Geller voltage reference schematic ??

[Gyro]

Argon has different thermal properties, which you may like better or worse. Just FYI.

Yes, it has around half the specific heat capacity [of Nitrogen]. I've never been able to work out whether that's a good thing or not either. It may reduce the effect of convection currents but may also cause higher component temperatures (if dissipation is an issue). Again, it might improve the effect of physical thermal coupling between parts but at the expense of lower thermal coupling to the enclosure (if not physically clamped).

One of the reasons it [Argon] is / was used in filament lamps is to reduce thermal losses to the glass for a given filament temperature, as well as being non-reactive with the incandescent Tungsten.

[iMo]

We have to learn how to work with glass and feed through wires. Imagine a standard test tube (laboratory glassware) - its material could be melted with a common gas torch or even a kitchen gas cooker, imho. The only problem could be to find the suitable wires which would seal hermetic with the glass (it cannot be a rocket science, however).. No vacuum needed in our case..

[Overspeed]

We have to learn how to work with glass and feed through wires. Imagine a standard test tube (laboratory glassware) - its material could be melted with a common gas torch or even a kitchen gas cooker, imho. The only problem could be to find the suitable wires which would seal hermetic with the glass (it cannot be a rocket science, however).. No vacuum needed in our case..

Hello

metal ( alloy ) for glass seal is Invar but there is different I link an interesting doc on wire glass seal

for low positive pressure and ambient temp epoxy works nice , even at high pressure epoxy is impressive just need a Male / Female fitiing in SS304 some bare copper wire and epoxy putty

Or that possible to use feed through as sale by TCdirect or other maker



Regards
OS

[Overspeed]

Argon has different thermal properties, which you may like better or worse. Just FYI.

Hello

I Agree regarding thermal property between Nitrogen and Argon , question is quantity will be low as cavity is small and pressure is low , I plan to stay in the 150KPa so around 20 PSI
Nitrogen is cheaper Argon is more common in industries as that use for welding

Regards
OS

[Gyro]

Argon is also easily available in small quantities, it is used for TIG welding non ferrous metals (select pure Argon rather than Argon/CO2 mix, both are sold).

[Overspeed]

Argon is also easily available in small quantities, it is used for TIG welding non ferrous metals (select pure Argon rather than Argon/CO2 mix, both are sold).

Hello

Yes a thing to know is the price of any welding gaz increase a lot when quantity is small , I use 3 m3 cylinder for welding and some smaller I have Argon but not laboratory grade argon , that not a real problem for preliminary tests as I suppose I will . can be in the case to reopen the enclosure and flush again

I will first make a ''seal proof test '' at 6 PSI to check the possible leaks

Regards
OS

[Conrad Hoffman]

For sealing wires in glass, look up Dumet wire and Housekeeper's seal.

[CalibrationGuy]

I'm all for a challenge,  but there's still the option of using a waterproof enclosure and a bit of desiccant.....

Just saying....

TomG.

[3roomlab]

that is a really chunky piece of metal box

it reminds me of the unit below

https://www.researchgate.net/publication/319890302_10_V_1_O_10_kO_high_accuracy_standard_setup_for_calibration_of_multifunction_electrical_instruments_and_for_inter-laboratory_comparisons

It's a shame the supply of NOS former Soviet hollow hermetic glass seals has dried up. They were ideal for running copper wires into sealed enclosures with no interruptions.

Hello

This kind of products is made  in USA from companies as https://www.completehermetics.com/hermetic-products-and-services/glass-to-metal-seals-and-feedthroughs

problem is the price $$$$$ as new and even on Ebay

Can be done with SS304 1/8 or 1/4 made in china and epoxy ... no so pretty but OK for a DIY or hobby purpose even  for a professional use as that a part for a show ...

I link my in process design with two valve one for flush on for nitrogen filling and a '' housing '' for desiccant 

Regards
OS

[dietert1]

I think this is not about buying some parts but about know-how. Just happened to read about "compression seals" and "matched seals". Reminds me of my early days at PTB. It can take a life to  build up a serious physics setup. There needs to be some clean room level and still humidity can evaporate from metal surfaces.
Looking at the humidity controlled bags we get when ordering certain parts from Farnell or Digikey, i think epoxy seals should be good enough and way better than using connectors. Desiccant inside ovenized hermetic enclosures still has its surprises for me. For example some people believe they should insert old used desiccant bags in order to enforce stable humidity right from the start. But that just isn't true. At least we now have those I2C sensors to monitor humidity inside a hermetic enclosure.

For the AD587 you just get the ceramic package and that's it. In comparison to a LTZ1000 it's a pretty noisy reference, even when you connected a good noise suppression cap. Geller had a patent for running the AD587 gain stage with a low noise zener.

Regards, Dieter

[Andreas]

Hello dietert1,

the Patend from Joe handles more about reducing the  T.C. of the AD587 than reducing noise.
https://patents.google.com/patent/US7382179B2/en

The LM399 which he uses with AD587 to get a temperature stable output has comparable noise (~0.6 ppm peak peak / V).
The advantage is having a temperature stable output voltage divider well below 1 ppm/K (in best case in a hermetically package).

see also here:
https://www.eevblog.com/forum/metrology/ad587lw-10v-precision-travel-standard/msg4142020/#msg4142020

with best regards

Andreas

[Overspeed]

I think this is not about buying some parts but about know-how. Just happened to read about "compression seals" and "matched seals". Reminds me of my early days at PTB. It can take a life to  build up a serious physics setup. There needs to be some clean room level and still humidity can evaporate from metal surfaces.
Looking at the humidity controlled bags we get when ordering certain parts from Farnell or Digikey, i think epoxy seals should be good enough and way better than using connectors. Desiccant inside ovenized hermetic enclosures still has its surprises for me. For example some people believe they should insert old used desiccant bags in order to enforce stable humidity right from the start. But that just isn't true. At least we now have those I2C sensors to monitor humidity inside a hermetic enclosure.

For the AD587 you just get the ceramic package and that's it. In comparison to a LTZ1000 it's a pretty noisy reference, even when you connected a good noise suppression cap. Geller had a patent for running the AD587 gain stage with a low noise zener.

Regards, Dieter

Dieter

Humidity is a tricky problem , oxygen is another

To reach a very low level of humidity there is some parameters to consider

The enclosure
The ambient atmosphere
The trapped humidity in the board ( under the IC , under the compoents , inside / between the wire ...)

The process I use on measuring equipment
 
I place all the parts in a pressure cooker , I warm at 75 deg C
I connect the pressure cooker to a vacuum pump
I open the valve to shallow the air / humidity inside the pressure cooker
I close the vacuum valve
I open the Nitrogen injection valve to fill the pressure cooker
I close the Nitrogen valve

At this level the parts are dry and in a Nitrogen atmosphere supposed with very low Oxygen level ( IF Nitrogen is over 99.99 ) be aware of lw quality Nitrogen as used to inflate tyres

I place the pressure cooker in a glove box ( keep dry with desiccant )
I open the pressure cooker
I can complete the assembly in a very dry place
After assy completion
I can flush and fill the enclosure with argon or nitrogen and remove from the glove box.

Right now I will not use this way I will just flush the enclosure on preliminary dried parts just to check if ''all is OK '' as a complete process is quite time consuming and my glove box is in use for another project

From a general view : humidity is a true nightmare and very difficult to measure at low level

Regards
OS
 

[CalibrationGuy]

@OS: In my home workshop where temperature is very stable but humidity is not, here is a result measured from a $100 reference bought on eBay.

The setup: eBay reference connected to Keysight 34420a meter next to Agilent 34970a with K thermocouple wire measuring room temperature. Everything is calibrated.

The result: Today's reading is 10.000097. Reading all winter was 10.000020. Almost 8ppm difference.

What changed? The heating system kept humidity at or below 30% RH. Now it is 55%.

Intrigued, I took the reference apart. It is based on the LT1021-10 in an SOIC8 package with the blue multiturn blue potentiometer for voltage trim, nothing special.

Yes, humidity matters. So for fun, I can place the reference in a plastic ziploc bag with some dessicant if anyone's interested.

Here's a picture.

TomG.

[CatalinaWOW]

This topic is perfect for volt nuts.  There are multiple possible reasons to care about humidity.

Physical stresses due to expansion/contraction of PWBs and other parts due to moisture content.  Presumably this reason only demands a constant humidity, absolute dryness isn't necessary.  There is the possibility of "tuning" for optimum humidity.

Moisture also is a factor in corrosion, leakage resistance and other factors.  In this case the lower the better. 

Attaining low absolute humidity is a major process.  Overspeed has touched on many of the points and described a relatively attainable home process.  One thing I think was not emphasized enough is how long it takes to remove moisture.  Captured moisture leaves via two pathways, I will call them leakage and diffusion.  Leakage is direct expansion through open passages, and can be slower than expected because those paths can be very small and convoluted.  Diffusion is passage of the water molecules through solid matter.  It takes much longer, in the case of metals essentially forever.  But diffusion rates through epoxy, rubber, plastic and other materials can vary widely and range from times measured in hours through times measured in years.  They obviously also depend on thickness of the diffusion path.

Sustaining a dry environment over time is another topic of its own.  Again leaks and diffusion are involved, along with the capacity (moles or other measure of water that the amount of desiccant can absorb).  Careful drying can reduce the how much of the desiccant capacity is consumed by residual moisture in the dry volume contents.  The modern moisture sensors are convenient as a rough check, but are literally orders of magnitude away from being able to measure a truly dry environment.    The interval is an interesting topic.  There is an argument for the remaining life of the owner of the device.  In any case times on the order of five to ten years seem appropriate.  The drying process is painful enough that doing it more often than that should be avoided.  Assuming that the initial dryness is good the life of the dryness can be calculated based on desiccant capacity and the total "leak" path, composed of the elements I called leakage and diffusion earlier.  I know of no way to estimate leakage.  In principal you could pull a vacuum on the chamber and watch the pressure rise (or pressurize it with clean dry gas), but for long life the rate of pressure change is too low to measure.  There is equipment to measure such leak rates (the helium leak rate detector mentioned above and the like) but most will not have this capability.  Diffusion can be estimated given appropriate material data on seals and their thickness along the diffusion path.  A practical solution is the very real uncertainty in the values involved is massive overkill on the desiccant supply.  Something like 80 percent of the total volume of the reference would seem like a good start.

All in all a very big side rathole for volt nuts to dive into. 
 

[Andreas]

The result: Today's reading is 10.000097. Reading all winter was 10.000020. Almost 8ppm difference.

What changed? The heating system kept humidity at or below 30% RH. Now it is 55%.

Hmm,

so roughly 0.32 ppm/% rH. Which is in the typical observations of 0.2-0.5 ppm/% rH for plastic packages.

If you want to do the test: the time constant is usually 3-7 days so it needs at least 2 weeks to settle to the new humidity level.

with best regards

Andreas

[Overspeed]

Hello

This equipment is use as a reference

https://us.flukecal.com/products/humidity-calibration/humidity-generators/5128a-rhapid-cal%C2%AE-humidity-generator?quicktabs_product_details=4

I have not succeed to download a service manual . problem is to generate and control humidity
OR
make reading and measurements  following mother nature goodwill .

Regards
OS

[dietert1]

In autumn 2022 i put some dry desiccant with a moisture sensor and some precision resistor arrays into a bottle with rubber seal. During the first 6 months the meter showed nothing (negative?). Recently i found that the coin cell of the meter needed replacement (working rapidly). Anyway, now after about 20 months the meter indicates 7 %RH. Will decrease another one or two % during the next week.
In an ambient of about 50 % average humidity this means a leakage & absortion time constant of 100 months, or 8 years. In other words, seasonal variations will get attenuated roughly by a factor 16 to 24. So inside the enclosure humidity will remain constant to about +/- 1 %RH once it has reached ambient average.
If the enclosure is an oven, the average inside will finally be below ambient humidity. The same water content gives less relative humidity at higher temperatures.
The IP66 aluminum enclosures i am using for precision electronics behave similar to the bottle due to their rubber seals and the electronics is made to last, with supply via epoxy sealed wiring. No need to open it once it is ready. The art is how to start the process in order to reach that final average humidity as fast as possible. Certainly not using dry desiccant!

Regards, Dieter

[Overspeed]

...
I am not paranoiac but I usually overkill problem when bomb proof solutions are cheap or available in the warehouse as surplus
I have no problem to machine a sealed container and flush it with Nitrogen
I am very upset regarding humidity trapped and that very difficult to get full dried parts and install them in an enclosure
Nitrogen is nor always Nitrogen
https://nigen.com/nitrogen-gas-purity-grade-specification-industrial-medical-food/#SnippetTab
So if that not OFN that does solve 100 % of the problem

If you're going to this trouble, why not use something completely non-reactive like Argon.

Hello

I just dont want to purchase ( or rent ) a 400 Euro cylinder + 100 Euro of Nitrogen fill just to use one or two liters of gaz for test . So Argon can do the job as
I have Argon available and right now I try to fight again humidity first .

I can add another pass through connector to connect a PT100 or even use a Pelletier element I have temperature controller Eurotherm available  and several Pt100 thermometer too .

Regards
OS

[CatalinaWOW]

For many of the effects of water it is absolute humidity, not relative humidity that matters.  In this sense it is like oxygen where the quantity matters, not how close it is to a state change condition.  So the reduction in relative humidity as temperature goes up is giving a false sense of security.

[dietert1]

My measurements indicate that higher temperature also "activates" the desiccant to absorb once more. I mean the reduction in relative humidity due to temperature increase is immediate and included in the sensor firmware, while the observed effects are slow and much larger than expected.
There could also be kind of poisoning of the sensor by the desiccant. I started some oven experiments to understand this better.

Regards, Dieter

[Overspeed]

My measurements indicate that higher temperature also "activates" the desiccant to absorb once more. I mean the reduction in relative humidity due to temperature increase is immediate and included in the sensor firmware, while the observed effects are slow and much larger than expected.
There could also be kind of poisoning of the sensor by the desiccant. I started some oven experiments to understand this better.

Regards, Dieter

Hello

I agree with yous statement on poisoning as some desiccant are not stable under high temp even ambient temp
The best neutral desiccant I use is wood charcoal for water filtering packed in coffee filter to avoid dust , silica gel is also an option if that pure product with additive .

https://www.agmcontainer.com/product-category/moisture-control/bulk-desiccant/?msclkid=5e853b16fedc1acada465d439b70d207&utm_source=bing&utm_medium=cpc&utm_campaign=**LP%20DSA%20-%20Desiccant&utm_term=desiccant&utm_content=Dessicants

I far prefer heat + vacuum as the NO chemical risk for preliminary desiccation , That important to not saturate the desiccant inside the enclosure but use it as a ''safety belt ''

Regards
OS.

[iMo]

The result: Today's reading is 10.000097. Reading all winter was 10.000020. Almost 8ppm difference.

What changed? The heating system kept humidity at or below 30% RH. Now it is 55%.

Hmm,

so roughly 0.32 ppm/% rH. Which is in the typical observations of 0.2-0.5 ppm/% rH for plastic packages.

If you want to do the test: the time constant is usually 3-7 days so it needs at least 2 weeks to settle to the new humidity level.

with best regards

Andreas

I can see that nicely with my 34401A (while measuring the ADR1001 10V ref, always on). After say 3-4weeks the meter is off, firstly the voltage it shows goes up by some 1.5-2.5ppm (after first 24hours on), then slowly decreases to the final state (takes aprox 10-14 days). That is due to the humidity, imho, as the temperature inside the meter goes from ambient to ambient+18.5C, moreover I've been using a thin "foam sleeve" beneath the top cover which "seals" the inside of the meter a little bit too (and keeps the moisture as well).. 

[CalibrationGuy]

Just to rule out temperature, I allowed the space to get as hot as it ever would when the heat was on in winter. And here are the results: 10.000106Vdc at 77.2 DegF. Humidity is the same, 55% RH. So 5.6 degrees temp rise gives 0.9ppm voltage drift. So most, but not all of the drift is humidity related.

Picture below.

PS - I suggest that we start another thread devoted to humidity effects on the various reference circuits because this thread didn't start with that topic.

@iMo: Meter was on long before I turned the reference on to stabilize. This meter tends to drift down a little from a cold start.

TomG.

[guenthert]

@OS: In my home workshop where temperature is very stable but humidity is not, here is a result measured from a $100 reference bought on eBay.

The setup: eBay reference connected to Keysight 34420a meter next to Agilent 34970a with K thermocouple wire measuring room temperature. Everything is calibrated.

The result: Today's reading is 10.000097. Reading all winter was 10.000020. Almost 8ppm difference.

What changed? The heating system kept humidity at or below 30% RH. Now it is 55%.

Intrigued, I took the reference apart. It is based on the LT1021-10 in an SOIC8 package with the blue multiturn blue potentiometer for voltage trim, nothing special.

Yes, humidity matters. So for fun, I can place the reference in a plastic ziploc bag with some dessicant if anyone's interested.

Here's a picture.

TomG.

   I bought the same VRef about 2 years ago.  Initially I thought it performed alright (just whining about the unfortunate choice of connectors), given its price.  However I recently noticed that it takes an awful long time (multiple days) to settle to a semi-stable output when off for a prolonged time (weeks in my case).  I assumed this to be humidity related, slowly driven out when running (a little warmer than the environment).  I haven't opened mine yet, but would be surprised, if it's not a plastic packaged LT1021.  I think this unit is particularly sensitive to humidity.  I have an older (presumably) AD588 based unit off eBay ("KKmoon") which settles within minutes after turning on.

In the graph red and blue are readings of a temperature sensors, purple is the difference voltage to a Fluke 515A calibrator (which I leave running).

[iMo]

It has no sense to mess with plastic references. It had been said many times in past here.
Otherwise you will waste huge amount of time with rather questionable results.
Of course, you may continue with epoxy experiments, it is a nice area of basic research on humidity effects.
All those epoxy Vrefs out of ali and ebay are simply a joke, nobody from this section would trust to their performance.

PS: all those plastic/ceramic references were designed ~45y back mainly for at that time mainstream high end 12bit ADC/DACs. Even an epoxy reference can work fine in that application. The "high-end ovenized hermetic LM399" was developed also at that time for up to 14bit (16bit??) ADC/DACs, afaik.