[time-nuts] how many seconds out does GPS discipline being to improve Rubidium stability?

Bob Camp kb8tq at n1k.org
Sun Feb 12 13:53:14 EST 2017


Hi

You likely need to do this with something like a TimePod to get the resolution. Setting up 
three of them (or anything similar) is going to cost you way more than the Rb.

If you have a 100 ps counter, you will get 1x10^-10 at 1 second for ADEV data. That will
go to 1x10^-11 at 10 seconds and 1x10^-12 at 100 seconds. You only hit 1x10^-13 at
1,000 seconds. Again it’s a “resolution 5X better than measurement” sort of thing. You
don’t want to read the floor of your test gear, you want to see the device under test. If
you have a 20 ps counter the numbers above would be your 5X limits rather than the floor. 

Again, this gets back to the “what do you want to do?” and “how short is your shortest tau?” 
questions. In the actual use case, there may be no interest at all in Tau’s shorter than 
1,000 seconds. It’s also possible that the OCXO’s have an expected floor at 1x10^-10 from
1 second to 1,000 seconds. There are lots of variables. 

Bob

> On Feb 12, 2017, at 7:12 AM, Peter Reilley <preilley_454 at comcast.net> wrote:
> 
> Could you do all three tests in parallel?   One unit under test driving three counters.
> Each counter using a different reference signal, one on a OCXO, one on a rubidium,
> and one on a GPS disciplined oscillator.   At each point in time during the test simply
> choose the one that gives the best ADEV?
> 
> Pete.
> 
> On 2/11/2017 6:52 PM, Bob Camp wrote:
>> Hi
>> 
>> Using ADEV as an example (the other stuff will have it’s own curves, but the result is the same):
>> 
>> A typical Rb should have a stability at short tau that goes as 1/ square root(Tau). If you are at 2x10^-11 at 1 second, you
>> will be at 2x10^-12 at 100 seconds and 2x10^-13 at 1,000 seconds. Somewhere in the parts in 10^-13 that relation will
>> start to diverge from reality.
>> 
>> A fairly normal low frequency OCXO has a stability that is fairly flat with tau in the 1 to 100 second range. If they have been
>> on power constantly that “flat zone" may extend to 1,000 seconds. Floors should be in the low parts in 10^-12 to mid parts
>> in 10^-13 range.
>> 
>> A good OCXO *may* beat a normal Rb at 1,000 seconds. That may or may not be an issue in your case. It depends a lot
>> on what you are trying to do.
>> 
>> Simple solutions:
>> 
>> 1) Run something better than an Rb. A hydrogen maser is one alternative (simple if you don’t have to pay for it).
>> 2) Do all your measurements as three corner hats. You run two references and one DUT into gear that will do that sort of test.
>> 3) Segment the measurements and use carefully selected references for those ranges.
>> 
>> None of those are actually simple. Number 3 sounds cool until you realize that you are switching test setups around a lot and
>> the devices you are using still need a setup like 2 to figure out which ones to use.
>> 
>> So do you need a GPS? What are the limits on your MTIE tests? (MTIE on an OCXO is highly dependent on several
>> things so there is no simple number there). A very normal quartz based GPSDO might be a fine reference for your test.
>> 
>> How much shorter are the other tests? Is ADEV at 1,000 seconds even of interest? If the answer is < 1,000 seconds a
>> Rb may not do you much good at all.
>> 
>> Lots of twists and turns.
>> 
>> Bob
>> 
>> 
>> 
>> 
>> 
>>> On Feb 11, 2017, at 5:52 PM, gkk gb <modjklist at comcast.net> wrote:
>>> 
>>> Thanks Bob,
>>> 
>>> 
>>> 
>>> I should clarify the MTIE measurement extends 100000 seconds (the others are less time). Is it a reasonable question to ask if GPS is needed? Or are there other variables that are involved?
>>> 
>>> 
>>> 
>>> Good point about the temperature stability, I hadn't considered that. Can I place in a temperature chamber to provide a better thermal environment, or does that cause other issues (vibration from blowers, EMI noise, etc.)? Other ways to mitigate temperature changes?
>>> 
>>> 
>>> 
>>> It seems a Rubidium is good after a timescale of 100 s. What do people do below 100 s to characterize quartz oscillators. Do they simply try to find the most stable parts they can afford and break the x-axis (tau) into two regions using difference references for each? If so, are there generally accepted "gold" standards anyone can recommend for crystal products with the best stability to use as a reference between 0.1 and 100 seconds, for example?
>>> 
>>> 
>>> 
>>> On February 11, 2017 at 6:29 AM Bob Camp <kb8tq at n1k.org> wrote:
>>> 
>>> Hi
>>> 
>>> Backing up a bit here.
>>> 
>>> On Feb 10, 2017, at 7:35 PM, gkk gb <modjklist at comcast.net> wrote:
>>> 
>>> Hello experts, I need a Rubidium frequency reference for my company, and wonder if I also need to GPS discipline it.
>>> 
>>> I characterize crystal-based OCXOs for ADEV, MTIE, and TDEV, and my longest measurement time is 100,000 seconds (28 hours).
>>> 
>>> If your longest measurement is a 100,000 second ADEV, then your measurement time will be out in the
>>> 1,000,000 to 10,000,000 second range. Is that really what you are doing?
>>> 
>>> If 100,000 seconds ADEV is your longest measurement, what is the shortest tau you are interested in?
>>> A Rb is not going to be much use for testing a good OCXO at shorter tau. Where the crossover happens
>>> depends a lot on the grade of OCXO you are working with. By the time you get to 1 second
>>> most OCXO’s will be noticeably better than most Rb’s.
>>> 
>>> I'm looking at this graph from SRS for PRS10,
>>> 
>>> http://www.thinksrs.com/assets/instr/PRS10/PRS10diag2LG.gif
>>> 
>>> I would suggest that plot is probably not the best one to depend on for GPS performance. In a GPSDO setting
>>> the cut over points are all over the place depending on which design you look at.
>>> 
>>> and thinking that as long as I calibrate a Rubidium source annually, there's no need for a GPS (since it only appears to degrade stability). Is this true in general, or is the graph misleading me because it may be true here, but not always.
>>> 
>>> The big issue is going to be temperature stability. If you have a Rb that is (say) 5x10^-10 over 0 to 50C, that is likely 1x10^-11 / C (or maybe more). A 2C delta in
>>> your lab as the HVAC cycles will give you a 2x10^-11 “hump” in your ADEV plot.
>>> 
>>> Also consider that if you want an “easy” measurement of the devices you are testing, the reference source probably should be
>>> 5X better than what you expect out of the DUT. You probably will not have that luxury in this case. That gets you into multiple
>>> references and things like three corner hat testing.
>>> 
>>> So my question, is a GPS necessary to discipline a Rubidium standard to characterize the best crystal oscillators for stability, or can I do without it (and just calibrate the Rubidium annually to maintain accuracy) and actually get better stability?
>>> 
>>> How many seconds out is a GPS generally needed to improve accuracy from a Rubidium standard?
>>> 
>>> If you really are running 1,000,000 to 10,000,000 second long tests, you need the GPS.
>>> 
>>> Lots of variables
>>> 
>>> Bob
>>> 
>>> _______________________________________________
>>> time-nuts mailing list -- time-nuts at febo.com
>>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>>> and follow the instructions there.
>>> 
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at febo.com
>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
>> 
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.



More information about the time-nuts mailing list