Using GPS 1pps to steer PRS-10

[tkamiya]

I am having a rather large collection of Rb time bases.  One of which, PRS-10 can take a 1 PPS INPUT from another device and sync its own time base.  I also  have a Trimble CM-3 GPS board with 1 PPS output.  Seems to me, I can join them together and create a GPSDO with Rb which in turn sync oven Crystal (in PRS10).  As far as what I understand from going over manuals, this is possible.  PRS-10 manual says time constant of this "steering" is rather long at 10 hours or so to ride over jitters and jumps. 

Is it?  Have anyone done it?
Trimble CM3 being rather old, is there newer with supposedly better performance? 
Wouldn't this be like continuous and automated calibration of Rb steered crystal oscillator wouldn't it?

[edpalmer42]

Both PRS-10 and X-72 Rb standards have the 1 PPS input, but it's an option.  None of the few units that I have are equipped with the option.  But you would think that it would function just as you described.   

Ed

[DaJMasta]

I thought the PRS10 had PPS disciplining standard in the configuration.... you feed it the signal and it will trim its performance to match, outputting a PPS signal of its own which will converge on the input signal.

There are some PRS10 options, but it was my understanding that it was swapping resistors on one internal board for alternate configurations, but I don't think there's any additional hardware required.  It's got an internal micro and already uses the Rb source to discipline the internal OCXO.


I've got one going in my source and it's worked just fine hooked up to a Trimble Resolution SMT.  Either because of the architecture of the GPS module or the relative insensitivity of the antenna input compared to modern modules and my mediocre antenna/placement, I will eventually lose the GPS PPS signal after a month or so of uptime, but the PRS10 just keeps on chugging.  Feeding it 24V and the PPS signal and monitoring the serial output.... which is functional but in my configuration I get a fair amount of desync, so it's not usable without multiple attempts.  My version is here, but the basics are just feeding it power and PPS and it will do the rest, very straightforward to integrate.  https://circuitmaker.com/Projects/Details/DaJMasta/Rubidium-monitor-and-glue/

[edpalmer42]

DaJMasta, you're correct!  I don't know how I missed that.

I guess I never seriously thought about disciplining a Rb because the drift is so low that I never saw the point.  The only time I thought it might be useful is if your GPS reception was really crappy and your GPSDO spent a lot of time in holdover.  Then, a disciplined Rb would start to make sense.

Ed

[DaJMasta]

With the architecture of the PRS10, the GPS is just ensures it doesn't drift at all, even when it's a very low drift option.  I think their algorithm acknowledges that the PPS signal is noisy and that the oscillator is very stable on its own, so the settling time for optimum performance on the PRS10 is about a week.  Gives you excellent holdover (I pulled my unit from a Stratum 1 time server), the very clean output of the OCXO, and ensures that you don't have to adjust the Rb source in the long term, at least, until it drifts far enough that it gets outside of the self adjustment range at the end of its lifetime.

[edpalmer42]

With the architecture of the PRS10, the GPS is just ensures it doesn't drift at all, even when it's a very low drift option.  I think their algorithm acknowledges that the PPS signal is noisy and that the oscillator is very stable on its own, so the settling time for optimum performance on the PRS10 is about a week.  Gives you excellent holdover (I pulled my unit from a Stratum 1 time server), the very clean output of the OCXO, and ensures that you don't have to adjust the Rb source in the long term, at least, until it drifts far enough that it gets outside of the self adjustment range at the end of its lifetime.

Agreed.  But remember that SRS specs the annual drift at 5e-10, i.e. 0.005 Hz.  Seriously, 5 milliHertz per year.  And if that's a conservative spec the reality could be even better.  And that's for the standard unit, not the 'low drift' version.  I guess if a GPSRb fell into my lap I'd use it, but I just don't have any need for that level of accuracy.  I use a stand-alone Rb as my house standard and occasionally compare it to GPS.

Ed

[tkamiya]

I am having difficulty distinguishing which is drifting, my GPSDO with Oven Xtal or Rubidium.  But drift is so small, it's non issue either way.

We have came long way from using HU-6 xtal as standard!

[RobBarter]

annual drift at 5e-10, i.e. 0.005 Hz.  Seriously, 5 milliHertz per year.

If any one would like to donate their overspec'd GPS-Rb unit to a good cause, please send them to me.  I'm looking at the viability of an Amateur Radio Astronomy Interferometer at 1420MHz.  By my calcs I need a drift of better than 0.05 milliHertz per year, at both scopes. e.g. 1/10 of a wavelength over 15mins.  (My back of a fag-packet calcs could be wrong).

[tkamiya]

I want to make one of mine a GPSDO. 

I know this particular model has a very low drift rate but initial tuning is rather tricky.  Also, having GPS steering will give me a piece of mind.

Question is, then, what GPS should I use.  I currently have GPSTM by Nortel, which is a close cousin of Thunderbolt.  I also have Trimble but it is an old board.  I see plenty of modules but they are all consumer grade.  I would think PPS functionality is really an afterthought. 

Any recommendations? 

[tkamiya]

https://www.u-blox.com/en/product/rcb-f9t-timing-board

This one looks new and made for timing use.

[cdev]

How much $$$?

Do you know?

Ublox's own dev boards are often very expensive for what they provide. Unless you have some need for some feature only they provide in their new multi-gnss system lineup. Seems great for navigation, multi-gnss systems - especially different bands offers very real advantages there - but what does it offer for timing besides having multiple sources of time which may not always agree?

Maybe I'm missing something though!? Lots of their new features I don't know much about.

[texaspyro]

https://www.u-blox.com/en/product/rcb-f9t-timing-board

This one looks new and made for timing use.

It does not yet exist.  If it is like the F9P it will be at least $200.

The F9P sawtooth error seems to be in the +/- 4 ns range.

[cdev]

Maybe I am missing something,

Couldn't you just use a good GPSDO?
Maybe I am missing something, though, again.

I suppose some observatories might be underground?

Monitoring for quarks or muons or whatever particles - I mean the ones that go right through the planet-

annual drift at 5e-10, i.e. 0.005 Hz.  Seriously, 5 milliHertz per year.

If any one would like to donate their overspec'd GPS-Rb unit to a good cause, please send them to me.  I'm looking at the viability of an Amateur Radio Astronomy Interferometer at 1420MHz.  By my calcs I need a drift of better than 0.05 milliHertz per year, at both scopes. e.g. 1/10 of a wavelength over 15mins.  (My back of a fag-packet calcs could be wrong).

[cdev]

Thats very good, how do they get it that low?

It does not yet exist.  If it is like the F9P it will be at least $200.

The F9P sawtooth error seems to be in the +/- 4 ns range.

[texaspyro]

Thats very good, how do they get it that low?

It does not yet exist.  If it is like the F9P it will be at least $200.

The F9P sawtooth error seems to be in the +/- 4 ns range.

They use a fast clock and Unicorn tears.

The Furuno GT87 has sawtooth in the +/-1ns range.   The measured jitter on the PPS output is around 5 ns.

[DaJMasta]

For what it's worth, I've measured several ubox timing modules (LEA 6T, 5T, M8T) and their PPS performance was all basically the same.... even as the non timing variants (7N, for example).

Perhaps they're doing something new on the newest version, or maybe I've overlooked some need-to-enable software feature, but I get the impression that the T is primarily for the use of their internal oscillator disciplining and output, not the PPS.

[edpalmer42]

For what it's worth, I've measured several ubox timing modules (LEA 6T, 5T, M8T) and their PPS performance was all basically the same.... even as the non timing variants (7N, for example).

What did you measure and how did you measure it?  The average output frequency will be identical.  Differences between navigation and timing receivers should show up in things like performance under challenging conditions (multipath, poor signal levels, etc.), maximum deviation of the 1 PPS from nominal, overall jitter, etc.  I've never had a chance to make that comparison.

Historically, the 1 PPS output of a navigation receiver was provided on a 'best effort' basis.  Some receivers listed the error as 'up to 1 microsecond'.  Perhaps we're reaching the point where there are so many processing cycles available that there's no need to optimize the code for timing vs. navigation.

Ed

[DaJMasta]

Fluke PM6690 12 digit counter referenced to a PRS10 disciplined by a Resolution SMT stabilized for at least a week.

Measured period variation over >12 hours per unit tested, stabilized for at least a day.  Yes, they all converged towards a mean of exactly 1s as they should, but all the ublox modules had similar peak to peak variation, adev, and minimum adjustment step size.

Am traveling, but I took pictures of the screen that I can probably dig up again.  The PM6690 defaults to graphing and the visualizations on the small time scale and the large time scale looked very similar between all ublox modules I tested.

[FriedLogic]

https://www.u-blox.com/en/product/rcb-f9t-timing-board

This one looks new and made for timing use.

It does not yet exist.  If it is like the F9P it will be at least $200.

The F9P sawtooth error seems to be in the +/- 4 ns range.

Did you test how accurate the timing pulse on the F9P is compared to a 6T or 8T/8N, or if there is any hint of ionospheric corrections to it?

[tkamiya]

I have this unit by Trimble.  I know it's old but I don't know how old.  Would something like this still provide respectable performance to steer SRS-10?  I can also pull 1 PPS from GPSTM Nortel.  Would that be better?  SRS-10 being very stable unit, I don't want to make it worse by pulling it with something inferior.  (yes, I know about using long time constant to ride out the bumps)

[DaJMasta]

Close to 20 years, it seems?  Don't know it myself, but seems to be a Trimble SVeeSix CM-3, and apparently it interfaces with Lady Heather.  Only 6 satellite tracking and I see no real sensitivity specification... so you probably need a good outdoor antenna to get a decent signal with it.

I presume the requirements are actually pretty low, since it's long term stability that's needed for steering and probably a lot of the shorter term noise is ignored by the PRS10's internal disciplining algorithm (this is actually configurable, if I remember right).


That said, GPS modules are REALLY cheap, and going with a modern module, even if it's not timing specific so long as it has a real PPS output, the sensitivity, time to lock, and power consumption should all be much improved.

[texaspyro]

https://www.u-blox.com/en/product/rcb-f9t-timing-board

This one looks new and made for timing use.

It does not yet exist.  If it is like the F9P it will be at least $200.

The F9P sawtooth error seems to be in the +/- 4 ns range.

Did you test how accurate the timing pulse on the F9P is compared to a 6T or 8T/8N, or if there is any hint of ionospheric corrections to it?

I have not tested that yet...

[metrologist]

I have this unit by Trimble.

If your scope triggers on the pps, you can see the pps jitter on the second channel displaying your rB output.

[floobydust]

I had problems with old Trimble GPS modules still generating 1PPS despite the loss (dropout) of enough signal/satellites for time/position fix.
Sometimes it would stop making 1PPS after a few seconds, other times it would keep going.
It was a Trimble firmware issue, to cover up satellite dropouts the MCU generated the 1PPS and that S/W PLL drifted. It caused me grief in a timesync product line.

Check your module's behavior with loss of signal, and use a very good antenna if you want 10 hours of no issues.

[SoundTech-LG]

I had problems with old Trimble GPS modules still generating 1PPS despite the loss (dropout) of enough signal/satellites for time/position fix.
Sometimes it would stop making 1PPS after a few seconds, other times it would keep going.
It was a Trimble firmware issue, to cover up satellite dropouts the MCU generated the 1PPS and that S/W PLL drifted. It caused me grief in a timesync product line.

Check your module's behavior with loss of signal, and use a very good antenna if you want 10 hours of no issues.

Can you be any more specific about "old" Trimble modules? Resolution T ? Resolution SMT?
The Ublox M8T seems far superior to those.