KZZ82 Voltage Reference

[iMo]

Hi, as I cannot afford the ADR1000 or ADR1399 refs, I decided to try the hard way 
Here are the 4x KZZ82 voltage references (TESLA Czechoslovakia), some 60+ years old, afaik, I got from a friendly expert in past.

The datasheet claims interesting numbers like Tc <10^-6/degC from 0-50degC..

I quickly tried to measure their Vref by applying the required current (depicted on the box), but I was not able to find their knees 

I think Jaromir did some measurements on his parts in past (he told me as we met together in Monte Carlo on the Metrology Congress some years back) and he saw some 4-5ppm/degC.. Maybe he can comment then..

I was thinking to send a sample to Noopy for a shot, but then I decided not to do as it would create a big mess under his microscope. A friendly expert told me inside the box are large metal-glass diodes and a wire wound resistor, box fully filled with a fine sand..
 

[Kleinstein]

The data on the references look sufficient to operate them. However a current of some 60 mA at some 8 V is about 500 mW and thus quite some heat.  With the relatively large case this will take quite some time to stabilize. Chances are they want to be mounted on a heat sink to at least speed up the termal response a bit and keep the temperature reasonable.

Providing a stable current is also a slight challange, as the interal resistance specs are not that good: 8,2 / 65 mA is some 125 Ohms of effective resistance. Compared to this the <15 ohms specs are not good and up to 1/8 of the current variations would get through to the ouput voltage. Depending on how bad the differential resistance really is, it would need rather stable resistors in the external circuit too.

This does not look very practical for normal use, more like rather special lab use. 
I don't think this is a practical divice today, more like a curiosity from the past.

If one can get one, the ADR1399 is not that expensive - just rather poor availabilty.

[maxwell3e10]

I wonder if with such large zener current and big thermal mass it could have low 1/f noise...

[jaromir]

I think Jaromir did some measurements on his parts in past (he told me as we met together in Monte Carlo on the Metrology Congress some years back) and he saw some 4-5ppm/degC.. Maybe he can comment then..

Yes, a few years ago I measured them using lab power supply and series resistor to set the needed current, the results were unsatisfactory, so I gave them another chance.

I remeasured 6 specimens of those reference diodes. It was done in two runs, they were arranged as a series string of three references, powered by DIY SMU in constant current mode, at the current prescribed by the reference sticker. Voltage on each diode was measured using Keithley 2010 switched by Prema 2040 switch, with frequency one sample per minute.
The reference string was inserted into the oven and temperature swept from 37 to 57 degrees C, in order to examine both tempco and drift/hysteresis.

Observe attached graphs, please. Each graph captures the voltage history of one diode - expressed as deviation from initial value, in ppm of initial value -  over two temperature sweeps (actually, staircases). One sample per minute.

I've never seen voltage references as random as those. In theory, the tempco is somewhere between 4ppm/C to 400ppm/C (one to three orders of magnitude worse than datasheet values), but the references are drifty as hell. Not sure if letting them more time to settle down will improve anything.
Perhaps the specimens I have were damaged or improperly handled over the years.

[iMo]

Congratulation to your new measurement setup! Interesting findings.

I was told those refs include the old bulky xNZ70 type zeners in metal and the two diodes in similar metal package (and a wirewound resistor).

Even though the diode packages are hermetic and the outer box is hermetic too it could be the contacts under the nuts have been oxidized, or something like that.

Out of curiosity I've quickly wired one of my sample under 66mA nominal current and the voltage is 0.45V higher than on the sticker. That is too much, imho.
The nominal voltage I get is with 42mA. The ref's internal dynamic resistance around the 66mA is 12ohm (at least this fits the datasheet).

PS: I could not resist the temptation to make a sim 
Here it is something like the KZZ82 with modern diodes, the zener is a modded diode I used to use in past. It shows TC min at 54mA current, the dyn resistance close to KZ, the TC close to 4ppm/K.

[iMo]

The sample above was broken - a much higher voltage than expected and it was quite sensitive to knocking at the box, and a pressure on the box created weird jumps in the voltage.

The other my three samples are better - a quick measurement shows the difference against the sticker voltage from 60mV to 10mV (after 4 hours of operating under the nominal current).

Therefore I decided to open the mystery box and look into the broken reference.

Warning: Do not do it at home!!   

Thanks a friendly ham we did it in a controlled environment, wearing respirators (really)..

The red hermetic box is made of a thick aluminum, coated with a black plastic isolation from inside.

The components are potted into a shiny white extremely (!!) fine powder, almost like a smoke. I knew that already, therefore we were pretty careful with handling it from very beginning.

The white powder was pressed inside such it looked as a ceramic, when we started to dig into it started to disintegrate. Finally we washed the components in water.

There are 2 bulky diodes and the zener, plus the wire wound resistor mounted on a 3mm thick plate (aluminum), no markings on the components. There is the box serial number written on the plate as well.

See below the pictures. A real archeology, indeed 

[Kleinstein]

My suspicion for the bad contact would be the ends of the wire wound resistor. Many of the resistor wires are pretty difficult to solder and may very well result in a kind of "cold" solder joint.