Proto 10.000000 0.1PPM with Ref102

[eurofox]

I'm experimenting with the Ref102.

I got very surprising results using low PPM resistors and 10K high precision NTC.
Short term: 1 day 0.1PPm!

The NTC is located under the Ref102 IC.

2 trimmers, 1 for the voltage and 1 for the NTC.

I will make more test on long term but this is surprising, normally it is specified 2.5PPM/°C!

[Rafael]

I was trying to save a project like this... Not tested and capacitors are wrong in the image.

NTC under the REF...

[Andreas]

Hello,

cirquit diagram?
photo?

The 2.5 ppm/K are a spec over the whole temperature range. (box method).
Local T.C. around room temperature can be lower or larger than the "average" value.
Sometimes you have maximas and minimas within the temperature range due to T.C. trimming of the reference.
(If you have luck it is around room temperature).

If you have trimmed T.C. the larger stability problem will be humidity (%rH) with plastic packages.
I have measured around 0.5ppm/%rH on 2 LT1027CCN8 (DIP8) devices as seasonal change.

with best regards

Andreas

[HighVoltage]

I got very surprising results using low PPM resistors and 10K high precision NTC.
Short term: 1 day 0.1PPm!

With what instrument did you measure this?

[eurofox]

I got very surprising results using low PPM resistors and 10K high precision NTC.
Short term: 1 day 0.1PPm!

With what instrument did you measure this?

This one:

[eurofox]

Hello,

cirquit diagram?
photo?

The 2.5 ppm/K are a spec over the whole temperature range. (box method).
Local T.C. around room temperature can be lower or larger than the "average" value.
Sometimes you have maximas and minimas within the temperature range due to T.C. trimming of the reference.
(If you have luck it is around room temperature).

If you have trimmed T.C. the larger stability problem will be humidity (%rH) with plastic packages.
I have measured around 0.5ppm/%rH on 2 LT1027CCN8 (DIP8) devices as seasonal change.

with best regards

Andreas

Before posting schematics I will do more testing, is just the second design with the Ref102.
I'm testing as well LTZ1000 design and I'm surprised what you can do with the Ref102 and a few low temco components.
 
Humidity can be solved with an hermetic metal box filled with argon.
Maybe soak the IC in epoxy and of course leaving the pins protected, I don't no if someone already try this.
Could be as well in heated box.

[cellularmitosis]

My understanding is that epoxy just increases the time constant for humidity to affect the part (though I'm working on a way to actually get some data on this).

If you don't mind a bit of periodic maintenance, you could always rotate out a fresh set of silica gel beads every X months in a resealable container.

[eurofox]

My understanding is that epoxy just increases the time constant for humidity to affect the part (though I'm working on a way to actually get some data on this).

If you don't mind a bit of periodic maintenance, you could always rotate out a fresh set of silica gel beads every X months in a resealable container.

I got several experiences in mind but the condition is that epoxy "seal" the IC for humidity.

Solder the IC on a "tulipe" socket and completely soak into epoxy, could add a NTC under the IC, this way to should be thermal and mechanical bonded.

With the NTC I will continue my test to change the correct voltage with temperature.

I will start an other test if epoxy work for humidity by bonding a NTC and a resistor and warming the IC to the "sweet spot temperature" where the Ref102 is the most stable (36-37°C).

[rhb]

I think a soldered metal enclosure would be more suitable for controlling humidity.  Maintain a fairly hard vacuum for several hours or more and then seal.  Use silicone thermal grease to reduce temperature gradients.  In the case of epoxy encapsulated parts, it might be desirable to hold under vacuum for several days at elevated (40 C) temperatures to dry out the epoxy prior to final assembly.

Standard  practice for drying out refrigeration systems is to hold a 0.5 micron pressure for an hour.  This is both to verify seal and to allow moisture to evaporate.  That's probably a bit extreme for this application.

[Conrad Hoffman]

AFAIK, there is no coating or potting material that's humidity proof, only humidity resistant, and that just changes the time constant. Back filling a sealed metal box with dry gas isn't that hard and you should also include a desiccant, probably a well baked 4A molecular sieve.

[Vtile]

I wonder what could be done with 8" long glass tube. Covar and reference in one end and then turned to vacuum tube. 8" should be long enough to prevent the reference end to heat excessively while the another end is melt to seal the vacuum.

[Gyro]

You'd do much better (easier) with copper or brass foil, soldered, with hermetic feedthroughs. Hollow solderable feedthroughs are available on ebay, you can pass your copper wires through without interruption and so avoid thermocouple effects. Soldered seams count as hermetic.

[mycroft]

Before posting schematics I will do more testing, is just the second design with the Ref102.
I'm testing as well LTZ1000 design and I'm surprised what you can do with the Ref102 and a few low temco components.
 
Humidity can be solved with an hermetic metal box filled with argon.
Maybe soak the IC in epoxy and of course leaving the pins protected, I don't no if someone already try this.
Could be as well in heated box.

Are the schematics available?

[arcnet]

There is actually a voltage standard using the REF102. Many of them
140 REF102 in four blocks of 35. Each block in an oven and heated to 44 °C +- 0.03 °C.
The circuit seems simple: Just 35 references in parallel (100 Ohm 0.1% resistors) and a 15 V +- 10 mV programmable supply for each block.
Result: "A proposed multi-zener based voltage-standard-source was analyzed during 11 years of operation. Its
internal noise was evaluated in 0.03 µV/V at one year averaging time, and 0.005 µV/V at 4 h
averaging time. The maximum variation during 11 years was 0.16 µV/V, with a drift of -0.009 µV/V
per year which is well below the calibration uncertainty." 1)
Four calibrations/comparisons were done against Josephson standards.

1) "Eleven years of monitoring an ultra-stable 10 V zener-based voltage standard", Souza et al., 2015
https://www.researchgate.net/publication/305800682_Eleven_years_of_monitoring_an_ultra-stable_10_V_zener-based_voltage_standard
2) "FUENTE PATRÓN DE TENSIÓN BASADA EN MÚLTIPLES ZENERS" Slomowitz et al., 2005
http://iie.fing.edu.uy/publicaciones/2005/STD05/STD05.pdf

[mycroft]

arcnet, thank you for the links. From this thread, eurofox's circuit seems simpler and very stable. Using a single NTC for compensation to 0.1 ppm is intriguing.

Regards.

I'm experimenting with the Ref102.

I got very surprising results using low PPM resistors and 10K high precision NTC.
Short term: 1 day 0.1PPm!

The NTC is located under the Ref102 IC.

2 trimmers, 1 for the voltage and 1 for the NTC.

I will make more test on long term but this is surprising, normally it is specified 2.5PPM/°C!