Using GPS 1pps to steer PRS-10

[floobydust]

I can't find the family or product name for example old board Trimble 39818-C. Just a surfer icon.
Point is, our embedded S/W was not expecting a missing pulse or hiccup, or pulses slowly drifting out. That caused timesync problems that were hard to track down, and I flamed Trimble for losing fix/timesync yet still outputting 1PPS.

[ZigmundRat]

That’s a Trimble Ace-lll. Used in Datum/Symmetricom TS2100 and similar timing devices. I think from around 2000 +/- a few years.

[texaspyro]

It could also be a SV6...

[tkamiya]

Texaspyro,

Was that you who made a fan-out board for PRS-10?  Do you have any left?

[texaspyro]

Was that you who made a fan-out board for PRS-10?  Do you have any left?

I have on laid out and all the parts, but have not built any.   I was tied up on a couple of projects for the last year.  Also both my reasonably priced local PCB assembly shops have disappeared.  A friend is building a pick and place machine and when (if?) that is done I'll build them myself.

[tkamiya]

Would you consider selling them as kits?  I'm very interested.  If so, PM me, please.

[DaJMasta]

I had something like that, nothing fancy, but it's gotten the job done for me:
https://workspace.circuitmaker.com/Projects/Details/DaJMasta/Rubidium-monitor-and-glue

Output is 10MHz sine but is a little limited in level with the quad driver ICs and them not being rail-to-rail, but it's done the job for all my equipment.  Used a Mean Well supply for the power input and put the thing in an aluminum box.

[tkamiya]

Nothing fancy?  Maybe so, but how fancy can you get on frequency standard?  I studied your design for few hours today.  I also have a "Mean-Well" power supply.  It always cracks me up when I say its name.....  "So sorry your project blew up....  we meant well...."   hahahaha.  I think something got lost in translation.  (I'm Asian.  I can make fun of our own!)

On my first stage, my primary goal is to protect PRS10 and have 1V output.  (which seems to be a semi-standard for clock input)

All I want is nothing as elaborate as yours.  PRS-10 puts out 0.5mVrms.  I want 1Vrms.  That requires 6db gain.  But that also means Vp-p at output will be 2.8Vpp.  I never knew this will be SO hard.....

[DaJMasta]

If you go the opamp route, you can probably manage on a single 5V supply, or if your amp is rail-to-rail the 3V supply for a GPS module or similar could do.  2.8Vpp at 10MHz needs 176+ V/us slew rate, then you need enough GBP to get your 6dB of amplification, so, 50MHz GBP would be plenty.  Not unreasonable demands for an opamp, and 75 ohm line drivers can usually manage that and cost fairly little if you want that amplifier instead.  I included a distribution amp so I didn't need an external one, then basic monitoring and a header for a GPS module - not much in terms of active parts, but some extra regulators/passives and I tried to pick some stuff to reduce any potential EMI issues.

[texaspyro]

Here is a quick look at the F9P 1PPS output.   It varies around +/-8 nsec which agrees with the sawtooth error message.

[SoundTech-LG]

Would you consider selling them as kits?  I'm very interested.  If so, PM me, please.

Also very interested. Kits would be great!

[texaspyro]

Would you consider selling them as kits?  I'm very interested.  If so, PM me, please.

Also very interested. Kits would be great!

I don't have any circuit boards yet and have not tested the design...

[tkamiya]

I with Stanford Research publishes circuit diagrams.  On PRS-10 manuals, they make references all over the place on circuit diagrams that are included (on page X) but there aren't actually any.  The interface board is also explained and refers to a circuit diagram that is no where to be found.  Very weird. 

I am envisioning making my own.  Take 10MHz output and split into two, then amplify.  1 pps in and out buffered, and everything else buffered.  It can't be all that difficult can it?  What I am not sure of is how to handle different configurations possible by jumpers.  I always thought there should be at least two output, so one can go to another circuit and the other can go to test instrument.

I am not familiar at all with new ways of making printed circuit boards though.  I'm from tapes and "rounds" generation. 

[texaspyro]

Try this...

[tkamiya]

No, you try it!  I'll hold your beer.

WOOOHOOO!  Thank you, thank you, thank you!!!!!!

ps.  When I see your forum name, I still read it as "Tax Pro."  After all, it's a tax season.

[tkamiya]

I am having second thoughts about using GPS 1 pps to phase lock a PRS-10.

It took me 3 days to completely stabilize PRS-10 and calibrate it to 3 x 10^-11.  Reference source is GPSTM running continuously in my lab.  The rubidium source just do not drift!  Take it offline and on, wait 3 days and it's back to how it was before.  Physical lock alone gives me about 2 order of magnitude worse. 

Two surprises....
Physical lock alone is not that great.
It takes 3 days to stabilize.
Once stable, it's so darn accurate and precise.

(ok, that's three surprises)

Add in the steering by GPS, I have no way to validate the result.  That bothers me.  Accurate by design but there are so many unknowns.  Rubidium will be pushed and pulled by GPS 1 pps, internal temperature sensor, and whatever else compensation taking place.  It's nice when it all works together but if  and when it doesn't, I have no way of knowing.

Hum...  I'm starting to think, it's not a good idea to try to DIY the most accurate thing in my lab.   

[DaJMasta]

If you want to validate it, you need a bunch of other Rb sources or go up to a Cesium beam.......


Don't think the GPS can offer better short term (within a day or two) performance, as the timing from it is only so precise anyways, you need a pretty long integration time before the noise in the GPS PPS signal is low enough to use to discipline a source with Rb stability, and the PRS-10 module does know that.  I wouldn't DIY the science package and module yourself unless you had another to compare it to, but it's essentially its own closed loop for the disciplining, so even if you DIY the surrounding parts, I don't know if that has much bearing on its accuracy provided you can supply it with the right voltage/etc.

Worth mentioning that you can do some long term analysis - compare the PPS input to the PRS-10's PPS output in the long term (a week or more) and see if you can see the fluctuations in the signals relative to each other (mostly caused by the PPS input after it's settled) and verify that they aren't drifting apart (which would mean the PRS-10 was successful in its disciplining).

But it's true, Rubidium sources are pretty good on their own.  Had a SA with an internal one that I calibrated to my GPSDO using a PRS-10 and after a couple months powered down, in less than 10 minutes from power on it was less than 1Hz out at 1GHz input locked to the GPSDO.

[metrologist]

Anecdote: I'm glad you said 1GHz because I was pretty happy with a decent 10Mhz OCXO powered up 1.2 Hz off my GPSDO, after more than 5 years on the shelf. It was calibrated somewhere else before I got it. I was amazed it was so close.

[FriedLogic]

It takes 3 days to stabilize.
Once stable, it's so darn accurate and precise.

You need a bit of patience with Rb oscillators. Some of the newer ones do seem to settle quite well in a few days, but others and most older ones can take weeks at least.

There are assorted PIC based 10MHz to 1PPS etc. dividers on LeapSecond.com:
http://www.leapsecond.com/pic/picdiv.htm

Using 1PPS makes it easy to check the long term drift. I was thinking of doing a multiplexer to allow a number of 1PPS inputs to one counter, but never quite got around to it.

If you have a few rubidiums it would be worth running them all for a while to see how well they behave. Some will have better ageing, lower temperature sensitivity, etc. You will also have one as a reference when you adjust another.

They perform better when run continuously.

The environment is a big issue for most rubidiums - particularly temperature, but often air pressure too. Ageing may settle to 1E-13 per day, but temperature sensitivity might be 2E-12 per degree centigrade, and pressure sensitivity near 1E-13 per millibar.

One slight issue with the PRS10 is that all the frequency control is done by a DAC which gives at best about 1E-12 resolution.

What to do all depends on the performance you want to get. It's a rather deep rabbit hole....

[tkamiya]

I now have PRS-10 running with 1 PPS from a GPS, and I am monitoring Rb status with LadyHeather. 

The GPS unit I'm currently using is rather unstable and it loses its lock too often.  Also, Rb is just sitting on a heatsink, so heat transfer is not optimum.  I ran it like this for an hour and everything looks reasonable.  I have a few GPS board heading my way, so I will have to redo this test after new boards arrive.  On top of that, power supplies are not ideal.  But proof of concept is working.  Any longer run to access performance will have to wait until heat sink is attached and other issues resolved.

Since the GPS is a part of GPSDO, I was comparing OCXO output with Rb output.  They are absolutely in sync.  No relative movement at all.  Which actually begs the question, what is this complication really for??  Maybe my calibration error will be large enough when accumulated that GPS 1pps will pull it correct every now and then?  Also, PRS-10 manual talks about PPS filtering with a side note that it is important particularly for GPS and needing to be few hours at minimum.  There will be some tuning necessary.

Does anyone have Cs time source I can borrow? 

[texaspyro]

The WS keyboard command (or clicking on the second column of data at the top of the screen) will give a much better screen dump.  Also clicking on the status info on the left side of the screen will zoom it to show more data (the only way to dump a zoomed screen is to click where the second column of data is (or would be) at the top of the screen since any keystroke will un-zoom the screen).

Disciplining a Rb needs very long time constants (like 10,000 seconds) and it can take like 4 time constants for things to fully stabilize.

[DaJMasta]

While at heatsink may seem appealing, it's probably not helpful.  the PRS-10 has a heater to keep the science package at the right temperature, so sinking it when you're otherwise in fairly open air, may just be making it heat more often to maintain it.

I certainly wouldn't put it in an insulated box, but putting it an additional metal enclosure without active cooling is probably better for power consumption.

[metrologist]

Since the GPS is a part of GPSDO, I was comparing OCXO output with Rb output.  They are absolutely in sync.  No relative movement at all.  Which actually begs the question, what is this complication really for??  Maybe my calibration error will be large enough when accumulated that GPS 1pps will pull it correct every now and then?  Also, PRS-10 manual talks about PPS filtering with a side note that it is important particularly for GPS and needing to be few hours at minimum.  There will be some tuning necessary.

Does anyone have Cs time source I can borrow? 

I find this a challenge of patience. How are you comparing and over what time period? I'd think you'd want to watch it for a month, after a month's settling...

[tkamiya]

TexasPyro, I feel stupid.  I had it working and accidentally lost the command line setup.

What is the flag for LH startup for PRS-10?  I know I found it somewhere..... but it's GONE!

[texaspyro]

TexasPyro, I feel stupid.  I had it working and accidentally lost the command line setup.

What is the flag for LH startup for PRS-10?  I know I found it somewhere..... but it's GONE!

/rxpr

If you use the /? command line option or the ? keyboard command you will get a list of all the command line options.

I also highly recommend reading the comments at the start of the heather.cpp file.  Those are the documentation of how to use the program.