Ultra Low Noise Reference 2DW232, 2DW233, 2DW23x

[Alex Nikitin]

Hi Alex,
from yours first graph I see two dip in zener voltage and its related to increase in temp but then when temp get down also the zener also lowering. did I missing something ?

Look at the second graph - the zener voltage crosses a "zero tempco" point about 21.6C and on both sides from that point the voltage will go down as the tempco changes sign.

Cheers

Alex

[Gyro]

Look at the second graph - the zener voltage crosses a "zero tempco" point about 21.6C and on both sides from that point the voltage will go down as the tempco changes sign.

That first plot is a really neat graphical demonstration of the TC turnover characteristic, I bet you couldn't have come up with a better temperature profile if you'd tried. 

[Kleinstein]

The data at 1.4 mA look quite good, but to get really good stability, temperature stabilization would be needed. The curvature is quite large and thus a low TC is only there in a rather limited range (e.g. +-5 C). As the ref. is low power one could go for something like 50 C. I am not so sure the internal sensor alone is really useful for regulation, due to a relatively slow response.

The second point that might be worth a test is to see if there is hysteresis after larger temperature excursion (e.g. heating to something like 60 C).

A third point would be to check how much the compensating point varies from sample to sample - worst case other samples might have compensation at to low a current.

[zlymex]

I suggest to perform an accelerate aging before actual test, which can be done in three ways as described by a manufacturer:
1. high temperature storage, 150 degree C for 120 hours
2. room temperature storage, for a month
3. power aging: run a 40mA to 50mA current through the device for 24 hours

[Kleinstein]

To get an estimate how much aging is expected, it would be a good idea to have a measurement before accelerated aging too.

The three steps might not be alternatives, but separate steps to do the aging. I don't think just a month a room temperature should do much alone.

For the 150 C step is might be important how fast or slow the units are cooled down. To get a stable state is might be better to do slow cooling in the 120 - 80 C range. In this range there is the glass transition of epoxy glues, and fast cooling can leave it in a less stable state.

[pmcouto]

I received my 2DW232 yesterday.

Thanks to Zlymex for providing the parts and to VintageNut and Plesa for handling the distribution. 
As Plesa suggested, a donation to Unicef has been made.

[branadic]

Thanks from my side too. Have received the parts yesterday.

[SvanGool]

Me too.

Thanks Zlymex, VintageNut and Plesa ! 

[bertik]

Yes, me too got a few. Thanks so much!

Now I am preparing some temperature scanning, after noting down initial "not yet aged" voltages at fixed temp and current.

Now the question: are the two diodes back to back meant to be equivalent?  The voltages do differ open swapping pins. If not, which diode is the "right" one to choose?

Somehow the physicist inside me would conjecture, based on symmetry,  that stability is max when the two voltages are identical.....

Thanks,
BK

[Kleinstein]

I would expect the diodes to be nominally the same. Just like there is scattering between chips, there can be some between the two on one chip. Though I would expect not that much difference for the two on one chip.

[Alex Nikitin]

As we have two zeners in one package, it would be good to use both and to use the case as a screen, I plan to do a proper layout for the circuit attached below. R6-R9 are from a single  LT5400-B1 array, providing a good tracking and long term stability, R1 and R2 are selected to pull apart the "zero tempco" points, providing a lower combined tempco, and the voltages from both zeners are added together, further lowering the noise. R5 is selected/adjusted to get exactly 10V output and the tempco/stability contribution of R3-R5 network is about 1/20 of individual resistors, so a quality metal film type should be OK.

Cheers

Alex

[Kleinstein]

I don't think it will help much for the TC to have two diodes combined with slightly different compensating temperature. They will both have a second order component of similar size and just slightly different linear TC. So the sum will be essentially the same as the curve at intermediate current. If at all the diodes would need more coupling to get any nonlinear effect - but than a difference in absolute value would be important.

Due to the relatively large curvature, the best option for a really stable output is in using temperature stabilization. So the second diode would be valuable as a temperature sensor. Even without temperature regulation, just a temperature reading would make current adjustment much easier and allow for later compensation. For use without temperature stabilization, the series version with higher current might be more useful.

Judging from the DS curve, the second order TC seems to be smaller there. Also adjusting the current could be less critical.

For the circuit around it, I do not really like the idea of starting with negative reference and than add an inverter. It needs even more stable resistors. I could be Ok if you want a +-10 V reference of cause.

The more natural way would be to use two of the 5.25 V refs in series to get a 10.5 V output.  A possible higher output would be only for biasing. The alternative bridge version does not even need the higher voltage. The appealing point is that a divider 10.5 to 10 V is not that sensitive to resistor drift or TC - so even with 20 ppm/K resistors the output would only chance by 1 ppm/K.

[Alex Nikitin]

I would rather get a lower noise. The tempco is not very high around the "zero point" anyway, so the temperature stabilisation does not require to be very accurate and can be done with an external sensor. The dissipation is low (total ~15mW) and a temperature around 50C with 0.1C stability should be fine. Two zeners in series is a good idea (using both zeners in each package you would get a half of a single zener noise), but it would require isolation between two cases and a good thermal contact at the same time. It is much simpler if the case is "grounded", and I could use my favourite LM35D in a metal can as a temperature sensor sitting in the same copper bar, providing a very quick and stable regulation.

Cheers

Alex

[Fixup]

Hi,

I also received the references on friday
Thank you Zlymex, VintageNut and Plesa.
Also nice of you to suggest the compensation to Unicef!

Now I need to start aging the references while I'm waiting for my "new" meter to arrive (Solartron 7075)     

[edavid]

I would expect the diodes to be nominally the same. Just like there is scattering between chips, there can be some between the two on one chip. Though I would expect not that much difference for the two on one chip.

Do we know that both diodes are on one chip?

[Kleinstein]

The very first post in the thread has a picture from the inside. So at least the units on the picture look like 2 more or less identical parts on one die. However other manufacturers might have two dies of even two separate capsuled diodes.

[plesa]

I'm going to make die photo of 2DW233, so we will see. It seems to be covered by some epoxy, which require some nasty etching process by hot nitric acid or sulfuric acid.
Thanks to all of you for donation to UNICEF

[klaus11]

Hi,

I also received the references and donate Unicef.
Thank you Zlymex, VintageNut and Plesa.

[bingo600]

I also received mine , and donated to Unicef.

Thanx goes to all that made this happen, Zylmex Vintagenut Plesa et all. 

/Bingo

 

[Squantor]

Also adding my thanks here

Zylmex, Vintagenut and Plesa, Thanks for all your hard work and effort on making this possible. Had a nice weekend with playing with these devices .

[Fixup]

Hi,

Today I received a book (Calibration philosophy in practice second edition).
In that they had described little about zener references that confirms that the other diode needs to be in series with the other zener diode.
There is also some info about the 732B.

Pics attached

[2N3055]

I received diodes today too...
I wasn't home so I had my wife open it..
She goes .." wow, this is something packed like some kind of sweets, a desert... Did somebody send you holiday sweets ? ...".
I was chuckling.. " ... better than that honey, better than that..."

Thank you Zlymex, VintageNut and Plesa.

I also donated to Unicef..

[Alex Nikitin]

Hi,

Today I received a book (Calibration philosophy in practice second edition).
In that they had described little about zener references that confirms that the other diode needs to be in series with the other zener diode.
There is also some info about the 732B.

Pics attached

I have this book too - it is a great read. I should however note that the book describes zeners with 6.2-6.3V voltage with a diode in series. The 2DW232 has a zener voltage about 6V with a diode in series (5.25V without a diode) , much lower, and so the tempco of a zener only is closer to zero unless you put a lot of current in and elevate the temperature.

Cheers

Alex

[Seekonk]

In looking at that 7-8 picture I thought about some old General Electric RA1 that I have.

[Alex Nikitin]

I've connected the second zener in the package, and it has a slightly higher current for the zero tempco point - about 1.7mA . Here are two short term (15 min each) measurement runs with HP3458A at NPLC1 and NPLC10. Vertical scale is 1uV (~0.2ppm) /div. Lab temperature was reasonably stable at about 24C. The trendline on the second graph is 10 per Mov Av, so should be equivalent to NPLC100, same as the trendline on the first graph. Looks different though... .

Cheers

Alex