[time-nuts] Improving ocxo temp control

Bob kb8tq kb8tq at n1k.org
Mon May 21 13:12:02 EDT 2018


There are a lot of reasons an OCXO drifts. Temperature control is rarely the issue. 
More likely you are looking at the drift / wander characteristics of the crystal ( and 
components) in the OCXO. The simple answer is to leave it on for a while ( like weeks)
to allow things to settle out a bit. 

The paper that Rick tossed up last week is a pretty good one in terms of temperature
control issues in an OCXO and what the issues are. 

This all assumes you are in a fairly benign environment. If you have lots of drafts, put a 
cardboard box over the unit to shield it. If you room temperature is all over the place, then
there are a lot of ways to get to ~ 1 or 2C sort of stability in a lab. What you do depends
a *lot* on what your situation is.


So, assuming you *do* want to improve the temperature control:

First you need to take out what’s in there now. If it’s wandering around get rid of it. This
involves tearing apart the OCXO.

Now take a look at Rick’s paper and redesign the thermal enclosure. Get the heater placement
and sensor placement right and feed that into a simple controller. 

Run some tests over temperature and check out the data. It’s likely your first guess at things 
is not going to be correct.

Try to optimize the heat sources and sensors and re-test the result. Everything interacts so
this is not a quick process.

Once you are reasonably happy with where things are, start looking at a more fancy controller.
A simple approach would be feeding thermistor voltage into some 24 bit ADC’s and then
processing the result with an MCU. 

Ok so that’s all a bit much.


What happens if you mess with the OCXO from the outside of the package? 

You change the heat loss out of the package. This increases the thermal gain. ( less power
to increase the oven temperature by 1 C ). Assuming the original circuit was balanced 
out, you have made things worse rather than better. 

Ok so you do an enclosure with a fan it it so the heat loss doesn’t get less. 

You now have more heat loss and the same issue applies. In addition the fan and it’s
nonsense probably haven’t done the poor little OCXO any good.

When one designs a double oven, the inner oven is optimized for performance *with* 
the outer oven present. Equally, the outer oven is optimized for performance with the
heat load (and dynamics) of the inner oven. 


Assuming you still want to head down this road, temperature controllers are no different 
than any control loop. The first place to start is a textbook on control loops and control 
theory. The basics of what a loop does and the terminology are what you are after. Anything
advanced will assume you understand this part first. 

Next up are temperature sensors. Simple answer here is that a glass bead thermistor is
the way to go. For heat, transistors are the normal go-to device. The controls loop takes
in the thermistor output and spits out a voltage to change the current through the transistor.

If you have the money for software licenses, the next stop is some good mechanical CAD
that will feed into thermal modeling. From that you can work out a proper heat flow and
gradient design. Assuming that is a bit to expensive, you are back to trial and error. There 
are no “just duplicate this” designs that I know of. 

Once you have the structure, sensors, heaters, and control you toss it into a temperature 
test chamber. That may be something fancy or something you put together. You run the
gizmo over temperature and observe what it does. You then optimize the P,I,D coefficients
in your control loop. Indeed you may not have all of them or you may have an extra one. 


Of course one could simply shop for a $20 OCXO on eBay. Even if you have to buy a
dozen before you find a good one, it’s still cheaper / faster / easier / more likely to succeed 
than all the nonsense above. If this is a commercial design for a product you are going 
to sell, that does not work very well. The same fundamental answer applies. If you need
better performance, shop for a better oscillator. 

Lots of fun !!!!

> On May 21, 2018, at 12:23 PM, Club-Internet Clemgill <clemgill at club-internet.fr> wrote:
> Thanks for your interesting replies.
> What I am actually trying to do is the following: 
> I bough a small ocxo (size of half a ping-pong ball) that performs well (Abracon / AOCJY3_B 10Mhz)
> Reaching about 5*10E-11 kind of MDEV at low point ("kind of"… because a I use an HP52132a as input to Timelab)
> But it’s frequency is slowly drifting with time, with a quasi linear slope. 
> I wondered if placing it in a third ovenized enclosure could improve things. 
> I tried a few experiments but seems that the temp needs to be very accurately controlled. 
> Any similar experience ? 
> Could you suggest papers describing high performance analog or digital controllers ? 
> Thx, 
> Gilles.    
>> Le 19 mai 2018 à 16:09, Bob kb8tq <kb8tq at n1k.org> a écrit :
>> Hi
>> One key point about the need for “zero gradient”: 
>> Crystals and many other components are quite sensitive to thermal gradients. Very small 
>> fractions of a degree (as a gradient ) can have significant impact on the frequency of an 
>> oscillator. 
>> One of many “interesting things” about fiddling about OCXO’s. 
>> The equally frustrating thing about this is that unless you can tease kind paper authors
>> into posting things ( thanks Rick !!) the papers are behind pay walls. I pretty much despise
>> that practice. Referencing papers that send people off to spend money ….not so much.
>> Bob
>>> On May 19, 2018, at 12:03 AM, Richard (Rick) Karlquist <richard at karlquist.com> wrote:
>>> In my experience, the oven temperature controller is rarely
>>> the determining factor for static oven performance.  This article
>>> explains what the real determining factors are:
>>> http://www.karlquist.com/oven.pdf
>>> An analog oven temperature controller will be limited in
>>> its dynamics by how much capacitance you are able to
>>> design with.  Digital controllers get around this as well
>>> as having the capability of double integration for much
>>> better transient response.
>>> Rick
>>> On 5/18/2018 11:03 AM, Gilles Clement wrote:
>>>> Hi,
>>>> I am trying to improve performance of an OCXO.
>>>> Could you point me at a good design of a high resolution oven temperature controler please ? Preferably analog.
>>>> Thx much,
>>>> Gilles.
>>>> _______________________________________________
>>>> 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