[time-nuts] Designing an embedded precision GPS time

Bob kb8tq kb8tq at n1k.org
Wed Nov 1 09:34:45 EDT 2017


> On Nov 1, 2017, at 9:17 AM, jimlux <jimlux at earthlink.net> wrote:
> On 11/1/17 6:01 AM, Bob kb8tq wrote:
>> Hi
>> Unfortunately not all TCXO’s are created equal. It depends a bit on the
>> original intended use. I’d bet it also depends a bit on the original target
>> price. Perturbations  (frequency jumps) over temperature are one “feature”
>> that might be present. Hysteresis at half the temperature spec is another
>> “feature”.
>> Even within the same batch or same test run, some will be much better
>> than others. You stop the compensation process when they get “good enough”.
>> That will mean that a few are right at whatever the production target is and
>> others exceed the target by quite a bit.
>> While crystal curves are indeed cubic, there are higher order terms in the
>> curve. The “why” is something people get to write papers on.If you are trying
>> to compensate to tight specs, you will see all sorts of stuff. It is not at all uncommon
>> to see >9th order curves residual curves. Indeed some of that is from residuals
>> in the compensation circuit as well as from the crystal.
>> Why all this yack? A lot of people come from a background using OCXO’s. An
>> OCXO generally has a low order temperature characteristic. It also is rare to see
>> things like frequency perturbations in an OCXO. Moving from one to the other
>> can be a bit interesting.
> Indeed - I was looking at algorithmically compensating some cheap TCXOs and there's an amazing spread in the "details" of the curves - sure, they all met the spec (several ppm, as I recall), but it was clear after very little testing that there was no "one algorithm to fit them all"
> As you say, good grist for a paper or thesis project.
> That's why I wish they'd sell OCXOs, cheap, without the oven. Or maybe look for regular XO (no TC).  Those might have a more "pure" (read lower order) freq vs temp characteristic.

The issue there is that the crystal is where the money is. The oven circuit is actually pretty far 
down the list cost wise. Poke at this spec, poke at that spec and you have a $50 crystal (in volume). 

> The problem I see with regular XO is that they tend to be designed to a cost point and there might be more of the hysteresis and mechanical effects - if you're not claiming better than 100ppm, then 50 ppm of hysteresis isn't a problem.
> A 1ppb OCXO is going to have to be a better mechanical design - so that it can hit that 1ppb every time when you turn the oven on and go from cold to hot.

It very much has a crystal that spent more time on the production line (somewhere) being processed than 
it’s lower spec cousins. Time is money and that equipment isn’t cheap either. There then is the minor 
issue of yield. Toss in a more expensive package while you are at it ….

> Maybe this is just griping in general - why don't mass production manufacturers make exactly the niche part I want to buy (that is of no use to anyone else)for $3 each
> I suppose if you were going to build little algorithmically compensated modules, you'd bite the bullet and design a crystal oscillator and then YOU get to choose what crystal in what mount etc.

If you are buying a few million of this or that a month, you most certainly can get people’s attention. Toss 
in a willingness to pay a few dozen bucks a piece on top of that and you will get a *lot* of people’s attention. 

Crystals are made and sold with characteristic data on them. The same is true of just about any type of
oscillator. Getting to the point that the data is *useful* takes a lot of engineering on both ends of the process. 
There are a number of companies that have set up to do the characterization once the device is in the
end end product. To a great extent that gets done to speed up the engineering process ….

> When all is said and done, the production cost for a design that uses a crystal in a can plus half a dozen discrete devices to make an oscillator is probably not a lot different than the production cost for a design using an oscillator in a can.  it's the "other stuff" in the design that will add up.

If the crystal is at some odd frequency or in an odd package … be careful. Experience counts in terms of making a good crystal. 
Experience at 5.000000 MHZ in an HC-40 is unfortunately not the same as experience at 5.000005 MHz in the same package. I have a lot
of data on this ….


>> Bob
>>> On Oct 31, 2017, at 10:42 PM, jimlux <jimlux at earthlink.net> wrote:
>>> On 10/31/17 1:47 PM, Bob kb8tq wrote:
>>>> HI
>>>> TCXO is a very loosely defined term. A part that does +/- 5 ppm -40 to +85C
>>>> is a TCXO. A part that does +/- 5x10^-9 over 0 to 50C may also be a TCXO.
>>>> Dividing the total deviation of either one by the temperature range to come
>>>> up with a “delta frequency per degree” number would be a mistake. You
>>>> would get a number that is much better than the real part exhibits.
>>>> Working all this back into a holdover spec in an unknown temperature
>>>> environment is not at all easy.
>>> Very much so - most of the TCXO curves I've seen tend to be "much" better than the spec over the central part of the frequency range (which makes sense, the underlying crystal is a cubic with temp, most likely)
>>> Retrace and hysteresis might be your dominant uncertainty.
>>> I've attached a typical TCXO data plot for your viewing pleasure..
>>> (that's an expensive oscillator, because it's for space, but I don't think space or not changes the underlying performance)
>>>> Bob
>>> <TCXODataVectron 47.pdf>_______________________________________________
>>> 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