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

[Andreas]

it seems all long term  stable zeners in use technique called buried zener ...

no,

buried zener is generated by a special ion implant method
(which is more costly than a planar zener).

with best regards

Andreas

[branadic]

Corrected the table in the translated datasheet in post #507

Currently I do some zero t.c. current measurement in the intended configuration on one device, while I'm still not sure if the zener is connected correctly or if I really understood the datasheet. Pin 1 (red dot) is used as the output, while pin 2 is connected to ground and pin 3 is left open.
I use a QH40A crystal heater mounted on a flat of an aluminium rod. The rod has a through hole of 8.2mm in which the diode fits in. Further I use a SMT172 sensor in TO18 package, that is thermally connected  to the reference diode. Attached are the first results of my measurements.

-branadic-

[branadic]

For me it looks like the datasheets available on the web have nothing to do with the 2DW232 devices on my desk. I can't find a resistor value that leads in a t.c. as low as 5ppm/K.
It's the first device I'm analyzing, but it leaves a bad tasting. As far as I can see Wolfgang investigated on 2DW233 and 2DW234 and found also values, that have nothing to do with the datasheets.
Maybe I get the datasheet wrong? Any comments on that?

Added first results of different specimen @ approx. 5mA.

-branadic-

[Kleinstein]

One problem with the data sheets is that there seem to be several manufacturers (and additional fakes) for the nominal same part. However the parts seem to be quite different, with higher and lower noise parts.

Other here in the thread have have also observed that it would need quite a bit more than the nominal 5-10 mA to get a zero TC. So the shown curve are kind of typical for the part seen so far, though there where a few to reach zero TC below 20 mA.

At least with some other parts that had such a high current in series mode, there is a second option to use the reference. When using the zener diode alone, chances are the TC can cross zero at some 1-3 mA. So those diodes that would need to much current in series mode might still be useful in single diode mode.

Even with getting zero linear TC by choosing the right current, there is quite some second order part. So for a really constant voltage it would likely need temperature stabilization anyway. Still a low linear TC reduces the requirements.

The single diode more would give some 5.3 V and this can also be an advantage: with 2 such diodes in series (or as a bridge) one would get some 10.5 V and thus a less sensitive divider down to 10 V.

[branadic]

Well, I will try the rest of my devices, there is still a bunch of diodes left. But I will also try to parallel the zeners and to connect a silicon diode to form a temperature compensated zener. Don't know about what this results in, but it's worth a try to see what happens. Therefore I cut the common lead and solder a smd silicon diode to the case.

-branadic-

[Kleinstein]

Using a normal diode together with the zener will likely not work very well: The forward voltage in the 2DW232 is rather high and thus the TC of the diode is low. So if at all it would need a similar high forward voltage diode (e.g. BAV199, maybe a zener) to get a comparable result. To get a lower current for zero TC it would need an even higher forward voltage and thus less TC for the diode.

The second problem it thermal coupling that is not that good and could lead to extra thermal induced noise.

[Andreas]

Maybe I get the datasheet wrong? Any comments on that?

Hello,

there are only 2 test-temperatures specified in the data sheet. 25 deg C and 75 deg C
the slope through these 2 temperatures should be less than 5 ppm/K
(nothing specified in between).

so between the 25 deg C and 75 deg C point the difference should be below 50K * 5ppm/K * 6.2V = 1.6 mV at 5 mA test current.
provided that you use pin 1 (with dot) as negative pin and both diodes in series.

with best regards

Andreas

[branadic]

Andreas, two different datasheet give two different connection schemes:

1. "Polarity arrangement: output 1 indicates the negative terminal with color point, 2 is positive terminal, and 3 is empty. "
2. "The color point of the pin "1". "2" is the negative side and the other side is the positive side. At the end, the pin "3" is the spare pin."

One the other hand there is a diagram in one of the datasheet showing the output voltage over the whole temperature range (-55 ... + 125°C). Since now I used pin 1 as the output pin. I will change the connection to see if there is any difference.

-branadic-

[Andreas]

Andreas, two different datasheet give two different connection schemes:

overlooked that: so which of the 2 data sheet corresponds to your devices?

with best regards

Andreas

[branadic]

By now? None of them.

-branadic-

[branadic]

Meanwhile I measured three devices and the third device also in reverse connection. 750R (General Radio Type 1434-G) resistor on 9.984V from a LT3042 board that is powered by a lab supply, so almost 5mA.
I also plotted voltage vs. current for this measurement. Interesting to note, the values fit to a straight line.
Next step is to measure the individual zener on the third device.

-branadic-

[SZA263]

My devices (Technix) differs from the datasheet. The divice is composed by two 5,36V@5mA zeners connected in Common Kathode.
So with the connection suggested in DS, color dot (pin1: Yel in my 232's Red in my 233's) on the negaive side and pin 2 on the positive side, one diode is used as a true zener and the other as a temperature compensating diode.
My device (random pick from a batch of 20) is mounted alone in a small cardbox and is connected to a quite stable power supply (Tek PS501-1) via a 770 ohm 50 ppM resistor.
The voltage of the compensated zener (two diodes in series) @5mA is 6.06236V (Datron 1062) and the Tc (with the normal room temperature variation) il less than 5 ppm/°C.
(DUT or measuring instrument drift?)
Heating it with a finger, shows a little positive Tc, but it fully agrees with the DS data: the zero Tc point is obtained at about 60°C @5mA, with a positive Tc under this temp and a negative Tc over.
I am running the device at 25..28°C so it is operating on the positive Tc side of the curve. A brief test a 3.3mA shows a better behaviour in this range of temperature.
I need to do a more extensive and accurate test on the device, but at first sight it seems to me very good.
No test on noise and long term drift (WIP).

[branadic]

I got my devices from VintageNut, not from Technix. They are common cathode too. Currently I'm running measurements on the single diodes.

-branadic-

[SZA263]

If you want to use a single diode, you should run it with a much lower current (say few hundreds of uA), but this will increase the noise and the differential resistance of the zener.
But why? The device is intended to run with both of the diodes in series, like any other Tc compensated device. LTZ1000, LM399, SZA263, LTFLU and the 1N821..9 devices are based on a diode-zener series.

[branadic]

Just for investigation to understand, what is wrong with the devices I have.

-branadic-

[Kleinstein]

If you want to use a single diode, you should run it with a much lower current (say few hundreds of uA), but this will increase the noise and the differential resistance of the zener.
But why? The device is intended to run with both of the diodes in series, like any other Tc compensated device. LTZ1000, LM399, SZA263, LTFLU and the 1N821..9 devices are based on a diode-zener series.

The zeners inside seem to be quite different. For a device close to the DS performance the series mode with a current in the 5-10 mA range is good, if this gives a low TC.
However the ones Brandic has would need too much current (e.g. > 50 mA) to reach the zero TC point. So the series mode is not a real option. Chances are they might be useful with just the zener. Because of the different zener the zero TC current in just Zener mode is expected to be also higher, so more like 2-5 mA.  There seem to be a larger 2nd order effect though - so the series mode has some advantage, but needs diodes to get low TC at a reasonable current.

[SZA263]

Please don't forget the self-heathing effect due to the power dissipation of the device.
The thermal resistance Junction to ambient (Rth-ja) of 2DW23x is not defined in DS, but i think it is about 400K/W. A DO35 case (1N4148) has 300K/W but the chip is bonded to two huge (referred to chip dimension) terminals, while  2DW has only one side bonded to case.
One of the Branadic's plots shows a working point at about 6.2V 30mA., very close to the power dissipation limit of the device (200 mW), and in this case the junction is at a temperature about 80 °C higher of the ambient. So the temperature plot (Branadic's 26-38 °C) is seen by the point of view of the junction as a 106-118 °C span, rather lower if seen in K or %.
If the device could survive with a power dissipation of 2.5 W, the junction will be at 1000 °C, and the 12 °C ambient variation will cause a very,very low drift...
My 2DW232 DUT used as datasheet @5mA without any exotic medium has a Tc less than the declared 5 ppM/K.

[branadic]

Again, picked up a random device. It's again common cathode with:
pin 3-1: 5.27175V @ 5mA (942.5R on 9.984V)
pin 3-2: 5.2688V @ 5mA  (943R on 9.984V)

Did the same measurement as before, thus 790R (GenRad) on 9.984V and changing temperature, with the reference connected as follows: pin 1 to ground and pin 2 to the resistor. I don't see any difference to the results before.

-branadic-

[branadic]

So I have characterized the single diodes of reference #4 in forward and backward direction @5mA. I found the t.c. as follows:

Pin 3-2: 464.37µV/K
Pin 3-1: 476.5µV/K
Pin 2-3: -1.5925mV/K
Pin 1-3: -1.589mV/K

So maybe the package with the additional pin3 for the common connection is not by accident but for an extra resistor to control the current through the compensation diode?

-branadic-

[Andreas]

So I have characterized the single diodes of reference #4 in forward and backward direction @5mA. I found the t.c. as follows:

Pin 3-2: 464.37µV/K
Pin 3-1: 476.5µV/K

Hello,

so the zener voltage is too low.
(zener tempco usually increases with zener voltage).

did you test the dynamic resistance of the zener too?

with best regards

Andreas

edit:
mhm from this diagram:
https://www.eevblog.com/forum/metrology/ultra-low-noise-reference-2dw232-2dw233-2dw23x/?action=dlattach;attach=510608
you have more than 10 ohms dynamic resistance (more than 10mV/mA near 5 mA).

[branadic]

did you test the dynamic resistance of the zener too?

No. How would you measure it? Typically to measure it, a small AC signal is superposed on a Zener current and measured as the differential impedance. But how should I do that?

-branadic-

[Andreas]

see edit above

[branadic]

So if I get you correct you come to the same conclusion, 5mA is to small for that parts and I need to get to much higher currents?

-branadic-

[Kleinstein]

I would not expect the thermal resistance to be that high. The normal TO5 is around 200 K/W. Self heating can be a problem, but it still makes sense to use the external temperature to define a TC. Self heating would be mainly a constant shift in temperature.
A rather high current and thus high power can be still a problem as trying to stabilize the temperature would about double the power consumption. So a current much above 10 mA is likely not practical.

Reducing the current through the forward diode would make the TC even worse.  Though not very attractive because of 2 extra resistors involved it would be possible to use only about 1/4 to 1/3 of the forward diode voltage. This way one could adjust the linear TC at a given current.

As the operating current does not match the DS, it might be worth checking if the diodes are at least very low noise. If the noise is not really low, it might not be worth looking for a way to compensate the TC.

My best guess would be using these "high current" diodes in a zener only mode, at a current lower than 5 mA. Maybe 1-2 mA could lead to a low TC and the noise might still be acceptable. With a slightly temperature dependent current (use the other diode), the zero TC point could be shifted to a little higher current.

The Zener impedance would be from a superimposed AC current and measuring the AC voltage. With typical a few 10 Ohms the 1 K resistor to provide the DC current could be ignored. Its not about precision but order of magnitude. A good zener would be < 10 Ohms, > 50 Ohms could be a problem.

[SZA263]

No. How would you measure it?

If you want calculate the dynamic resistance of a zener @ 5mA simply take the zener voltage at a little lower current Imin (say 4.9 mA): call it Vmin; then take the voltage at a little higher current Imax (say 5.1 mA): call it Vmax. The dynamic resistance , linearized around 5 mA is  Rd = (Vmax - Vmin)/(Imax - Imin)