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

[Gyro]

Classic, I wonder how they perform!    How desperate do you need to be to do that, given the factory price?

It looks as if gold plated leads might be a givaway, the 17th Radio Factory ones have stiff tinned silver ones.

Edit: Given the taller than normal TO5 package, I wonder if this is standard practice for the other (noisier) manufacturers, or just fakes? I can't see any purpose in putting standard DO?? glass packages inside a can aside from maybe either thermal matching or keeping the light out???

[Andreas]

Whats this? Seems like a bit of a crack in between the bottom and lid of the TO-5 package?

OMG: they are shurely more than double the price of the standard thing:
Triple hermetically sealing !!!
So nearly infinite long term stability.

With best regards

Andreas

[Kleinstein]

This might not such a bad construction if you want two thermally coupled diodes. Could be one of the less good versions from an other manufacturer. Could be still good enough to test something like an oven for temperature stabilization

[Andreas]

At least they have the "right" polarity for a ground sensed temperature junction.
But I fear the chip is too small to have a good noise figure.

With best regards

Andreas

[Gyro]

I've done a bit more work on my voltage divider TC compensation from last week. This time setting for 2mA. Compensation performance seems similar to before. There are clearly fixed end-stops to the compensation range (see attached diagram):

Alex's Zener-only @ Iz=1.45mA:        R1=[edit]0R,  R2=0R             (Zener only TC turnover point)
Last week, Zener+Diode @ Iz=5mA  R1=765R,       R2=1k3
Today, Zener+Diode @ Iz=2mA          R1=1k6,        R2=460R
Zener+Diode @ Iz=20mA approx.      R1=0R,          R2=infinite      (Zener+Diode TC turnover point)

These set the bounds of passive voltage divider compensation, with a (possibly linear) ratio range for different operating currents. One thing that isn't clear to me... Can some (small) amount of divider compensation improve on the TC of Zener alone?

EDIT: The Diode isn't Present in Alex's Zener only configuration. In the voltage divider configurations, regardless of the divider ratio, the entire Diode forward voltage is in series with the Zener supply. This will change the Zener current with temperature and is another factor that the divider needs to compensate (so maybe not linear).

[Kleinstein]

Using the divider is just a way to adjust the TC part from the "normal" diode. This allows to operate the zener diode at a different current - not just the about 1.4 mA with zero TC or the about 20 mA (could be less with other samples) point where the TC of the zener and diode compensate.

The normal diode has a pretty linear dependence on temperature. The curvature for the 1.4 mA and 5 mA current looks comparable so not much change there too. I don't think there is a really magic point, that would give a significant better performance than using the "right" current for either only the zener or the series connection. There is only the disadvantage, that one needs two more stable resistors. Also the output impedance is a little higher.

For many applications I would prefer the zener only operation: the noise already seems to be very low (e.g. comparable to LTZ1000 or OP07).  Also 5.3 V is rather attractive to divide down to 5 V (or 10 V from two of the refs in series).

One disadvantage I suspect is, that at low currents the TC could be quite sensitive to the exact current. So choosing the right current could be more important. Towards higher currents it looks like the TC is less dependent on the current - so less critical choice of current. Here some more data are still missing. It is not clear that the zero TC will be near 1.4 mA for all samples, though it looks like the two from Alex and Gyro could be similar.

[Gyro]

I agree, Zener-only operation looks preferable if noise is low enough at that current (which it seems to be). Adding divider resistors does introduce more complexity and opportunities for drift (or best case, opportunity for fine trim  ).

The only reason I can see for persuing Zener+Diode voltage divider configuration now is if it can be shown to widen (or re-position maybe) the flat TC area, which I am not convinced about having seen Alex's plots.

[Kleinstein]

The results from VintageNut around post#180 suggest that using the series configuration a current slightly higher than for zero TC still gives good stability. He got zero TC around 8 mA and still low drift and thus low TC at 10 mA. So the TC seems to be not that sensitive to the current at higher currents.

There is a second order temperature effect, and thus the TC changes with current and temperature. Through the current one can choose the temperature, where voltage versus temperature has it's flat part. The curvature might be a little larger at low current - but this is comparing data from a limited range and the data-sheet.

To get a very stable voltage one would likely need temperature regulation - and this is very effective to compensate for 2nd order effects. Compensation of the second order effect is likely more complicated. One may not even have to use the internal diode as a sensor, though one has the option. The LTFLU circuits also work with an external temperature sensor.

For using the zener only, it might be still Ok or even good to let the current also flow through the other diode. This eliminates the one pin resistance and also slightly works against the curvature. My first estimate would be a about a 10% reduction, if current comes from a 10 V source. So likely not practical to get second order compensation, as this would need a rather low resistance and low voltage. It should also shift the current to even lower currents, as it would add a little positive TC.

[TiN]

I'm backski, so ideas of what to measure next are welcome.
Will have 2002s and 3458 measuring tempco's with TEC box , but after few days should have time for 2DW love.

[Alex Nikitin]

I did run another logging session overnight with a slightly more current, to see what changes. Below are the results from both runs, in a form of V(t) curves. 10uV (~2ppm) /div for the Volts scale, and the temperature derived from the internal diode voltage, taking 0.715 V = 22C and -2mV/C tempco (not very accurate, but gives an idea of what is happening) . Visible "hysteresis" effects are most likely due to an imperfect wiring and temperature gradients.

Cheers

Alex

[Kleinstein]

The curvature looks quite strong: -5E-6 V/K² is considerably more than the DS gave for 5 mA (about -1.5 E-6V/K²). This could be a drawback for the low current operating point, if used without temperature regulation.

The difference in current is quite small, so it's a little hard to tell how much the TC changes with current. Adding 4µA shifts the flat portion of the curve by about 0.7 K. So if one wants this point to match a given temperature by something like +-1 K the current would need to be adjusted down to the 0.5% range. So current adjustment might need to quite accurate, even with temperature regulation.

Edit: The relatively large sensitivity of the TC to the current might be useful to do 2 nd order compensation: If the current increases with temperature this would counteract the curvature. With a relatively small resistance (e.g. 500 Ohms range) and thus not so high voltage to drive the current through both diodes in series this might work. However this comes at the price of being more more sensitive to resistor drift. The optimum current will also shift to slightly smaller values in this case.

[Gyro]

This looks as if it could be a helpful paper:

Developing for 2DW_(232-236)~(14-18) Type Silicon Reference Voltage Diode
SONG Chuan-li, WEI Dong, MA Rui-fen 
This article introduces the development process and result of 2DW_(232-236)~(14-18) type silicon reference voltage diode.

Unfortunately the download site seems to need some sort of login, and I don't have the language capabilies. Anyone able to access it? Maybe zlymex, blueskull or technix?

http://en.cnki.com.cn/Article_en/CJFDTOTAL-SDDZ200003019.htm

[technix]

This looks as if it could be a helpful paper:

Developing for 2DW_(232-236)~(14-18) Type Silicon Reference Voltage Diode
SONG Chuan-li, WEI Dong, MA Rui-fen 
This article introduces the development process and result of 2DW_(232-236)~(14-18) type silicon reference voltage diode.

Unfortunately the download site seems to need some sort of login, and I don't have the language capabilies. Anyone able to access it? Maybe zlymex, blueskull or technix?

http://en.cnki.com.cn/Article_en/CJFDTOTAL-SDDZ200003019.htm

You need to either buy access to the article, or if you have connections to Chinese academia and the school have bought access for students and faculty. If this thread came up a few months earlier I would use my Donghua University student identity to grab that document for you.

[zlymex]

This looks as if it could be a helpful paper:

Developing for 2DW_(232-236)~(14-18) Type Silicon Reference Voltage Diode
SONG Chuan-li, WEI Dong, MA Rui-fen 
This article introduces the development process and result of 2DW_(232-236)~(14-18) type silicon reference voltage diode.

Unfortunately the download site seems to need some sort of login, and I don't have the language capabilies. Anyone able to access it? Maybe zlymex, blueskull or technix?

http://en.cnki.com.cn/Article_en/CJFDTOTAL-SDDZ200003019.htm

I have the account for the site, and I had downloaded that article before, but it is in Chinese. Attached it anyway.
The article was published in 2000 by three authors apparently from Shandong province(not from Shanghai). They claimed that the zener was sub-surface, and use surface passivation process to achieve high stability, low dynamic impedance, low TC and high reliability(no mentioning of the low noise). They describe the manufacture process such as lithography, dispersing and evaporation of aluminum etc to which I'm not familiar.

[Gyro]

Thanks zlymex,

I guess it was too good to be true that the abstract was in English. Thanks for the brief summary though, interesting that they may be 'single component' (as opposed to IC) burried zeners, the only ones in captivity maybe? I guess there's nothing that can easilly tie this in with the 17th Radio Factory apart from maybe the low noise, which would hopefully result.

[Cerebus]

... They describe the manufacture process such as lithography, dispersing and evaporation of aluminum etc to which I'm not familiar.

I spot a likely mistranslation. Was "dispersing" perhaps "diffusion"?

[technix]

Maybe someone of us that have a microscope can saw one can open and post some die shots?

[plesa]

All EU members should receive agreed quantities (12/8/4) next week.
People in distribution chain make it without any profit. In case someone want to compensate I asked to send donation to UNICEF.
Thanks zlymex and VintageNut!

[zlymex]

... They describe the manufacture process such as lithography, dispersing and evaporation of aluminum etc to which I'm not familiar.

I spot a likely mistranslation. Was "dispersing" perhaps "diffusion"?

That is correct.

[eliocor]

I want to thank Zlymex, VintageNut and Plesa (and all of the others) for the efforts they have spent for all of us.
As politely and rightly suggested, I provided to make a donation to Unicef.

[CalMachine]

Any more of these legit "diamond" 2DW232s around I can purchase to play with?

[technix]

Any more of these legit "diamond" 2DW232s around I can purchase to play with?

I have struck an agreement with the factory. 75 US cents each, plus $5 shipping to most parts of the world.

[Gyro]

@technix:

Have you had any opportunity to study the temperature coefficients of the factory ones yet? I'm just wondering if they show the same TC turnover characteristics as the ones that we have.

[Alex Nikitin]

Another run, this time 96 hours, at 1.406mA. If the zener temperature can be stabilized +/-0.1C near the "zero tempco" point, than stability and noise could be quite good, from that data. According to the measurements of 0.1-10Hz peak-to-peak noise zlymex has posted at the beginning of this thread, even at 1.4mA the noise should be lower than that of the LTZ1000 at 4mA.

Cheers

Alex

[Theboel]

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 ?