xo aging after the time frames in datasheet

[metebalci]

I have a device with an ocxo inside. Aging spec is 50ppb in 1 (first) year, 25ppb in 2 years. After 2 years, lets say 5, what would be the aging ? How is this estimated ?

[mendip_discovery]

Year 1 +(Year 2 x 4) would be my guess.

The first year is when you see the most drive then after that it is what year 2 is for each year. If it's been adjusted then year 2 would be would be the expected annual drift from there on.

[metebalci]

My understanding is it is not linear and getting more and more stable. So I expect it to be close to 2 years number even for 5 years (at least better than 4x). But I dont understand if there is a way to roughly estimate this, if there is an approx. model or it is just impossible. I think they normally test this by speeding up the aging, but I dont know why they particularly stop at 2 years for example. For the numbers they measure I saw there are simple models (eg what happens at 6 months), but what to do if there is no number lets say for 5 or 10 years.

[MegaVolt]

I have encountered the need to predict the aging of an alternator after a long period of time. The company that sold us the generators agreed to give a guarantee for a long time. To do this, they measured the aging for 1 month and approximated it with a logarithmic function. And they gave us only those generators that remained in tolerance for the time we needed. Of course, there are still random events that cannot be so predicted. But this is the best they could do for us.

[metebalci]

OK I understand. It seems that is the best that can be done. I find this paper quite helpful: https://www.isotemp.com/wp-content/uploads/2011/04/Crystal-Oscillator-Aging.pdf

I made a simple calculation like in the paper above, for an Rb clock, last calibration ~16 years ago. It has 0.5ppb/year after 2nd year. If I do it right, it makes ~12 mHz change after 16 years. I see 6 mHz in measurement against a GPSDO.