[time-nuts] Re: 20230122: Please help me understand my OCXO

Thu Feb 16 16:44:43 UTC 2023

Hi

Should have used a different browser. The first one I used cut off the 
note about power going out on the 24th 

Bob

> On Feb 16, 2023, at 11:37 AM, Bob Camp <kb8tq at n1k.org> wrote:
> 
> Hi
> 
> So, right at day 23 the nice upward trend takes a step down. It does it suddenly 
> and the rest of the curve pretty much keeps on going. You might wonder how
> abrupt the change was. It could have (and likely was) very abrupt. Sample at
> whatever speed you wish, itâ€™s a jump between two readings (maybe). 
> 
> Various folks have various names for this. Frequency hit, frequency jump, and
> phase hit are three. There are many others. They all are talking about an instantaneous 
> change. Needless to say these changes can be a bit of an issue in some systems. 
> 
> Back on topic: Do they count as aging? Since they are random in direction and 
> random in spacing the hits *should* average out to zero. (Yes, there is more than
> a little debate about that â€¦). If aging is simply the â€œsmooth trendâ€ (a *very* 
> arbitrary definition) then they donâ€™t count. 
> 
> Even when it is clear â€¦. itâ€™s not :) :) :)
> 
> In some cases an OEM might have a spec on the jumps *and* a spec on the 
> aging â€¦.
> 
> Bob
> 
>> On Feb 16, 2023, at 11:01 AM, Andrew Kalman <aekalman at gmail.com> wrote:
>> 
>> Just a follow-up if anyone is interested ... I ran my test for over a month, and I conclude that the OCXO's drift has settled to around 0.25-0.33ppb/day. Plot attached.
>> 
>> --Andrew
>> 
>> --------------------------------
>> Andrew E. Kalman, Ph.D.
>> 
>> 
>> On Tue, Jan 24, 2023 at 8:05 AM Andrew Kalman <aekalman at gmail.com <mailto:aekalman at gmail.com>> wrote:
>>> Hi Bob.
>>> 
>>> That helps a lot -- gives me a lot of insight, esp re measurement resolution vs time.
>>> 
>>> (I'm slowly being dragged into long-time duration measurements, and I'm not enjoying the associated energy costs :-) ).
>>> 
>>> --Andrew
>>> 
>>> --------------------------------
>>> Andrew E. Kalman, Ph.D.
>>> 
>>> 
>>> On Tue, Jan 24, 2023 at 6:15 AM Bob Camp <kb8tq at n1k.org <mailto:kb8tq at n1k.org>> wrote:
>>>> Hi
>>>> 
>>>> Looking at the data, past day 10 you get into the â€œunclearâ€ range. Is the data
>>>> limited by the test resolution? maybe. It temperature an issue? who knows. Is
>>>> there something else going on â€¦..
>>>> 
>>>> After a month or so, you can likely draw a box around data with wobbles in it and
>>>> make a rational guess about the aging rate. Take the diagonal through that box
>>>> and itâ€™s a pretty reasonable guess at the max rate. 
>>>> 
>>>> Back in the 70â€™s 1 ppb per day was a pretty good spec on an OCXO. As time went
>>>> on, things got better and better. Going from AT to SC crystals helped a bit. Mostly,
>>>> crystal processing just improved in many small steps over many decades. 
>>>> 
>>>> To really know what category the one you have falls into aging wise, you need more
>>>> data. A month or three of monitoring will give you a better idea. Best guess is it
>>>> will fall away to the â€œneed that boxâ€ level. You simply will be guessing with the current
>>>> test setup. 
>>>> 
>>>> Hermetic (or reasonably hermetic) OCXOâ€™s tend to do pretty well after storage. They
>>>> donâ€™t soak up humidity like the open designs do. What you are seeing is not unusual 
>>>> on a part like you have.
>>>> 
>>>> How good can it get? Iâ€™ve seen groups of OCXOâ€™s from back in the 1970â€™s that stayed
>>>> well below 1x10^-8 per year. There are TBoltâ€™s with OCXOâ€™s on them that have even 
>>>> less change per year over the 10 to 30 years since they left the factory. 
>>>> 
>>>> Does aging when powered off count? How long has this or that example been on 
>>>> power vs off power? Thereâ€™s not a lot of way to get that data on most of these parts. 
>>>> One would *guess* that this or that device did spend significant time on power before
>>>> it got scrapped out.
>>>> 
>>>> Take the yearly numbers and do the division. You can get a daily rates.  1x10^-8 / 365 
>>>> gets you 3x10^-11/day. Iâ€™d suggest that while itâ€™s a valid use of math, the result isnâ€™t 
>>>> really significant for a â€œper dayâ€ estimate. 
>>>> 
>>>> The good news: Your OCXO seems to be running ok and itâ€™s not broken.
>>>> 
>>>> Bob
>>>> 
>>>> > On Jan 24, 2023, at 12:53 AM, Andrew Kalman via time-nuts <time-nuts at lists.febo.com <mailto:time-nuts at lists.febo.com>> wrote:
>>>> > 
>>>> > I recently revived an instrument with a non-small OCXO in it, and I'd like
>>>> > to better understand what I've measured in terms of the OCXO's accuracy and
>>>> > aging.
>>>> > 
>>>> > Background & Test setup:
>>>> > 
>>>> >   - The OCXO is an Isotemp Model OCXO36-44, date code 9552.
>>>> >   - It originally came from an XL Microwave Model 3060 -- I dunno if it's
>>>> >   the Option 112, 115 or 120 OCXO.
>>>> >   - It has just three connections -- 12V, GND and RF output.
>>>> >   - I moved it into an XL Microwave frequency counter Model 3080 (10Hz to
>>>> >   8GHz).
>>>> >   - In 2017 I adjusted the OCXO (using its coarse and fine adjustment
>>>> >   screws) to an indicated 10,000,000.00MHz using the
>>>> >   two-traces-on-an-oscilloscope method and a GSPDO.
>>>> >   - Around 2019 this Model 3080 stopped working (it blew a transistor on
>>>> >   its power PCB).
>>>> >   - A few weeks ago I repaired this Model 3080 (a Schottky power rectifier
>>>> >   had failed, and took out a transistor). I replaced three rectifiers and two
>>>> >   transistors in the power supply.
>>>> >   - Once repaired, the Model 3080 started up immediately and the OCXO
>>>> >   measured 9,999,999.66MHz.
>>>> >   - My measurement setup is an HP Z3805A GPSDO providing a 10MHz reference
>>>> >   that feeds into an XL Microwave Model 3120's external 10MHz reference
>>>> >   input; the Model 3120 is in turn measuring the Model 3080's 10MHz reference
>>>> >   output. My recorded data is strictly from the Model 3120's display (max
>>>> >   digits). All of this equipment is just in my lab, temperature swings are
>>>> >   small. The Model 3xxx frequency counters have no fans.
>>>> > 
>>>> > I've attached a picture of the OCXO and a plot of its behavior over 17 days
>>>> > of uptime since the day I repaired it.
>>>> > 
>>>> > My questions:
>>>> > 
>>>> >   - *Is my measurement setup one that provides reasonably accurate
>>>> >   measurements of the OCXO's absolute accuracy and aging?*
>>>> >   - *Is the "warm up behavior"of this OCXO typical for a nearly 30-year
>>>> >   old OCXO with unknown total uptime?*
>>>> >   - *Does my test setup allow me to conclude that the OCXO is
>>>> >   demonstrating roughly 0.5ppb aging from day 11 onwards?*
>>>> >   - *From this limited amount of data, is this representative of a great /
>>>> >   good / average / poor OCXO, and why?*
>>>> > 
>>>> > Thanks for your responses.
>>>> > 
>>>> > --Andrew
>>>> > 
>>>> > --------------------------------
>>>> > Andrew E. Kalman, Ph.D.
>>>> > <20170512_IsoTemp_OCXO36-44.jpg><20230109_XL_Microwave_3080_OCXO_Aging.pdf>_______________________________________________
>>>> > time-nuts mailing list -- time-nuts at lists.febo.com <mailto:time-nuts at lists.febo.com>
>>>> > To unsubscribe send an email to time-nuts-leave at lists.febo.com <mailto:time-nuts-leave at lists.febo.com>
>>>> 
>> <20230109_XL_Microwave_3080_OCXO_Aging.pdf>
> 