OCXO unusual application idea

[IanJ]

Hi all,

I've a need for a low wattage resistor, no initial accuracy (0.1% will do) but with a low TCR. My main requirement is a really low TCR.
Problem is I need specific value and rather go for a custom trimmed resistor (at cost!)............I hit upon the idea of using an old OCXO, i.e. opening it up and embedding a 10ppm/degC resistor in there and achieving the equivalent of 1ppm/degC or better. The 4 terminals of the OCXO will be the power and of course the 2 terminations of my resistor.

I'm looking at 56.7k, but if it works I may even embed a series trimpot in there also so I can externally adjust.

I've not got much experience of OCXO's, I'm assuming the temperature in there is extremely constant/stable....................any thoughts?

PS. It's to add additional stability to my much modified EDC501J DC Voltage Standard.

Ian.

[macboy]

Have you thought about just buying a string of 100 or more 57.6k resistors and measuring the TC of each one? Usually, they fall on a spectrum including above and below zero, with a few close to zero. Of course this won't work for manufacturing, but we are talking about a one-off here.

You can also try to find two 115k with equal but opposite TC and put them in parallel. You will find this technique in some very expensive gear.

[Zucca]

Surely not a stupid idea on the paper. I have no clue how it will end in reality.

I was wondering how can you tell the warm-up phase is ended. It would be nice to have a "Thermal Locked" LED confirmation on the front panel when the oven OXCO is reached a stable temperature.

Another thought is the power supply requirement, I remember devices like a Agilent 531xx always keep the OXCO oven on even when the device is turned off to keep the temeparure as stable as possible (yes it will keep sucking current at the mains even in power off state). Maybe they don´t want the sensitive thermal component in the OXCO will be thermally stressed by cooling down and warming up again when you turn off/on the unit.

A brocken Rubidium Frequency Standard device from ebay could be another good starting point for your hack circus (more space in the enclousure and better pinout fantasy).

Just my $.02.

[IanJ]

Hi,

I was actually going to embed a temp sensor inside the OXCO also, as well as an ambient temp sensor inside the case of the EDC also. There's plenty of real estate on the LCD for both.

I've ordered a couple of cheap used OCXO's via Ebay.

Ian.

Surely not a stupid idea on the paper. I have no clue how it will end in reality.

I was wondering how can you tell the warm-up phase is ended. It would be nice to have a "Thermal Locked" LED confirmation on the front panel when the oven OXCO is reached a stable temperature.

Another thought is the power supply requirement, I remember devices like a Agilent 531xx always keep the OXCO oven on even when the device is turned off to keep the temeparure as stable as possible (yes it will keep sucking current at the mains even in power off state). Maybe they don´t want the sensitive thermal component in the OXCO will be thermally stressed by cooling down and warming up again when you turn off/on the unit.

A brocken Rubidium Frequency Standard device from ebay could be another good starting point for your hack circus (more space in the enclousure and better pinout fantasy).

Just my $.02.

[dannyf]

I'm assuming the temperature in there is extremely constant/stable....................any thoughts?

It will work but too complicated.

All you need is a heater that generates more heat when the temperature is low and vice versa. A heater can be power resistors or the heatsink on a to220 transistor, for example. Something like that can be done with a temperature sensor (ptc or ntc for example) + this heater, with your resistor wrapped around it.

[Vgkid]

It should work pretty good. If you have ever looked at the efc output on an ocxo, you will notice that they are extremely stable.

[German_EE]

Current drawn by an OXCO will drop once it reaches the set temperature as the heaters turn off.

[jpb]

Geller labs used a thermister (negative coefficient) in their voltage reference to compensate for the temperature changes in the rest of the circuit (resistors mainly but also the reference itself):

http://www.gellerlabs.com/voltage%20references.htm

This is similar to the suggestion above of comparing the temperature coefficients of several resistors I guess.

[Dr. Frank]

The idea is not bad.

The main problem will be to couple the resistor to the oven mass correctly.

The resistors case, or generally its construction is critical, i.e., if heat is transferred more via the resistors wires, or through its housing..

Therefore, the appropriate resistor technology has to be chosen, or the heat coupling has to be taken into account, and solved.

Please check the thread about the determination of T.C.s of different precision resistors.. you will see different hysteresis effects due to heat coupling mechanisms between the resistors and the heater mass:
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/


The resistor itself should also be stable enough, in terms of low T.C. and low ageing rates.
It's hardly possible to get a stable reference, if the T.C. w/o oven is too high, or if the resistor has already other too big drift effects, like ageing.
These different instability parameters go hand-in-hand in most cases.

Check also this publication, as an example:
http://www.imeko.org/publications/tc4-2008/IMEKO-TC4-2008-137.pdf


One last key problem will be, how to connect the resistor to the outside, without transferring heat over this connection.. that would disturb the oven, or even would change the resistors temperature..

The frequency of the XTAL in an oven can be easily coupled to the outside by amplifiers, that's of course not possible for a resistor..

Frank

[macboy]

Don't shrug off the idea of measuring the tempco of jellybean metal film resistors. You will probably find some low single digit ppm ones in a batch of 100 or more (and some very bad ones too). Just because they are not sold as low tempco, doesn't mean that they can't be. They are simply not tested and binned for tempco, as that is an expensive process.

Get some good quality test grabbers, preferably with beryllium copper hooks; forget nickel plated steel crocs. Connect to the resistor, zero the meter, hit the resistor with hot air from your SMD rework station, heat gun, or a butane soldering iron with a hot air tip. Note the delta reading. You are not going for precise measurement here, just a moderately repeatable procedure to find the best 5 or 10 from the batch. Then you can properly measure those few using a controlled method like hot and cold mineral oils of known temperatures.

[coppice]

Ovenising chunks of electronics used to be quite a common practice. Dave stripped down something with an ovenized voltage reference. We used to use entire ovenized ADCs, when people couldn't get the various parts of a board sized ADC to track well over temperature. Will ovenizing a resistor work? In the general case no. Self heating of the resistor will defeat what the oven is doing. However, if the current you put through the resistor is small enough that self heating is negligible it should work well. Of course, you want to start with a resistor who's temperature coefficient is not too bad, otherwise your oven's stability could be unreasonably critical to the overall result. A copper resistor, for example, with its 0.4%/C temperature coefficient might be tough for even a tightly controlled oven to tame. 

[IanJ]

Cheers guys for the input...........no show stoppers so I'll build this and see how it goes. I'll report back (with measurements).

Ian,

[Dr. Frank]

...

Will ovenizing a resistor work? In the general case no. Self heating of the resistor will defeat what the oven is doing. However, if the current you put through the resistor is small enough that self heating is negligible it should work well

.....

Hi coppice,

these statements are definitely wrong, especially in the case of heat dissipating resistors!

In contrary, that's a very interesting application of an ovenized resistor.. as an example in the Fluke 5440B/5442A calibrator...

In this instrument, there are 10 resistors of 200k in an oven @55°C, which measure the 1000V DC output.

Therefore, each resistor is loaded with 50mW, which would normally heat up each resistor by several 10°C, and therefore would change its resistance by its T.C., several 10ppm, as you predict...

That's the general problem of each high voltage divider..

But in this configuration, the oven controller detects this temperature rise, caused by dissipation, and therefore reduces the heater power, so that the resistor temperature is again kept constant, which means, that he resistance does NOT change at all.

Parts of the heater power is exactly replaced by the power dissipation in the resistor.

This way, you get an extremely stable resistance divider. Please examine the 5440B service manual for details..

You may see that assembly here: http://www.amplifier.cd/Test_Equipment/other/Fluke5442A.html  (last picture)

Frank

[coppice]

...
Will ovenizing a resistor work? In the general case no. Self heating of the resistor will defeat what the oven is doing. However, if the current you put through the resistor is small enough that self heating is negligible it should work well
.....

Hi coppice,
these statements are definitely wrong, especially in the case of heat dissipating resistors!
In contrary, that's a very interesting application of an ovenized resistor.. as an example in the Fluke 5440B/5442A calibrator...
In this instrument, there are 10 resistors of 200k in an oven @55°C, which measure the 1000V DC output.

Therefore, each resistor is loaded with 50mW, which would normally heat up each resistor by several 10°C, and therefore would change its resistance by its T.C., several 10ppm, as you predict...

That's the general problem of each high voltage divider..

But in this configuration, the oven controller detects this temperature rise, caused by dissipation, and therefore reduces the heater power, so that the resistor temperature is again kept constant, which means, that he resistance does NOT change at all.

That's not the general case. That's a carefully engineered case. You need several things to make it work properly:

• The self heating has to be fairly limited
• The resistors have to be heat sinked well enough that they are close to the temperature of the surrounding oven
• The oven needs to be poorly insulated (most are well insulated), so it can dissipate whatever heat the resistors produce in a sustained way, with the main heater idling.
  

[Dr. Frank]

That's not the general case. That's a carefully engineered case. You need several things to make it work properly:

• The self heating has to be fairly limited
• The resistors have to be heat sinked well enough that they are close to the temperature of the surrounding oven
• The oven needs to be poorly insulated (most are well insulated), so it can dissipate whatever heat the resistors produce in a sustained way, with the main heater idling.
  

Well, in general, you always HAVE to account for external heat sources in such oven assemblies.
You always have to adapt your special requirements to the oven design, i.e. if you have relatively low or relatively big dissipation values.. but that's a trivial statement..

In case of most OCXOs, which are a VERY common used case, the oven controller, oscillator and driver circuitry are also situated inside the oven enclosure, and these dissipate several 10 mW, despite the fact, that these ovens are relatively well thermally insulated.
As an example, which is well documented, take the HP 10811 OCXO family.

Edit: Yep, in the manual of the 10811, it's specified, the oscillator circuitry consumes 12V * 30mA (typ.), that are 360mW!! Compared to 2W steady state typ. for the oven. My other example was 500mW (1kV over 2MOhm), so that's even in the same ballpark..


In other applications, like oven stabilized zener references (LM199, LTZ1000), the reference circuitry itself also dissipates some mW, and the whole component is also quite well insulated.

Anyhow, an oven assembly always requires some degree of heat leakage to the environment , otherwise the bare oven mechanism would not work at all, instead heating up indefinitely, or not regulating properly, but that's also a trivial fact.
 

So you see, in general, heat dissipation has to be accounted for ..

I don't know any oven application, where this is NOT the case..

Maybe you know such real products?

Frank

[Dr. Frank]

On this occasion, I made further investigations about the stability of resistors in an oven..

Elevated temperatures strongly increase the ageing drift, and I have found a document from Vishay, which makes some quantitative statements for thin film resistor technology (which might be chosen for 10ppm/K) :
http://www.vishay.com/docs/28809/driftcalculation.pdf

There are different modus operandi, and the precision mode, which relates to a film temperature of 85°C, would lead to a drift of 500ppm/1000h, or 12ppm/day!

Assuming that the TO cannibalizes an OCXO, the internal, stabilized temperature usually runs on 82°C (e.g. the HP10811).

Therefore, there is no advantage from the low T.C. (<1ppm/K) in an oven, if the resistor itself drifts an order of magnitude more, per day.

Using ordinary (jellybean?) thin film resistors may really be a showstopper, depending on the requirements for this application..

In comparison, precision resistors (PWW, metal foil) specify 20ppm/year for shelf life (no load, at Ta = 25°C), and about 50ppm/1000h at 80°C, that's 1/10 of the drift for the Thin Film technology.

Therefore, either the oven temperature has to be set to much lower values, 45°C for example. That would give a better stability - factor of 2 for each 10°C (Arrhenius law).

Or higher grade resistors (PWW, metal foil) have to be used, which may already have a low T.C. around 1ppm/K.

Frank

[IanJ]

Hi all,

I managed to pick up some nice resistors for this test......

Vishay VSMP series, 1206 package
300 to 750mW (depending on value)
Tolerance = +/-0.02% or +/-0.01% depending on value
TCR = 0.05 ppm/degC (for range 0 to +60degC)
TCR = 0.2 ppm/degC (for range -55 to +125degC)
Load life stability = 0.005% (50ppm), 2000hr at rated power

To get the final value I need from the value's I've aquired I'll have to string 5off in series.

If I can I'll lower the temp of the OXCO down to below 60degC per the TCR above.
I'll also try without the OCXO given the specs above.

Ian.

[Vgkid]

Iwould try to get the ocxo to go as low as it can temp wise. I would shoot for a a few (10 or less)degrees above your hottest room temps.

[babysitter]

Gut feeling: The low-temperature approach sounds fine.Combining selected resistors to compensate for TC drift is well worth it. But regarding the IMECO design, I wonder if it is better to have the resistors in the same horizontal plane? Multiple layers dont seem perfect to me:

At least in air when stacked there might be big gradients, how about resistors embedded in thermal mass where conduction is the mode how heat behaves instead of convection and radiation?

My only experiments with ovens had a cylindrical brass tube in horizontal position, with both reference and device under test going into the same center drill from two sides.

(And my boss is going to replace my personal HP6632B which died during these experiments finally, yeah!)

@IanJ: Let me plug again my favourites for heating tiny parts: QH60A or QH40A from Kuhne electronics!
http://shop.kuhne-electronic.de/kuhne/en/shop/amateur-radio/accessoires/crystal-heater/Precision+crystal+heater+40%C2%B0+QH40A/?card=724

[IanJ]

@IanJ: Let me plug again my favourites for heating tiny parts: QH60A or QH40A from Kuhne electronics!
http://shop.kuhne-electronic.de/kuhne/en/shop/amateur-radio/accessoires/crystal-heater/Precision+crystal+heater+40%C2%B0+QH40A/?card=724

I'm still awaiting the OCXO's, so have given these a go in the interim. I removed the 49K resistor and replaced with 56.7K, the change in size was so I could use an LM399H instead of the original 1N829A ref diode.
I also replaced the original EDC chopper amp modules with one of my own design, not much in it to tell you the truth but mine being a bit more stable as the EDC's own chopper moule zero offset pot was simply across V+ & V- supply!

Have started testing and the unit certainly gets to operational temperature/stability quicker. However, there's is a crucial trimpot in the circuit which could also do with being stabilized, so thats next for the heater. Lucky I bought two of them.

Heater spec: 40.8degC +/- 1.5degC.

Great fun..............

Ian.

PS. The OCXO when it gets going will encompass the trimpot in question also.

Original ref diode and 49k resistor (old pic!):-


New heater (prior to heatshrinking) with a couple of 5ppm resistors. Also LM399H in background:-


New chopper amp module. You can see one of the original potted modules in the photo above:-

[splin]

Or you could dig or drill 5m or more into the ground where, depending on location and the soil conditions the temperature may be stable within 2 degrees Celsius or less throughout the year. No carbon 'footprint' and you wouldn't run the risk of suffering 10ppm or more of hysteresis shift after a power failure.

You'll need a good transfer standard for calibrating the long term drift as sending it to a cal lab could be a bit a drag 

And not so good if you live on the third floor.....

Splin.

[babysitter]

Actually, reasonable depth burying is not as good as it sounds... yearly temperature swings still there.

[IanJ]

I am guessing that one thing I will have to watch out for with these wee heaters is noise inducement.....i.e. onto my circuit. I haven't looked to see how they operate i.e. static load, PWM etc etc. Possibly some shielding wouldn't go amiss. I am down in the uV level afterall.


Info: The precision crystal heater provides temperature stabilisation for crystals. The circuit, which is built on AL2O3 ceramic substrate, has to be mounted on the 40 °C thermostat crystal. Then, the crystal is heated to 40.8 °C with a regulation accuracy of better than 0.1 °C. This provides high frequency stability.
 The precision crystal heater is a in expensive alternative to completely heated oven oscillators (OCXOs). However, the stability values of an OCXO can not be reached.

Ian.

[IanJ]

Hi all,

Wee update.

Got a Trimble 65256 OCXO.



Cut into it exposing the heater circuitry and the oscillator can.



Cut open the oscillator can.



Removed the oscillator pcb.....it's not needed.



I put a temp sensor on the floor of the oscillator can and measured it at 80degC (setpoint). In-rush current from cold for first few seconds is 1.2A.

I then modified the heater circuit by changing a couple of resistors to first limit the in-rush to 600mA (which slows down initial heating), and also to adjust the temp setpoint to 44degC.

So, the plan now is to build my custom circuit into the oscillator can housing and solder it all up.
The only thing I am deliberating over is whether to leave my LM399H external, or build it into this new device (and not bothering to use the LM399H's own heater). Need to look up the LM399H temp spec.

Great fun!

Ian.

[Vgkid]

I would leave the lm399 as is, just insulate it against temp changes. It runs toasty.