Two GPSDO to get more accuracy?

[FriedMule]

I had one GPSDO and I imagine I am not the first one to get one more? :-)

To no surprise for me do they not agree upon what 10MHz exactly is, but still does this give the question: "which one is most accurate?"
I remember having read an article translated from Russian, about combining two GPSDO in some way to let them help each other to gain 2-3 orders of magnitude better accuracy, but I have not been able to find it again.

Is there any truth to this? What is this configuration called?

Thanks in advantage. :-)

[trobbins]

I'd recommend wading through the many GPSDO threads on this forum and extracting out pertinent posts that discuss GPSDO accuracy to get a better awareness of how the accuracy of one GPSDO can be assessed, especially those GPSDO that are aiming for much better than 1ppb, and how the various parts of a GPSDO can influence accuracy.

I'd suggest that there have been quite a few methods used in GPSDO's, and each has its benefits/limitations, and beyond that then the GPS receiver can also be a significant influence - especially for attainment of high accuracy.  So your query may well be too open-ended, as you haven't identified the GPSDO you use.

[tkamiya]

I have more than half a dozen GPSDO in my lab.  If I average each for 24 hours, they are accurate and precise, but instantaneous reading is all over the place.  With two, you can potentially average.  With three or more, you can do N-cornered hat.  But verifying each method is actually working in home lab is entirely different matter.  You'll need a reference guaranteed to be accurate and precise than DUT.  Still, I think about putting dozen of them and do some kind of scheme, but I doubt there is any practical value in such attempt.

[FriedMule]

Long time is fairly precise, down to a few nanoseconds, but my biggest problem is when I am trying to measure a crystal and do not have days or weeks to measure.
How can I let say be more sure a crystal is accurate in 1-2 hours of measurement time?

Would a combination with another type of oscillator crystal, I mean to remember that this fellow had combined 3-4 GPSDO's to a cesium but said 2 would be fine in most cases. I think he measured down to 13 digits.

Is it right that he used two long term precise crystals and connected them to a short term accurate crystal?

[edpalmer42]

Long time is fairly precise, down to a few nanoseconds, but my biggest problem is when I am trying to measure a crystal and do not have days or weeks to measure.
How can I let say be more sure a crystal is accurate in 1-2 hours of measurement time?

Put some numbers on this.  What level of accuracy do you require?  Is 'accuracy' the correct word to define your requirements?

Would a combination with another type of oscillator crystal, I mean to remember that this fellow had combined 3-4 GPSDO's to a cesium but said 2 would be fine in most cases. I think he measured down to 13 digits.

Is it right that he used two long term precise crystals and connected them to a short term accurate crystal?

Do you remember where you saw this?  We need more info to comment intelligently.

The only reason to use GPS is if you need to confirm absolute frequency accuracy over a long period.  As you've found, GPS isn't particularly good for short-term stability.  A really good quartz oscillator is better for short-term stability, but finding one is not a trivial task.  In fact, a good Rb is probably better than most GPSDOs for short-term stability.

But expecting to make a high resolution measurement of a crystal (?) in an hour or two is almost doomed to failure.  If you meant oscillator instead of crystal, you'll need to start with an oscillator that's worthy of consideration and then age it for a month or three to get the aging down to something approaching its long-term potential.  If you actually meant a bare crystal, it's even harder.  The performance of a bare crystal is heavily influenced by the oscillator and heater circuits that it's connected to.  You can't really measure the crystal in isolation.  And then you have to age it for a month or three before you can measure it.

Ed

[thinkfat]

but still does this give the question: "which one is most accurate?"

Simple answer: The one that you trust the most.
Right answer: The one that has passed the performance verification in a trusted laboratory, with the best result.
Timenuts answer: No one! Only H-Maser is accurate.
Long answer: Into ordinary home laboratory you can't measure accuracy of GPSDO. You only can try measure stability into homelab, but it need timesource with better stability.

As usual, it depends entirely on the precision you're interested in, but if you want to measure the stability of a GPSDO, a good counter with a stable time base is enough to judge it.

A HP53131A with a Rb oscillator as time base is stable enough to judge the stability of GPSDO over 10s of thousands of seconds if not more. It will not tell you a lot about the second-to-second stability, but for 10s or 100s of seconds you can get a good idea and you'll be able to compare your devices. Even with a good OCXO timebase it can do that.

You'll have to shed some money, but above setup is entirely within a hobbyists budget; or at least it was, before the pandemic hit and drove prices up.

[David Hess]

Long time is fairly precise, down to a few nanoseconds, but my biggest problem is when I am trying to measure a crystal and do not have days or weeks to measure.
How can I let say be more sure a crystal is accurate in 1-2 hours of measurement time?

Would a combination with another type of oscillator crystal, I mean to remember that this fellow had combined 3-4 GPSDO's to a cesium but said 2 would be fine in most cases. I think he measured down to 13 digits.

Is it right that he used two long term precise crystals and connected them to a short term accurate crystal?

Use a timing GPS to discipline an OCXO crystal oscillator over a period of days and then you can make the measurements in seconds to minutes, or discipline a rubidium oscillator.

[guenthert]

Now with a bit of fantasy and out-of-the-box thinking, I might be able to imagine some applications for a high stability oscillator, but honest question, what do you need an extremely accurate clock for?  How accurate can you measure the position of the GPS antenna? (if you care for local time).  And the length of wire connecting your GPSDO and said antenna?  And what time do you actually care for -- GPS time, UTC, UT ... https://stjarnhimlen.se/comp/time.html

[David Hess]

Now with a bit of fantasy and out-of-the-box thinking, I might be able to imagine some applications for a high stability oscillator, but honest question, what do you need an extremely accurate clock for?  How accurate can you measure the position of the GPS antenna? (if you care for local time).  And the length of wire connecting your GPSDO and said antenna?  And what time do you actually care for -- GPS time, UTC, UT ... https://stjarnhimlen.se/comp/time.html

Radio interferometers require an accurate clock.

[martinr33]

David is dead on. Get a really good ovenized oscillator driving your timer/counter. Morion, Oscilloquartz and Trimble are all good in my experience. I have one in a 53132A

Then, adjust it until your GSPDO reads with balanced deviation on either side of 10MHz. This process takes a few hours to a couple of days.

I have a suspicion that a high dollar GPSDO (EG Fluke) will do this work for you. Seems that the cheap ones could, but mine don't There's a new crop of maybe more sophisticated GPSDOs showing up from China, they might be better.   

[Zenwizard]

GPSDO's are inherently accurate with long term averaging. It is the instantaneous to short term that is the issue. This is based one the update rate of the oscillator. Their short term stability is usually good for a few hours even with the antenna unplugged. the best thing for GPSDO is get everything good and warmed also stable. Then unplug the GPSDO and let it run in hold over for measurements This is fin for short term measurements even in to hours but not days.

For the best long term stability that I have found at what I would consider reasonable cost. (all things are relative) is GPSDO 1pps in to Rb unit with a LONG integration 12 hours or longer up to 48. This keeps the Rb from getting yanked around by the 1pps to much. to squeeze out more stability than this I would need to go to caesium which would increase the already descent cost by a factor of about 10X for reference this is how some calibration labs do the 10 Mhz ref GPSDO in to Rb with a 48 Hour warm up time.

Zen

[guenthert]

Now with a bit of fantasy and out-of-the-box thinking, I might be able to imagine some applications for a high stability oscillator, but honest question, what do you need an extremely accurate clock for?  How accurate can you measure the position of the GPS antenna? (if you care for local time).  And the length of wire connecting your GPSDO and said antenna?  And what time do you actually care for -- GPS time, UTC, UT ... https://stjarnhimlen.se/comp/time.html

Radio interferometers require an accurate clock.

Radio interferometers as used in radio telescopes?  I would have thought that at the frequencies of interest (400MHz up to a few GHz), clock accuracies needed for long distance interferometry (VLBI) would be unfeasible and instead they use local clocks and calibrate the images after the fact while on smaller scale they use optical signals distributed via fibers for synchonization.

[David Hess]

Radio interferometers as used in radio telescopes?  I would have thought that at the frequencies of interest (400MHz up to a few GHz), clock accuracies needed for long distance interferometry (VLBI) would be unfeasible and instead they use local clocks and calibrate the images after the fact while on smaller scale they use optical signals distributed via fibers for synchonization.

As far as I know the processing still requires the different local clocks to be timestamped.

I do not know if it has changed, but cell towers relied on GPSDOs for synchronization for a long time which is why so many rubidium oscillators were showing up on the used market.

[mendip_discovery]

Maybe some of the info in this document might help,

https://www.researchgate.net/publication/320798359_Evaluating_the_Frequency_and_Time_Uncertainty_of_GPS_Disciplined_Oscillators_and_Clocks

[4thDoctorWhoFan]

I now have two different GPSDO units and they agree with each other down to 1mhz.  This is good enough for me.

One is the BG7TBL type Chinese made GPSDO with the 2 line blue LCD screen which I bought about 3 years ago and the other is the Leo Bodnar made unit with the dual output which I have had for about 8 months.  I leave them on 24/7 and as stated previous, they agree with each other down to 1 milli hertz.

I only purchased the Leo Bodnar model for a sanity check and I'm glad I did. 

[Zenwizard]

The Leo is also good as the output can be adjusted, it is not just 10Mhz it can be changed in software over a fairly wide range.

Zen

[MIS42N]

Long time is fairly precise, down to a few nanoseconds, but my biggest problem is when I am trying to measure a crystal and do not have days or weeks to measure.
How can I let say be more sure a crystal is accurate in 1-2 hours of measurement time?

A good GPSDO will be locked to the GPS - i.e. if it has a 10MHz output, it should *on average* produce 10,000,000 cycles per second. However, in any one second it may produce a bit less or a bit more. For argument sake, say it is lagging or leading the average by 2 cycles at the start of an hour, and is lagging or leading or leading by 2 cycles at the end. Then in the hour it has produced 36,000,000,000 cycles ±2 cycles. The maximum error in the number of cycles is 4, so the maximum error is a little more than 1 part in 10^10.

I don't know of any hobby application that would need to determine the accuracy of a standalone oscillator better than this. What did the OP have in mind?