Empathy List Archives

DM

Demetrios Matsakis

Fri, Oct 21, 2022 4:17 PM

I have three F9T’s, which I set up to output PPS on one port and 10 MHz on the other. All three showed periods of super-excellent consistency punctuated by several-ns jumps. I saw these on an oscilloscope, and then put two F9T’s on a Keysight counter, with 50 ohms impedance, so I could plot them. Their patterns were not the same in detail, and I’ve attached the one example. This one shows much cleaner signals than the other one I looked at. There has been some discussion of impedance on time-nuts, but this seems more like an internal frequency conversion issue.

Have any of you seen such things?

Also, it appears there is no way, with u-center2 ver 22.10.39687, to alter the PID values that go into generating these signals from GNSS. Given that their internal oscillator can’t be of high quality, I guess it’s not very important. But I’m curious.

I posed the question to U-blox a question 2 weeks ago via a form, and they haven’t answered yet. (I might not have approached them the right way, though). Anyhow, I’d like to ask if any of you know how to control F9T’s PID values?

Thanks, Demetrios

I have three F9T’s, which I set up to output PPS on one port and 10 MHz on the other. All three showed periods of super-excellent consistency punctuated by several-ns jumps. I saw these on an oscilloscope, and then put two F9T’s on a Keysight counter, with 50 ohms impedance, so I could plot them. Their patterns were not the same in detail, and I’ve attached the one example. This one shows much cleaner signals than the other one I looked at. There has been some discussion of impedance on time-nuts, but this seems more like an internal frequency conversion issue. Have any of you seen such things? Also, it appears there is no way, with u-center2 ver 22.10.39687, to alter the PID values that go into generating these signals from GNSS. Given that their internal oscillator can’t be of high quality, I guess it’s not very important. But I’m curious. I posed the question to U-blox a question 2 weeks ago via a form, and they haven’t answered yet. (I might not have approached them the right way, though). Anyhow, I’d like to ask if any of you know how to control F9T’s PID values? Thanks, Demetrios

SH

stefan.hegnauer@gmx.ch

Fri, Oct 21, 2022 6:52 PM

To the best of my knowledge there is no PID in ublox GNSS receivers for the state prediction, including the Z9T. There is a Kalman filter, though. It generates the (x,y,z,t) prediction and thus the time pulse, plus a number for the expected quantization error in the UBX-TIM-TP message due to the fixed frequency of operation (TCXO). I.e. every single hardware timepulse is off by the calculated respective quantization error. For best performance the timepulse quantization error must be taken into account.

The main influence you have on Kalman filter parameters is selecting the correct ‘dynamic platform model’ for your application, see the (F9T) integration manual for details. Yes, this may seem a bit corse.

-Stefan

To the best of my knowledge there is no PID in ublox GNSS receivers for the state prediction, including the Z9T. There is a Kalman filter, though. It generates the (x,y,z,t) prediction and thus the time pulse, plus a number for the expected quantization error in the UBX-TIM-TP message due to the fixed frequency of operation (TCXO). I.e. every single hardware timepulse is off by the calculated respective quantization error. For best performance the timepulse quantization error **must** be taken into account. The main influence you have on Kalman filter parameters is selecting the correct ‘dynamic platform model’ for your application, see the (F9T) integration manual for details. Yes, this may seem a bit corse. \-Stefan

FC

Forrest Christian (List Account)

Fri, Oct 21, 2022 9:33 PM

Your attachment didn't come through, but I think I can guess what it looks
like.

An F9T is just a GNSS receiver, although one optimized for timing. It
isn't a GPSDO. The typical architecture of these receivers is to use an
internal fixed oscillator, and all of the timing signals which come out of
the receiver will be aligned to the clock edges of that oscillator. Based
on the time pulse jitter spec of +-4ns, my guess is that the internal clock
of the F9T is likely 125Mhz.

The GNSS receiver, when outputting the 1PPS signal, will calculate whatever
clock edge is closest to the actual 1PPS alignment and will output the 1PPS
on that edge. Because it can only output to the nearest clock edge, and
the clock edges are 8ns apart (assuming a 125Mhz oscillator), what you'll
see is that as the actual 1PPS drifts in relation to the 125Mhz clock, the
1PPS will exhibit what appears to be 8ns (+-4ns) phase jumps. As I don't
know the stability of the internal oscillator I can't predict how often
these will occur.

Timing GNSS receivers know how far the pulse is off - so they also can emit
phase correction data. These messages basically say that the 1PPS they
just emitted (or about to emit) is off by a certain amount from ideal.
That way, something using the 1PPS to discipline an oscillator can apply
this correction as appropriate in their application. A google search for
"sawtooth correction" provides more data on how this is done and also more
data about what a typical GNSS receiver will output.

Note that since there is no disciplined oscillator, no PID values are
likely to be used.

On Fri, Oct 21, 2022 at 10:58 AM Demetrios Matsakis via time-nuts <
time-nuts@lists.febo.com> wrote:

I have three F9T’s, which I set up to output PPS on one port and 10 MHz on
the other. All three showed periods of super-excellent consistency
punctuated by several-ns jumps. I saw these on an oscilloscope, and then
put two F9T’s on a Keysight counter, with 50 ohms impedance, so I could
plot them. Their patterns were not the same in detail, and I’ve attached
the one example. This one shows much cleaner signals than the other one I
looked at. There has been some discussion of impedance on time-nuts, but
this seems more like an internal frequency conversion issue.

Have any of you seen such things?

Also, it appears there is no way, with u-center2 ver 22.10.39687, to alter
the PID values that go into generating these signals from GNSS. Given that
their internal oscillator can’t be of high quality, I guess it’s not very
important. But I’m curious.

I posed the question to U-blox a question 2 weeks ago via a form, and they
haven’t answered yet. (I might not have approached them the right way,
though). Anyhow, I’d like to ask if any of you know how to control F9T’s
PID values?

Thanks, Demetrios

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com

--

Forrest

Your attachment didn't come through, but I think I can guess what it looks like. An F9T is just a GNSS receiver, although one optimized for timing. It isn't a GPSDO. The typical architecture of these receivers is to use an internal fixed oscillator, and all of the timing signals which come out of the receiver will be aligned to the clock edges of that oscillator. Based on the time pulse jitter spec of +-4ns, my guess is that the internal clock of the F9T is likely 125Mhz. The GNSS receiver, when outputting the 1PPS signal, will calculate whatever clock edge is closest to the actual 1PPS alignment and will output the 1PPS on that edge. Because it can only output to the nearest clock edge, and the clock edges are 8ns apart (assuming a 125Mhz oscillator), what you'll see is that as the actual 1PPS drifts in relation to the 125Mhz clock, the 1PPS will exhibit what appears to be 8ns (+-4ns) phase jumps. As I don't know the stability of the internal oscillator I can't predict how often these will occur. Timing GNSS receivers know how far the pulse is off - so they also can emit phase correction data. These messages basically say that the 1PPS they just emitted (or about to emit) is off by a certain amount from ideal. That way, something using the 1PPS to discipline an oscillator can apply this correction as appropriate in their application. A google search for "sawtooth correction" provides more data on how this is done and also more data about what a typical GNSS receiver will output. Note that since there is no disciplined oscillator, no PID values are likely to be used. On Fri, Oct 21, 2022 at 10:58 AM Demetrios Matsakis via time-nuts < time-nuts@lists.febo.com> wrote: > I have three F9T’s, which I set up to output PPS on one port and 10 MHz on > the other. All three showed periods of super-excellent consistency > punctuated by several-ns jumps. I saw these on an oscilloscope, and then > put two F9T’s on a Keysight counter, with 50 ohms impedance, so I could > plot them. Their patterns were not the same in detail, and I’ve attached > the one example. This one shows much cleaner signals than the other one I > looked at. There has been some discussion of impedance on time-nuts, but > this seems more like an internal frequency conversion issue. > > Have any of you seen such things? > > Also, it appears there is no way, with u-center2 ver 22.10.39687, to alter > the PID values that go into generating these signals from GNSS. Given that > their internal oscillator can’t be of high quality, I guess it’s not very > important. But I’m curious. > > I posed the question to U-blox a question 2 weeks ago via a form, and they > haven’t answered yet. (I might not have approached them the right way, > though). Anyhow, I’d like to ask if any of you know how to control F9T’s > PID values? > > Thanks, Demetrios > > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe send an email to time-nuts-leave@lists.febo.com -- - Forrest

time-nuts@lists.febo.com

Ublox F9T tick to phase