Here is a graph of
the time difference between PPS outputs of two RCB-F9T boards on the bench.
Each had a U-blox dual-band ANN-MB-01-00 antenna, they were 1 metre apart,
on ground planes, with excellent sky view.
They were set for fixed position timing mode with a 24-hour survey-in.
PPS outputs were measured with a Pendulum TIC and compensated.for sawtooth.

First the whole dataset:
[image: 2xf9t-48hours.png]
Just the second 24 hours:
[image: 2xf9t-24hours.png]
Histogram of the above - every PPS, only once the positions were fixed:
[image: 2xf9t-histogram.png]
Standard deviation of 1.5 ns. So the claim of sub- 5 ns relative timing
checks out, at least for short baselines.

A few more notes for interest:
1) this was GPS L1/L2 only
2) I set one GPS with a 100 ns "user delay" to keep the TIC readings
positive, I've removed that again here. So there's still a 5 ns fixed
offset.
3) No RTCM communication between the devices. I think it helped a bit when
turned on, but the real improvement was in having fixed positions.
4) the sawtooth TIM-TP messages were buggy (or maybe it was me), Some runs
I struggled with a few points coming out the "wrong side of the wrap", and
had to write an algorithm to eliminate suspect points or add/subtract 10 ns
from them. Some days every second point was wrong, then I was lost.
Comparing two GPSs made it harder.
5) This was mid-2019 firmware, I know there have been updates. I bricked
one and and had to send it back to the agent for a "remote glovebox
upgrade" - even they weren't given the file.
6) The receivers were pretty insensitive to large temperature fluctuations,
I tried them at 20 C and 40 C, no difference to the time delta
7) but they were sensitive to vibration. Moving one around gently lead to a
big increase in the errors.
8) The F9s came programmed with an antenna delay of 50 ns, which matched
pretty well the delay I measured on the antenna. Only one delay field for
both bands though.

I'd love to know how stable their actual output time was, and how much
common mode error. As one expert put it, "now you know that the errors in
these units are well correlated."
I never got around to running one of these against our masers or national
laboratory UTC before covid and then moving away, the best reference I
could find was the other F9T.


Thomas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 2xf9t-48hours.png
Type: image/png
Size: 66401 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20210228/ad80ac5b/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 2xf9t-24hours.png
Type: image/png
Size: 72013 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20210228/ad80ac5b/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 2xf9t-histogram.png
Type: image/png
Size: 12127 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20210228/ad80ac5b/attachment-0002.png>