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

[chuckb]

Many of my older 2DW23X diodes with 2014-2016 date codes had 0.72v diode drop with a 1ma test current. The latest batch from Technix are measuring 0.66V. They were tested at the same temperature. If your circuits are sensitive to that voltage you may want to evaluate the new chips. I have not had time to fully test them out yet. I'm off on a 1 week vacation now.

One very good change is the older chips seemed to have plain steel leads that would not solder well. The leads on the latest chips take solder very well.

[GigaJoe]

Interesting , got a package, ...233 model:

diode has very high internal impedance, this lead it to very sensitive  to a current flow.  I'm still puzzle about why 2 diodes in a package, noy a single, or let say 4, or 6 .... and common cathode, seems, due-to manufact. process. Anyway:

2 internal diodes in parallel, just to decrease the diode internal resistance, hook to a precision voltage source, + 2 variable multi-turn resistors, 10K + 100, 100 ohm - give an option to dial up a fraction of microamp. So heating, cooling, adjusting current, I received 2-3 ppm/C  in a range 20-30C.   Thinking , of 6-8 in parallel, with 50 ohm multi-turn trimpot to adjust current for optimal temp stability, and possible to dial each to diff rising or falling temp. trend, it may allow me to compensate temp. further

With total 3.7 ma current, it observe a self heating.   

while I waiting trimpots ( takes weeks, i think) will do an aging\torching  process  ... will see ...

[Kleinstein]

The zener impedance also depends on the current. Especially at low current the impedance can be relatively high. An important figure of merit is the number  R_zener * I_zener / U_zener. This gives approximately how much drift / TC of the current setting series resistor will be attenuated. To get a good long time stable reference it is important to get a low value here otherwise one would need more and more expensive resistors.

With just the zener diode, the current where the TC crosses zero is relatively low - like in the 0.5 - 2 mA range. So a high R_Z is kind of normal. Also zener diode at around 5.2 V usually have a higher differential resistance, compared to the 6.2 V ones. So it would be a surprise if one could get good values at the low current and around 5.2 V voltage.

[GigaJoe]

2 hours: 50 ma, + 140 C , ( 2 diodes in parallel ) - increase output +2 mV , comparing before torture, wonder , is it a good exercise ? 

[2N3055]

2 hours: 50 ma, + 140 C , ( 2 diodes in parallel ) - increase output +2 ma , comparing before torture, wonder , is it a good exercise ?

140°C is way too high... die seems to be attached to case with epoxy .. 140°C will damage die attach... 65-85°C is probably better choice.. And best would be to let it age on working temp. for 500 hours...
It will take a long time to settle back to working temp after excessive heating.. 

[GigaJoe]

nothing shaking inside    ,   maybe it survive. One more thing, usually after thermal shock all references a bit shifted, and this one, not an exception, a few thermal shocks, had eliminated this shift, heat to + 100 it back precisely to the same point.

BTW, found,  this forum:
REDUCING OF THE TIME DRIFT OF ZENER STABILIZATION VOLTAGES, famous DATRON, Thanks zlymex !
https://www.eevblog.com/forum/metrology/t-c-hysteresis-measurements-on-brand-new-lt1027dcls8-5-voltage-reference/?action=dlattach;attach=239750
need to heat it up at 100 C for 4 hours, then cool slowly, and ,they heating it in an active state,  so i running to solder 10 on board , run max Amps, and heat it a few hours ....
upd, didn't read with attention ( as usual)  its a single cycle , around 70 - 100 cycles need to complete aging process .... ops ...

[2N3055]

I didn't say it's gonna fall of.. 

I know about that document.. It is good read, but not so much relevant here in terms of temperatures and such, completely different diode and technology.. Principle is true, of course..
Read first 10 posts in this thread.. Zylmex boiled them to be sure they won't go over 100°C.

Such high and short thermal shocks are just shocks that will damage it.. You need to make temperature high enough to soften epoxy and than keep it there long enough that it creeps and plastically deforms to release all tensions on die..  If you do it on a temp that's too high, it will relax for that temp, and when it cools it will again be in tension.. Trick is to do it somewhere closer to work temp so residual tension is not too much and that will fall in place during later work at nominal temp.

Accelerated aging only helps with first few hundreds hours, where changes are huge.. Last few ppms will take years and you can't rush it.
It is more helpful with less soft die attach materials (metals). Epoxy is softer and flows more easily on room temp..
If you overdo heating, chances are you will put so much tension in die attach that it will take additional years to relieve.. Or you can fracture the die or put tears in die attach. That will act as nonlinearities that will manifest itself in all kinds of stochastic behaviour..

[Spikee]

I'm thinking about using this reference in one of my designs. Zero tempco is not a must, low noise has the higher priority.
Altrough a low-ish tempco is preferred. At this moment The LTC6655 is planning to be used.

I have a few questions. What is the recommended mode for this zener to be run at ? I see al kind of numbers flying by.
I would assume that his reference is made to be run / connected in a certain way in which it would perform best.
The vref voltage that is needed would be about 0.5V. So it will need to be divided (exact value does not matter that much).

I also saw talk about regulating the temperature of the vref accurately by measuring the internal temp using the second diode. Have other people tried this ?

[Kleinstein]

There are essentially two possible modes of operation: one using only the zener diode at a relatively low current (e.g. 0.5-2 mA range). This operating point shows more curvature in the U vs T curve and also needs a rather well adjusted current to get a small TC. Due to the low current, the noise is not that extremely low, but usually still good, compared to something like an LM329 / LM399. In some cases the voltage in the 5.2 V range can be an advantage, as this can make dividing towards 5 V relatively stable.

The other operating point is using diode and zener in series. Here the current adjustment is less critical, but as a downside the current for a low TC can vary quite a lot between units, somewhere in the 7 mA - 40 mA range. So some units might not be practical to run in this mode. With the rather high current, self heating (and thus a critical thermal design) can be a problem. Usually the noise is very low and chances are that this circuit is also less sensitive to resistor drift.

Depending on how the current is set (e.g. using a current that changes with temperature), one can slight shift the zero TC temperature to slight low / higher currents. However this only a small change. It can be interesting for the zener only version to slightly increase the current at very little effort.

The diodes also seem to be different. The zero TC current can vary quite a lot - so some (lower current ones) are more suitable for the series mode and other (higher current ones) may not be practical in series mode but the better ones for the zener only mode. My guess is the diodes are made for the series mode - however many of the diodes measured so far seem to need a rather high current to achieve zero TC. At the nominal 5-7.5 mA current the TC is still relatively high with most units.

With an adjusted linear TC, there is no need for a super stable temperature regulation, however some regulation can be a real advantage, as the second order TC is relatively high, especially in the zener only mode. A sensor directly at the chip is a slight advantage, however for a fast regulation one would need sensor and heater close together. So using only the internal sensor can be tricky. Temperature regulation tends to about double to quadruple the power requirement. The rather high power in the series connected mode makes the version with temperature regulation really high power.

For 0.5 V reference one would need a divider in essentially any case. The divider could also very well limit the stability. The following amplifier might already be limiting and thus super low noise might not be such a big advantage.

[GigaJoe]

had a prototype, use just a 2 zeners in parallel (a single package),   then 2 zeners as 2-nd packages with averaging, adjust current for first to positive curve, second as negative, need a few hours and patient, for narrow temp range, result less then 1 ppm, main trouble a current can drift , disturbing the result

[vindoline]

OK, so I've got 30 of these fancy new 2DW232 & 233 Zeners on my bench and I'm anxious to build something. What's the consensus on how to "age" them? Should I actually drop them in boiling water? For how long? How many times? Should I set up a board to burn them in? What current, etc? Should I just build up a 10V circuit and run it at some arbitrary current until it begins to settle down, and then try and find the zero tempco current?

Any advice is welcome!

[Kleinstein]

I won't expect a gross effect of aging. So one could likely just build a reference circuit and adjust the current for a low TC. I would start without extra aging before building the circuit - alone to get an idea of the order of magnitude for the drift.

A first thing might be looking for the zero TC point for just the zener diode (around 1-2 mA). This would give an indication if the diodes are more suitable for series configuration of just using the zener diode(s). To a certain degree the diodes determine which circuit is possible. There seem to be some diodes that do not reach TC compensation for the series configuration at a feasible current.

The first circuit would likely not be a 10 V reference but just the unscaled reference: so either around 5.2 V or 6.5 V, maybe double if 2 chips are used. The second step would than be likely some temperature stabilization (especially the zener only version likely needs it) and only after some drift tests one could decide if it is worth to build a stable scaling to 10 V. For a 10 V scaling the zener only configuration with 2 in series has some benefits. For a really stable reference the resistor(s) needed around the zener are very likely more expensive than the diode. This is especially true for a 10 V scaling even if one could start at 10.5 V.

I would prefer accelerated aging in circuit, so one could see how much the diode changes. Part of the aging would be the likely epoxy glue. These glues usually have a glass transition in the 100-140 C range. So the logical stabilization of the glue would be by heating so something like 80-100 C, holding it there for something like 1-3 hours and than slowly (could be days: e.g. 0.5-1 hour for the first 10 K and than doubling the time for the next 10 K and so on) cool it down.

To get a curve on aging one could do repeated equal cycles to maybe 80 C or so. It depends on the part how many cycles are needed. An upfront estimate is rather difficult. How practical such a stabilization curve is, depends on the instrumentation.

[chuckb]

I'm monitoring a setup with the diode in series mode with 20ma of bias current. I have temporarily added insulation around the case so the case is operating at 100 deg C, the die will be warmer. I have measured -120uV / day (at 6V) for the last three days. I hope it slows down soon!

[Kleinstein]

100 C is already quite high in temperature. So more interesting for testing, maybe accelerated aging (if it slows down).

Is the -120 µV/day measured at constant temperature ? Without an individually adjusted current the TC of the diodes can be relatively high (e.g. 200 µV / K would not be such a surprise to me).
It might help to get separate measurements for the voltage of both diodes (forward and zener mode).

[vindoline]

I'm monitoring a setup with the diode in series mode with 20ma of bias current. I have temporarily added insulation around the case so the case is operating at 100 deg C, the die will be warmer. I have measured -120uV / day (at 6V) for the last three days. I hope it slows down soon!

Wow, your Zener is really toasty at 20 mA! Are you sure it's not 100 F? I just ran one of my 2DW232 overnight at 15 mA. It's insulated in foam inside a box and the thermocouple is physically on the metal can. It has stabilized at 31.9 C (89.4 F)

[chuckb]

The zener and electronics are in a small thermal chamber with no forced air movement. The case is operating at 100 deg C. I have seen TO-39 junction to case thermal resistances ranging from 15-45 deg C / watt in the literature. That would indicate a chip temperature of up to 105 deg C.

Yes, this is an accelerated aging test. The Zero TC point for this diode is around 78 deg C. I run the diode for a day at 100 deg C then lower it to 78 deg C and record the voltage. I also check that the turnover temperature has not changed. It has not.

I am in the middle of automating the process of finding the zero TC point. To do it correctly takes HOURS of recording voltages by hand. Good thing I'm retired now! The spreadsheet of data for this setup has 500 rows of data. I will add a scanner to check several more voltages during the next group of tests. The bias voltage could be moving. The bias resistor could be drifting. Or one of the diodes inside the 2DW234 could be changing voltage.


BTW I use a - SciGene Hybex Microsample Incubation System for a small thermal chamber (heat only). You can get them surplus for less than $100 and it contains a 200 watt heater and a Watlow SD31 Digital temperature controller. The heater is very large for the chamber size so I have some metal plates on the bottom of the chamber to add thermal mass and slow down the peak rate of temperature change. It also minimizes temperature overshoot.
I typically change the chamber from 1-10 deg C, wait 30 minutes for everything to stabilized and record the voltages. You start a temperature sweep in 10 deg C steps till you find the turn over temperature then go back and do 1 deg C steps (or smaller) to find the exact temperature.

This whole process really makes me appreciate the LM399 and the LTZ1000 with heated chips!

[Andreas]

Hello,

Lymex suggested soft thermal shocks:

https://www.eevblog.com/forum/metrology/t-c-hysteresis-measurements-on-brand-new-lt1027dcls8-5-voltage-reference/msg981794/#msg981794

other hints for reference zeners (selecting for low noise) can be found here:
http://volt-nuts.febo.narkive.com/9QIiUfpN/some-questions-to-zeners-1n823-1n829

with best regards

Andreas

[Dave]

I am now in queue for the aggregated order, but I'd like to do something ahead of time while I'm waiting for these to arrive.
I don't want to reveal my plans before I have something to show, but I would need measurements of the device cans to be able to pull it off. The two datasheets that I found in this thread both show slightly different dimensions, so I don't know which to trust.

Would someone that has already received the genuine diamond brand diodes be so kind and grab some measurements for me?

I need the diameter of the lip, the diameter of the can (the wider part just above the lip), the height of the package, the thickness of the lip and the size of the little tab (width, length).
These measurements must be from the biggest one you can find in your bin.

If anyone would be so kind to take the time to do it, I would really appreciate it.

I will, of course, share my whole project once I manage to bring it to life.

[cellularmitosis]

• Diameter of can: 0.3215 in
• Diameter of lip: 0.3635 in
• Height: 0.2565 in
• Width of tab: 0.0315 in
• Diameter of lip + tab: 0.395 in

Caveats: I just measured one (didn't look for the largest one), and I used a cheap Harbor Freight set of calipers.

Curious to see what you come up with

Edit: oops, wrong units.

[Dave]

Thank you, cellularmitosis! 

[Dave]

There is one more measurement that I still need to get this done: the thickness of the lip. Thanks!

[cellularmitosis]

This one is a bit more difficult to get an exact measurement, because the lip has a bit of a chamfer, but I'll call it 0.030"

[floobydust]

I found another datasheet for the part; posting because it was hard to find: "Guiyang 873 plant (Guiyang Zhongguang), 873 factory is 2002 version of 2DW23X military standard drafting unit"
It's very close to TO-39 package.

I also see this part in gold lead/brass base, and black paint-dipped offerings.

[Dave]

Thanks for the input. I suppose I can give you a little teaser, but I won't say much.

[zhtoor]

Thanks for the input. I suppose I can give you a little teaser, but I won't say much.

hello,

an isothermal block with temperature control maybe and then averaging for still lower noise reference?

regards.