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

[amspire]

I had got some 2DW232's from here:

https://www.aliexpress.com/item/10PCS-2DW232-quality-assurance/32589219690.html

Wasn't sure if they were the genuine "diamond" references, so I decided to see what was inside anyway.


Similar to the 2DW234 photo (taken in 2013) posted in the first page of this thread, but it has an opaque white enamel/epoxy/gunk instead of the brown gunk in the older photo. The chip layouts are similar, but it looks like my chip is about 12% longer then the 2013 chip.


Seems to work with the zero TC point at about the correct current and voltage, but I haven't done any serious testing.

[amspire]

This is the best image I can get of my zener chip

[julian1]

Nice pictures at @amspire!. How did you remove the case jacket - with an angle grinder or hacksaw?. I would like to open one of mine, but lack the tooling.


I have populated another board - with the best parts I have on hand.  The divider is now 5k (z201) + 2k (s102) + 1k (z201) / 10k (z201) giving a current around 25mA and t-set of 47C.

The majority of the divider voltage drop and therefore tempco sensitivity will be across the z201 (0.2ppm) resistors instead of the 102 (2ppm).

It is curious that noise appears lower (even despite my terrible cabling) - with obs varying around 5uV over quite a few minutes. At the 7th digit - that must be near the limit of my lm399 based daq to measure. I'm not sure if that could be attributable to the resistors, or lower operating temp, different 2dw232 part or just a rearrangement of my cables.

[BravoV]

Similar to the 2DW234 photo (taken in 2013) posted in the first page of this thread, but it has an opaque white enamel/epoxy/gunk instead of the brown gunk in the older photo.

Richard, have you try to clean the epoxy to reveal the die using strong solvent like acetone or MEK ?

[amspire]

Nice pictures at @amspire!. How did you remove the case jacket - with an angle grinder or hacksaw?. I would like to open one of mine, but lack the tooling.

My Lindstrom wirecutters. The metal is fairly soft. Just did a cut on part of the top corner to make a hole, and then worked my way down.

[CalMachine]

Well it seems as if my very first prototype is more of a thermometer!   

Each X axis hash is ~6 hours.

Time to make adjustments to the zener current to find the TC

[amspire]

Richard, have you try to clean the epoxy to reveal the die using strong solvent like acetone or MEK ?

Done my best to clean my "2DW232". This is taken on a stereo microscope (45x magnification) with my camera plugged straight into one of the lens sockets with a $5 adapter.



Here is another 2013 photo from the first page of this thread:

This is a precision 0.2mm grid over the chip, if anyone wants the size.

[amspire]

Well it seems as if my very first prototype is more of a thermometer!   

One of the ways I was thinking of doing a temp controlled reference is to run the second diode at a current perhaps half the first diode zero TC current along with a resistor in series to bring the voltage drop up to the first diode. Then you just have a op-amp based temp controller that just keeps the two voltages identical.

[julian1]

Nice pictures at @amspire!. How did you remove the case jacket - with an angle grinder or hacksaw?. I would like to open one of mine, but lack the tooling.

My Lindstrom wirecutters. The metal is fairly soft. Just did a cut on part of the top corner to make a hole, and then worked my way down.

Good thanks. wirecutters and tin-snips are pretty effective. For reference, here's one of my 2dw232 sourced by zlymex.

[amspire]

For reference, here's one of my 2dw232 sourced by zlymex.

That looks like exactly the same white stuff that I have. It is looking like I did get genuine Diamond zeners.

It is good to know.

[Kleinstein]

The overall drift of the reference circuit is set be the zener itself and that of the resistor(s) scaled with a certain factor.
That factor depends on the zener resistance: with the LTZ1000 (and good quality 6.6-7 V compensated zeners) it is about 0.01. So for ultimate stability good resistor(s) are important, but still in many cases the zener itself is the larger contribution.

It might be interesting to see on how sensitive the 2DW232 reference is to resistor change. I would expect is to be more sensitive, as usually the zener resistance is high for a 5.7 V zener compared to a 6.2 V zener. I would expect the experts constructing the zeners refs know that and choose the 6.2 V for lowest R_Zener despite of higher noise. I could be that by some means the 2DW232 also got low R_Zener, despite of using the low voltage. Lower noise at the lower voltage is not such a surprise. A suitable test could be intentionally increasing the current by something like 1 % (e.g. add 20 K in parallel to the 200 Ohms) and observe the increase in voltage.

How much the temperature setting resistors contribute depends on the TC compensation - with adjustment this should be better than the more of less fixed TC of the unheated LTZ1000. So with reasonable adjustment of the current, the temperature setting resistors (and the temperature regulation itself) should not be that critical. This is the advantage one gains from individual adjustment of the TC. Of cause one has to pay the extra effort for the adjustment.

So with the 2DW232 it should be only drift in the current setting that really matters.  The resistor ratio used for amplification can be measured. I don't think the OPs offset should be so important - so reading the OPs output and the zener output should be good enough for this. To resolve an 1 ppm change for the reference would only need to resolve something like a 20-100 ppm change in the voltage ratio.

[technix]

I wonder if I should saw one of mine own open and take a picture? Mine came directly from the factory and should be the same as the one sent to folks that bought through me.

[bertik]

It seems that there are at least two different batches of the original diamond brand, the one that has a fatter font and a date code below the 2DWxxx (like pictured here in this thread), and one with thinner font and no date code (and white epoxy).  It seems that they differ quite a bit at least as far as the noise is concerned. I am collecting statistics while waiting for more samples to arrive from various sources.

[amspire]

My 2WD232s have a 16-4 date code and white gunk that is like a thin silicone sealant. Also although the chip looks identical to the ones sealed with the brown gunk, it seems to be slightly larger.

[janode]

My 2WD232s have a 16-4 date code and white gunk that is like a thin silicone sealant. Also although the chip looks identical to the ones sealed with the brown gunk, it seems to be slightly larger.

I got some from the same source. Are you sure those are 16-4? Printing quality isn't great on all of them. I got 15-4. I cut one open and it looks exactly the same on the inside. They also look identical to what I got from zlymex.. Even the date code is the same!

[bertik]

Here a summary of some findings:   http://hololaser.kwaoo.me/electronics/2DW23x.html

Relatedly: http://hololaser.kwaoo.me/electronics/LowNoise.html

Cheers,
BK

[zhtoor]

Hello gazelle,

could you tell me how and where did you order these?

thanks and regards.

[Muxr]

These look like they have the "Diamond" logo and the date code: https://www.aliexpress.com/item/10PCS-2DW232-quality-assurance/32589219690.html?spm=2114.13010608.0.0.RXTW5A

[SvanGool]

Well, at least that shows that Technix got them from the same factory with QA label, as you did.
This is the QA label I got with my purchase from Technix:

[chuckb]

2DW234 and 2DW236

I got a few 2DW236 zeners from China but the noise performance was obviously worse than the 2DW234. I notice the metal can had a different finish (matt vs bright), the part number lettering was a different font, there was no date code and on closer inspection the zener die was 0.5mm square verses rectangular with a 0.72mm length.

In the future if I have a concern about a batch of chips I will use diagonal cutters, remove the case and just look at the die. Based on my samples and other comments on this thread, the low noise chips are 0.72mm long and rectangular.

The 2DW236 sample I received were connected common anode. In a series configuration that means you will measure 0.7v on the case, pin3. The 2DW234 rectangular chips seem to be common cathode with 5-6 volts on the case, pin 3.

The 2DW236 zeners came with Factory QC paperwork also. But keep in mind they meet all the data sheet specs. It would be great if these special low noise zeners had a unique part number so they did not get mixed up with the normal 2DW23X zeners.

I'm evaluating the 2DW234 and they are very low noise. It's a challenge to measure it. I don't do the 0.1Hz to 10Hz measurements I build two identical systems and measure the difference with a nanovoltmeter. This gives a better picture of the overall worst case performance. I need to temperature regulate the parts before I will be able to get good stability numbers. Maybe I'll have that running in the next month.

[bertik]

This is entirely consistent with my findings: the ones without date code typically have a lot more noise - irrespective of 2DW234 or 2DW236. Actually I couldn't find any consistent electrical difference between the various 2DW23x if from the same source. See http://hololaser.kwaoo.me/electronics/2DW23x.html#noise for a plot.

[chuckb]

There is an easy test to determine if you have the special die inside your zener package. The special die is configured as common cathode. The cathode is connected to the case.

Therefore I set up a test with my Fluke 289, or K2000 on the diode scale with the black lead on the case. When I connect the red lead to pin 1 or 2 the meter displays about 0.7v for the good rectangular die. A normal zener will read open in this test.

The test does not guarantee a low noise device, however it is an easy first test if you have concerns about a group of parts you have purchased.

[uwezi]

As a latecomer to this discussion: is there still a possibility to join in for ~20 pcs to Sweden?

[technix]

As a latecomer to this discussion: is there still a possibility to join in for ~20 pcs to Sweden?

There is an aggregate thread. I just placed an order for a batch of diodes so you may need to wait a while for the aggregate to build up again.

[Muxr]

This is the agregation thread: https://www.eevblog.com/forum/metrology/factory-2dw23x-order-aggregation-thread/