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

[branadic]

Cracked diode #3 open and found the same white stuff covering the silicon as shown here before in post #420. So this devices I received from VintageNut are not the ones zlymex presented here. 

-branadic-

[Andreas]

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?

the 6.05 V at 5mA are obviously too low to get a good compensation.
So you might have more luck if you select for devices with 6.2 ... 6.4 V @ 5 mA. (both diodes).

with best regards

Andreas

[branadic]

the 6.05 V at 5mA are obviously too low to get a good compensation.
So you might have more luck if you select for devices with 6.2 ... 6.4 V @ 5 mA. (both diodes).

Well, then none out of 12 devices do match this criteria, as 11 out of 12 devices are below 6.1V. I currently test the one with the highest voltage of around 6.12V @ 5mA over temperature, but as I alway say: Hope dies last, but it dies.

-branadic-

[branadic]

As stated before, the diodes I have are far away from what is presented in the datasheets. Zero t.c. current for diode #7 is in the order of 23 - 24mA, still searching for the exact point. However, far far away from 5mA. An the rest of the diodes will need even higher currents to get to zero t.c. point.

-branadic-

[Magnificent Bastard]

As stated before, the diodes I have are far away from what is presented in the datasheets. Zero t.c. current for diode #7 is in the order of 23 - 24mA, still searching for the exact point. However, far far away from 5mA. An the rest of the diodes will need even higher currents to get to zero t.c. point.

-branadic-

In engineering, "Just because you CAN, does not always mean that you SHOULD".  Yes, you can get compensation by operating these diode pairs in series, but you would be far better off running only ONE diode-- and then finding the current that gets you zero-TC at lab temperature.  This will produce an upside-down parabolic voltage curve with temperature.  By adding a thermistor (e. g. silicon type PTC) with a shunt resistor and with that in series with a larger resistance to equal the value you have found-- you can flatten-out that parabolic voltage curve almost exactly.  This "hybrid" resistance would be a current source for the Zener, the other end of which is connected to your 10V output.  Then, follow that by a zero-drift amp (e.g. opa189), and voilà, now you have a 10V reference.

[Magnificent Bastard]

buried (subsurface) Zener diodes, and so their LF noise time-drift should be inline with anything else you can buy with one of these type of diodes. 

I have a doubt about it, if the process would be so simple, then anyone can reproduce it.  But problems to grow a prefect semiconductor lay in  millions layers. starting with a regular process parameters, and quality of raw materials, and finish with fancy problems like unwanted injection of atoms that disrupting a structure, or metal atoms diffusion from conductor  that connecting atoms to output legs .... So it not soo simple ....

The variation of crystals may indirectly indicate an overall quality - that means a significant effort for selection, and even this doesn't guarantee a long term stability for a simply reason, during manufacturer process crystal was contaminated, and defects slowly spreading ....

Go to http://translate.google.com (or use the chrome or chromium browser), and then type in this URL: http://bbs.38hot.net/thread-120731-1-1.html
( >>>HERE<<< is an "all in one" link if you are lazy ).


About 3/4 the way down in the page, we have a post by user Archwang:

"The discussion Ni originally "high power stability," page 68, PN zener diode emitting end portion work described herein to a breakdown occurs; surface Breakdown causes the device to be noisy. This is the case with the first two samples.  The new 2DW234 does not emit light when it is working, suggesting that the device is likely to have a buried structure. Moreover, according to the longshort altar friend's previous reminder, the two PN junction directions in this device are exactly opposite to other products, indicating that the production process has undergone major changes. The buried Zener in the literature report is also the formation of P-type impurity diffusion, which may be related to this change."

[branadic]

In engineering, "Just because you CAN, does not always mean that you SHOULD".  Yes, you can get compensation by operating these diode pairs in series, but you would be far better off running only ONE diode-- and then finding the current that gets you zero-TC at lab temperature.  This will produce an upside-down parabolic voltage curve with temperature.  By adding a thermistor (e. g. silicon type PTC) with a shunt resistor and with that in series with a larger resistance to equal the value you have found-- you can flatten-out that parabolic voltage curve almost exactly.  This "hybrid" resistance would be a current source for the Zener, the other end of which is connected to your 10V output.  Then, follow that by a zero-drift amp (e.g. opa189), and voilà, now you have a 10V reference.

At a certain point, when playing with LMx99 and LTZ1000 there is a moment when you want to do it the way old engineers did, because you can learn a lot and develope a gut feeling for the things that you do. On the other hand, there are a lot of volt nuts out there, that started with 1N829 and the sort of references. I started with the game when such references became unobtanium or expensive. So going along this way is not a waste of time.
Specimen #7 won't be a travel reference, but a sample to learn from. So after I found the working point I will now go on. Thus I need to measure noise and that sort of stuff.

-branadic-

[Kleinstein]

Measuring noise can be an important point. While some of the 2DW232 can be very good with noise, not all of them are. So there is little use in finding the exact zero TC point if turns out too noisy.

Sorting out the noisy ones could be faster than checking for low TC.

For those that don't work well (no low TC) in series mode, there is still the option of using only the zener (at much lower current) or using something like 1/3 of the forward diode.  It's not as good in many aspects, but still a possibility.

[Gyro]

As stated before, the diodes I have are far away from what is presented in the datasheets. Zero t.c. current for diode #7 is in the order of 23 - 24mA, still searching for the exact point. However, far far away from 5mA. An the rest of the diodes will need even higher currents to get to zero t.c. point.

-branadic-

@branadic,

You're beginning to replicate some for the findings from the earlier days (wow 2016  ) of this thread, you might want to go back and do some digging from about page 10. I also found the Zero TC turnover point (zener+diode in series) in the 20mA region, again well off 'published' spec. Alex found a Zero TC turnover point (zener alone) at about 1.4mA.

I also found that a potential divider across the zener, in the series zener + diode configuration yielded some promising results (Reply #264) when running at lower currents (~5mA) that you may want to try replicating with your better test gear. Diagram attached for reference.

Chris.

I've been playing around with the divider approach today and it looks as if it is possible to achieve very low TC values over a limited temperature range even with a simple hook up. I've attached the schematic for clarity but it's basically just taking the output voltage from a divider tap across the forward biased diode. I simply used a 2k cermet preset, it took a little tweaking but it looks as if the flat portion of the TC curve can be adjusted to any temperature and operating current, in this case about 23'C and is very flat over a few degrees. TC is sensitive to both divider ratio and supply current as you would expect.

Operating the device at 5.010mA with the divider (pot) measuring 765R / 1k3 the output of my sample is 5.878V  Temperature profile came out as follows (values relative to 23'C case temperature):

20'C     -12uV
20.5'C  -4uV
21'C     -3uV
21.5'C  -2uV
22'C     -1uV
22.5'C   0
23'C      0
23.5'C   -1uV
24'C      -5uV
24.5'C   -7uV
25'C      -14uV

Temperatures were measured during device warm-up (it finally settled at about 25.3'C) using a thermocouple taped to the side of the can. I was using a Geller SVR-T and Muirhead decade divider with Datron 1041M, null meter style. I repeated 3 times, each warm-up taking less than 5 mins, with consistent results (obviously the readings were a bit noisy at this level).

[branadic]

You're beginning to replicate some for the findings from the earlier days (wow 2016  ) of this thread, you might want to go back and do some digging from about page 10. I also found the Zero TC turnover point (zener+diode in series) in the 20mA region, again well off 'published' spec. Alex found a Zero TC turnover point (zener alone) at about 1.4mA.

I also found that a potential divider across the zener, in the series zener + diode configuration yielded some promising results (Reply #264) when running at lower currents (~5mA) that you may want to try replicating with your better test gear. Diagram attached for reference.

Thanks for those comments. I currently found the time for playing with the diodes, so I'm a bit late. I already read through the complete thread and also found your approach, but denied it because of to many resistors being involved. I currently go on with #7 and if I stuck I maybe take a look at the option using one zener only.

Found the zero t.c. current at 23.2mA. I keep the reference mounted to the crystal heater. I received some small dewar type 00 C but found they are still quite big. So I will put everything inside a small flask together with some foam and cotton as thermal isolation.

-branadic-

[Gyro]

That's good - I just didn't want you 'running out of steam' on them before being able to get some decent low TC measurments. I wouldn't be too concerned about the resistors - those were just roughly measured values from the 2k pot that I had strapped to it. Obviously they would become more critical for a stable reference but not particularly for TC evaluation.

[Kleinstein]

The resistor part over the forward diode would be responsible for something in the 500 mV range and thus around 10% of the total voltage. So these resistors would be still less critical than scaling the voltage from 6 V to 5 V or 10 V.

Due to the relatively high power at a little over 20 mA of current, the thermal insulation should not be too good, other wise the temperature would be to high, from self heating alone.

If thermal regulation is used, the larger second order TC for the zener only version would not be such a problem.

[branadic]

No problem Kleinstein, the thermal loss seems big enough so that the crystal heater is sourcing ~30 - 40mW.
If you like I can send you one sample so you can experiment by yourself and post and share your results over here. You are then able to show how you would operate this device.

-branadic-

[Andreas]

Nice build 

[branadic]

Thanks Andreas.
I started a first 24h measurement the day before but had an issue with my wiring that was a little bit odd. So I rewired everything, connected the crystal heater from a seperate lab power supply to the 10V output of the LT3042 board and started the measurement again yesterday to see if everything is stable. But due to an error of the temperature system (SMTAS04usbmini) the record stopped after 2.3h. Damn 
However, the reference looks stable within that time at around 6.25025V and is still sitting there from what I can read on the front panel of the meter. The measurement seems to be dominated by the noise of my meter, so time to do the next step, noise measurement.

-branadic-

[branadic]

Repeated "longterm" measurement, this time I recorded about 18h but also measured low frequency noise (0.1 - 10Hz, replace micro by nano in the plots). Looks promissing, even though the current through the zener is high.

Furthermore I received some 2DW232 and 2DW233 from the netherlands today. Thank you very much. They differ by the fact, that they don't have a date code printed on top. So I will measure them within the next days to see how they perform.

-branadic-

[iMo]

Why do you consider those diodes ultra low noise when they do ~400uVpp?

[Echo88]

"replace micro by nano in the plots"

[branadic]

Why do you consider those diodes ultra low noise when they do ~400uVpp?

The scope can't be configured for a gain of 80dB of the low noise amplifier, so I configured it as x0.1 and the remaining 60dB is just a replacement of micro by nano

-branadic-

[Kleinstein]

The noise data look really good. One has to consider that this should include the amplifier's noise as well. So only part of the noise if from the reference. So one should check the amplifiers own noise, with a short, or maybe a 7.2 V battery so that the input cap does get discharged)

I would suspect the temperature effect is more from the meter than from the reference.

The remaining unknown is the longer time drift. However this needs time and maybe a better meter.
 

[branadic]

The noise data look really good. One has to consider that this should include the amplifier's noise as well. So only part of the noise if from the reference. So one should check the amplifiers own noise, with a short, or maybe a 7.2 V battery so that the input cap does get discharged)

Did that already, but can repeat that and add it. It's the low noise amp from pipelie with 100nVpp noise floor.

I would suspect the temperature effect is more from the meter than from the reference.

Correct.

The remaining unknown is the longer time drift. However this needs time and maybe a better meter.

Not necessarily, the multimeter is okay for that, you just have to measure long enough.

-branadic-

[branadic]

The references I received from the netherlands draw a different picture. I now get a positive t.c. at 5mA (6.37V reference voltage) so need to decrease the current through the diodes. Nice to have some different parts on the desk.

-branadic-

[Andreas]

Hello,

which approximate T.C. at 5 mA?

with best regards

Andreas

[branadic]

Added the results in the post above. The t.c. of this 2DW232 is to high even though the reference is running at 5mA. So maybe this is an outlier? I need to measure the rest of the references so we will see how they compare soon.

Another 2DW232 received today, this time from Italy and it was already characterized. Thank you very much.
To confirm the results I measured this reference with 770R. The current of 5.1mA in this configuration is a little bit to high, so the t.c. doesn't perfectly match but it's close to what is given in the datasheet. Next measurement is already running.

-branadic-

[iMo]

OT: what would happen if you feed the diode with a simple 2xPNP 5mA current mirror?