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

[branadic]

While it got pretty quiet here and since I finished some of my projects, I started again playing with 2DW23x. Last time I wasn't successful with the temperature compensated and temperature stabilized version. This time I use the single zeners only.
The sample under test is a common cathode one, which means the case is connected to cathode. First, I've measured the t.c. of the individual zeners, with the second zener shorted, using my AN3200 set to 10 V as a source and General Radio 1434-G to vary zener current. I used a relatively fast temperature slope to speed up things (1K/min) and to make some progress.
Afterwards, I paralleled both zeners, halved the resistor and varied the zener current in small steps. I use a DIY TEC mount with a 8.2 mm drillhole in a brass plate, in which the TO case fits, but also a GA10k3A to control it via Arroyo Instruments 5305.

Attached are the first results of the t.c. measurement. Currently, I monitor stability. My plan again is to use QH40A crystal heater to temperature stabilize the zener. ADA4523 is planed for the gain stage. For the resistor divider I'm yet unsure as a gain of 1.9 is required. Many different option come into mind, although LT5400-8 could come in handy (9k:10k), but I would have to order some.

-branadic-

[dietert1]

In order to get 10 V you could use two of those in series and then divide 10.5 V down to 10 V, with less requirements for the divider and 40 % less output noise.
Already for a long time i wanted to test an opamp driven bridge arrangement with two 5 V zener - resistor pairs. I ordered some glass zeners to find out how stable they are. Didn't know the DW232 can be used to make this.

Regards, Dieter

[analityk]

Hi,
it is never ending story.

Currently i have some free time (i have looking for job now) so i let start to experiment with my 2dw236 diodes.

I will show you my approach to stabilisation of temperature of 2dw23x. You can see top layer PCB, U4 is ref to 2dw23x but zener is placed upside down on this two heating resistors, between them is NTC for temp. stabilisation. Between U4 and ntc, as top silk show, is placed thermal pad for better heat exchange.
On bottom side of PCB was placed 6 300R resistors for some heat up all resistors and initial heat up of U4.

Firs task was set current for U4 to achive tempco in resonable high, but not to high temp. Then i perform seting up of divider for obtain 10V on output. Lastly i have change R24 and R30 resitor to achive temp. with zero tempco of zener.

It cost me a lot of time ofcourse but i think it is possible to build some better PCB and some tester/calibrator for it. Maybe in future.

My 2DW236 have unusualy big zener voltage, something like 6.45 V even with some small current like 2 mA. If i good remember my diode work with current in between 2 mA to 3 mA.
Also power supply isn't super duper, it is ordinary transformer (230 V -> 14 V) and full bridge rectifier, then is 7815 for heaters, 33mH choke (100R, it act like fuse/current limiter btw. Removing it was a bad idea, output was noisy) and 7812 for powering OPA189.

All of this stuff was enclosed in aluminum case, with good thermal foam guled to wall although it is not hermetic.

Last question is when voltage stop dropping. This one diode is in operation since 200h in temperature close to 70 degree of Celsius.

Sincerely.

[Kleinstein]

The drift looks rather large and with so much drift / noise to start with I would not expect it to get very stable over any sensible time. A question may be if the drift is from the zener itself or maybe the resistors or maybe coupling from the heater current at the ground side of the amplifier.  It may be worth looking at the raw zener votlage and not only the 10 V output.

[iMo]

Also your 10V output is not short proof, imho. When 10V output is shorted out the opamp will push large current through your low value resistors and the base-emitter of your Q4 transistor.. I would add a reverse biased diode in series with the opamp's output at least..
Better to also add the transistor for the current limiting (like aprox 30mA below).

[analityk]

Thx, I will change a bit my power supply. I will add capacitance multiplier, some choke etc.
Today I test 18 diodes and I find zero tempco for each with 4.1 mA but some diodes reach this zero plateau in 60 degrees of Celsius but one diode will work up 102 degrees of Celsius. I suppose this one need smallest current in given temperature to work with constant voltage. I find that in 2.3 mA I need stable temperature equal 60 degrees of Celsius and it will be easy to test. I will post some graph in near future.

[Birb]

Hi, it has been a while. I just found out about this voltage reference. Does anyone still have the datasheets?
Specifically, I obtained some 2dw231, and would like to find out the zeroTC current.
I also want to know if the 2dw231 is similar to the others. It seems to feature a higher impedance, with a lower voltage.
Also, does parallelizing voltage references reduce drift? (Some may drift upwards and some downwards, so on average it should balance out?)
Though the zeroTC current may be different for each zener diode.
I did manage to find a small datasheet, which seems to claim the default TC of the 2dw23x to be 5ppm/K, except 2dw230.
The max power dissipation is 200mW, max current is 30mA, and max temperature is 150C.
However, notice how the Iz part seems to have 6 entries despite the 7 items on the left.

[Gyro]

Hi, I've attached the three datasheets that I have, covering the 2DW23x. They are a mixture of Chinese and English.

[Kleinstein]

Parallel connection helps with noise, but little with long time or thermal drift. The long term drift is likely similar and thus would not average out. The thermal drift depends on the current setting and should follow a similar curve.  As the diodes are usually at least low noise, there is little need for a parallel combination. Multiple references would be more a thing for a plausibility check and detection of outliers that drift more than others. This would be more like a later step if really needed.

The Zero TC current seems to scatter quite a lot. There is a chance to have a slightly lower current for the DW231 than 232. I would still not count on it to be close to 5 mA.
For a quick test one can look at the voltage for the first few seconds after connecting the current. The self heating would cause a drift in the voltage, if the TC is not zero.
This at least works OK for the coarse part (e.g. +-10 %) to see if the diode is at least reasonable (e.g. current between 1 and 10 mA) and which approximate range.

[dietert1]

The setup to determine the zero TC operation current for a given zener reference needs a precision meter, an oven with temperature control or at least temperature measurement and an adjustable precision current source. Using the setup one can determine the operation current for each part. This will be more reliable than using numbers from a datasheet. Such work was described in this forum before. As far as i remember getting better than a LM399 required quite some effort.
Only after having executed the procedure for a set of parts one can tell whether it was necessary or not for the intended use of the parts. I can't recommend using parts from unknown sources without characterizing them.

Regards, Dieter

[Birb]

Would using a dual op amp based current source work? (like the attached photo)
However, a stable 10V is needed. Could a lab bench supply suffice?
I am thinking of using op07 instead of op77. Resistors with lower tempco will be used, and placed externally.
For the oven a peltier with a peltier controller will be used. Hopefully sub 0.1C accuracy.
Thanks

[Kleinstein]

The normal solution for the current source is an amplifier stage from the 6.6 V to some 10 V  (e.g. a gain of 1.5) and than a simple resistor from the 10 V to the 6.6 V zener level. So only 1 OP needed. There is no real need for an extra current source with high output impedance, as the zener voltage is rather low impedance and a stable voltage.

A lab supply may not be good enough, the quality can differ quite a bit however.  For the tests a rather simple amplifier on a bread board should be OK for the first test however.

If the current is adjusted to get close to the zero TC point, the temperature does not have to be super stable. The idea than is having may be a TC of less than 1-2ppm/K at the set temperature and a temperatuer stability of some 0.1 to 0.5 K.  The better the current is adjusted the less critical the oven gets. Ideally the oven would mainly fight the 2nd order TC and this is relatively easy.
If needed the oven (with an adjustable set point) could also be used for the final chaeck on the zener TC.

[iMo]

For example a schematics for your initial experiments.. The OP07 may need an output transistor buffer to boost the 2DW231 current..

[dietert1]

Yes, the proposal to use a circuit similar to the final one can save on precision equipment. One could use a good DAC to adjust the 22K 10K divider in order to make the current adjustable with a given fixed R4. Voltage measurement would be at the zener, not on the "10 V" bootstrap output.
One could also look at the forward voltage of the diode inside the reference during the temperature sweep in order to improve the temperature measurement.

Regards, Dieter

[iMo]

.. or a 10k 10T WW pot..

[GigaJoe]

btw ....  does anyone got a suitable result as a precision reference ...  all my experiments ended by result of sporadic drifting up and down randomly. up to 50ppm/month depending on quality i guess ....

considering that some my samples diode was burned for a year or more ...

even some zeners after burning had some funky behavior of jumping to a 20-50 microvolts than stay at this level for a while (couple hours , or day ), than randomly jump again to another level up or down and so ...

[Birb]

Supposedly, in this datasheet, it claims a long term stability of 120uV per 8hours. So it probably isn't that stable.
I guess it can only be used as a low cost short term reference

[iMo]

Those diodes are ordinary zeners, imho, not the buried zeners like the 1000/1001/399/1399/1021/1236 etc.
So much more sensitive to environmental effects, thus the jumps most probably coming from the surface contaminants..

[GigaJoe]

120uV per 8hours.  -  can be if brand new  ....

i have around 30 that run over the year ....

from set of 8 (if i remember) i pick 3 , under condition:  2 internal in parallel, current flow - minimal Temp-drift.
it simply lying on the bench and for  2-3 days didn't shift at all,  basically 5.4XXXX stay the same. no PPM shift
so i made a low noise PS , like 0-15V 200mA,  with Zener adjusted to min tempco

the result was so good , what after warmup , i can measure , fraction of microvolts using that PS as reference by diff. measure for a day it was stable enough .....

unfort ... had a design mistakes , and pre-regulator bad, that are diff story ....  so i going to make another one as a low noise voltage source.

[Birb]

Hi, random question regarding the pinout.
For some reason, the pinout for the zeners I received have the + at the pin connected to the case, - at either two terminals. This is the only way I can get a ~5.7V output.
Based on resistance measurements, the two non-case connected pins read open, whereas ~20M ohm from either terminal to the other two pins.
Is this normal?
Thanks

[chickenHeadKnob]

Hi, random question regarding the pinout.
For some reason, the pinout for the zeners I received have the + at the pin connected to the case, - at either two terminals. This is the only way I can get a ~5.7V output.
Based on resistance measurements, the two non-case connected pins read open, whereas ~20M ohm from either terminal to the other two pins.
Is this normal?
Thanks

Yes, there are data sheets showing both common  anode or common cathode bonded to the case. It is a part made by multiple different manufactures, each with their own idea of a pinout. Only select parts made in one factory are good, the rest are junk.

[Birb]

Main oddity is that this pinout is consistent across several sellers.

[Birb]

.. or a 10k 10T WW pot..

Does the circuit still work if both cathodes of the zener diodes are connected together? I say this because for some reason all of the 2DW23Xes I have appear to share a common cathode.
Thanks.

[Kleinstein]

The intended use is with the 2 diodes in sereies with one in forward and the other in zener mode. So the normal circuit does not care wether the diodes are connected at the cathodes or anodes. It would only be circuit versions that use the forward biased diode also as a temperature sensor that do care about the direction.