crystal aging and specifically OCXO aging

[codeboy2k]

I'm thinking of buying an OCXO on eBay, specifically this one:

http://www.ebay.com/itm/170923748997

I have not designed with an OCXO before. The most I've ever designed with is the bog standard canned crystal oscillator or smt crystal, and I didn't really care about the accuracy or stability, because it was just running a micro or providing USB timing or Ethernet timing, and it didn't really matter, so I am not sure how to properly read the datasheet.

http://www.isotemp.com/wp-content/uploads/2011/10/143-1000.pdf

The datasheet has a section on aging:

2.2. Aging
   a. At time of shipment         < ±5x10-10/day
   b. After indefinite storage      
        i. Daily                  < ±5x10-10 after 30 days
        ii. Yearly                < ±1x10-7
        iii. 10 years             < ±4x10-7

My question is, is this aging data cumulative? being that it's on eBay should I just assume that I can't know it's true age and
assume it's at least 10 years old or more (worst case)?  and then does that mean it's < ±4x10-7 + (10yrs * ±1x10-7/year ) ??
i.e < ±14 x10-7 or ±1.4ppm ??  If so, It's starting to become a pretty bad clock .

or is it just saying after 10 years its never going to be worse than  ±4x10-7, final. (0.4ppm)

What if it's actually 8 years old? should I use the Yearly aging figure, then it's (8yrs * ±1x10-7/year) = ±8x10-7.
This is actually worse than the 10 year data... so I am confused.

thanks !

[alm]

The specs mean that after 10 years, it should be within 0.4 ppm. The specs mention indefinite storage, so after 8 years in the warehouse it should still meet those specs. I would expect it to be better after being turned on for 8 years, but there is no way to tell this from the specs.

After 8 years it's probably going to be somewhere between 0.1 and 0.4 ppm worst case. Probably closer to 0.4 ppm. Aging is often specified as ppm drift * sqrt(time). This assumes its a truly random process (like diffusion). From the 1 year figure, you would estimate the 10 year figure to be 0.1 ppm * sqrt(10) = 0.3 ppm, which is quite close to the datasheet value.

[AlfBaz]

Does it age if it hasn't ever been powered up?

[alm]

Yes, though probably at a different (probably lower) rate than when turned on. The temperature is much lower, for example.

[codeboy2k]

The specs mean that after 10 years, it should be within 0.4 ppm.

That's what I was expecting. Thanks.

...Aging is often specified as ppm drift * sqrt(time). This assumes its a truly random process (like diffusion). From the 1 year figure, you would estimate the 10 year figure to be 0.1 ppm * sqrt(10) = 0.3 ppm, which is quite close to the datasheet value.

Aha. The drift*sqrt(time) is an important piece of the puzzle to read the datasheet properly.
So my theoretical 8 year old clock would be about



Given that it's an eBay part I'll probably just assume worst case 0.4 ppm and see how that fits into my application needs.

[codeboy2k]

follow up... so I bought a few OCXOs on eBay, they arrived today, I've not had a chance to power them up yet, but here's a pic of what I bought.

Does size matter?  These three are different sizes, and each one has different significant figures in the output value.
Does anyone  have any comments on which of these are known to be any better than the others?

Thanks!

The large one:
DESC: OFC/McCoy SC-CUT SINE 10 MHz OCXO CRYSTAL OSCILLATOR
MODEL:  OSC92-100B
FREQ: 10.000000 MHZ

middle one:
DESC: CMAC 10MHz ocxo oscillator square wave sc-cut
MODEL: STP2390C
FREQ: 10 MHZ

small one:
DESC: SC Cut Crystal 10MHZ ISOTEMP OCXO OSCILLATOR SQUARE WAVE
MODEL: ISOTEMP OCXO143-141
FREQ: 10.000 MHZ

[ve7xen]

OFC: This is reported to be made by Isotemp and equivalent to OCXO134-10 from them. Not a bad part, 5ppb vs. temp from -30 to +60C, -105dBc @ 10Hz. It's also the oldest, so probably has aged the most, which means its aging rate will likely be very low now.

CMAC: Can't find any information on this.

ISOTEMP: Hard to say, the OCXO143 series spans a couple decades of performance, and the -141 seems to be a custom-spec part. It could be a little better than the OFC (1ppb vs. temp, 115dBc @ 10Hz), or it could be a lot worse. Given that it's square wave output it might be harder to implement well.

I've received several OCXOs from eBay that were well off frequency (some 20ppm on a 5ppb part, and the EFC range insufficient to correct for it) and had poor stability as well (drift of some ppm with temp / time). Make sure you check them out before using.

[saturation]

Oscillators vary in quality as you demonstrate, even among a type such as the ovenized set.  For academic details, lots here and plow through the website, much is free:

http://www.ieee-uffc.org/frequency_control/teaching.asp?name=vigtoc

The OFC seems to be the best of the breed, and older aged crystals are better than young among the best makers, assuming it still works and was not abused in the reclamation.  The downside is older ones [ decades] are closer to death too, but at an unknown time, but until then it functions as good as it will ever be.

[codeboy2k]

OFC: This is reported to be made by Isotemp and equivalent to OCXO134-10 from them. Not a bad part, 5ppb vs. temp from -30 to +60C, -105dBc @ 10Hz. It's also the oldest, so probably has aged the most, which means its aging rate will likely be very low now.

CMAC: Can't find any information on this.

ISOTEMP: Hard to say, the OCXO143 series spans a couple decades of performance, and the -141 seems to be a custom-spec part. It could be a little better than the OFC (1ppb vs. temp, 115dBc @ 10Hz), or it could be a lot worse. Given that it's square wave output it might be harder to implement well.

I've received several OCXOs from eBay that were well off frequency (some 20ppm on a 5ppb part, and the EFC range insufficient to correct for it) and had poor stability as well (drift of some ppm with temp / time). Make sure you check them out before using.

thanks for the info, ve7xen.. (xen.. good call sign !) I saw your blog, I'm also planning to make a GPSDO like you wrote on your blog too.

Along with these few OCXOs, I also ordered and received a Navman Jupiter TU60 GPS receiver.  This one has the 10Khz correlated output so I am hoping to use that with the OCXO to make a GPSDO.  provided, as you say, it's not out so much that it can't be pulled in.

I wasn't pleased with what I received , the picture on the ebay sale shows a nice clean unit (of course!)



what I got was sufficiently rusted to give me some initial worries. It would seem that it was previously in a pretty harsh environment outdoors.  I just hope it still works and has life left in it.



My goal is to have a decent 10Mhz reference on my bench, that I can use everywhere, so I can build a dual-slope ADC and perhaps get 7 or 8 digits from it, so I can have an accurate 10-12 digit frequency counter, so I can build an accurate DDS based function gen from it, so I can know my scope sweeps are timed accurately, etc.. etc.. there are so many uses for a good clock !

after 25 years, I am retired from my day job as an engineer (hoping to freelance now) and just starting out with an empty home bench, so I am on a limited budget, and I'm planning to bootstrap my bench with stuff I build myself. The first thing I need is a decent clock. After that it's all gravy

[codeboy2k]

Oscillators vary in quality as you demonstrate, even among a type such as the ovenized set.  For academic details, lots here and plow through the website, much is free:

http://www.ieee-uffc.org/frequency_control/teaching.asp?name=vigtoc

The OFC seems to be the best of the breed, and older aged crystals are better than young among the best makers, assuming it still works and was not abused in the reclamation.  The downside is older ones [ decades] are closer to death too, but at an unknown time, but until then it functions as good as it will ever be.

Thanks saturation. Good link, I'll read it up.

Seems like the OFC is the one I will start with.. It looks bent out of shape, that worries me a bit.

My plan is to sync these 3 oscillators to a GPS receiver (as I said in my previous post, I have the Jupitor TU60 which has a GPS correlated 10Khz output).

What I'd really like to do, but cannot because I don't have the equipment, is to warm these 3 oscillators up for 24 hours, then the next day take variance measurements for about tau=10^4 or 10^5 or so, and plot their Allen deviations over these tau. This will give me a baseline. Then, I want to connect them up to GPS and do it again. Hopefully they get can get down to better than 10e-13 or so while disciplined, as has been seen on the web blogs of many others who have tried this.  Then I want to unplug them from the GPS signal for a week and see how they drift over 1 week.  Maybe I'll know in only 24 hours..

Alas, I don't have 1ps accurate counters at home   I could perhaps rent one good counter and do it sequentially, but it would be far better to be able to do them all at the same time with the same GPS signals.. but I don't want to rent 3 counters.

[marshallh]

See Agilent's various white papers on various subjects that may be related:
http://www.home.agilent.com/agilent/application.jspx?cc=US&lc=eng&ckey=2232321&nid=-33080.0.00&id=2232321&cmpid=zzfindmetrology

[codeboy2k]

See Agilent's various white papers on various subjects that may be related:
http://www.home.agilent.com/agilent/application.jspx?cc=US&lc=eng&ckey=2232321&nid=-33080.0.00&id=2232321&cmpid=zzfindmetrology

Thanks marshall I will check it out.
Also found some good practical information here http://www.ko4bb.com/Timing/index.php
and of course there's timenuts

[saturation]

You're welcome.  If you intend to do any time-nuts type stuff you'll need a good counter, time-nuts archives have many recommendation for  models.  Quartz tend to  a wider spread short term vs Rb, but overall  average is improved if ovenized, so you'll need to a counter at your lab to check the status of your oscillators routinely.  Adjusting a quartz oscillator can create a deal of instability that could take many days to settle.

http://www.ieee-uffc.org/frequency_control/teaching.asp?vig=vigcomp

Thanks saturation. Good link, I'll read it up.

Seems like the OFC is the one I will start with.. It looks bent out of shape, that worries me a bit.

My plan is to sync these 3 oscillators to a GPS receiver (as I said in my previous post, I have the Jupitor TU60 which has a GPS correlated 10Khz output).

What I'd really like to do, but cannot because I don't have the equipment, is to warm these 3 oscillators up for 24 hours, then the next day take variance measurements for about tau=10^4 or 10^5 or so, and plot their Allen deviations over these tau. This will give me a baseline. Then, I want to connect them up to GPS and do it again. Hopefully they get can get down to better than 10e-13 or so while disciplined, as has been seen on the web blogs of many others who have tried this.  Then I want to unplug them from the GPS signal for a week and see how they drift over 1 week.  Maybe I'll know in only 24 hours..

Alas, I don't have 1ps accurate counters at home   I could perhaps rent one good counter and do it sequentially, but it would be far better to be able to do them all at the same time with the same GPS signals.. but I don't want to rent 3 counters.