LTZ1000 building dilemma

[eurofox]

Hello voltnut friends, 

Input and ideas are welcome, question is simple does it make really sense to control the temperature in my design? 

I have a kind of dilemma building my enclosure for a double LTZ1000 10V reference.

My lab temperature is never lower than 21°C and maximum 25°C.

I have 2 fully working and very stable 10V references based on LTZ1000, they will be connected to 2 different set of good quality binding post on the front.
My aim is to build the complete system including power supply in a block of aluminium with different areas completely isolate and hermetic for power supply part, 2 LTZ1000 PCB’s and binding posts.

It should be totally hermetic and the enclosure filled with Argon to avoid ant humidity problem.

To improve the design it should be heated, I already developed and tested the heating system but if I implement heating I have to isolate the enclosure and put it in another enclosure and instead of having my binding post on the aluminium I need to have it on the front panel and wiring …

The design is done to be able to power it from 5VDC to 12VDC, I use a very stable and clean power supply for the 2 LTZ1000.
I have a 10A 12V battery that I can use externally but if I use heating it will dramatically decrease the off grid function.

Since I have to start milling the aluminium block I need to decide ….

 

[TiN]

I fail to see what is the practical benefit to go into all that trouble, so seems like answer would be "don't listen anyone and do what you think is right, as we are here for fun anyways". 
You don't say what is your topology in 7V - 10V stage, so the "mileage may vary" of all those design challenges (hermetic box, argon filling cavities, super-puper power regulation, etc).

[Dr. Frank]

I think your approach is completely over-engineered.

Especially the aimed stability figures and specifications are lacking, or have no visible connection to your build.
You claim that your circuit is already 'very stable', so why all this additional effort, and what does 'very stable' mean in numbers, and what are your goals, in numbers, again?

The LTZ1000 circuit can easily  be made stable (w/o any fancy gimmicks) to < 1ppm over 1 year, and < 0.05ppm/°C over temperature, and humidity changes to my knowledge are not really well understood / specified compared to the foregoing stability figures.

The 7 => 10V step-up circuit is a completely different story, and should be separated content wise and also physically from the LTZ1000 circuit.

Again, the amount of humidity stability effects might also fall below the timely and temperature driven instability effects.
And especially this depends on the case technology used, i.e. molded versus hermetically sealed resistors
So I would avoid that Argon gimmick, before you are fully aware of the expected effect, and possible other countermeasures.

At last, an oven around the 7=> 10V circuit only, similar to the 732B / 57x0A reference hybrids, that would makes sense, but you should physically separate / isolate the LTZ1000 circuit, as an additional oven would only de-stabilize the reference.


Frank

[eurofox]

I think your approach is completely over-engineered.

Especially the aimed stability figures and specifications are lacking, or have no visible connection to your build.
You claim that your circuit is already 'very stable', so why all this additional effort, and what does 'very stable' mean in numbers, and what are your goals, in numbers, again?

The LTZ1000 circuit can easily  be made stable (w/o any fancy gimmicks) to < 1ppm over 1 year, and < 0.05ppm/°C over temperature, and humidity changes to my knowledge are not really well understood / specified compared to the foregoing stability figures.

The 7 => 10V step-up circuit is a completely different story, and should be separated content wise and also physically from the LTZ1000 circuit.

Again, the amount of humidity stability effects might also fall below the timely and temperature driven instability effects.
And especially this depends on the case technology used, i.e. molded versus hermetically sealed resistors
So I would avoid that Argon gimmick, before you are fully aware of the expected effect, and possible other countermeasures.

At last, an oven around the 7=> 10V circuit only, similar to the 732B / 57x0A reference hybrids, that would makes sense, but you should physically separate / isolate the LTZ1000 circuit, as an additional oven would only de-stabilize the reference.


Frank

LTZ1000 and buffer from 7 to 10V are on the same PCB, all resistor are top quality, lowest PPM/°C reasonable money can buy but molded, reason of my concern about humidity, enclosure will be hermetically sealed, option of argon is silica gel.

My target is 1PPM/year or better.

[Dr. Frank]

LTZ1000 and buffer from 7 to 10V are on the same PCB, all resistor are top quality, lowest PPM/°C reasonable money can buy but molded, reason of my concern about humidity, enclosure will be hermetically sealed, option of argon is silica gel.

My target is 1PPM/year or better.

Again, you should at first create a clear target specification concerning humidity, temperature and time, then measure the as-is specification of time / temperature, and then calculate the influence of these three parameters, especially the humidity onto the stability of the circuit.
Humidity effects have been measured by Andreas and /or Branadic, I think.

It might turn out, that to achieve your target specifications, humidity effects can be neglected, but an oven might be needed as the only measure to achieve a certain T.C.

'1ppm/year or better' does not fit at all to temperature or humidity stabilization measures, that's dependent mostly on the quality / timely stability of your step-up resistors .

So if you did not use Vishays oil filled, hermetically sealed resistors  in first place, you will definitely or very probably get >1ppm/year stability.
But again, don't confuse that with the temperature coefficient (stability)

It's a big disadvantage, if you can't separate LTZ1000 and boost circuit.

Frank

PS: I have built my 7 => 10V step up by 10 bulk metal foil resistors, green molded Z201, or so, and that 10V tracked its LTZ1000 voltage relatively within < 1ppm over many years.
Temperature or humidity effects I did not observe, or did not test it explicitely.
As an orientation, the left-over 49.99k Z201 I have, has a T.C. ~ 0.2ppm/°C, but I don't know about the 10 others.

[TiN]

My target is 1PPM/year or better.

Woohoo Now we talking numbers. Before we begin, please note, that below is just my (rightfully biased) opinion on the subject.

If you hoping that you build one or two devices and get 1 ppm/annual stability, then you asking for big luck. No amount of engineering wankery in PCB, enclosure, power regulation and shielding will help you to get LTZ1000 chip that is <1ppm/annual itself. Out of my modules (over 10pcs already, since 2014)/meters with LTZ, I have only two units that <1ppm/year. Rest are in 2-3ish region. And they ALL running lowish oven temperature. So okay, let's imagine you bought 20+ LTZ chips like I did, and spent few years with aging jig to confirm and sort few chips that are 1ppm/year. By this time you are already well known in your nearest calibration lab, that is always happy to see you. Since you must have at least 3 calibrations every year with uncertainty below 0.4 ppm on 10VDC to verify that your LTZ magic box is indeed reaching stability you want. Expensive, you say? Sure, but it's cheaper than buying 4+ Fluke 732A/B/Cs and/or Datronski 491x , getting these calibrated and verified for same 0.4ppm/year aging figure. Hmm, where we are by now? Ah, we need also 1ppm/year stable 7V -> 10V. Well, we repeat the exercise again, but this time buying tens of hermetical precision resistors or networks, to ensure ratio is indeed stable over years. Since you have concern about humidity (valid one, grant you that), you'd indeed want hermetical package, so at least you don't worry about hysteresis from soaking water into resistive elements.

There you go, after 3+ years and x-xx K$ , now you getting to a point when to worry if to use argon or silica and what aluminum block to CNC . Sorry if that sounds too grim, but this is my 5c. 

[eurofox]

My target is 1PPM/year or better.

There you go, after 3+ years and x-xx K$ , now you getting to a point when to worry if to use argon or silica and what aluminum block to CNC . Sorry if that sounds too grim, but this is my 5c. 

I will try but I don't want to spend K€ on it, is just an hobby and since I have the machinery to build something different and Argon lets wait for the results. 

I will put Argon, it is an inert gaz that should not arm any component, it is more heavy than the air and this way I will be able to tune the voltage by a little hole closed by a screw and o-ring, this way I should solve the humidity issue.

[try]

LTZ1000 and buffer from 7 to 10V are on the same PCB, all resistor are top quality, lowest PPM/°C reasonable money can buy but molded,

My target is 1PPM/year or better.

Resistor names, please!

[eurofox]

LTZ1000 and buffer from 7 to 10V are on the same PCB, all resistor are top quality, lowest PPM/°C reasonable money can buy but molded,

My target is 1PPM/year or better.

Resistor names, please!

They are all Vishay branded.

[try]

Eurofox,

you are holding back the information necessary to answer your question. 

Hello voltnut friends, 

Input and ideas are welcome, question is simple does it make really sense to control the temperature in my design? 

LTZ1000 and buffer from 7 to 10V are on the same PCB, all resistor are top quality, lowest PPM/°C reasonable money can buy but molded,

My target is 1PPM/year or better.

Resistor names, please!

They are all Vishay branded.

[eurofox]

Information below of resistors used:

RESISTOR PRECISION 10K 0.01%
Resistance : 10kohm
Voltage Rating : 300V
Resistor Case Style : Radial Leaded
Power Rating : 400mW
Resistance Tolerance : ± 0.01%
Product Range : S Series
Resistor Element Material : Metal Foil
Temperature Coefficient : ± 3.5ppm/°C
Automotive Qualification Standard : -
No. of Pins : 2Pins
Resistor Mounting : Through Hole

[Andreas]

Hello,

So for the S102 resistors even Vishay states a 20 ppm seasonal change due to humidity:

http://www.vishaypg.com/docs/63171/TN104.pdf

The measurements from Lars that he sent me indicate around 25ppm change
for a humidity change from 35/65 %rH for a 10K S102 resistor.

So humidity matters for non hermetically sealed resistors.
Partly this will be compensated out since the epoxy of 2 equal devices reacts similar.
But dynamically there may be different time constants (3-7 days) and slightly different behaviour.

With best regards

Andreas

[David Hess]

Again, you should at first create a clear target specification concerning humidity, temperature and time, then measure the as-is specification of time / temperature, and then calculate the influence of these three parameters, especially the humidity onto the stability of the circuit.

And barometric pressure or in this case if a rigid hermetic enclosure is used, pressure inside which will vary with temperature unless a bellows is used.  Did anybody ever build a reference inside of a metal bellows?

I might use extreme construction not for the LTZ1000 itself but for the supporting parts.

If I did something like this, I might use a vacuum instead of an inert gas fill but maybe it is a bad idea.  A vacuum would remove pressure effects, contamination, and thermometric effects from convection.  Also it is possible to continuously verify the condition of the vacuum unlike an inert gas fill and I have a vacuum pump but not an argon tank.

[floobydust]

The LTZ itself has some volume of gas inside at atmospheric pressure, on the day of packaging?
Placing it in a vacuum would stress the can, I think it's a stamped seal. Not sure how a delta of 14psi would affect the LTZ can or what the pressure rating is.
Makes me wonder about hermetic-packaged semiconductors in space.

I have seen TO-39 packages crushed from pressure in down-hole oil tools. But that was 100's psi.

[ap]

Hermetic packages (with cavity) are used in space. There are challenges using parts in space (does not make sense here to go into details), but not the cavity by itself.

[David Hess]

A change of 14psi will not do anything to metal, ceramic, or plastic packages.  Induced strain has an effect but with a vacuum it would be constant.  The enclosure on the other hand can have considerable stress on it due to large surface area.

I brought it up because it is one of the things I have not tried.

[iMo]

In space: ~0,5kg from inside the package to the LTZ's package top/bottom/sides..

[floobydust]

I see metrology components do not like change - in temperature, strain, or moisture ingress.

I think the vacuum would stop moisture from getting in, but maybe create new problems- bending the IC can and adding strain to the die. The LTZ "A" version adhesive might be soft and make this a non-issue.
Also the vacuum stops convection heat loss so the LTZ heater gain is high now and the heater transistor would need heatsinking. It would take even longer for the ref amp parts to reach equilibrium.

The part I think might be a deal breaker is if you have any leaks, at gas-tight feedthroughs and such, (large) changing pressure I'll bet causes a shift.

[borghese]

And maybe the oven goes out of regulation due to the limited heat loss

[IRFP]

Hello,
I use dielectric oil …
https://www.eevblog.com/forum/metrology/lm399-under-temperature-controlled-environment-in-oil/

[David Hess]

The part I think might be a deal breaker is if you have any leaks, at gas-tight feedthroughs and such, (large) changing pressure I'll bet causes a shift.

One of the reasons I like the vacuum idea is that if there is a leak, it would be unambiguous.  If an inert gas fill leaked, how would you know?

[branadic]

Vacuum has the disadvantage of outgassing printed circuit boards and other plastic materials.

-branadic-

[David Hess]

Vacuum has the disadvantage of outgassing printed circuit boards and other plastic materials.

They would but I expect that in a production environment, a bake out under vacuum would be done during burn in.

It only proposed it as an idea to see what others thought before I try it.  The stress on any precision IC packages strikes me as the largest problem.

[branadic]

I a production environment they would use ceramic boards instead of FR4. Same for integrated circuits in hermetic metal and cermamic packages if not bare die in a big hermetic case for the full ceramic board.

-branadic-