Lowest drift, lowest noise voltage reference (ADR1000AHZ)

[branadic]

Hi folks,

after I played with the simplified version of W/F7000 schematic with LTZ and achived -0.0106 ppm/K after t.c. trimming for the bare ovenized zener, I now played with a very new, lowest drift, lowest noise voltage reference I had on the desk. Actually I already had it for some months already, but it always takes some time to get all the components and board design ready.

I treated it the same way as mentioned in the linked thread before, thus first setting the z.t.c. temperature to a specific temperature with the oven turned off. Next step was to turn on the oven and adjusting the oven to that specific z.t.c. temperature.

I then captured some low frequency noise data with a 0.1 ... 10 Hz LNA (80 dB, 100 nVpp or 16 nVrms) and an Advantest R9211E used as a DSO.

Afterwards I've measured the remaining t.c. of the board and trimmed the t.c. in multiple steps. The last trimming was slightly overcompensating things., turning the t.c. into opposite direction. The measurement is already limited by the lab temperature stability and the t.c. of the meter. Thus, it needs another run of adjustment and another run in the thermal chamber plus comparison to F7000-2 in the next run, but results are looking good so far.
Compensating the ambient temperature influence, I can find the t.c. already being at -0.03713 ppm/K, but I'm sure we can do better than that.

The board is currently missing the gain in the boost stage, but uses the amplifier as a buffer only. This is part of the next steps that will be performed soon.

-branadic-

[branadic]

After some tweaking I'm now down to about -10 ppb/K. With some more iterations I'm sure I could get it even smaller, but hey the result is not too bad, or is it?

-branadic-

[Kleinstein]

The temperature effect looks good. However this is the easy part - expecially for a rather limited tempearture range.

The noise performance also looks really good, though the time window is relatively short to really tell. Sometimes popcorn noise happens on a long time scale and some 100 seconds without a jump are well possible even for a LM399. The standard to LF noise charactrization is 0.1 to 10 Hz, because this is a window that can still be measured with AC coupling. However the LF noise that really matters is often the even lower frequency part, like 1-100 mHz. This is especially the case for a precision DC voltage reference. This is really hard to measure - so it is very understandable to measure the sharter time scale. It could still help to have a longer window.

[branadic]

At least in the observed time frame, a t.c. measurement takes about 10h 40min, no popcorn noise was present.
The only real measurement setup for very low frequency noise I know of is propably a JVS and a nV-meter. As the LNA already limits the bandwidth to 0.1 ... 10 Hz, longer time frames with this setup don't make any sense at all.

-branadic-

[Kleinstein]

I know the problem with the very low frequency noise. One could do a measurement with 2 identical reference, but it is quite some effort.

The TC measurements take quite some time, but the vertical resolution is not as good and the reference in the meter is involved too. So hard to tell which ref. is causing the change.

[ramon]

I usually liked to make measurements for 1 month at ambient temperature.

[branadic]

I repeated the measurement with my F7000-2 sitting at room temperature in parallel. I can state with some confidence, that the reference shows no obvious t.c. in the range 15 ... 31 °C. Both, the LDLN reference in the thermal chamber and the F7000-2 at room temperature, track each other.

-branadic-

[branadic]

I recognized that I've never published results of the final reference. After my last post I've trimmed the output voltage to 10.000 0x V and finally trimmed t.c.
The 10 V boost stage added some small higher order portion to the t.c. as already seen on my similar LTZ reference.

-branadic-

[Castorp]

I have taken noise measurements down to lower frequencies, fully in the 1/f region. ADR1000 seems to be at least 2x quieter than good samples of LTZ1000. There's also very little spread among different units. But this is all too preliminary and there are still a number of things that prevent me from publishing the results.

So far I only have two units of HPM7177 with ADR1000, but there will be more soon. What I can tell for sure is that in those units the LF noise at full scale is no longer dominated by the Vref. With LTZ1000 it was, with ADR1000 it's some extra noise coming either from the ADC chip or the AZ op amps (but definitely NOT from the resistor networks). I also have a hint for long-term drift - both units have drifted less than 0.5 ppm in 9 months. They have been powered on and off many times, and one of them was severely abused with EMC tests (conducted RF, EFT, LF magnetic field). The 0.5 ppm/9months was taken with a rock-solid 732B that we calibrate at METAS every year.

I'll mention some of these findings in the coming MM talk.

[Castorp]

Snapshot of an ongoing test. Once again - that's with a 732B and only a small fraction of the LF noise comes from the ADR.

[Dr. Frank]

Hello,
Thanks to ScoobyDoo and branadic, I got 5 samples and some PCBs.
Here's the mockup board, where I made the first tests.

The lower zener voltage compared to the LTZ1000 now allows this Reference Amplifier to be trimmed to zero T.C. for a given oven temperature, by individual selection of the collector resistor R2. Compare this to the +50ppm/K for the unregulated LTZ1000.

That trimming process is the same procedure as for the SZA263 and LTFLU in the different FLUKE instruments.

I measured U_ref vs. temperature, which always gives a negative, flat parable. The zero T.C. point varies from 45 to 60°C for a fixed value of R2, and the T.C. varies from about +5 .. -5ppm/K over 24 ... 80°C only.

At a fixed divider ratio R4:R5 for around 52°C oven temperature for all five samples, the individual T.C.s are well below +/- 2ppm/K.
It follows that the attenuation factors for R3 and the most critical R4, R5 are at least 10 times greater than in the datasheet. These resistors now can have much relaxed stability specifications, only R1 and R2 should be PWW or BMF types.

I also measured the low frequency noise 0.1 .. 10Hz, like branadic and Castorp.
I use an LNA from Andreas, and always made "typical" measurements of 10sec long samples, so to have a direct comparison to my LTZ1000s, but also to the measurements done by Andreas.
100s measurement time or averaging give no essentially different results.
Other LNAs will give different absolute noise figures, so these results are not directly comparable .
 
The noise of the ADRs RefAmp evidently is about 50% of the LTZ1000, which several others also have confirmed.

ADR #1 would be the most quiet sample, but it suffers from  noise, or frequent dips, but with a very moderate amplitude of maybe 400nV_pp.
The ADA4522 buffer contributes a lot to the overall noise, so I have to investigate if it's only this specific unit, or what can be done to improve this.

I will add the noise figures of these samples, when measured with my 3458A, maybe now I can see a pronounced difference.

As my 34465A showed much higher stability / lower noise when I replaced its LM399 reference by a LTZ1000 clone PCB, my idea is to replace the LTZ1000 in my 3458A with an ADR1000 (*).
Maybe its noise figures will also go down, best by a factor of 2.
Castorp found that in his DVM other components overwhelmed the noise budget, so let's see what happens.

I would like to remind that TiN some time ago made a similar experiment on one of his 3458A, by implementing 4 parallel LTZ references, which would also yield 50% noise of a single LTZ.
I'm not aware what his final result was.

Anyhow, I think this new device will require quite a big time slot during the Metrology Meeting 2021, and I'm excited to meet the volt-nuts community personally and we'll have some fruitful discussions and comparisons   

Frank

(*) my 3458A already has a modified FW, with CAL? 2,5 = 6.50000V (7V reference lower limit).
Thanks to Poul-Henning Kamp for your advice how to correct the checksum.   

[Kleinstein]

Directly buffering the reference with an AZ OP can cause problems from input current spike of the OP effecting the reference circuit that uses a relatively slow OP. The usual solution is to have some filtering of the reference voltage. The OP is high input impedance and one can thus use a simple RC low pass, at least to filter out the spikes (MHz range) but also a little more (e.g. 5 K and 1 µF).

I don't think it would be really needed to fine tune the unheated TC. It is enough to start with a typical much smaller value like 5 ppm/K instead of 50 ppm/K. A factor 10 less sensitive resistors is already a big step forward and should be good enough. For the tests it is of cause good to have the unheated TC measured.

Using the ADR1000 in the 3458 may not be so easy, as the voltage is lower and the meter may very well have points to complain about too low a votlage and CAL constants out of range.

[branadic]

Directly buffering the reference with an AZ OP can cause problems from input current spike of the OP effecting the reference circuit that uses a relatively slow OP. The usual solution is to have some filtering of the reference voltage. The OP is high input impedance and one can thus use a simple RC low pass, at least to filter out the spikes (MHz range) but also a little more (e.g. 5 K and 1 µF).

There is some R-C-R filter in front of the AD4522-1, though Frank uses slightly different values than I do

I don't think it would be really needed to fine tune the unheated TC. It is enough to start with a typical much smaller value like 5 ppm/K instead of 50 ppm/K. A factor 10 less sensitive resistors is already a big step forward and should be good enough. For the tests it is of cause good to have the unheated TC measured.

Once the unheated reference is fine tuned the demand for the resistors with heated references decreases dramatically, same as for all t.c. compensated references (LTFLU, LTZ, 1N829, 2DW23x, ...) with only one exception, LMx99 which is more than just a zener. This is true for ADR either way.

-branadic-

[branadic]

I'm constantly monitoring the ADR for 26 days now. Every morning the skript first performs an ACAL, measures the amplified ADR reference with R6581D, the difference to F7000-1 with K2182A-1 and the difference between between F7000-1 and F70002 with K2182A for an hour.
The reference is currently drifting downwards. Allthough some people might think this is due to the resistor network in the boost stage, I tend to believe it's the reference itself. But I don't want to bother the setup and add additional equpiment to prove that.

-branadic-

[ScoobyDoo]

Hello BRANADIC - You have probably observed the LTD graph from its datasheet.

If you power the ADR1000 without an accelerated ageing process - it will take at least 5-6 months before it settles to an acceptable level. It is possible to age the ADR1000 (as done with LTZ) but ageing process needs modification to get it working ... - the rest is proprietary recipe from the Mastercook.

cheers
ScoobyDoo

[Kleinstein]

The curve for aging is with 75 C set point temperature. With a lower temperature the time scale is expected to be even longer.

[branadic]

You have probably observed the LTD graph from its datasheet.

I'm aware of that diagram, but since the recipe of accelerated aging is a secret, it doesn't help a lot to know that there is one. The only recipe available by now for LTZ was given by Cern.

https://xdevs.com/doc/CERN/1101699_V1_Burn_in_of_LTZ1000.doc
https://cal.equipment/doc/CERN/1101699_V1_Burn_in_of_LTZ1000.pdf

Modifying this receipt requires knowledge of the die attach being used and its glass transition temperature, an info that is not available at ADI website.
Aging means cycling around its glass transition temperature to reduce internal strain and allow further curing.

-branadic-

[ScoobyDoo]

Hello Folks - this is our ADR1000 EVB - up and running.

We did some basic STDEV tests using 3458A - but running against limits (of 3458A) - 120nV STDEV (K=2) with an output voltage of 6.659V typical for the ADR1000.

The 120nV STDEV (0.018ppm) is just an indication - we expect it to be better using a setup with a null meter - but this would require an averaged ADR1000 buried zener bank or JJA/JVS.

Best regards
ScoobyDoo

[Kleinstein]

A 2nd ADR1000 refrence and than a low noise meter would be enough. This would measure the sum of the noise of the 2 ADR references plus the noise the meter in a low range (e.g. 100 mV). In the lower range the 3458 could be good enough. It should be good at least for the lower frequency part, as the DMM amplifier has little 1/f noise, but the reference is expected to have 1/f noise in the very low frequency part. A low noise amplifier (no need to be very high impedance) would be an option too.

[dietert1]

Yes, what is the numerical resolution of the HP 3458A in that measurement? 100 nV? If you want to go beyond what you have as resolution or stability of your meter, you setup a difference/bridge type  measurement.

For example with the two LTFLU references i made last year, i am observing a standard deviation of day-to-day averages of about 270 nV. This is the difference of the two references and comparing one day to the next day. So there are two factors sqrt(2) involved and i estimate the noise of each reference to be about 130 nV. The combination of both LTFLUs should be a reference with 90 nV standard deviation for day averages. I am getting similar numbers for hourly averages.

Of course i would also prefer a direct comparison to a much better reference, but that will be expensive.

Regards, Dieter

Edit: The LTFLU references i mentioned include 10 V gain stages, so it's a difference between two 10 V reference voltages.

[Dr. Frank]

Hello,
I've performed the first two stability measurements on the ADR1000, and updated the 1h noise figures as determined with my 3458A in the table above.
It's been placed inside an aluminum case with proper jacks.

The references show a strong negative drift in the first hours already, so confirming branadics findings.
Room temperature and internal temperature of the 3458A was stable to about +/- 0.1°C, so that's fully the ADRs drift what we see here.

I have then chosen 1h of mostly constant output voltage and determined the noise (SDEV) for 1h and for several 10 minute periods.
I have to emphasize again, that these figures are always a convolution (quadratic sum) of the different noise sources like ADR1000, plus 3458A amplifier, plus its LTZ reference and plus the A/D noise.

The 1h noise (185nV for ADR#3, 168nV for ADR#1) is remarkably lower than measurements on other LTZ based DUTs.
Of course it can't be as low as 50%, as the 3458A contributes a lot to the noise budget.

This also confirms the SDEV~ 120nV measurement by ScoobyDoo, which were probably made on a shorter sampling period.
I've even seen 80..90nV r.m.s. over 1min (16 samples).

In the 2nd graph you can recognize quiet and more noisy 10min periods, ranging from 133nV to 237nV, latter showing sort of small bumps.
This seem to indicate different instability mechanisms, the first being the usual zener noise.
So what might be the 2nd mechanism?

Qualitatively, you'll find that picture identically in differential mode measurements, of course on a bit lower level, as only the instabilities of both references come into play. (10V level, 146nV @10 min., 189nV @ 1h)

So it would be interesting to run two ADRs against each other over 24h, to confirm that the noise is cut by 50%, and to investigate, whether these small bumps vanish.
In this case, the ADR could really improve the short term stability performance of the 3458A, when replacing its LTZ1000A.

(I need to change about a dozen further ACAL limit parameters inside the FW, so that the lower reference voltage is accepted.)

I guess I will also bring this mockup board and Andreas LNA to the Metrology Meeting, on next weekend!

Frank

[TiN]

I got some 0.5 ppm jumps from first few chips runs, sadly. In these Log 1, Log 2 two socketed ICs tested starting from power on. Circuit and PCB is very same xDevs FX with 13k/1k Tset and 120R.

Logged by 3 x 3458A + 4th (red line) on 10V boosted output.
DSV-files : Log 1, Log 2
Downward drift also present in logs, as others show.

We had crazy heat here in August, so ambient temperature stability was also very poor.

[Kleinstein]

There seem to be some popcorn / RTN type noise ( sudden jumps by some 0.05 ppm).  With rare event one can get intervals without and intervals with such jumps. I consider these jumps problematic, especially if they are rare, as averaging ober some 10 minutes is not very effective against such noise.
For short time tests this can be OK, but for a long time experiment you never know whether one is more on the upper of lower values.
It is still the question, if this from the ADRs or maybe the LTZ1000 in the 3458. I don't think it is the amplifier or ADC of the 3458 - though at the 0.05 ppm level INL errors may reach this, much and still be in spec. Still the main suspects are the 2 references.

edit:
TiN's cuves show the jump, at least some comig from the DUT - don't expect 3 times 3458 to jump in sync, even with something like EMI.

[quarks]

thanks a lot for sharing information

[MegaVolt]

I seem to have missed the big ADR1000 sale )) Where do you get them?

[YetAnotherTechie]

I seem to have missed the big ADR1000 sale )) Where do you get them?

They are already available in alibaba 

[ScoobyDoo]

Hello everyone - I have an update for all potential ADR1000AHZ buyers:

They can buy through ADI’s distribution arm.   It is called ADX
Analog Devices Express (ADX) | Design Center | Analog Devices
https://www.analog.com/en/design-center/device-programming.html

Also any customer can reach out to Rui Zhou or Carlos Sanchez to inquire about product availability and sales of the ADR1000AHZ. 
We can provide them with a datasheet but no extensive technical support.  This is why it is not on the web.

Carlos.Sanchez@analog.com
Rui.Zhou@analog.com

Best regards
ScoobyDoo

[ScoobyDoo]

Hello Folks

                good news is coming - Carlos, Rui and ADI team are adding the ADR1000AHZ in the ADI webshop online - part will be visible within a day or so.

You will be able to pay with credit card - no credit letter is needed

To be updated soon !

Best regards
ScoobyDoo

[Andreas]

They are already available in alibaba 

but who wants the SMD plastic package?

Mouser germany also lists the device (0 Stock limited availability)
https://www.mouser.de/ProductDetail/Analog-Devices/ADR1000AHZ?qs=%2Fha2pyFadugtPWjaDttya%2FKT5U0JA9UT00v1%252BPGnc%252Bo%3D

has someone got already a offer from them?

with best regards

Andreas

[quarks]

not yet

EDIT: today Mouser wrote me they cannot make me an offer for ADR1000

[ScoobyDoo]

Hello folks
                The ADR1000AHZ is now available from the ADI online webshop.
The MOQ has been set as low as one to reach all Voltnuts - (MOQ = 1).
The part will be monitored for a period of at least six months to understand if this is a viable (sustainable) path for ADI - so pls. order as much as you can ...
ADI might raise the MOQ to 5 coming days ...
You need to be logged in in order to view the part and price

Best regards
ScoobyDoo

[TiN]

Inspired by chuckb and his light show experiments I've reattached camera onto microscope over my decapitated ADR1000 chipie and turned off all lights in lab. And after some pondering at camera settings (ended up ISO1600 30", so kinda grainy) we got stray photos escaping zener effect region. ADR1000 is very poor LED in that regard. And yes, when I triggered battery operated photoflash (to get countour exposed) I can see a big jump in output 6.611850V voltage on K2002. So it's photodetector now too.

[martinr33]

ADR1000AHZ is at MOQ of 5 in the US. I may try and call them tomorrow.

[chuckb]

TiN
Was the glowing chip at 4 or 8ma?

[Noopy]

Of course I had to take some pictures of the ADR1000 too. 
Special thanks to harerod (https://www.harerod.de/) for donating the part. 








In the ADR1000 pin 4 is connected to the case. In the LTZ1000 (https://www.richis-lab.de/REF03.htm) the case wasn´t connected to pin 4 directly. There was just a connection through the substrate.




The die attache is done with some glue like material.
LTZ1000 was soldered which gave you a temperature resistance of 80K/W.
LTZ1000A was glued with some special fluffy polymer you can see in the LT1088 too (https://www.richis-lab.de/LT1088.htm). The temperature resistance was risen to 400K/W.
For the ADR1000 216K/W had to be good enough.




A very clean edge! Perhaps it´s normal. Perhaps they expended effort to get cleaner edges since impurities due to sawing of the wafer can worsen the stability?




Analog applied two layers of metal.






ADR1000 is quite similar to the LTZ1000 but the heater structure is much simpler, no more spare structures.




Let´s take a closer look at the circuit. There are two additional bondpads leading to the reference circuit.








With the two metal layers the circuit is a little harder to read...




The active parts are the same as in the LTZ1000 but here we have the possibility of a 4-wire-connection! VREF and GND are connected a second time directly at the zener-transistor-combo so the bias current and the traces conducting the current don´t interfere with the reference voltage. The sense traces are connected to the two unused bondpads.




Trying to clarify the connections.






And of course I had to light it up (5mA). 
It looks like the light is more uniform than in the LTZ1000. Perhaps here we see one reason for the lower noise figure? A more uniform structure in the buried zener?


https://www.richis-lab.de/REF19.htm

 

[branadic]

Thanks for the pictures. I'm somewhat of two minds, on one side I like to see some of the insides, on the other hand all that parts are lost forever

In the ADR1000 pin 4 is connected to the case. In the LTZ1000 (https://www.richis-lab.de/REF03.htm) the case wasn´t connected to pin 4 directly. There was just a connection through the substrate.

The die attache is done with some glue like material.
LTZ1000 was soldered which gave you a temperature resistance of 80K/W.
LTZ1000A was glued with some special fluffy polymer you can see in the LT1088 too (https://www.richis-lab.de/LT1088.htm). The temperature resistance was risen to 400K/W.
For the ADR1000 216K/W had to be good enough.

That's a mistake, LT used ICA aka conductive adhesive in the LTZ1000, the die isn't soldered. In the LTZ1000A the adhesive is filled with glass beads, a material that hasn't many suppliers and was thought of for ADR1000 too, as said by Eric Modica at MM2021.

A very clean edge! Perhaps it´s normal. Perhaps they expended effort to get cleaner edges since impurities due to sawing of the wafer can worsen the stability?

Looks like a normal wafer saw cut with diamond blade.

ADR1000 is quite similar to the LTZ1000 but the heater structure is much simpler, no more spare structures.

Ah, you missed that part of explaination on that by Eric Modica.

The active parts are the same as in the LTZ1000 but here we have the possibility of a 4-wire-connection!

I already spotted that force-sense connections on the pictures provided by TiN. Unfortunately the package is missing two more pins to make use of them. Maybe at some point we will see ADR1000 in a different 10-pin package too?

-branadic-

[dietert1]

At least they could have bonded the second connection to the existing pins. The weak points are the two wire bonds on each wire and redundancy would already help. I'd consider that half ready.

Regards, Dieter

[Kleinstein]

I would not consider the lack of kelvin sensing a real problem. The temperature is quite constant and thus a constant resistance for the trace and bind wire.
A slightly wider trace my be possible if the extra one for kelvin sensing is made thinner or left out. Even the pin temperature is still quite stable. The current is also constant and thus only a rather constant part of maybe 1 mV or so.

[Noopy]

Thanks for the pictures. I'm somewhat of two minds, on one side I like to see some of the insides, on the other hand all that parts are lost forever

You are right but after all the ADR1000 is still in production.

LT used ICA aka conductive adhesive in the LTZ1000, the die isn't soldered. In the LTZ1000A the adhesive is filled with glass beads, a material that hasn't many suppliers and was thought of for ADR1000 too, as said by Eric Modica at MM2021.

Thanks for the info!
That means that in the LT1088 they used something different:
https://www.analog.com/media/en/technical-documentation/application-notes/an22.pdf
"air impregnated polymer die attach"

ADR1000 is quite similar to the LTZ1000 but the heater structure is much simpler, no more spare structures.

Ah, you missed that part of explaination on that by Eric Modica.

That´s right. Can you enlight me?

[iMo]

Ask ADI for chips/dies only. I bet you get them for a small fraction of the ADR1000 price.
And you may experiment - I think bonding 10 wires cannot be a big problem in Germany..

[Kleinstein]

Just massive glass beads would not help much with thermal conductivity. It would need some hollow glass beads to really make a difference.
Massive beads would only help to get a thicker layer, which may be enough.

For a low power use, there would be the option to use the power of the external heater transistor too. Just spread it out evenly to make good use of it instead of putting the transistor in the most distant corner of the circuit.

[harerod]

Thanks for the pictures. I'm somewhat of two minds, on one side I like to see some of the insides, on the other hand all that parts are lost forever
...

I had a similar emotion, when I randomly picked that poor little lamb out of its flock.
Noopy contacted me to buy an ADR1000, with the express intention to provide die pictures for the EE community. The reason why I decided to support his work, is because noopy has been sharing the excellent results of his considerable efforts with the community for years.
I also hope that the high quality of his pictures will keep serving as a basis for technical discussions and keep many others from cracking components open, since they would have a hard time generating better results.

[Dr. Frank]

...
And of course I had to light it up (5mA). 
It looks like the light is more uniform than in the LTZ1000. Perhaps here we see one reason for the lower noise figure? A more uniform structure in the buried zener?


https://www.richis-lab.de/REF19.htm

 

Hello Noopy,
thank you for the die pictures and your analysis.

In the session, I posed the question to Eric Modia, how they achieved these 50% noise level compared to the LTZ.
His answers were
1. "Pure luck"
2. "Due to the lower reference voltage, the zener structure of the ADR1000 operates in the Zener effect mode, compared to the LTZ which operates in the more noisy Avalanche mode"

That sounded reasonable at first, but I can't judge his statement.

Anyhow, after my next question Eric seemed to be surprised, that the ADR1000 can be (and should be) operated with an un-heated TC = 0ppm/K by choosing appropriate zener and collector currents.
His advise though, to use 8mA for the zener, is definitely a wild guess and not necessary at all. See my measurements in this thread.

Frank

[magic]

Comparing the images side by side, I see that dimensions and layout are almost the same as LTZ1000.
Several people who attended the MM mentioned that the chip was under development for many years. Is this a rebranded "LTZ1001" from Linear or a clone that AD worked on before the merger?

[branadic]

Eric Modica (ADI) mentioned, that ADI wanted to have their own LTZ1000-like reference for products such as AD5791, that's why they started the development.

-branadic-

[ScoobyDoo]

Hello all Voltnuts,

Before there was any internal noise of the LTC M&A - the ADI VP responsible for Voltage reference devices spotted the AD5791 board - he was not thrilled - and wanted to have his own premier Voltage Reference circuit on the board.

The legacy LTZ1000(A) was designed by Carl Nelson and Mark Glanville  (mask designer) hence the initials CN MG on lithography/die.

The ADR1000 was designed by Eric Modica, Paul Henneuse (retired), and Mark Reisiger and had many R&D pitstops - this design initiated 4-5 years before the M&A with LTC.

The VP of ADI would never have launched this project if he had been aware about the M&A talks with LTC.

After the acquisition of LTC the ADI team had access to Carl Nelson (LTC) and heavily leaned on his wide shoulders to bring it to a successful ending ...

Best regards
ScoobyDoo

https://readingjimwilliams.blogspot.com/2011/08/app-note-22.html

[Noopy]

ADR1000 is quite similar to the LTZ1000 but the heater structure is much simpler, no more spare structures.

Ah, you missed that part of explaination on that by Eric Modica.

That´s right. Can you enlight me?

Would you please share your knowledge?

Ask ADI for chips/dies only. I bet you get them for a small fraction of the ADR1000 price.
And you may experiment - I think bonding 10 wires cannot be a big problem in Germany..

Worth a try. 
Bonding such small wires is no easy task. If you look too long at the pictures you often think a little probing would be no problem but without special equipment that is no fun.

It looks like the light is more uniform than in the LTZ1000. Perhaps here we see one reason for the lower noise figure? A more uniform structure in the buried zener?

In the session, I posed the question to Eric Modia, how they achieved these 50% noise level compared to the LTZ.
His answers were
1. "Pure luck"
2. "Due to the lower reference voltage, the zener structure of the ADR1000 operates in the Zener effect mode, compared to the LTZ which operates in the more noisy Avalanche mode"

Answer 2 seems wrong. In the zener mode the tempco would move into the negative area, wouldn´t it?

[iMo]

@Noopy:
1. in the presentation EricM explained why they went with a single duo heating ring..
2. bonding - not easy but in DE there is a lot of manufacturers/r&d_centers where they can bond a chip into a package. They usually use ceramic packages with N-pins for prototyping/measurements, afaik.

[Noopy]

@Noopy:
1. in the presentation EricM explained why they went with single heating's ring..
2. bonding - not easy but in DE there is a lot of manufacturers/r&d_centers where they can bond a chip into a package. They usually use ceramic packages with N-pins for prototyping, afaik.

Can I download this presentation somewhere?

[Dr. Frank]

It looks like the light is more uniform than in the LTZ1000. Perhaps here we see one reason for the lower noise figure? A more uniform structure in the buried zener?

In the session, I posed the question to Eric Modia, how they achieved these 50% noise level compared to the LTZ.
His answers were
1. "Pure luck"
2. "Due to the lower reference voltage, the zener structure of the ADR1000 operates in the Zener effect mode, compared to the LTZ which operates in the more noisy Avalanche mode"

Answer 2 seems wrong. In the zener mode the tempco would move into the negative area, wouldn´t it?

Hi Noopy,
please carefully read my question. Here I reproduced his answer concerning the noise, but not about the T.C.!
Or what's your point?

The T.C. of the zener diode is indeed depending on its absolute value, or in other words, if it's in either zener or Avalanche mode, I guess.
The trick with LTFLU, SZA263 and ADR1000 (unintentionally?) is, to inversely match its T.C. with the -2mV/K of the transistor BE diode.
Therefore you have a limited choice of zener voltages which will do the job.

Frank
 

[branadic]

If I got Eric right he said:
"...First I believe Jim Williams used a similar arrangement to make the LT1088 rms to dc converter, but also when you start to look at the stability of any thermal feedback system the transit delay from the heater to the sensor plays a big role in determining the effective loop phase margin of that loop. So in theory a trade-off exists between the thermal gradient and the thermal stability, that is the closer the rings are to the core the less phase delay is introduced in the feedback, but that would cause a larger gradient between Q2, the zener and Q1. So the fact that this was a kit part makes me think that both, Jim and Carl were using this for the wrong purpose, Jim was using it to create an rms to dc converter, so he just wanted the thing to run fast, so he wasn't using the zener, he was using the temp sensor and the heater. However Carl had to be very concerned with this phase delay effect and also gradients. If you move the heater infinitely far from the sensor what you get is basically a bang bang controller. The heater heats up, it takes to long for the sensor to pick it up and consequently the control loop bangs back and forth inbetween the rails..."

What I hear out of that "kit part" is, both Jim and Carl basically used the same set of masks to create their individual solutions, Jim his thermal converter and Carl his heated zener reference. So that could explain why there are these additional heater rings, we again find in the LT1088.

Answer 2 seems wrong. In the zener mode the tempco would move into the negative area, wouldn´t it?

It's always a mixture of Zener and Avalanche effect (breakdown voltage <4V is kind of a pure zener effect, while with breakdown voltage > 6V the Avalanche is dominant). So decreasing the breakdown voltage by accident, as Eric said by "pure luck", its a little more of a Zener than Avalanche effect.

-branadic-

[wutieru]

Test board：

[wutieru]

Test whole night and datas:

[wutieru]

1000 sample data

[Noopy]

It looks like the light is more uniform than in the LTZ1000. Perhaps here we see one reason for the lower noise figure? A more uniform structure in the buried zener?

In the session, I posed the question to Eric Modia, how they achieved these 50% noise level compared to the LTZ.
His answers were
1. "Pure luck"
2. "Due to the lower reference voltage, the zener structure of the ADR1000 operates in the Zener effect mode, compared to the LTZ which operates in the more noisy Avalanche mode"

Answer 2 seems wrong. In the zener mode the tempco would move into the negative area, wouldn´t it?

Hi Noopy,
please carefully read my question. Here I reproduced his answer concerning the noise, but not about the T.C.!
Or what's your point?

The T.C. of the zener diode is indeed depending on its absolute value, or in other words, if it's in either zener or Avalanche mode, I guess.
The trick with LTFLU, SZA263 and ADR1000 (unintentionally?) is, to inversely match its T.C. with the -2mV/K of the transistor BE diode.
Therefore you have a limited choice of zener voltages which will do the job.

You asked him about noise and for that question his answer is ok. Avalanche breakdown is a rude process.
I was wondering if that wouldn´t destroy the T.C. compensation. The transistor gives you the negative T.C. and the zener has to supply you with a positive T.C. otherwise you won´t get near zero T.C.
Moving the zener diode from avalanche to zener breakdown moves its T.C. from positive to negative. Because of that I was wondering if that is even possible. But probably branadic´s point is important: In this range it´s a mixture of avalanche and zener and probably the lower voltage still gives positive T.C.
Of course in the LTZ1000 and the ADR1000 the T.C. compensation of zener and transistor isn´t as important as in other references but in my view it´s not irrelevant.

If I got Eric right he said:
"...First I believe Jim Williams used a similar arrangement to make the LT1088 rms to dc converter, but also when you start to look at the stability of any thermal feedback system the transit delay from the heater to the sensor plays a big role in determining the effective loop phase margin of that loop. So in theory a trade-off exists between the thermal gradient and the thermal stability, that is the closer the rings are to the core the less phase delay is introduced in the feedback, but that would cause a larger gradient between Q2, the zener and Q1. So the fact that this was a kit part makes me think that both, Jim and Carl were using this for the wrong purpose, Jim was using it to create an rms to dc converter, so he just wanted the thing to run fast, so he wasn't using the zener, he was using the temp sensor and the heater. However Carl had to be very concerned with this phase delay effect and also gradients. If you move the heater infinitely far from the sensor what you get is basically a bang bang controller. The heater heats up, it takes to long for the sensor to pick it up and consequently the control loop bangs back and forth inbetween the rails..."

What I hear out of that "kit part" is, both Jim and Carl basically used the same set of masks to create their individual solutions, Jim his thermal converter and Carl his heated zener reference. So that could explain why there are these additional heater rings, we again find in the LT1088.

OK, it´s just the LT1088 thing.
But that´s an interesting background story. 


Thanks Dr. Frank!
Thanks branadic!

[miro123]

1000 sample data

Thanks for sharing. Do you have some pictures of full assembly?

[Kleinstein]

Reading  an external reference directly with a DMM will give the combined noise / drift of the DUT and the meter internal reference. With the ADR1000 based reference chances are most of the noise and a significant part of the drift would be from the meter.

[branadic]

With the ADR1000 based reference chances are most of the noise and a significant part of the drift would be from the meter.

I disagree to some extend. I'm observing my ADR1000 for 2 months now by directly reading it with a R6581D, differential measurement with a K2182A against my F7000-1 and differential measurement of the F7000-1 against F7000-2 with a second K2182A. Consequently, I take readings for an hour each morning. You can clearly see that the DMM can detect the very same amount of drift that is also visible in the differential measurement.
The ADR reference is not pre-aged, neither by an 168 h bake&burn @150 °C that was mentioned by Eric Modica nor a cycling receipe as proposed by Cern for the LTZ.

-branadic-

[Andreas]

Test whole night and datas:

Hello,

- which (model of) instrument did you use?
- do you have any clue where the negative -3 ppm spikes are from?

with best regards

Andreas

[syau]

Test board：

Did you leave any gap between the ADR1000A and the PCB ? Also, what are the value of those film resistors (1K/13K, 120, 70K x 2) ?

I am think of  making up a test board based on Dr. Frank’s PCB using ww resistors.

[wutieru]

I use HP 3458A for testing.
and the -3 ppm spike maybe come from Vref Pin current ，I made a mistake when wiring.

[wutieru]

I use 1K:12.5K / 70K /70K/ 100R for testing.

[iMo]

OT: @wutieru: I downloaded the MeterKnife Lite (as in your picture above) - is there an english version available too?

[wutieru]

not only the software but also a hardware named "metercare" you needed. this is a DIY GPIB device by funs.

[syau]

I use 1K:12.5K / 70K /70K/ 100R for testing.

May I know if you leave any gap between the PCB and the ADR1000A ? If so, what is the distance between the PCB and the ADR1000A ?

[syau]

My version:

ADR1000AHZ



LTZ1000ACH



Those WW resistors were result from a group buy in this forum. Will burn in for few days before taking measurement.

[Noopy]

I found a ADR1000 with datecode 2108. 

Wanna see what´s inside of this batch? 

[wutieru]

Quick start to stable state，only need 30 seconds.

[Andreas]

Quick start to stable state，only need 30 seconds.

Hello,

you should zoom more in (into the ppm level)
The heater voltage on my scope still changes after 7-8 minutes.

with best regards

Andreas

[Kleinstein]

The heater power can change slowly even if the temperature is stable. It is normal that this takes longer to stabilize as the whole PCB heats up a little and not just the refrence chip.  The point to look for the actual temperature would be the voltage over R1 (the resistor setting the Zener current).
The chip is small and the temperatur regultion can thus be quite fast.

The longer part is settling in the reference, e.g. thermal stress between the die an the case. This what can cause hysteresis and slow settling over weeks and more. Epoxy glue and similar polymeres can also show internal relaxation in the structure. Above or near the glass temperature the internal structure reaches a kind of equilibrium (temperature dependent) fast (e.g. a few seconds at the glass temperature). At lower temperature this gets slower (like double the time every 3 to 10 K). So with more than some 50-100 K below the glass temperature the structure tends to lag behind the equilibrium.

[Noopy]

I stripped the layers of the ADR1000.
Here you can see how I proceeded: https://www.eevblog.com/forum/projects/decapping-and-chip-documentation-howto/msg3733501/#msg3733501




The Linear Technology Application Note 83 shows how a buried zener can look like.
Breakdown occurs between the n+ emitter doping and the p+ isolation sinker.
In my view the picture is not ideal. Under the n+ doping there is no p- but a p+ due to the addition of the isolation sinker doping and the base doping. Furthermore the doping decreases to the edges of the sinker. Only because of that the breakdown occurs in the middle of the bottom of the n+ emitter doping.




In my view the ADR1000 is built like this:
The blue temperature sensing transistors are easy to analyse. There is a buried n+ collector connector. In the outer area there is a sinker connecting the buried collector connector. On top of the collector there is a base area in which you can spot the emitter area. 
The collector of the green inner transistor is connected by four long buried n+ collector connectors. The emitter area is the circle in the middle of the die. The base is a wide p doped circle between the collector and the emitter.
The base circle is connected to the anode of the zener too. Different to the AN the anode is smaller than the cathode. We can be sure about that because the wider ring is connected to the metal layer that carries the cathode potential.






Cutting through the die I assume the structures look like this.


https://www.richis-lab.de/REF19.htm

 

[magic]

Mostly the same kind of structure I came up with.
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg3078768/#msg3078768

One difference: if the connection from pin 4 to the anode/base element were made at the surface, there would be no need to join this element with the P isolation wall and the substrate. I speculate that there is a gap between the anode/base and the P isolation wall, and that the deep anode diffusion is not isolated from the substrate except for those four buried Q1 collector links and that's where the connection to pin 4 lies.

The Linear Technology Application Note 83 shows how a buried zener can look like.
Breakdown occurs between the n+ emitter doping and the p+ isolation sinker.
In my view the picture is not ideal. Under the n+ doping there is no p- but a p+ due to the addition of the isolation sinker doping and the base doping. Furthermore the doping decreases to the edges of the sinker. Only because of that the breakdown occurs in the middle of the bottom of the n+ emitter doping.

I agree, that drawing seems inaccurate.

[Noopy]

if the connection from pin 4 to the anode/base element were made at the surface, there would be no need to join this element with the P isolation wall and the substrate. I speculate that there is a gap between the anode/base and the P isolation wall, and that the deep anode diffusion is not isolated from the substrate except for those four buried Q1 collector links and that's where the connection to pin 4 lies.

I agree with you that this configuration would be possible and would provide us with a Kelvin connection but I don´t see a gap... 

[Noopy]

With the input of magic and a second look at the pictures I would like to change the picture a litte.

[Noopy]

i am guessing i got the pin numbers right?
pads A and B, any guesses what are the uses? a special ADR with 10 pins?
there is a square with 25 dots, is that some kind of via?

Let´s take a closer look at the circuit. There are two additional bondpads leading to the reference circuit.








With the two metal layers the circuit is a little harder to read...




The active parts are the same as in the LTZ1000 but here we have the possibility of a 4-wire-connection! VREF and GND are connected a second time directly at the zener-transistor-combo so the bias current and the traces conducting the current don´t interfere with the reference voltage. The sense traces are connected to the two unused bondpads.

 

[chuckb]

Special thanks to Noopy for braving the dangers of Hydroflouric Acid to show us more details of the chip.   

[Noopy]

With the input of magic and a second look at the pictures I would like to change the picture a litte.

With some background information I know can tell you that this picture shows the right arrangement (in principle of course the length and width are not true to scale). 

Special thanks to Noopy for braving the dangers of Hydroflouric Acid to show us more details of the chip.   

 

[magic]

With some background information I know can tell you that this picture shows the right arrangement (in principle of course the length and width are not true to scale). 

Hmm, what sort of secret information is that?

I agree with you that this configuration would be possible and would provide us with a Kelvin connection but I don´t see a gap... 

That's a fair argument. I can't really find such gap on your photos now.

there is a square with 25 dots, is that some kind of via?

On pin 2? Not sure why there are dots but it's definitely connection between the upper (south) and lower (north) layer of metal interconnects. The two layers look slightly different and there are several other connections, but without dots.

[iMo]

With some background information I know can tell you that this picture shows the right arrangement (in principle of course the length and width are not true to scale). 

Hmm, what sort of secret information is that?

EricM from ADI presented (MM2021) the cross sections of the heater rings and the reference breakdown region with details (the dopants profiles)..
PS: there are at least XXX people who saw the presentations.. It could be all of them signed an NDA  and therefore everything is kept secret
PPS: I was told there are only 2 pages - n.13 and n.15 related to the ADR1001 (out of 17) of Eric's presentation marked as "ADI Confidential". His presentation is not marked as "Confidential".

[Noopy]

With some background information I know can tell you that this picture shows the right arrangement (in principle of course the length and width are not true to scale). 

Hmm, what sort of secret information is that?

EricM from ADI presented (MM2021) the cross sections of the heater rings and the reference breakdown region with details (the dopants profiles)..
PS: there are at least XXX people who saw the presentations.. It could be all of them signed an NDA  and therefore everything is kept secret
PPS: I was told there are only 2 pages - n.13 and n.15 related to the ADR1001 (out of 17) of Eric's presentation marked as "ADI Confidential". His presentation is not marked as "Confidential".

That is the source of my knowledge. 
I don´t want to make someone sad or angry. I won´t upload the presentation or the cross section but I´m happy I got quite close in the first place and with a small correction my picture shows the right structures. 

there is a square with 25 dots, is that some kind of via?

On pin 2? Not sure why there are dots but it's definitely connection between the upper (south) and lower (north) layer of metal interconnects. The two layers look slightly different and there are several other connections, but without dots.

You are right in the heater supply there are a lot of small vias while the other vias are big squares. Interesting... Perhaps the smaller vias are better for higher currents. Perhaps there is just no reason for the difference. I can´t tell.

[CDN_Torsten]

You are right in the heater supply there are a lot of small vias while the other vias are big squares. Interesting... Perhaps the smaller vias are better for higher currents. Perhaps there is just no reason for the difference. I can´t tell.

If memory serves me (sorry haven't routed an IC in decades)...I believe the current carrying capability of a via (in an IC) is a function of the perimeter.    So the multiple small vias would have more perimeter than one big via while occupying a similar area.

[chuckb]

What has eight legs and an eerie green glow?
My friend Jim has a home lab filled with high end microscopes. Yesterday he discovered why the ADR1000 has such fantastic performance. It is actually a Radioactive Spider at heart. A little Black Light (UV Light) shows its true identity.  It’s a good thing AD keeps them locked up in the metal can. 

Nikon Z6, amazing low light performance
Nikon SMZ-U Trinocular, great depth of field.

[Noopy]

What has eight legs and an eerie green glow?

Looks dangerous! 

[ramon]

Let guess, Indium?

Not long ago I read some document about bonding stress free die attachment. Sorry, I have lost the source.

[antintedo]

Not long ago I read some document about bonding stress free die attachment. Sorry, I have lost the source.

Might be this one: https://zymet.com/new/TC601_1_070522.pdf

[wutieru]

Does anyone spot this problem on the picture?
First， I thought it was interference caused by the power supply or wiring problems.
I changed another PCB that I used before , but still exists.

[rigrunner]

If I saw that here I would also be thinking wiring, supply, local disturbance etc.

I have only got one ADR powered up currently, but I haven't observed any jump or spike larger than 5uV so far.

[Kleinstein]

Mobile phones can sometimes produce quite some high power interference for some sync calls, that happen not very often (e.g. 10 x a days). The output side of the LTZ/ADR1000 circuit is a bit sensitive to loading. Depending on the circuit details it may be more or less sensitive to EMI.

[iMo]

The capacitor at the adr1000 output (100nF in the datasheet) should be a ceramic (TH), imho. The foil capacitors do not work well at higher EMI frequencies..

[Andreas]

Hello,

can be many root causes.
One is popcorn noise (either your instrument or the reference).

The other is EMI influence from outside (USB-cables, refrigerator, mobile phone ...)

I have the impression that the ADR1000 is more sensitive to EMI as the LTZ1000.
But I have at the moment no comparable cirquit where I can prove the impression.

I had (during 1/f noise measurement) 2-4 uV negative spikes all 15.4 seconds
which were related to a unused FTDI RS232 cable which was laying (unintentionally)
on the same desk where I did the 1/f noise measurement.

with best regards

Andreas

[Noopy]

In the ADR pin 4 is connected directly to the case.
In the LTZ pin 4 is just connected to the die and there it connects to the substrate.
Does the special polymer die attach electrically connect the substrate and the package?
Either way EMI coupling to the ADR package adds directly to your reference voltage. LTZ looks a little more robust.

[iMo]

I think operating a mobile phone close to the vref is a good check.. For such frequencies even 100nF ceramics do not work well. You would need something like quality 1nF ceramic smd. Mind the ceramic smd capacitors generate voltage under mechanical stress..

[iMo]

In the ADR pin 4 is connected directly to the case.
In the LTZ pin 4 is just connected to the die and there it connects to the substrate.
..

I can see a bond wire from the pin 4 to the case in the LTZ..

[Noopy]

In the ADR pin 4 is connected directly to the case.
In the LTZ pin 4 is just connected to the die and there it connects to the substrate.
..

I can see a bond wire from the pin 4 to the case in the LTZ..

Really? 

[iMo]

On 1 o'clock..

[Noopy]

On 1 o'clock..

Sorry, I don't get it. 

[iMo]

..

[Noopy]

..

 
That's just a part of the package that originated due to the grinding.
Pin 4 is at 4 o'clock.

[iMo]

  Now, what would the experts say on the "grounding vs. not grounding the ref case" ? Would be interesting to know..

[Noopy]

That's an interesting difference between the LTZ and the ADR.
In an ideal environment that doesn't matter but in real life... 

[iMo]

In the ADR the 100ohm resistor (wired to its case) is somehow blocked against EMI via the opamp's feedback capacitor (100nF in the DS). That would not work for vhf/uhf frequencies, imho. I would try with an 100pF-1nF ceramic smd parallel to the 100ohm resistor.

[wutieru]

I think it is caused by insufficient aging.

[Kleinstein]

With the 100 nF capacitor (C5 in ADR1000 DS) directly at the OPs FB, adding capacitance parallel to the 100 Ohms would act like capacitive load to the output. Depending on the OP used this may be bad for the stability.  AFAIR Andreas added more to the circuit (AFAIR an additional resistor between pin4 and the OP) in his EMI enhanced version.

With such changes it is a good idea to check the stability / tendency for oscillation in a simulation. With multiple feedback paths and some gain from the transitor the stability is not so easy.

For a practical compact circuit is may be interesting if C3,C4,C5 could be made somewhat smaller (e.g. 33 nF) to make C0G capacitors more feasable.

I would not expect aging to cause sudden jumps. Aging is mainly about the glue / die attach and this is more like smooth, slow drift.

[iMo]

For, say 800MHz - 5GHz (wireless), even a 10pF ceramic smd would help..

[wutieru]

see picture as below:

[MiDi]

I think operating a mobile phone close to the vref is a good check.. For such frequencies even 100nF ceramics do not work well. You would need something like quality 1nF ceramic smd. Mind the ceramic smd capacitors generate voltage under mechanical stress..

Indeed, especially if you are making a call.

You talk about the piezoelectric effect of ceramic capacitors?
It is unrelated to smd, different ceramics show differently pronounced piezoelectric effect, for e.g. C0G type it is negligible.

[iMo]

There is a relation to smd, indeed. Typically the stress on an smd part is higher than with a TH part.
PS: example: the knocking at the bnc connector of an o'scope is a standard test Dave is showing in his teardown videos..

[Kleinstein]

Most THT parts have some extra decoupling from the board stress. However it is not only capacitors that react to stress. Also resistors react to stress and change there value. With resistor arrays in a plastic case it may be easier to get a change in ratio from stress than from temperature.

[3roomlab]

see picture as below:

so this zener seem to only perform best after high temp age "treatment" ? anybody else doing high temp age vs not-high temp aged comparison?
could you talk a little about your 168hr "treatment". like the heat cycling you did?

[branadic]

Eric Modica mentioned in his talk at MM2021 - that's where this graph is coming from that wutieru linked in the post above - that them (ADI) but also their customers prefer a bake&burn for 168 h at 150 °C and observed that it removes low frequency noise. No cycling, no secret recipe, just bake&burn.

-branadic-

[harerod]

Eric Modica mentioned in his talk at MM2021 - that's where this graph is coming from that wutieru linked in the post above - that them (ADI) but also their customers prefer a bake&burn for 168 h at 150 °C and observed that it removes low frequency noise. No cycling, no secret receipe, just bake&burn.

-branadic-

branadic, to make that statement crystal clear: are we talking about baking the powered-down ADR1000 in an oven for one week @ 150°C ?

[Kleinstein]

The graph with the spikes on the low curve for the non baked parts is suspicious. The extra spikes and excursions all happen at the same time for the 5 DUTs. Chances are this is more an effect of the measurement system or EMI and has nothing to do with the treatment. Hard to tell if there is some effect on the noise. That part is hidden under interference.

The drift changes quite a bit. For my feeling it looks like the 168 h at 150 C may be a bit too much / too high, as the dirft changed the direction.
Some high temperature bake is definetly a good idea to let the chemical reaction of a 2 K glue really finish. It help to be above the glass temperature for this.

If the die attach glue stays glassy / amorphous  the rate of cool down from some 140 C to 80 C may have some effect. Not sure if fast or slow cool would be better, both have there merits. A fast (e.g. less than 10 min) cool cold help with stess relieve as the stress can relax before the structure. Slow cool could result in a denser structure and thus slower creep afterwards.

[branadic]

Here is a small snippet of the Q&A session:

Question: Is a burn-in possible with the integrated heater?

That's a very good question. I don't know. It is one of the reasons why we build in the capability to raise the chip temperature all the way up to 120 °C, it's not quite as hot as we would run our typical burn-in, but we will certainly try it and if it works we will publish something in the datasheet. Because yes, we did think about that.

Question: At which temperature did you make the burn-in?

I believe it's a 150 °C.

Question: So simply powering the zener and have the ambient at 150°C and then - what was it? - 168 h or something like that?

Yeah, you know and we haven't tested the bounce of that, we basically just copied the bake and burn of the LTZ1000, which is a 168h both bake and burn. We have at least small sample size evidence that you maybe able to get away with just doing an unpowered bake and that's good enough to disipate the charge but I have to say I haven't looked at it very closely.

Question: Simply elevated temperature or cycling during burn-in? Because, some rumor says cycling around glass transition temperature might improve the aging.

Our burn-in process doesn't utilize any cycling.

So you can see, it's a bake&burn process with the circuit powered, otherwise it would be a bake only. Hope that answers your question.

-branadic-

[Andreas]

In the LTZ pin 4 is just connected to the die and there it connects to the substrate.
Does the special polymer die attach electrically connect the substrate and the package?

Hello,

if you measure the package against 0V you will read about 0.5-0.6V.
So yes: the substrate is somehow connected to the case also in LTZ1000.

with best regards

Andreas

[Noopy]

In the LTZ pin 4 is just connected to the die and there it connects to the substrate.
Does the special polymer die attach electrically connect the substrate and the package?

Hello,

if you measure the package against 0V you will read about 0.5-0.6V.
So yes: the substrate is somehow connected to the case also in LTZ1000.

with best regards

Andreas

Thanks! I have donated my (destroyed) LTZ1000 so I can´t measure the path.
Nevertheless 0,5-0,6V seems to be no hard connection so the ADR can be a little more robust against EMI than the LTZ.
Do you know whether there is a different between the LTZ1000 and the LTZ1000A? This strange polymer die attach with the glass beads doesn´t sound very low ohmic.

[TiN]

Substrate connected to case only on LTZ1000CH. ACH have it isolated.

[Noopy]

Substrate connected to case only on LTZ1000CH. ACH have it isolated.

That would explain a better EMI behaviour. 
The package of the ACH is still no shield since it isn´t connected to anything but it´s better than connecting it to the reference output.

[3roomlab]

The drift changes quite a bit. For my feeling it looks like the 168 h at 150 C may be a bit too much / too high, as the dirft changed the direction.
Some high temperature bake is definetly a good idea to let the chemical reaction of a 2 K glue really finish. It help to be above the glass temperature for this.

If the die attach glue stays glassy / amorphous  the rate of cool down from some 140 C to 80 C may have some effect. Not sure if fast or slow cool would be better, both have there merits. A fast (e.g. less than 10 min) cool cold help with stess relieve as the stress can relax before the structure. Slow cool could result in a denser structure and thus slower creep afterwards.

maybe after 168hr HT-age, it needs the hot/cold modulation cycling.

[branadic]

The graph with the spikes on the low curve for the non baked parts is suspicious. The extra spikes and excursions all happen at the same time for the 5 DUTs. Chances are this is more an effect of the measurement system or EMI and has nothing to do with the treatment. Hard to tell if there is some effect on the noise. That part is hidden under interference.

I was thinking the very same and what we can see there seems to be some common mode effect, not low frequency noise.

Some high temperature bake is definetly a good idea to let the chemical reaction of a 2 K glue really finish. It help to be above the glass temperature for this.

Eric argued, that the bake&burn removes some surface charges. However, I think bake&burn helps for two major effects:

1. curing / hardening the die attach and remove stress introduced by it --> needs to be around/above above glass transition temperature of the die attach and cycling around glass transition temperature would help even more
2. "healing" of the silicon latice from the doping process --> normally rapid annealing with neon lamps is done after doping steps, which doesn't mean that the latice is already back in perfect condition

I can barely remember some website entry / article from an institute / university discussing such investigation for the LTZ1000, but can't seem to find it at the moment.

-branadic-

[iMo]

An annealing process (ie after an ion implantation) is done at around 1000degC for about an hour.
The surface charges might be somehow removed by 150degC (ie gasses/vapors) but they will condense back on the surface afterwards.
So the baking as discussed with Eric is mostly about the epoxy glue, imho.

[Kleinstein]

A burn in at 150 C would not help much with latice defects in silicon. The temperature is just way to low. For annealing the silicon, think about 700°C and up, so nothing to be done with the ready made part.  Some of the process steps to grow a new oxide layer already works as nealing for much of the earlier damage.
If at all it may be some impurities (e.g. hydrogen) to still move inside the silicon, the silicon itself is fixed.

Surface charges can indeed by a problem, but here it would be more like having a surface that is not susceptiple to catch new one. The glow from hot electrons visible with many zeners or transistors in avalance mode is a thing to avoid, as this could also excite deap states in the oxide.
Water can bond relatively strong to the oxide surface: In vaccum it takes some 100 C to get it off in reasonable time. This water causes stress, kind of making the oxide swell.  I have seen the water in/at the surface oxide bend the silicon (though with thin silicon parts). With the sealed case the water is kind of fixed and nothing to get it out or in (not sure which case is better). The water at the surface could cause some hysteresis, like a new relaxation after oven turn on after off for some time.

I would expect the burn in to be mainly about the epoxy / glue. Here there are 2 processes:
1) stress from mismatch in thermal expansion, of the glue, die and case . At the glass temperature or slightly below the relaxation should the fast enough, that there is no more stress from processes before.
2) relaxation in the structure of the amorpous polymer. Well below the glass temperature the material wants to get a denser structure, but the relaxation gets slow and the actual structure is usually less dense than the equlibrium. The degree of order in the strcuture also effects the strength and speed of the relaxation. Initially it can be fast, but once relaxed it can slow down the process quite a bit.

Chances are a good process would be cure at a high temperature, like 150 C (I don't think this would need a full week and I don't think cycling would help), than cool down relatively fast and to a relatively low temperature (maybe even room temperature) to let much of the stress relax with a still relatively soft structure. This step may take quite some time.  Than heat up to a temperaure a little (e.g. 20-30 K) above the later oven temperature (but below Tg - 50K) for some time to get the structural relaxation done, without adding much new stress. The last step may well be with just the internal heater and the zener running.

[wutieru]

Noise Test (using 120R),Magnification：100K
Converted to actual value： about 0.9uVpp

[branadic]

A burn in at 150 C would not help much with latice defects in silicon. The temperature is just way to low.

For a pure bake your statement is right, but for a powered circuit at 150 °C things are different.

-branadic-

[Noopy]

Now let´s take a look into an ADR1000 out of the new batch.








Looks the same like the 1839 ADR1000. 


https://www.richis-lab.de/REF19.htm

 

[Andreas]

Noise Test (using 120R),Magnification：100K
Converted to actual value： about 0.9uVpp

Hello,

can you do a FFT on the aquired data?
there seems to be much non statistical noise on the screen:
either from AC mains frequency or too much bandwith of the scope.
(BW-limiter 20 MHz is active?)

with best regards

Andreas

[wutieru]

Niose with FFT

[Andreas]

Helllo,

ok this is with bandwidth limiter.
So the fuzzyness of the signal is much better now.

FFT reveals that there is one frequency outside the passband of the LNA which is in amplitude similar than the passband signal.
I am not familiar with this scope. For me it looks like you did a linear frequency display with 157 Hz center frequency and 313 Hz span.
So roughly 31 Hz / div.
I guess that the (red marked) discrete frequency is the mains frequency with 50 Hz.

I usually use a metal cookies box to shield against fields from outside operating the DUT on batteries. But even with the shield I have to put all transformers away with a minimum distance of 1 m. Even my HP34401A internal transformer (shielding around transformer and the housing of the Instrument) has to be at a minimum distance of 0.5 m.

On a quiet place I have the 50 Hz hum  ~18 dB below the pass band for a ADR1000A.
See attachment with double logarithmic display.

with best regards

Andreas

[wutieru]

Yes, there is 50Hz noise Exist, it comes from the test environment, I changed another oscilloscope, it disappeared, it seems that the shielding of KS oscilloscope is not very good.

[Andreas]

So you can see, it's a bake&burn process with the circuit powered, otherwise it would be a bake only. Hope that answers your question.

So according to that what is published the burn in cirquit would look something like this.
(the resitors of course cemented wire wound for the high temperature).
The unused heater is connected to substrate.

any comments?

with best regards

Andreas

[branadic]

I can't really comment on that as I simply don't know. However, with respect to the question at MM2021 if the internal heater could be used for a bake&burn, Erics answer was related to the ADR1001, not to the ADR1000.
So one could propably set the oven temperature to a value of 150 °C, which would translate to a divider of 17.74k : 1k with the numbers given for the datasheet circuit and use the build in heater for the bake&burn of the ADR1000?

-branadic-

[rigrunner]

I've been experimenting with heater temperature on the ADR1000.
Running the ADR heater at 85C for 48hrs, 110C for another 48hrs and then 150C for a further 48hrs briefly changed the drift direction on my sample of 1.
I'm still playing around with heating and will post should I discover anything useful.

[wutieru]

I think the burning@150° process is irreversible.  Test it several times and see how the curve changes？

[wutieru]

Power off-on testing，first interval 3 hours，second interval 30 minutes.
It seems that the voltage has changed a little  after a long time of power off.

[Andreas]

Running the ADR heater at 85C for 48hrs, 110C for another 48hrs and then 150C for a further 48hrs briefly changed the drift direction on my sample of 1.

The more interesting things would be:
- does the noise actually decrease.
- does the ageing drift (after a settling time) decrease.

Question: how did you check that the temperature was actually 150 deg C?

with best regards

Andreas

[Andreas]

So one could propably set the oven temperature to a value of 150 °C, which would translate to a divider of 17.74k : 1k with the numbers given for the datasheet circuit and use the build in heater for the

Hello,

I fear this is not true with my sample.
when calculating from my measurements I get a slope of -2.24 mV/K for the sensing transistor with 70K and pin 6+8 connected. (offset 648.9 mV)
To get 150 deg C temperature I would have to set around 312 mV. which translates with a zener voltage of 6600 mV to a ratio 20.15K : 1K.

But will the Zener really still have 6600 mV at 150 deg C?

with best regards

Andreas

[branadic]

I fear this is not true with my sample.

... which would translate to a divider of 17.74k : 1k with the numbers given for the datasheet circuit

Using the temperature sensor as a diode requires for sure a different divider setpoint.

-branadic-

[Vtile]

If there is any analogy to metallurgy, glass working and mineral manipulation by annealing one should not only apply heat, but also control the cooldown. Rule of thumb on those mentioned before is that slower is better.

[rigrunner]

The more interesting things would be:
- does the noise actually decrease.

No noise reduction yet.

- does the ageing drift (after a settling time) decrease.

It is still drifting at about the same rate.

Question: how did you check that the temperature was actually 150 deg C?

Measuring the diode voltage across Q2 using -2mV/C 

If there is any analogy to metallurgy, glass working and mineral manipulation by annealing one should not only apply heat, but also control the cooldown. Rule of thumb on those mentioned before is that slower is better.

Stepped temperature tests are something I had considered on a fresh ADR. I may try that when I have finished poking ADR#1.

[rigrunner]

when calculating from my measurements I get a slope of -2.24 mV/K for the sensing transistor with 70K and pin 6+8 connected. (offset 648.9 mV)
To get 150 deg C temperature I would have to set around 312 mV. which translates with a zener voltage of 6600 mV to a ratio 20.15K : 1K.

But will the Zener really still have 6600 mV at 150 deg C?

With Q2 configured as a diode I measure around 574mV @ 25.2C for a 71K5 R3. ADR#1 and a random used LTZ are in the same ball park.

[Andreas]

With Q2 configured as a diode I measure around 574mV @ 25.2C for a 71K5 R3. ADR#1 and a random used LTZ are in the same ball park.

The question is: how this is measured.
I need sampling with 1 NPLC (faster would be better) and measuring the very first sample (with heater off)
after power on to get 592.3 mV @ 25.3 deg C and 69.8K.
Due to self heating the voltage changes quickly during the first seconds
and finally stabilizes at around -20 mV lower level (+10 deg C).

... which would translate to a divider of 17.74k : 1k with the numbers given for the datasheet circuit

Ok, in datasheet the pull up resistor is only 62K so we have a slightly higher collector current.
But which datasheet numbers do you refer to? For Q2 I see only the table 7 with some values.

Measuring the diode voltage across Q2 using -2mV/C 

With my cirquit this would be roughly 10% too low. So only around 135 deg C.

This is one reason why I want to use 2*25W-aluminium wirewound resistors as heater
and sensing the case/environment temperature of the ADR with a PT1000 sensor.
The whole in a termo-isolated coffee mug or something like that.

with best regards

Andreas

[rigrunner]

With Q2 configured as a diode I measure around 574mV @ 25.2C for a 71K5 R3. ADR#1 and a random used LTZ are in the same ball park.

The question is: how this is measured.
I need sampling with 1 NPLC (faster would be better) and measuring the very first sample (with heater off)
after power on to get 592.3 mV @ 25.3 deg C and 69.8K.
Due to self heating the voltage changes quickly during the first seconds
and finally stabilizes at around -20 mV lower level (+10 deg C).

I can measure Q2 voltage with the opamp from the heater side of things removed.
I'll need to play around with my logging setup. I currently trigger a sample every minute.

    Measuring the diode voltage across Q2 using -2mV/C

With my cirquit this would be roughly 10% too low. So only around 135 deg C.

Best I can do at the moment for checking ADR can temperature is take a thermal image.

[Kleinstein]

I think the burning@150° process is irreversible.  Test it several times and see how the curve changes？

I would expect some reversible effect of the heating to 150 C.
The 150 C are likely above, or at least close to the glass temperature of the die attach glue.
So this would kind of reset the glassy structure to a relatively open (high free volume) state. How much of this is retained depends on the speed of cool down and later aging at lower temperature like 60 C during operation.

The 150 C would be enough for stress from thermal mismatch between the silicon and case to relax. So this would start again from the same level.

The curing of the glues part would be irreversible and probably well fisched after 1 week at 150 C.

[rigrunner]

The question is: how this is measured.
I need sampling with 1 NPLC (faster would be better) and measuring the very first sample (with heater off)

3457a setup with NDIG=4, NPLC=.1, AZERO=1,TARM=AUTO,TRIG=SGL, NRDGS=1.
Readings were taken as fast as the pi could manage in a python loop with query(3457,'?')

594.98mV start and drops 13.62mV in 190 seconds to 581.36mV with self heating.

[branadic]

But which datasheet numbers do you refer to? For Q2 I see only the table 7 with some values.

I'm refering exactly to that table, from which you can extrapolate the resistor values for a 150 °C oven temperature for the datasheet circuit.
However, things are different when using Q1 as a diode only. By the way, from what I measured on my first unit I get a -2.166 mV/K for Q1 used as a diode.

Running my reference for about 1850 h by now I don't observe any real settling, though not a steady drift either.
The linear drift rate is about -360 nV/day (~13 ppm/a), allthough a square fit matches much better, while a square-root fit shows large deviations and doesn't match well, at least by now.

-branadic-

[harerod]

I have been asking myself how a gang-burn-in of several ADR1000 might be set up. I have some Vishay RSSD resistors in the lab, which usually serve as loads during EMI tests of power supplies. Those resistors feature a cylindrical ceramic body, which can accommodate several TO-99.

Today's test was about general feasibility. I usually wouldn't touch a breadboard with a ten feet pole, but in these trying times of component shortages, I decided to go retro. It turned out that the only THT OPAMP, which I had at hand, was a veteran that has been with me for a very long time. No component on that breadboard is younger than thirty years.

Since I didn't want to risk an ADR1000 for tests, I decided to use an 1N4148 as target and a KTY81-122 as sensor for a temperature regulator.

I was a bit concerned about flammability, so I took care with the setup:
The two devices go into the RSSD 25x168 100R.
Eager to start the experiment I didn't even bother cleaning out the cookie jar.
The RSSD is suspended by some highly flame resistant steel mounts from the mill router. Some electrical insulation goes under the mounts, not shown in the picture.
The cookie jar is wrapped in cotton cloth. The cotton cloth rests on a thick towel and is covered by another towel.
The whole heap is placed on a metal step ladder next to my desk. 

It turns out that in steady state the setup burns 13W in the RSSD. It is autumn here and I need central heating to keep the lab warm, so I am not bothered by the power dissipation.
 

Now that everybody had a good laugh - could anybody tell how critical that 150°C burn-in temperature is, please?

In which temperature band should the die be kept?
What is the lowest temperature that makes sense?
At which temperature can we expect damage to the ADR1000?
After all, according to the datasheet, 150°C is the junction temperature limit for storage and operation.

[Kleinstein]

The main part to worry about, when going high in temperature is the die attach glue.  Epoxy may not like a temperature much above 170 C. With a little patience I would consider some 140 C and maybe even 130 C sufficient. This should still be at least close to the usual epoxy galss temperarure. It just takes longer at a lower temperaure.  If the target is an open structure and thus fast cool down, I would like to start at the glass temperature or above, but not much below.

I may make absolutely sense to later use a step at a lower temperature, maybe some 20 C above the later oven temperature.

[branadic]

What is the lowest temperature that makes sense?

It depends on what the mechanism for the initial drift actually is.

Eric Modica argued at MM2021: "... Another important aspect that carries over to the ADR1000 is the benefit of bake and burn process. So when I have first started working with buried zeners and Vrefs in general the conventional wisdom was that bake and burn speed up the aging of the part and that one could accelerate the early life drift to finish up sooner. That’s this area here, that’s the majority of your drift on parts like this, your uncertainty. So I’ve learned enough in my years working in analog design not to make broad generalisations, but what I’ve observed is that bake and burn is helpful primarily for low frequency noise reduction in this context. We have had tier 1 manufacturers tell us the same thing when we send them ADR1000s. Prior to instituting a bake and burn most of the parts look okay from a low frequency noise perspective, but if you look at it enough you find one that is extremly noisy and actually remedies with the bake and burn. So only being a process hobbiest myself, the only explaination I can give there is that you have these dangling hydrongene bonds at the surface of the SI and sometimes that amounts to surface charge and if that ends up in the vacinity of the zener, if you are unlucky enough, then this is the kind of thing that you get. And it’s really the only thing that would move at such low temperatures where you are running a bake and burn, say like a 150 °C...."

I don't know if this is the real deal and wouldn't argue against it, cause I can't prove him neither right nor wrong. However, I believe from what I've learned over the years on packaging of integrated circuits that two mechanisms are involved. One being the heavily doping in the silcon latice and the second coming from the assembly of the die within the package using die attach. All sorts of plastics, thermoplastic and thermoset, do suffer from stress after being (injection) molded or applied and cured. A thermal treatment afterwards can to some extend remove this stress in those and there are many examples where such thermal treatment process is being applied.
I don't know if there were ever studies to rule out what is actually the reason for the initial drift of voltage references in hermetically sealed packages and if they can be seperated from another and what is just stupid hypthesis, rumor or wisdom.
Nevertheless, the 150 °C treatment for 168 h Eric mentioned as well as the cycling that Cern performs on the LTZ1000 gives evidence that at least the stress due to the die attach is a major contributor. I do have build at least one LTZ reference, that was heat-treated using the cycling methode and what I observed on it was, that it settled within about two weeks. I know other forum members, not to mention any names here, that observed similar effects. However, I couldn't get enough people together to perform a double-blind experiment on that to give it more evidence.
So the answer to that die attach contribution would be: storing the Vref above or cycling around the glass transition temperature, whatever that is for the die attach used in the ADR. Chances are, it's in the range of 110 to 120 °C as for many epoxy materials in that field, but we simply don't know and Eric couldn't answer that either.

-branadic-

[Andreas]

In which temperature band should the die be kept?
What is the lowest temperature that makes sense?
At which temperature can we expect damage to the ADR1000?
After all, according to the datasheet, 150°C is the junction temperature limit for storage and operation.

Hello,

so what I have heard now 150 deg C is the oven temperature to go? (or is it more the 125 deg C of other processes?)
In a bake & burn configuration we will also have some self heating by the biased zener.
(at room temperature this is ~10-11 deg C).

As already mentioned I have 2 x 25W heater resistors which I will put into a thermally isolated coffee mug or thermo-jar. So I think I will get away with 10-30 W power (or less). A PT1000 temperature sensor just arrived today. I will use one of my ADCs with 4 channel MUX to evaluate temperature. (2 channels for 0 deg C and 150 deg C reference resistors).

The PC software will allow ramping the temperature (e.g. from 145-155 deg C with 0.1K/min slope).

with best regards

Andreas

[Dr. Frank]

Hello colleagues,
I recommend not to go near or even above 150°C, as this is the absolute max. rating for storage and chip temperature, and 125°C is the maximum operating temperature, surely for some reason.

I would also chose around 120°C, what branadic has suggested, being a typical glass temperature for die attach.

You might use external heating and additional internal heating.

Such burn-in circuits are often designed with a passive substitution circuit around, i.e. w/o using any OpAmps or other semiconductor components, but appropriate HT resistors instead, due to these high environmental temperatures which are involved.

Frank

[dietert1]

Which means the reference will be soldered after burn-in. No good, better try to run it in its circuit, maybe using an external oven at 85 °C and the internal heater to get the additional 35 °C.

Regards, Dieter

[branadic]

so what I have heard now 150 deg C is the oven temperature to go? (or is it more the 125 deg C of other processes?)
In a bake & burn configuration we will also have some self heating by the biased zener.
(at room temperature this is ~10-11 deg C).

I do understand that the engineer wants a specific number for an upper temperature limit, allthough there are also some physicists among us. But noone can give you such a number.

You have to understand that the numbers given in the datasheet such as "Specified for −40 °C to +125 °C operation" or "Junction Temperature Range −65 °C to +150 °C" means continuous operation over lifetime. It's not that at 125.1 °C the part goes pop because of some curious effect kicking in, it's not as steep as that. But instead some effects are starting to accelerate, such as electromigration. Having said that, there is no reason to believe that operating the part for a week at 150 °C, either oven temperature or chip temperature, would degrade it permanently. The structures on the silicon are rather large and the burn-in time is limited to cause some serious degradation. Long story short, I wouldn't worry to much about the absolute temperature.

-branadic-

[DavidKo]

Will not be the lifetime at higher temperatures smaller? LEDs have the lifetime only ~500h instead of ten thousands hours when the junction is at 150°C .

[Kleinstein]

The shorter life time limit at high temperature is due to slow diffusion or similar processes to be faster with high temperature (about double every 5 to 10 C higher). With the reference we want most the those slow processes to already settle before the refrence is used. For this reason there is an initial burn in / anneal phase at a relatively high temperature to lett those effects settle in reasonable time (e.g. 1 week). Later use of the reference is at lower temperature like 40 C, just warm enough to get good temperature control.

[rigrunner]

More torturing of ADR#1 

I decided to try something else with ADR#1. It takes about an hour to get stable with heater temperature at 150C (long legs, no baffle or insulation). With that I decided to see what happens if I power the ADR on for 60 minutes, wait 15 minutes to cool down and go again. This has given the largest shift so far. A little under 200uV upward in 31 hours.

There are a couple of data points missing or blanked out. My power controller doesn't talk to the logging part, so sometimes it screws up and powers down in the middle of a reading etc. Sorry

[Kleinstein]

Is the temperature regulated at the 150 C , or is it just the maximum possible to reach with the given supply ?

With regulation I would have expected much faster stabilization of the temperature. Normally it should be more like 10 seconds, maybe a minute for a stable temperature.

It may be interesting to also get a stable lower temperature (e.g. 50 C range) and maybe wait longer there to at least get a stable temperature there and get thus a comparable good reading to see changes to the chip.

[rigrunner]

150C is regulated. I'm not using the data sheet circuit. The heater is wired up differently. At a more sensible temperature  stability is reached around 20 ish minutes.

I'm planning on dropping the temperature in a little while to see what has changed at the lower settings.

[Kleinstein]

Is the 20 minutes the time to reach a stable temperaure or the time to reach a stable ouput voltage ?
These 2 can be different. If still in regulation the temperature should be faster, unless most of the time at full power still before actual regulation starts.

The difference betwenn stable temperature and stable voltage can be interesting, as the extra relaxation to effect the voltage would be one of the effects (e.g. stress between case and die) to look at with the burn in process.

[rigrunner]

Is the 20 minutes the time to reach a stable temperaure or the time to reach a stable ouput voltage ?
These 2 can be different. If still in regulation the temperature should be faster, unless most of the time at full power still before actual regulation starts.

Good point.
20 minutes is the time to reach stable output voltage.

The difference betwenn stable temperature and stable voltage can be interesting, as the extra relaxation to effect the voltage would be one of the effects (e.g. stress between case and die) to look at with the burn in process.

When I lower the heater temperature next I will hook up another DMM for the heater opamp and log a sample to show the heater settle time.

[branadic]

Here is an update of my drift measurement, almost 2000 h have past. Again, the reference wasn't pre-aged, only the z.t.c. temperature was measured, the oven adjusted to that temperature, the amplification adjusted to 10.000 00 V and the overall t.c. trimmed. The reference is sitting at the bench, supplied by a low noise lab power supply.
The reference is drifting downwards compared to F7000-1. By the eye the drift is not linear, but slightly curved, a sqr-fit matches best. Although, we are looking at 3.2 ppm of change, which is quite a lot. I'm tempted to stop it here and to perform a one week cooking, using the internal oven. On the other hand I could also build a second one. Not yet sure how to proceed.
Don't take the result too serious, this is only a single specimen and thus without any statistical significance.

-branadic-

[Kleinstein]

The drift still look pretty linear. There is a little curvature in the curve, but not really much.  Chances are, that just longer waiting would not change very much (excepts the continued linear drift).

Going up in tremperature for some time would be an option. Similar some time (e.g. 1-2 days) of power off may also be an option, to see if there is a hysteresis from power down.  The change to a higher temperature may need to power down anyway. So if possible, one could to the change in the configuration in a way so that one could do 1 day power off, than maybe 1 day with the current temperature (to see if there is a fast settling part) and only than cooking at a higher temperature (with the internal heater likely more 100-120 C). It could help to also get readings at the higher temperature - so keep the regulator active and just change the set point and maybe increase the external temperature for the whole board a little.

To see the change it would be good to get back to the current temperature / config anyway.

[Andreas]

Here is an update of my drift measurement

It would be interesting also to have the 6.6V output data.
So we do not know wether it is the zener or the output voltage divider.

with best regards

Andreas

[branadic]

As the same output stage was used on several other LTZ references before, without showing any drift, I'm convienced, that it's the ADR itself
However, equipment is limited, so you can't track it all at the same time.

-branadic-

[miro123]

The graph is almost linear
What about office temperature?
If I characterize reference I'd measure the reference. Re-start and mesayre for 1 week wil have you and idea if the trend is changed

[wutieru]

Noise testing，it seems that in right way.

[ScoobyDoo]

Hello Voltnuts,
                           
                           I received following communication from the ADI product marketing team in USA responsible for the ADR1000 release - In a nutshell they made following communication:

The ADR1000 will become available through various sales channels (Digikey - Mouser - Arrow) and an EV-board is in progress - but due to other related product releases (ADR100x) - with higher priority - this process is rolling out more slowly than anticipated - some patience is needed before this part will be available through your preferred distributor sales channel ...
The fact that Digikey and Mouser were mentioned means that for Voltnuts it will be much easier to order these parts in smaller quantitities & w/o VAT reg. hurdles

The ADI product marketing team is hoping to ‘release’ this part on the web by end of year.

Best regards
ScoobyDoo

[rigrunner]

Is the 20 minutes the time to reach a stable temperaure or the time to reach a stable ouput voltage ?

Thanks for making me take a closer look at the heater settle time.
The heater was making it to around the 85-90C mark pretty quickly and then taking the rest of the cycle to slowly reach 150C. My 11V on board supply was the problem. Running with a higher supply the heater settles much faster now.

[Andreas]

So according to that what is published the burn in cirquit would look something like this.

Hello,

after the discussion I plan to do a  up to 125 deg C burn in cirquit.
So I can use standard PCBs (non high TG) and standard (non PTFE) burn in sockets.

Some impressions of the mechanics with security switchoff at 140 deg C.

With best regards

Andreas

[Andreas]

Hello,

I am trying two different methods of finding the Zero TC point of the ADR reference.

method a)
via setpoint change (using a external DAC to influence the voltage divider for the setpoint) and active internal heater
in this case I can do rather fast temperature sweeps (1K per minute shown) which saves measurement time.

method b)
via external (environment) temperature change in this case the internal heater is de-activated.

In both cases the temperature sensing transistor (Pin 6 + 8 ) connected together is used as X-axis = internal sensor temperature voltage for the diagram.
The y axis is the zener voltage over temperature.

in case a) (heater on) I get 542.4 mV = 47.5 deg C as zero TC temperature
in case b) (heater off) I get 538.2 mV = 49.4 deg C as zero TC temperature

so around 4 mV temperature sensing voltage difference or 2 deg C which I cannot explain 

with best regards

Andreas

[Kleinstein]

The test with the environmental temperature also effects the resistors. So the resistor TC would be included. I have not calculated how much shift ( added linear TC) is needed, but it may not be much.  Another part is possible thermal EMF, that may come in from the heater power. This may cause a little and would be part of the residual TC even if the heater is active.

There is a small contribution from the base current, but this should be only some 0.2 mV of shift.

[dietert1]

Also reference voltage at zero TC point shifted by about 18  uV (3 ppm).
If you calculate the residual TC at the average temperature 48.5 °C, i'd guess it will be very small and much better than a LTZ 1000. I mean if you can't decide which oven method you will use.
Our LTFLU references run in small TEC ovens with the support circuit, including the 10 V amplifier. I know they run better when using a mixture of on-chip temperature measurement and oven temperature measurement, since the circuitry around the reference chip imposes its TC times small temperature differences (gradients inside oven). These effects can be separated by their response times. After a temperature step from another temperature to your zero TC point you may see a response with a fast and a slow component that finally compensate each other in steady state.

Regards, Dieter

[Andreas]

Hello,

still waiting for some parts for the burn in PCB.
But in the mean time I did some dry-runs with the little oven within the coffee mug.

All tests with ~15V supply voltage (unstabilized).
So with the two 10 Ohms resistors in series I have around 11 W heater power.
The heater is isolated with some cotton cloth.
At 130 deg C inside the outside of the coffee mug is around hand warm.

Startup time from 25 deg C to 110 deg C is less than 25 minutes.
There is enough heater power to do ramps with 2 deg/minute between 110 and 130 deg C.
So this will be also the possible burn in cycle which takes 2 hours. 10 minutes ramp and 50 minutes steady state.

Switch-off of overtemperature switch is done at ~137 deg C. ( Data sheet value = 140 deg C +/- 5 deg C).
But in this case we have a large hysteresis down to 115 deg C until the switch closes again.
Eventually a ramp between 115 and 135 deg is possible without interference of the overtemperature switch.

with best regards

Andreas

[RikV]

As of today, 21st of november, the ADR1000 nor ADR1001 are listed on the AD website (https://www.analog.com/en/products) (in Belgium), nor on the ADX site. Mouser.BE doesn't know the device (even not the ADR1399). So, what is all this?

[Andreas]

Hello,

first (still unadjusted) TC-Measurement of my ADR1000#1.
I am using long legs.

And the output voltage (10V) is done by a 2:1 statistical output voltage divider (nominal ratio = 1.500000)
Zero TC of the zener is relative low with ~49.5 deg C.

Calculated setpoint temperature is ~51.8 deg C (so slightly above Zero T.C.)

Result:
10V output voltage T.C. ~190 ppb/K.
6.6 -> 10V transfer ratio ~24 ppb/K and ~6 ppm too low against nominal ratio.
(so naked zener would be 166 ppb/K).

Did someone already check the influence of the length of the legs on T.C.?

with best regards

Andreas

[Andreas]

Hello,

another test.

the idea was: can I influence the slope of the T.C. by using a slightly higher (or lower) setpoint temperature as the Zero T.C. temperature.
Above the zero T.C. temperature the slope is falling so what when setting the setpoint somewhat higer (or lower?)
I adjusted the voltage divider for T.C. setpoint from 11.5 / 1 to values in 5 degree steps.

With ~10-12 deg C self heating of the device by zener current I expected to see the device falling out of regulation when decreasing the setpoint to 46.9 deg C.
Interestingly was that this was not the case. (obviously the zero T.C. is good enough to compensate this for environment temperatures up to 40 deg C.)

I made a further test  with 41.9 deg C which clearly showed that the temperature regulation is only valid up to ~32 deg C.
Result: I need to set the zero T.C. and the final oven temperature somewhat higher (e.g. 52-55 deg C) for a environment temperature up to 40 deg C.

Total result: unfortunately the slope of 190 ppb/K cannot be compensated by setting the heater setpoint slightly around the zero T.C. point.
(although the graphical representation seems to have a sweet spot near the zero T.C. temperature, but maybe this is only stray by the measurement uncertainity).

with best regards

Andreas

[Kleinstein]

The small residual TC for the ouput valtage can be from multiple sources.  The TC of the unheated reference and than a not so perfect temperature regulation is only one possible mechanism. As it looks this is more like a small part.

Other contributions are the resistors (though better than with the LTZ1000 as much less effect from the temperature setpoint). There is still the 100/120 Ohms resistor to set the current and the collector resistor(s).

Anther contribution is directly an effect of the temperature gradients on the chip or at the case and thus the heater power. One could check this part by adding / removing some insulation around the chip. The constant part of the gradients are not a problem, but the heater power changes about linear with the external temperature and temperature gradients about scale with the heater power. So the temperature gradients will change about linear with temperature.

Ideally one would get a stable temperature of the reference without much change in the heater power - like having a 2nd oven, even if not very stable.

The board temperature can also change the mechanical stress to the reference, though I don't think this would be a large contribution - at least we should hope it is not, as thus would also come with a humidity effect.

[dietert1]

If you run the reference with its own heater inside an oven made to keep the divider at constant temperature, you could use the reference heater signal for TC adjustment, as proposed in the LTZ1000 datasheet. Inside the oven the heater is well protected from ambient temperature changes and you have a fixed operation point, so higher order terms will be negligible. If you bring the heater signal out of the oven, you can fine tune the adjustment without opening the oven. In your case an extremely small contribution will be necessary (maybe 100 MOhm instead of 400 KOhm).

Why not use a TEC oven? You can keep temperatures lower and near usual ambient temperatures, so any kind of power failure or the like will cause less deviation. Our LTFLU ovens restart with less than 0.1 or 0.2 ppm deviation after 24 h without TEC control. I know because one of our Arroyos used to turn off every two or three months (for no good reason). Meanwhile i solved that problem: Send output on command once per minute.

Regards, Dieter

[Andreas]

Hello,

now I have built 2 burn in PCBs for the ADR1000A.
This one is intended for noise measurements before and after the burn in.

But I fear that T2 base emitter voltage is too low (0.41V) to give any useful noise measurements.

with best regards

Andreas

[Kleinstein]

The circuit is a bit tricky, as the current is driven from the external 1.8 K resistor on top, and not controlled by the transistor inside the ADR1000. So the TC and noise would not be the same as in the normal circuit. If not carefull there may even be change to damage (at least stress) the transistor from too much base current (e.g. at higher temperarure when VBE goes down). So I don't think the board is really good for burn in.

Why not use the "normal" circuit with maybe lower grade resistors ?

[Andreas]

Why not use the "normal" circuit with maybe lower grade resistors ?

Hello,

sorry I did not find a suitable Op-amp which is specced for up to 150 deg environment temperature.
And what do you mean with "lower grade" resistors. Can there be lower grade resistors than the 5% metal oxide resistors (with 235 deg C maximum temperature?)

In the mean time I measured the noise of the cirquit.

Unfortunately I have typical 7uVpp over a 100 sec time span with this cirquit.

with best regards

Andreas

[dietert1]

If you have a small p-channel mosfet, it can prevent excessive base current while hot. You connect source to reference output Pin3, gate to the collector of Q2 Pin5 and drain to ground Pin7.

Regards, Dieter

[Kleinstein]

For protection of the transisitor a resistor at pin 7 would be enough. The temperature / output measurement would than be not from ground, but pin 7 instead.

I did not consider that the whole circuit is planed to run so hot. In this case it would be a bit tricky to build a circuit to work with high temperature parts.
In principle one could consider a circuit with discrete transistors (e.g. 2N2222), maybe as a cascode and than a emitter-follower. 
 
The more logical way would be the OP and most of the resistors at a lower temperature and only the reference chip itself hot. This could possibly be some 60-80 C for most of the circuit and than the ref internal heater to reach the final temperature.

[dietert1]

The p-channel mosfet i proposed is a 150 °C capable substitute for an opamp. It brings the reference ciruit into an operational state very similar to the standard circuit and supports the intended noise measurements during burn-in.

Regards, Dieter

[Andreas]

Hello,

so your suggestion is this way - right?

This will shunt away current from the zener and base as soon as the base current increases.

I do not think that it helps for the noise but it should help against base current overload so I will implement it.
I still have some BSS84p or similar which are not ideal for the job but should work.

with best regards

Andreas

[miro123]

In the mean time I measured the noise of the cirquit.
Unfortunately I have typical 7uVpp over a 100 sec time span with this cirquit.

Interesting graphs. The noise spectrum flatten around 100Hz.
Is this typical ADR1000 behavior or limitation of measurement setup?

[Andreas]

Hello,

uVpp is always 1/f noise measured with a bandwidth limited amplifier (0.1-10 Hz)
Additionally I have a 1000 Hz math filter in the scope to reduce noise of the scope input.

The flattening should be around the quantisation noise of the scope (16 bit mode)

The reason why im showing above 10 Hz is just to see wether I have too much mains frequency (50 Hz) noise in the measurement.

with best regards

Andreas

[mawyatt]

Why not use the "normal" circuit with maybe lower grade resistors ?

Hello,

sorry I did not find a suitable Op-amp which is specced for up to 150 deg environment temperature.
And what do you mean with "lower grade" resistors. Can there be lower grade resistors than the 5% metal oxide resistors (with 235 deg C maximum temperature?)

In the mean time I measured the noise of the cirquit.

Unfortunately I have typical 7uVpp over a 100 sec time span with this cirquit.

with best regards

Andreas

Nice!! What scope and setup did you use for the spectrum plots, those are impressive indeed!!

Best,

[Andreas]

Hello,

Setup is a low noise 10000 fold amplifier 0.1-10 Hz bandwidth (cirquit attached as Filt1105w.pdf) here:
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg834013/#msg834013

together with the DUT (voltage reference) in a metal cookies box (all battery supplied).
Scope is a PicoScope 5444A. (the successor model sold today is a 5444D)

Settings:
20 MHz hardware bandwidth limiter.
10 mV/Div resulting in 1uV/Div with the low noise amplifier.
16 Bit hardware resolution.
Sample rate 20 kS/s 100 seconds (2 Million samples per scan, 19 scans as input for FFTs).
1 kHz digital filter. (far enough away from the 10 Hz hardware filter).
FFT with rectangular window 1 Million points (binning of 2 Million samples)
Averaging of 19 FFT outputs (not input samples) to reduce noise floor of the scope (not the DUT).

without 1kHz filter and without averaging the FFT looks like this:

with best regards

Andreas

[Kleinstein]

The simplified circuit can be sensitive to noise in the supply to provide the external current. If the P-MOSFET gets active some (on the order of 1/200) of the noise of the MOSFET can also enter the result. The TC would likely also be quite a bit higher than the normal reference circuit and there can thus be noise from temperature fluctuations.

In the ciruit with the extra feedback (P MOSFET or 2 x NPN) one may also want some capacitance (e.g. at pin 3) to slow down the loop at 1 defined point to make sure it gets stable and is not oscillating.

[dietert1]

Andreas, yes that's what i thought. According to datasheet the BSS84p operates up to 150 °C.

Without regulation the power supply noise gets divided by 1K/100R=10. With regulation it will be much less. And i think you get regulation already at room temperature, as (13.9 - 6.43V) / 1K gives about 7.5 mA, not 4.3 mA. So there should be base current at room temperature.

Yes, there is a chance for oscillation, as the two transistors have enough gain to make a nice oscillator. If it happens, it can be solved by removing C4. The mod poses no risk to the rare/expensive reference. But of course one can make a ADR1000 dummy from discrete parts to check a circuit for "fatal threats".

Regards, Dieter

[Andreas]

Hello,

its not 1K for the upper resistor but 1K8. (I wanted to stay below 5 mA)

with best regards

Andreas

[Andreas]

Hello,

pre-test for burn in cirquit with a temperature sweep from 60 to 130 deg C.
Above ~80 deg C the base emitter voltage of T2 decreases by the shunted current of the FET.
So the cirquit works.

But obviously the zener current decreases from initially 4.1 mA (410mV) to around 3 mA (300 mV) at 130 deg C.
So the cirquit is not ideal but should be ok for burn in.

Another effect is due to the massive heat sink capacity of the aluminium bar where the ADR is connected to.
The self heating seems to be only around 2-3 deg C.
(but this is measured with around 8 NPLC of my ADC#25 since my DMMs are occupied by ADR#1)

with best regards

Andreas

[dietert1]

While the p-channel mosfet works similar to the usual opamp, there is another variant with a p-channel depletion JFET shunt parallel to R1, i.e. S = U1pin4, G= U1pin5 and D=Gnd. That variant does not shunt zener current at high temperatures. The J175 datasheet says it can be used up to 150 °C.

Regards, DIeter

[Kleinstein]

In the normal operation with the closed loop the zener current is not constant, but going down. So the change in current is normal. The unusual part is more the constant current at lower temperature.

A problem with the circuit is that the noise and drift may well be more effected by the FET and not so much the reference itself. So one would not get a good real time measurement of the aging / burn in.

[dietert1]

From previous discussion it was clear that the loop amplifier in this application isn't critical. If one uses the output of the on-chip transistor, it becomes the input stage and determines noise and temperature dependent offset.

Regards, Dieter

[MaxTesla]

Hello,
Thanks to ScoobyDoo and branadic, I got 5 samples and some PCBs.
Here's the mockup board, where I made the first tests.

The lower zener voltage compared to the LTZ1000 now allows this Reference Amplifier to be trimmed to zero T.C. for a given oven temperature, by individual selection of the collector resistor R2. Compare this to the +50ppm/K for the unregulated LTZ1000.

That trimming process is the same procedure as for the SZA263 and LTFLU in the different FLUKE instruments.

I measured U_ref vs. temperature, which always gives a negative, flat parable. The zero T.C. point varies from 45 to 60°C for a fixed value of R2, and the T.C. varies from about +5 .. -5ppm/K over 24 ... 80°C only.

At a fixed divider ratio R4:R5 for around 52°C oven temperature for all five samples, the individual T.C.s are well below +/- 2ppm/K.
It follows that the attenuation factors for R3 and the most critical R4, R5 are at least 10 times greater than in the datasheet. These resistors now can have much relaxed stability specifications, only R1 and R2 should be PWW or BMF types.

I also measured the low frequency noise 0.1 .. 10Hz, like branadic and Castorp.
I use an LNA from Andreas, and always made "typical" measurements of 10sec long samples, so to have a direct comparison to my LTZ1000s, but also to the measurements done by Andreas.
100s measurement time or averaging give no essentially different results.
Other LNAs will give different absolute noise figures, so these results are not directly comparable .
 
The noise of the ADRs RefAmp evidently is about 50% of the LTZ1000, which several others also have confirmed.

ADR #1 would be the most quiet sample, but it suffers from  noise, or frequent dips, but with a very moderate amplitude of maybe 400nV_pp.
The ADA4522 buffer contributes a lot to the overall noise, so I have to investigate if it's only this specific unit, or what can be done to improve this.

I will add the noise figures of these samples, when measured with my 3458A, maybe now I can see a pronounced difference.

As my 34465A showed much higher stability / lower noise when I replaced its LM399 reference by a LTZ1000 clone PCB, my idea is to replace the LTZ1000 in my 3458A with an ADR1000 (*).
Maybe its noise figures will also go down, best by a factor of 2.
Castorp found that in his DVM other components overwhelmed the noise budget, so let's see what happens.

I would like to remind that TiN some time ago made a similar experiment on one of his 3458A, by implementing 4 parallel LTZ references, which would also yield 50% noise of a single LTZ.
I'm not aware what his final result was.

Anyhow, I think this new device will require quite a big time slot during the Metrology Meeting 2021, and I'm excited to meet the volt-nuts community personally and we'll have some fruitful discussions and comparisons   

Frank

(*) my 3458A already has a modified FW, with CAL? 2,5 = 6.50000V (7V reference lower limit).
Thanks to Poul-Henning Kamp for your advice how to correct the checksum.

Hello Frank,

in your post you mentioned that you modified your 3458a's firmware to accept the lower voltage value of the ADR.
Would you mind sharing the process? I like to try it as well...

BR,

Max

[Dr. Frank]

...
 
The noise of the ADRs RefAmp evidently is about 50% of the LTZ1000, which several others also have confirmed.

...

As my 34465A showed much higher stability / lower noise when I replaced its LM399 reference by a LTZ1000 clone PCB, my idea is to replace the LTZ1000 in my 3458A with an ADR1000 (*).
Maybe its noise figures will also go down, best by a factor of 2.
....

I would like to remind that TiN some time ago made a similar experiment on one of his 3458A, by implementing 4 parallel LTZ references, which would also yield 50% noise of a single LTZ.
I'm not aware what his final result was.


Frank

(*) my 3458A already has a modified FW, with CAL? 2,5 = 6.50000V (7V reference lower limit).
Thanks to Poul-Henning Kamp for your advice how to correct the checksum.

Hello Frank,

in your post you mentioned that you modified your 3458a's firmware to accept the lower voltage value of the ADR.
Would you mind sharing the process? I like to try it as well...

BR,

Max

Hello Max,

It was not sufficient to only change the minimum limit for the CAL? 2,5 parameter, i.e. the minimum internal voltage reference, to 6.5000V, but it was also necessary to modify all those lower ACAL limits, which the lower reference voltage also affects.
This can be identified by comparing the auto-calibration constants of a 3458A with two different reference voltages, like a regular LTZ1000, and one with a different LTZ chip, or with an ADR1000 on board, and slightly faked calibration reference values.
TiN kindly helped me by testing my first version on one of his 3458As, and sending me dumps of all his calibration constants. With the ADR1000, he found out, that the ACAL procedure gave system errors, due to several  lower limit violations.

So the procedure is to search for the location of these additional, crucial constants CAL? XXX,5 inside the firmware (=> 13 EA), and then lower them by the approximate difference by the ADR1000 voltage, about 6.667V, and the lower limit of the LTZ1000, i.e. CAL? 2,5 = 7.000V, i.e. each new minimum value has be calculated to about -6% each.
The displayed minimum values CAL? XXX,5 are not represented directly in the firmware, instead they are a calculated product of several internal cal parameters, so that's a bit tricky to find them in the hex file.

For that you need an appropriate hex editor, which is also capable of calculating the DOUBLE numbers back and forth, by taking care for the correct endianess and correct byte-order inside the single or 6 fold EPROMs, so that you can now determine the new lower limits.
I've done that successfully on my 2nd firmware version, which I also run in my own 3458A.

TiN was again so kind to test it and he confirmed that now the basic calibration processes, CAL 0, CAL 10 and CAL 10000 are successful, as well as the ACAL ALL procedure.
Next step would be to optimize the ADR1000 reference board for lowest noise performance, maybe also using less noisy OpAmps inside the 3458A, and then check against another external ADR1000 reference, or another 4x averaged zener reference, if the 3458A then shows a lower noise performance as well.

I don't know, how much further hints or spoilers you want to receive, just let me know.

Frank   

[MaxTesla]

...
 
The noise of the ADRs RefAmp evidently is about 50% of the LTZ1000, which several others also have confirmed.

...

As my 34465A showed much higher stability / lower noise when I replaced its LM399 reference by a LTZ1000 clone PCB, my idea is to replace the LTZ1000 in my 3458A with an ADR1000 (*).
Maybe its noise figures will also go down, best by a factor of 2.
....

I would like to remind that TiN some time ago made a similar experiment on one of his 3458A, by implementing 4 parallel LTZ references, which would also yield 50% noise of a single LTZ.
I'm not aware what his final result was.


Frank

(*) my 3458A already has a modified FW, with CAL? 2,5 = 6.50000V (7V reference lower limit).
Thanks to Poul-Henning Kamp for your advice how to correct the checksum.

Hello Frank,

in your post you mentioned that you modified your 3458a's firmware to accept the lower voltage value of the ADR.
Would you mind sharing the process? I like to try it as well...

BR,

Max

Hello Max,

It was not sufficient to only change the minimum limit for the CAL? 2,5 parameter, i.e. the minimum internal voltage reference, to 6.5000V, but it was also necessary to modify all those lower ACAL limits, which the lower reference voltage also affects.
This can be identified by comparing the auto-calibration constants of a 3458A with two different reference voltages, like a regular LTZ1000, and one with a different LTZ chip, or with an ADR1000 on board, and slightly faked calibration reference values.
TiN kindly helped me by testing my first version on one of his 3458As, and sending me dumps of all his calibration constants. With the ADR1000, he found out, that the ACAL procedure gave system errors, due to several  lower limit violations.

So the procedure is to search for the location of these additional, crucial constants CAL? XXX,5 inside the firmware (=> 13 EA), and then lower them by the approximate difference by the ADR1000 voltage, about 6.667V, and the lower limit of the LTZ1000, i.e. CAL? 2,5 = 7.000V, i.e. each new minimum value has be calculated to about -6% each.
The displayed minimum values CAL? XXX,5 are not represented directly in the firmware, instead they are a calculated product of several internal cal parameters, so that's a bit tricky to find them in the hex file.

For that you need an appropriate hex editor, which is also capable of calculating the DOUBLE numbers back and forth, by taking care for the correct endianess and correct byte-order inside the single or 6 fold EPROMs, so that you can now determine the new lower limits.
I've done that successfully on my 2nd firmware version, which I also run in my own 3458A.

TiN was again so kind to test it and he confirmed that now the basic calibration processes, CAL 0, CAL 10 and CAL 10000 are successful, as well as the ACAL ALL procedure.
Next step would be to optimize the ADR1000 reference board for lowest noise performance, maybe also using less noisy OpAmps inside the 3458A, and then check against another external ADR1000 reference, or another 4x averaged zener reference, if the 3458A then shows a lower noise performance as well.

I don't know, how much further hints or spoilers you want to receive, just let me know.

Frank   

Hey Frank,

thanks for your detailed answer. This sounds like a lot of changes to the firmware (especially for my limit skillset regarding this kind of stuff). I will definitly have a look into it.
THB I was more on the look for a higher stability spec then lower noise, but for that I first need to finish some boards to test and select some ADRs.

For me it would seem a lot simpler to redesign the A9 board with a small (stable) amp which boosts the ADR output just enough to get in the range of working (i.e. 7 Volts if i remember it correctly).
Obviously noise is concerne in this regard, but this should be manageable (at least from my perspective today) with the right selcetion of ADR/Circuit and would be a "upgrade" path for others without touching any firmware.
To compare agiant I would use another ADR reference, as I am currently designing my take on a 10V reference.

Opions on this "plan" are obviously welcome

BR,

Max

[Kleinstein]

Adding some 6% of gain the reference circuit would still require very stable resistors, not to increase the drift. Added noise would be the least problem.
It it is just about getting a long term stable reference the old type LTZ1000 reference may be the better choice, maybe with a reduced temperature.

So far the limited reports on the ADR1000 are showing very low noise, but also still some drift. There is still some hope to improve on the drift with burn in or annealing, but I would not absolutely count on this. The advantage of the ADR1000 is more that it can get away with lesser grade resistors and still get good stability. With really good resistors the LTZ1000 may still be more stable.

[branadic]

Speaking of drift, I reached 3000 h but I'm not happy with the result. Nevertheless, I guess it is good advice to share such results anyway.

-branadic-

[MaxTesla]

Adding some 6% of gain the reference circuit would still require very stable resistors, not to increase the drift. Added noise would be the least problem.
It it is just about getting a long term stable reference the old type LTZ1000 reference may be the better choice, maybe with a reduced temperature.

So far the limited reports on the ADR1000 are showing very low noise, but also still some drift. There is still some hope to improve on the drift with burn in or annealing, but I would not absolutely count on this. The advantage of the ADR1000 is more that it can get away with lesser grade resistors and still get good stability. With really good resistors the LTZ1000 may still be more stable.

Yes you might be right. But with the limited dataset I still plane on experimenting. I have treated myself with a set of VHP101 for a LTZ (70k,120,13k/1k) and the ADR (acc to the Datasheet). Plane is to put the LTZ in the mix with the ADR in with the same layout and test them side by side. This will take some time though as my ADR resistors are scheduled for march…goal is to compare drift and tempco. This will be done against my well aged 732a so may in a year from now I will have somewhat meaningful results, after that I will have a look at the A9 board project again I think.

[MaxTesla]

Speaking of drift, I reached 3000 h but I'm not happy with the result. Nevertheless, I guess it is good advice to share such results anyway.

-branadic-

Wow nice work

What temperature setpoint do you use on your ADRs?

[Dr. Frank]

Adding some 6% of gain the reference circuit would still require very stable resistors, not to increase the drift. Added noise would be the least problem.
It it is just about getting a long term stable reference the old type LTZ1000 reference may be the better choice, maybe with a reduced temperature.

So far the limited reports on the ADR1000 are showing very low noise, but also still some drift. There is still some hope to improve on the drift with burn in or annealing, but I would not absolutely count on this. The advantage of the ADR1000 is more that it can get away with lesser grade resistors and still get good stability. With really good resistors the LTZ1000 may still be more stable.

Yes you might be right. But with the limited dataset I still plane on experimenting. I have treated myself with a set of VHP101 for a LTZ (70k,120,13k/1k) and the ADR (acc to the Datasheet). Plane is to put the LTZ in the mix with the ADR in with the same layout and test them side by side. This will take some time though as my ADR resistors are scheduled for march…goal is to compare drift and tempco. This will be done against my well aged 732a so may in a year from now I will have somewhat meaningful results, after that I will have a look at the A9 board project again I think.

Hello Max,
unfortunately, reading every speculation from him, always the complement seems to be correct. No own experience from his own experiments, as far as one can deduce.

A divider being a few % off from "1" is extremely stable.. simply calculate the total differential! A 6% amplification should attenuate all drifts of the implemented resistors by about these 6%. I.e. a resistors T.C. of 2ppm/K will be attenuated to 0.1ppm/K of amplification drift. 5ppm/yr. => 0.3ppm/yr.
Added noise by an OpAmp IS the most severe problem for the ADR1000, as I have demonstrated upwards in this thread, in conjunction with the ADA4522 @ x1 amplification!! Any additional buffer amp might spoil the superior noise performance of the ADR1000.  Therefore the easy replacement of the FW is the best option.. if the 3458A circuit really benefits from the low noise ADR.
   
The long-term stability of the ADR1000 MIGHT initially be worse compared to the LTZ1000. Only branadic has presented so far a single drift result @10V.

My single sample experiment, measured directly at the raw ADR1000 output, w/o to 10V amplification, shows no remarkable drift so far after 500h. That's also no reliable result at all. Others in the background reported that the ADR needs a longer burn-in than the LTZ. So the speculation which chip is more stable, is at the moment completely baseless and speculative.
Spoiler alert: CERN obviously uses the ADR based circuit as an improved 10mA precision source! 

Lower oven-temperature is always a good way to improve timely stability, see P.J. Spreadbury publication from 1990. Chose about 55 .. 60°C in any case. Don't use the ADRs datasheet values for best performance of the ADR. See hints in my post above.

The T.C. of the circuit for both chips have to be trimmed to near zero. There's a lot of potential on the 3458A's LTZ1000A board. Mine has about 0.2ppm/K, half of the overall temperature drift of my instrument.
Use that infamous 200k resistor for T.C. trimming. (or not at all).
The LTZ1000A inside my 3458A drifted -2ppm over about 9 years only, due to the reduced oven temperature, but mainly because the 3458A is running only sparsely. Again, see Spreadbury.

My ADR #3 runs at about 55°C, just at its determined zero TC point of the Reference Amplifier.  Simple thing.

And no, the LTZ1000 does not necessarily need more stable resistors. The timely stability of the circuit is mainly determined by the chip itself, the T.C. drift can be trimmed to near zero. Exactly the same goes for the ADR. So, please, no Vodoo stuff here, again. The ADR has much better noise performance, definitely, and that's the main reason to go for it.

The 3458A is no voltage reference, by all means. Therefore, I might exchange its very stable LTZ1000A by the ADR1000 some day, using the very same PCB with different resistors, and adding Andreas capacitors, for higher noise immunity.

Frank

[MaxTesla]

Adding some 6% of gain the reference circuit would still require very stable resistors, not to increase the drift. Added noise would be the least problem.
It it is just about getting a long term stable reference the old type LTZ1000 reference may be the better choice, maybe with a reduced temperature.

So far the limited reports on the ADR1000 are showing very low noise, but also still some drift. There is still some hope to improve on the drift with burn in or annealing, but I would not absolutely count on this. The advantage of the ADR1000 is more that it can get away with lesser grade resistors and still get good stability. With really good resistors the LTZ1000 may still be more stable.

Yes you might be right. But with the limited dataset I still plane on experimenting. I have treated myself with a set of VHP101 for a LTZ (70k,120,13k/1k) and the ADR (acc to the Datasheet). Plane is to put the LTZ in the mix with the ADR in with the same layout and test them side by side. This will take some time though as my ADR resistors are scheduled for march…goal is to compare drift and tempco. This will be done against my well aged 732a so may in a year from now I will have somewhat meaningful results, after that I will have a look at the A9 board project again I think.

Hello Max,
unfortunately, reading every speculation from him, always the complement seems to be correct. No own experience from his own experiments, as far as one can deduce.

A divider being a few % off from "1" is extremely stable.. simply calculate the total differential! A 6% amplification should attenuate all drifts of the implemented resistors by about these 6%. I.e. a resistors T.C. of 2ppm/K will be attenuated to 0.1ppm/K of amplification drift. 5ppm/yr. => 0.3ppm/yr.
Added noise by an OpAmp IS the most severe problem for the ADR1000, as I have demonstrated upwards in this thread, in conjunction with the ADA4522 @ x1 amplification!! Any additional buffer amp might spoil the superior noise performance of the ADR1000.  Therefore the easy replacement of the FW is the best option.. if the 3458A circuit really benefits from the low noise ADR.
   
The long-term stability of the ADR1000 MIGHT initially be worse compared to the LTZ1000. Only branadic has presented so far a single drift result @10V.

My single sample experiment, measured directly at the raw ADR1000 output, w/o to 10V amplification, shows no remarkable drift so far after 500h. That's also no reliable result at all. Others in the background reported that the ADR needs a longer burn-in than the LTZ. So the speculation which chip is more stable, is at the moment completely baseless and speculative.
Spoiler alert: CERN obviously uses the ADR based circuit as an improved 10mA precision source! 

Lower oven-temperature is always a good way to improve timely stability, see P.J. Spreadbury publication from 1990. Chose about 55 .. 60°C in any case. Don't use the ADRs datasheet values for best performance of the ADR. See hints in my post above.

The T.C. of the circuit for both chips have to be trimmed to near zero. There's a lot of potential on the 3458A's LTZ1000A board. Mine has about 0.2ppm/K, half of the overall temperature drift of my instrument.
Use that infamous 200k resistor for T.C. trimming. (or not at all).
The LTZ1000A inside my 3458A drifted -2ppm over about 9 years only, due to the reduced oven temperature, but mainly because the 3458A is running only sparsely. Again, see Spreadbury.

My ADR #3 runs at about 55°C, just at its determined zero TC point of the Reference Amplifier.  Simple thing.

And no, the LTZ1000 does not necessarily need more stable resistors. The timely stability of the circuit is mainly determined by the chip itself, the T.C. drift can be trimmed to near zero. Exactly the same goes for the ADR. So, please, no Vodoo stuff here, again. The ADR has much better noise performance, definitely, and that's the main reason to go for it.

The 3458A is no voltage reference, by all means. Therefore, I might exchange its very stable LTZ1000A by the ADR1000 some day, using the very same PCB with different resistors, and adding Andreas capacitors, for higher noise immunity.

Frank

Hello Frank,

thanks a lot for the insight and the paper, didn't knew that one.
This is highly interesting, out of my guts i would have suspected stabiltiy to be more of an topic, but as the amplifer would be one rougly 1:1.06 you are right of course.
My idea was to implement the amp with a VHD200, so tempco/stability impact is very small.

Regarding the noise and your buffer setup: This suppries me a lot tbh, I mean the ADA4522 with the right kind of input lowpass should not contribute that much, with the assumption that the datasheet noise values are correct, but as you discribed your samples are much lower noise than the spec. Will be interresting to see what my samples do, I have enought of them to get a little bit more "meat at the bone" so to speak.

Thanks for the tipp with the temp setpoint, I don`t have the appropriate means of measuring it myself (yet). So revised plan for tempco/stability test is the following: 6 ADRs in total, 3 with 12K/1K divider (~60°C) and 3 with 11k6/1k (~53°C) setpoint to get a span, each mounted in a housing with the same board, no preconditioning and just accquiring data form the start. This should give me/us a somewhat good idea what is going on, as three is the statistical minimum. Other resitors are 61.9k and 95.3Ohms. Monitoring ambienten contiditions of course to see any kind of correlation during normal weather shifts, all done automaticly every day with my 1281 and 7168 scanner. One channel reserved for a short, one for the 732a. If all goes well we can see a difference in drift with setpoint temperature in two points.
Thats at least the plan so far

After the results comes the tuning...

BR,

Max

[Dr. Frank]

Hello Frank,

thanks a lot for the insight and the paper, didn't knew that one.
This is highly interesting, out of my guts i would have suspected stability to be more of an topic, but as the amplifier would be one roughly 1:1.06 you are right of course.
My idea was to implement the amp with a VHD200, so tempco/stability impact is very small.

Well, a VHD200 is technical overkill, expensive, and has a long leadtime.
Use two appropriate cheaper single resistors from stock, WW or MBF ones, like 6k8 + 470 Ohm, or so.

Regarding the noise and your buffer setup: This surprises me a lot tbh, I mean the ADA4522 with the right kind of input lowpass should not contribute that much, with the assumption that the datasheet noise values are correct, but as you described your samples are much lower noise than the spec. Will be interesting to see what my samples do, I have enough of them to get a little bit more "meat at the bone" so to speak.

I was also surprised, that the MAX4522 (of course with input LPF) adds so much noise. The ADA4522 has typically 115nV pp noise only, which should not affect the ADR1000 noise performance.
This combined noise level is about the same as an LTZ1000 w/o buffer. Maybe I have to look for a better sample of the ADA4522.
It's not only my 5 samples, which have 50% noise of that of the LTZ1000. Other volt-nuts also reported the same.

Inside the 3458A, both LT1001 on A5 board, U160 and U165, which generate the +/- 12V reference for the A/D will definitely spoil the game.
These have 300..600nV pp noise, i.e. at the same level what I've measured for the ADR1000.
These might have to be replaced also.

Thanks for the tipp with the temp setpoint, I don`t have the appropriate means of measuring it myself (yet).

Probably you already HAVE the possibility to measure the oven setpoint by yourself. A faster bench DMM is sufficient, to catch the initial diode voltage @ R.T., with oven disabled, and then extrapolate by using -2.2m/K for the temperature dependency. A 34401A is sufficent for that, and your 1281 will definitely do the job.

You may first test, at which collector current of the Reference Amplifier you get zero T.C. for the (not ovenized) RefAmp at your intended oven temperature.

So revised plan for tempco/stability test is the following: 6 ADRs in total, 3 with 12K/1K divider (~60°C) and 3 with 11k6/1k (~53°C) setpoint to get a span, each mounted in a housing with the same board, no preconditioning and just acquiring data form the start. This should give me/us a somewhat good idea what is going on, as three is the statistical minimum. Other resistors are 61.9k and 95.3Ohms. Monitoring ambient conditions of course to see any kind of correlation during normal weather shifts, all done automatically every day with my 1281 and 7168 scanner. One channel reserved for a short, one for the 732a. If all goes well we can see a difference in drift with setpoint temperature in two points.
That's at least the plan so far

After the results comes the tuning...

BR,

Max

Setpoint of the oven can also be trimmed by changing the collector resistor of the temperature sensing transistor inside the LTZ/ADR circuit.
So you may always use a  12k/1k divider and select the collector resistor instead. The stability requirement on latter is relaxed, especially if you run your circuit exactly at the sweet spot, i.e. at zero T.C.
I do not recommend 11k6/1k, that may be too low a temperature.
Select the zener resistor for a bit more than 5mA @ oven temperature, and select both collector resistors as described above.
Your monitoring capabilities sound pretty good.

Frank

[Kleinstein]

The resistors for the 6.6 to 7 V gain would still be the most critical one when it comes to drift. They enter with a 6% factor, or a bout a facto 17 attenuation. The resistors at the ADR have a larger attenuation, like 200 times and higher.
For the gain stage one may just get away with an LT5400-7 as a maybe better available resistor.

Az OPs can react to the input impedance and details in the supply decoupling. So there is a potenetial for them to show more than the datasheet noise. They are also a possible source for EMI so that other parts than give the extra voltage.

As a lower noise OP at the ADC I would consider the OPA202 as a possible candidate.

[e61_phil]

And no, the LTZ1000 does not necessarily need more stable resistors. The timely stability of the circuit is mainly determined by the chip itself, the T.C. drift can be trimmed to near zero. Exactly the same goes for the ADR. So, please, no Vodoo stuff here, again. The ADR has much better noise performance, definitely, and that's the main reason to go for it.

It is true that the LTZ1000 CAN be trimmed to ZTC, but without trimming the ADR1000 is much better in terms of TC. In one project exactly that was the reason to move to the ADR1000.

[MaxTesla]

Yeah the lead time on these VPG resistors is horrible but ok, metrology is no hobby for the hasty ones, trying to improve my patience here
I am gathering test gear for this project since nearly two years now (orignally planned with the LTZ of course)...

Thanks for the tipp, didn't thought on that one...in my head I needed a external source (of heat) to do it, not that I don't have a 2510, but I wasn't ready to put the efford in.
So I will think about a proper setup to do it (waiting on some new PCBs anyways which one of them will be modified for that).

Yes I am aware of the trimming, however i would like to get the drift characteristics first (and see how the tempco behaves) and then tweek the refs. Indeed i was wondering about the low setpoint, but I think its worth a shot.

Regarding the exchange of U160 and U165: I assume (not checked) there supply is bipolar +-15V, so a LT1128 would work fine and it has very very low noise...considered it instead of the LT1013 but the bipolar power supply needed to get to (or near) 0V keeped my from implementing it. Didn`t seem necessery at the time, but after this discussion about the really low noise floor this is worth a rethink.

BR,

Max

[Kleinstein]

There should be no absolute need to get really zero TC for the unheated reference. If possible it would help to get the TC without the heater to below 20 ppm/K. Something like the 10 ppm/K range should be perfectly OK with a reasonable stable temperature. A slightly higher current for the zener and/or transistor could do the job.

The LT1128 is low noise for a very low impedance ( e.g. 100 Ohms) source, but not with some 10 K like in the 3458 reference amplification. For the ADR1000/LTZ1000 reference the OPs see some 50-70 K and thus even more impedance. The OP is also behind the gain of the transistor and thus less critical. So the LT1013 should be good enough there.

So if at all the LT1128 may have a small chance to work in the 6.6 to 7 V gain stage, but even there I would prefer a lower noise current OP. The chopper artifacts from a chopper OP could be an issue.
The ADC in the 3458 should react a little to reference noise in the 150 kHz range. The noise in this range may depend on the compensation (including OP speed) around the reference. The curve in the DS looks good for the ADR1000, though not too sure which OP is used and there may be limit of the test system.

Trying to get even lower noise from the ADC in the 3458 is a different topic. Only U160 would see the same frequency band as the reference. For U165 and U110 the chopper action of the AZ mode is effective and thus a relevant frequency of more like 2.5 Hz with an often relatively low BW and some extra gain (around 2.25 for the more critcal U110, should be 1.25 for U165).  For the actual use reference noise at lower frequencies than 0.1 Hz may be relevant. The 0.1 to 10 Hz range is more like a range still easy to measure. It would be mainly U160 to worry about for low frequency noise.

[Andreas]

Hello,

working on alternatives to the statistical divider with a LTC1043 cirquit.
The main amplification is done by a 3 to 2 (factor 1.5) charge pump.

The remaining part can be easily done with standard resistors.
Either up or in a additional buffer stage with a voltage divider down.

Temperature compensation is done with a diode cirquit.
The typical +220 ppb/K can be compensated with the following cirquit.
R8+R9 are meant to be a 50K trim pot where the T.C. can be trimmed from around -300 to +300 ppb/K.

Unfortunately there is a unwanted offset along with the T.C. trimming where I am looking for a solution.
Any ideas?

with best regards

Andreas

[Kleinstein]

The TC adjustment would effect the absolute value in most cases. At best one can eliminate the effect at 1 temperature, but it can be made better than in the circuit shown. Both sides of the trimmer should see about the same voltage at a reasonable temperature.

One could get such a setup with a bridge with 2 diodes and 2 resistors on opposite sides to replace the 2 diodes in series. The main parts is shown in the attachment.

It looks like there should be a little more filtering for the switched capacitor part.
The final circuit should also have supply filtering for the AZ op and LTC1043 to keep the supply clean.

[Andreas]

Hello,

current results of ongoing burn in of ADR1000A#02 in burn in socket.

Procedure is temperature cycling from 110 deg C to 130 deg C within 2 Hours.
so 10 minutes rising with 2 deg C/min 50 minutes plateau and 10 minutes falling with 2 deg C/min.
(see previous post).

At around 200 hours I thougt output voltage would stabilize but obviously this is not the case.
Now at nearly 400 hours I am somewhat unshure:

- will there be a stabiisation point
- or will the device be destroyed (because 130 deg C is too high?)
- or is the temperature still too low (as on MM the temperature of 150 deg C was mentioned)

with best regards

Andreas

[branadic]

Hi Andreas,

I guess what you look at is an exponential drift curve. Multiple times I though myself the reference would now start to stabilize, but then drift moved on and it was just a plateau in the drift curve for whatever reason - such as variation of the meters reference and t.c..
Meanwhile I'm confident to predict the drift by an exponential function, as can be seen in the image below.

-branadic-

[Kleinstein]

If there is a stabilizing point is a good question. There is a chance that there is some process that us not stabilizing but contineously going one direction - kind of aging to wards destruction. One part seems to have stabilized after some 200 hours.

The parts still showing slow drift at some 120 C shoud be quite a bit (e.g. a factor of around 100, maybe more) slower when at a more moderate temperature, like 50 C. So chances are the part should be stable enough by now. So I would not worry too much about the still visible drift.

Another point is that the burn in circuit is more sensitive to the resistors (and maybe the MOSFET). The resistors are not the most stable ones. So the observed drift could also be in parts from the resistors in the burn in circuit and maybe the temperature control part.

It is hard to tell if the 130 C are too low and which annealing process is best. Usually the first and obvious part is to really finish the chemical reaction in the die attach glue. Here the 130 C should be high enough, especially with so much time. For a glassy epoxy I than see 2 possible routes:
A) Get a dense stucture in procedure like this:
 1) anneal near or above Tg to finish the chemical reaction
 2) anneal somewhat (e.g. 30-50K) below Tg to get a dense amorphous structure. This could also be with a slow coold-down, maybe in steps.

B) Get a more open glue stucture to allow better stress ralaxation in procedure like this:
 1) anneal near or above Tg to finish the chemical reaction , so more like short time at a rather high temperature
 2) cool down to room temperature fast, to suppress the densification , so keep the epoxy soft by freezing a more open structure.
 3) wait at room temperature or maybe even slightly below, to allow relaxation of thermal mismatch stress.
   (a lower temperature gives more thermal stress and may this way give faster creep, despite of lower temperture)
 4) anneal at an intermediate temperature (e.g. 70 C) to get the structural relaxation in the glue, but not much new thermal stress.
 
For a more definitive answer one would need to run the reference in a more normal circuit and with better resistors. To get a shorter time scale, this may be with a slightly elevated temperature (e.g. 20 K higher than normally planed for).  There can also be some hystereseis type effect and thus a new relatively fast part when going to a lower temperature. So a test at let's say 60 C may again show a fast initial part. So this test would again need some time.

[Andreas]

I guess what you look at is an exponential drift curve.

It looks like that:

When I make the x-Axis as sqrt(kHrs) and the y-Axis as ppm instead of mV then I get (nearly) a linear curve.
Regression is -62.4 ppm/sqrt(kHr) at 120 deg C.

So with Arrhenius and factor 2 every 10 deg C and 60-70 deg C (including self heating) we would get something like 1-2 ppm/kHr which is still too high for a voltage standard. I want to get down to better than 1-2 ppm/year.

with best regards

Andreas

[branadic]

I want to get down to better than 1-2 ppm/year.

Then run the aging process longer. As you can see with my non-pre-aged reference it takes some time. Fit your drift curve with an exponential curve, predict the drift for the future - say for 8760 h - and monitor, if your prediction changes/stabilizes over time. Once the drift prediction stabilized you will see how long it will take to run this procedure.
We can then compare how long it took your reference to stabilize with my one and see, if this process did anything good at all. 

-branadic-

[Andreas]

Hello Branadic,

I calculated roughly -2.3 ppm/kHr for your ageing curve at ~50-60 deg C which is around that what I theoretically would get if lowering the temperature.
Perhaps I will put my setpoint some 5 deg C higher to 115-135 deg C (the maximum I can use with my curent setup) to prove whether the drift increases.

with best regards

Andreas

[Kleinstein]

The drift in the burn in circuit can be more than just from the chip. This can also include other parts, like the resistors.
50-60 C would be usually the die temperature, including self heating. Depending on how the temperature is measured during burn in there may be some self heating to add there.

The doubling every 10 K rule is only a first approximation. The actual number can vary depending on the process. The difference can be especially large when it comes to epoxy close to the glass temperature. There is the Arrenius factor and an additional change in structure.

If an exponential fit is suitable depends.  Aging processes can also follow different laws like square root or stretched exponentials or teh suposition of several effects.
The relaxation of stress can be nonlinear as plastic processes often are. Pure viscose creep is more like the exception.
It probably takes more data to really decide if an exponential is suitable. An exponential fit normally corresponds to a single relatively simple process.
Especially the initial phase has usually more contributions, some faster will decay full to later only leave the slowest visible process and other processes that are too slow to see.  An exponential decay is more like an optimistic case.

I don't think the drift measured for the simple burn in circuit is good to estimate the later drift at some 60 C. The circuit is different (more drifting components) and the temperature is very different, and thus quite some extrapolation.

[Andreas]

Hello,

since the statistical voltage dividers are in plastic packages I also examined influence of humidity to the voltage ratios.

Setup is as follows.
ADR1000A#01 in my cooling box. Environment is kept to 27 deg C on the heat spreader. (so a little more on the ADR PCB).
Humidity is changed either near 30% rH or up to near 80% rH.
Divider Ratio is measured with the ratio functionality of my HP34401A (100 NPLC).

Results:
The output ratio (nominal 1500 mV/V = 1R/2R built from 6 resistors) changes up to 0.3 ppm for a near 50% rH change within 2 weeks.
The setpoint ratio (nominal 12500 mV/V = 11.5R/1R built from 8 resistors) changes up to 2.2 ppm for the near 50% rH change.
The second ratio drift is larger maybe because the ratio needs several "single" resistors/dividers and so is not really "statistical".
A ratio with 12:1 with 6 resistors in series and 2 resistors in parallel would have the chance to be better in this regard.

And it seems that 2 weeks settling time is still not enough for final settling.

with best regards

Andreas

[Kleinstein]

A 2.2 ppm change in the temperature set point would not be so bad for the ADR1000, as the drift is attenuated quite a bit (e.g. like 200 times - depends on the unheated TC).
The 0.3 ppm drift of the output gain would be the larger effect. This number is still not so bad, as 50% change in RH is already quite some change.
The actual RH change at the resistors would be smaller, as a higher temperature would reduce the RH (about a factor 0.7 for 5 K higher). It still makes sense to use the environmental RH, as the a similar situation would be common in applications.

80% RH is quite humide and may reach the range where surface film may form, adding extra surface leakage currents. This effect may be highly nonlinear, with a material dependent critical RH for the film to form. This part would be relatively fast and not take weeks, more like minutes.

[Andreas]

Hello,

news from the burn in cycling of ADR1000A#02

After  ~500 hrs I actually increased the temperature cycling from 110-130 deg C to 115-135 deg C.
But I recognized that the overtemperature switch was actuating at the end of the temperature rising
so I finally adjusted the cycle to 114-134 deg C.
Expectation was that with ~5 deg C increase of the average temperature the ageing would increase by ~factor 1.4
Or somewhat lower (1.32) for ~4 deg C.

The first new measurement trendline (in blue beginning at sqrt(0.5 kHrs) = 0.7) shows something like 200 ppm/sqrt(kHr).
Which is around factor 3 above the 4 deg lower temperature (was 62 ppm/sqrt(kHr)).
So something is different above 130 deg C.

Unfortunately I do not have a good noise measurement "before" with the ADR#02.
Fortunately I remembered that Jason (cellularmitosis) has done a good job (thank you for that) in publishing several LTZ1000 PCBs.
So overcome this for ADR#03 I built a noise measurement cirquit from one of Jasons PX PCBs.

with best regards

Andreas

[miro123]

Hello Adreas,
Thanks for sharing. I have two questions related to your last two posts.
1. How do you control the humidity by resistor test? - I want to do similar test. My goal is to create resistor divider model. I want to find certain pattern between humidity, temperature and resistor divider ratio. I expect to determine Humidity behavior like - humidity coefficient. humidity time constant & hysteresis. I hope that  a. Observed behavior is repeatable and model can by build b. Expect simple polynom fits the model.
2. At your last post mentioned T 114...134C Is this environmental temperature or Die / Transistor base emitter temperature? I'm afraid that you are going far beyond the specs. Epoxy is just one component in the whole picture. e.g. How SiO2 react to such temperatures, and what about chip multilayered structure, are you casing permanent damage with it? That are all questions that i dont have answer

[Kleinstein]

The higher temperature speeds up the termally activated porcesses. However there is more change in getting a different equilibrium state.
For the epoxy part, relatively close to or above the glass temperature the structure can reach an equilirium, with a more open structure at a higher temperature. The more open structure also makes things faster, in addition to thermal activation.

With a different equilibrium state (especially for stress) the question is a bit if the aging at the high temperatre actually helps with the later performance at a considerably lower temperature. At least the higher temperature aging can not replace all the burn in time needed also at the final temperature.
There are still hysteresis effects, that than start a new settling / ageing.

[Andreas]

1. Control of the humidity is done with a 1 kg Dry pack (Silica gel) for the ~35 % rH.
    For the ~80% rH I used acryl polymere crystals like these:
    https://www.amazon.de/GERMANUS-Humidor-Befeuchter-Kristalle-AcrylPolymere/dp/B0796GG83B

    All is within my cooling box. (isolated and humidity tight).

    Essential is a constant environment temperature (here ~27 deg C)
    PWM controlled by heater foils on a aluminium sheet as heat spreader.


2. it is the temperature of the PT1000 sensor which is bound (with strong copper wire) to one of the heater resistors.
   (unfortunately I did no photo). But perhaps you can imagine the setup from the linked pics.

https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3788051/#msg3788051

https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3758498/#msg3758498

   Connection is via the heat spreader (aluminium bar) to the housing of the ADR.
   All is cotton cloth isolated within the thermo mug.
   So the measured temperature is nearly case temperature.
   With self heating of the ADR I think that the chip temperature is ~3 degrees warmer.

with best regards

Andreas

[RikV]

All this is very nice, but I would like to play along! Where and how can I get ADR100's? Are they reserved for VIP's? No sign of life on AD's site.

[RandallMcRee]

All this is very nice, but I would like to play along! Where and how can I get ADR100's? Are they reserved for VIP's? No sign of life on AD's site.

Somewhere in this thread (or was it a different one?) it is explained that you need to type "ADR1000AHZ" directly into the shopping cart search bar. When you do this it pops up with an order for 100, you can then update it to a lesser amount. Not sure, if this will be honored of course since I dare not check out at this time!

[Grandchuck]

Tried it.

Some info here:

https://xdevs.com/article/adr1000order/ 

I guess you need an account.

[RandallMcRee]

Tried it.

 

No, that is the normal search bar, not the shopping cart search bar. The shopping cart search bar is visible in my screenshot; its near the bottom of the page.

Yes, thanks for the link: it does describe the correct procedure. I believe the backorder time to be accurate....nothing until June.

Edit: it allowed me to place an order. Will update y'all in June. 
Edit again: Order confirmation says "In stock - est delivery Jan. 10". So yay!

[Grandchuck]

They really want to sell these?

[Andreas]

One could get such a setup with a bridge with 2 diodes and 2 resistors on opposite sides to replace the 2 diodes in series. The main parts is shown in the attachment.

Hello,

after playing with the suggestion of Kleinstein here the results:
- the T.C. trimming is now symmetrically to a center value near 25 deg C.
- I had to fine trim the suggested 22K resistors which where already a good estimate to 18K4 to achieve this.
- what I had not expected is that the sensitivity is about halved. So I had to reduce R10 from 1 Meg to 470K.

with best regards

Andreas

[Andreas]

Hello,

news from my ADR1000A#01 (which is built on one of branadics PCBs. See first post).

I now changed the Zero-TC-temperature from around 48 deg C to 55.4 deg. (using method a)
(refer to https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3810545/#msg3810545)
The plan was a zero TC temperature between 52-55 deg C for up to 40 deg C environment temperature.
But since I had no suitable resistors below 15 ppm/K for trimming I finally got 55.4 deg C.

Also the setpoint temperature was changed to the corresponding 524.7 mV at the heater sensing transistor. (got 524.5 mV actually)
With this modifications another T.C. test showed +166 ppb/K temperature slope for the zener voltage.
Zener voltage is now increased by ~2.4 mV.

The next step will be trimming the T.C. by around -178 ppb/K (+166ppb for the zener and half of the 10V output stage T.C. of +20-24 ppb/K).

with best regards

Andreas

[Andreas]

Hello,

ADR1000A#01 after T.C. adjustment:
I got near linear T.C. values

6.6V Zener voltage ~     -6 ppb/K
10V Output voltage ~ +20 ppb/K
the 10V output voltage is still too low (~9993 mV) and needs further adjustment (after some ageing drift measurements).

with best regards

Andreas

[Andreas]

Update on Burn-In of ADR#02 after 1kHr total.

compare with:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3903542/#msg3903542

just after last shown measurement the ageing changed the slope.

At about the same time the output voltage got a much larger hysteresis, which partly explains the larger stray of the drift values.

Slope is now around -44 ppm/sqrt(kHr) (from initial -62 ppm/kHr over -200 ppm/kHr)

next step is to go down again to 110-130 deg C cycling to see if ageing also here has decreased.

with best regards

Andreas

[branadic]

Thanks for keeping us updated Andreas.
I almost hit 4000 h on my sample, with the setup untouched for the whole time. Attached is an updated graph.

-branadic-

[Andreas]

Hello branadic,

it would be good if we get the exact receipe of burn in for the ADR1000 at Analog Devices.
Is it really the 1 week at  constant 150 deg C under bias?

I now could tweak my burn in cirquit for slightly higher temperatures as I just got a
different overtemperature switch with ~170 deg C.
https://www.conrad.de/de/p/thermorex-tk24-t01-mg01-oe170-s160-bimetallschalter-250-v-16-a-oeffnungstemperatur-5-c-170-c-schliess-temperatur-160-1678193.html

Btw. I have a 2nd temperature sensor measuring the PCB-temperature which is around 30 deg C lower than the heat spreader temperature.
So I can use "normal FR4" for the burn in PCB.

with best regards

Andreas

[branadic]

Is it really the 1 week at  constant 150 deg C under bias?

That's at least what Eric Modica said at MM2021.

Question: At which temperature did you make the burn-in?

I believe it's a 150 °C.

Question: So simply powering the zener and have the ambient at 150°C and then - what was it? - 168 h or something like that?

Yeah, you know and we haven't tested the bounce of that, we basically just copied the bake and burn process of the LTZ1000, which is a 168h both bake and burn. We have at least small sample size evidence that you maybe able to get away with just doing an unpowered bake and that's good enough to disipate the charge, but I have to say I haven't looked at it very closely.

Question: Simply elevated temperature or cycling during burn-in? Because, some rumor says cycling around glass transition temperature might improve the aging.

Our burn-in process doesn't utilize any cycling.

-branadic-

[Andreas]

I almost hit 4000 h on my sample, with the setup untouched for the whole time. Attached is an updated graph.

So to summarize.
you have around -4ppm in 4000 hours / -2 ppm in 1000 hours
so -2 ppm/sqrt(khr) without burn in and @55 deg C heater temperature (assuming you use zero T.C. temperature as set point).
Extrapolating for the next year (13 kHrs in total) we will have -7.2 ppm total drift or -3.2 ppm for the next year.

The data sheet value of 0.5ppm/year after the first 3000 hours with a setpoint temperature of 75 deg C seems to be somewhat optimistic.
There is no burn in mentioned in data sheet for the ageing drift.

When I take my -44 ppm/kHr now at 124 deg C average for ADR#02 and translate it to 55 deg C setpoint temperature.
(factor 2^7 = 128 when doubling all 10 deg C)  = -0.34 ppm/kHr at 55 deg C.
This gives still -1 ppm/year.

with best regards

Andreas

[branadic]

I guess you have to think different. The datasheet states 0.5 ppm/year after 3000 h, not sure if this is with or without burn in, but running the oven at 75°C. It excludes the first 3000 h of stabilizing, that's at least how I interprete things.
So going from here the drift of my sample was far less than 1ppm within the last 1000 hours and from here we need to observe how the drift goes on for the next 7760 h (~11 months). And yes, I'm running my oven at 55°C which is the zero t.c. point of this sample. This means, stabilizing will probably take a bit longer than 3000 h @ 75°C oven temperature, but lets assume it reached its stable point already. We will know more after the next 1000 h, I guess.

-branadic-

[Kleinstein]

The doubling every 10 K is only a rough approximation. There are some effects (especially slow ones) with a stronger temperature dependence.

Similar for the drift over time there are several possible functions: exponential for simple relaxation processes and square root time for more random walk type processes (or superposition of many uncorrelated random effects) are just 2 examples. Another possible form would be a stretched exponential. So extrapolation is somewhat tricky.

For the high temperature part, I see mainly 3 effects to happen: fully finish the chemicial curing of die attach and possible stabilization (e.g. build up of a surface film, surface "reconstruction" ) on the chip. There may be a tiny bit of diffusion at the contacts, but hardly in the silicon itself.
The stress between the die and case should relax as the die attach would be relatively soft.
A later step would than include relaxation of thermal stress from the different thermal expansion of the silicon and the metal case. This would be near the final use temerature as the stress only comes back on cooling. The die attach can also have some structural relaxation when well below a glass temperature. The 150 C are more like too high for this, more like near the glass temperature where the structure is more like in a local equilibrium.

If I anderstand right the data for the high temperature anneal are for the simplified circuit, and this may show additional drift from the resistors and the FET.

The exponential fit to the data from brandic should give around 0.5 ppm/kh (5 µV for 1000 h) after 3000-4000 hours. Part of the drift could also be from the 6.6 to 10 V stage.  So this would not be so far off from the DS numbers.

[Neo2199]

Hi all, I am planning on building my first voltage reference. I have hard time getting standard LTZ1000 but I managed to order one ADR1000 from Analog Devices site which should perform similarly. It is quite hard to figure out good resistor values and set the operating point right where good long term stability is. But based on the info in this and other similar threads I think I have puzzled everything together. My target is 60C, 5mA zener current at 100uA collector current. I came up with R4 12k, R5 1k, R2 61.9k, R3 61.9k, R1 100ohm. Did I get this right, or should I change something?
I am also planning on designing my PCB, but I cannot find too much information about layout recommendation. It is interesting to me to see so many differences in layout of KX LTZ1000, reference Analog Design board and HP3458A LTZ1000 board for example.

[Andreas]

I have hard time getting standard LTZ1000 but I managed to order one ADR1000 from Analog Devices site which should perform similarly.

I came up with R4 12k, R5 1k, R2 61.9k, R3 61.9k, R1 100ohm. Did I get this right, or should I change something?

I am also planning on designing my PCB, but I cannot find too much information about layout recommendation.

Hello,

Each single device behaves different regarding ageing.
So you might want to have at least 3-4 stable references to sort out those with the largest ageing.
Whether ADR1000 behaves similar to the LTZ1000 still has to be proven in practice.
At least it seems that the ADR1000 needs longer time to settle than a typical LTZ1000.

The resistor values are those recommended from the data sheet so it should be ok to go with them.
Others try to set the temperature setpoint to the zero T.C. temperature of the ADR1000 to reduce influence from set poin divider resistor drift.
(this is not possible with LTZ1000 due to the higher zener voltage).

For layout it is essential to do star point wiring and  especially kelvin sensing of the zener voltage.
(no ground currents from heater or OP-Amps across the ground sensing line).
When you have to integrate the zener in a device like 3458A it might be partially easier to work with ground current compensation.
So the constraints (dual/single supply) and pre-ageing and selecting on device / module level may give different board layouts.

with best regards

Andreas

[Andreas]

Update on Burn-In of ADR#02 after ~1.2 kHr total.

compare with:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3942472/#msg3942472

After last reporting I reverted back to 110-130 deg C cycling hoping that the ageing slope would flatten.
But it increased instead from -44 ppm/sqrt(kHr) to -84 ppm/sqrt(khr) calculated for 120 deg C.

I wanted further reduce the temperature cycling to 100-120 deg and further steps down to see wether there is some hysteresis.
But unfortunately the controlling PC stopped working so I have to stop the bun in heating for a while.

with best regards

Andreas

[branadic]

Others try to set the temperature setpoint to the zero T.C. temperature of the ADR1000 to reduce influence from set poin divider resistor drift.
(this is not possible with LTZ1000 due to the higher zener voltage).

Why you think this is not possible for the LTZ1000? I did it here.

-branadic-

[Kleinstein]

The extra resistor in sereis with the zener diode mainliy shifts the problem to different resistors. One gets a reduced temperature effect, but quite some effect of the new resistor and also the 100-120 ohms resistor to set the current gets more important.
It can still help, as it reduces external effects on the oven part.

With the ADR1000 the zero TC point is much more accessible, but chances are the exact point depends on individual units. Here it is accessible by adjusting the zener current, set temperature and transistor currents with reasonable values.
The LTZ1000 would need the series resistor or very low currents for the Zener and/or transistors.

There is no need to get the unheated TC very low, but it helps to start in the +-5 ppm/K range instead of the 50 ppm/K like with the LTZ1000.

I don't think the ADR1000 circuit would need the 400 K (R9) resistor for compensation of residual TC (with heaster).

[branadic]

I for myself prefer real experience than guesswork for LTZ but also ADR.

@ Andreas
I can only encourage you to go on. Attached is the drift of my ADR reference, now after half a year and without any burn-in or pre-aging prior to the build and lower oven temperature (55 °C) compared to the datasheet (75 °C), but still at the reference's zero t.c. point. As you can extract from the diagram the reference has really stabilzed, which took a little longer than 3000 h, more like 3500 h, but again I'm running it at a lower oven temperature.

-branadic-

[Andreas]

After having adjusted the T.C. on my ADR#01 (which has no burn-in),
I started ageing drift measurement against 3 instruments (HP34401A, K2000 and ADC27).
From time to time I also do a comparison of the instruments against my most stable reference LTZ#4.

Unfortunately the K2000 shows nearly a +1 ppm Jump between day 8-10 of the measurement due to popcorn noise.
(this jump shows also up on the LTZ#4 measurement).

And also ADR#1 had a -0.6 ppm jump on day 5.
After 500 hours we have a ~-1.5 ppm drift compared to ADC27 and HP34401A.

with best regards

Andreas

[Neo2199]

So I have started with my schematic for my ADR1000 reference. I am wondering if my design will work as expected. Any recommendations?
For the Vishay precision resistors I went with what is available on digikey at the moment.
The opamps are temporary, but I am thinking of using ADA4522.

[Dr. Frank]

Sorry to say so, but these are too many bugs, and an unreadable schematic.

Why don't you simply use any of the working schematics/designs/layouts from the LTZ1000 thread?

Frank

[Neo2199]

Hmm, that's not great. I can tune schematic for readability, it is my first draft, but I won't learn much if I just reuse someone else schematic. Point of this project for me is to learn how to do good/decent design.
It is also true I was basing my design on xDevs design coupled with Marco Reps LTZmu rev 0.2 WIP schematic.

[Kleinstein]

The AZ OPs likely do more harm than good with a LTZ1000 / ADR1000 refrence due to the current spikes and possibly also the startup behaviour. There is nothing wrong with the LT1013.  The ADA4084 suggested in the ADR1000 datasheet may be a bit questionable.

Otherwise the circuit looks like the standard circuit (except for C2 that may be a bit on the high side).

There is no need to have the exact values for the 62 K collector resistors - some 50 K should be OK as well. Using 2 high value resistors in parallel is not such a good idea, as the higher values resistors tend to drift a little more.

The ADR1000 is even less sensitive to the resistor quality than the LTZ1000. So there is no real need to use highest grade resistors at all places.

For the ouput amplifier one should have some fltering action between the reference and OP if an AZ OP is used.  The resistor series strings with multiple 0 Ohms may not be a good idea here, as more junctions can add thermal EMF. Here a parallel trim for fine adjustment may be the better choice. For a 2:1 resistor ratio, I would consider a resistor network (e.g. LT5400 or ORN ), at least as on option.

[Neo2199]

Thank you for the advice. All makes sense to me. I will work on updates.
This design is very hard mainly because of all the parasitics that will make all the difference.

[Andreas]

Update on Burn-In of ADR#02 after ~1.5 kHr total.

compare with:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3976202/#msg3976202

After last reporting I reduced cycling to 100-120 deg C to see what happens.
After a interuption due to measurement PC failure the ageing inverted from
-84 ppm/sqrt(khr) to +63 ppm/sqrt(khr)

now I think that the "drift" measurements during burn-in with my simplified burin-in cirquit have nothing to do with a real ageing drift.
Perhaps I measured something like a introduced hysteresis?

I am now cycling down in 10 deg C steps to remove hysteresis
and will see on a proper cirquit if ageing drift / noise is lower than ADR#01.

with best regards

Andreas

[ap]

If I read your numbers and chart correct (and I have not studied each and every page of the whole thread thourougly, so correct me if I am wrong please), your graph shows a drift of 80ppm downwards within 1khr. This is supposedly after the first upward drift within the first couple hundred hours as shown in figure 4 of the ADR datasheet. So direction seems ok. To do a very rough and ballpark figure sanity check, at an average assumed say 125°C and assuming an Arrhenius Q10 factor of 2 (and there may of course be other drift contributors as well such as the circuitry resistor drifts, die attach...; so this is really a very rough check), and related to the 75°C Tset of the data sheet, that would translate to a drift rate at 75C of about 2,5ppm/khr. Assuming Q10 of 2,5, it would translate to below 1ppm. So the initial drift you have seen so far may actually be correct and not related to "hysteresis" as you state.
The drift direction change is a different story, it will be interesting to see how this develops and what your cycling results are.

[Andreas]

Hello,

it is slightly > 1200 hrs for the first -80 ppm drift at high temperatures. (note the x-axis is sqrt(kHrs) so a value 1.1 translates to 1200 hrs).
I am now in the downcycling phase (every 2nd day 10 deg C less) so currently a slope measurement does not make sense.
But up to now the 2nd day measurement is always higher than the 1st day after temperature change. (so still upwards drift).

I plan to set the zero-TC and temperature setpoint to ~55 deg C like on sample #1.
And then we will see (after a few khrs) in comparison what the burn in has done a pre-ageing.

with best regards

Andreas

[CurtisSeizert]

I have built up a couple boards that are similar to the HPM7177 schematic (the main difference from the datasheet schematic is QHTR is on the low side and the U2 feedback network is different, also there is a 20R resistor on the zener cathode) and have done some experimenting. I am trying to get a handle on the best strategy for tempco compensation, and I am hoping someone can correct me if I am mistaken.  Referring to the ADR1000 datasheet schematic component references:

-Adding resistance between the zener cathode and the zener+ sense output in the decreases the tempco by weighting the Vbe tempco of Q1 by a factor of (R1 + R_cathode)/(R1)
-Increasing the collector current of Q1 (by reducing the value of R2) increases the tempco by increasing Vbe and decreasing the absolute value of its tempco (see p.5 of http://web.mit.edu/klund/www/Dphysics.pdf for Vbe tempco dependence on Vbe)

I've attached a noise spectrum for one reference that was burned in at 130 for 1 week using the internal heater.  Zener current is 5.5 mA (90.9R for R1), 2.5 mA Ic for Q from 2.5k R2 (!). R_cathode is 20R.  Tempco ended up being around +12ppm/K for the heater setpoint.  I overshot tempco by dropping R2 because it crossed zero between resistor values that I had; it would have been easier with lower R_cathode, but that would have been a PITA to mess with given the design, so I didn't.  Gain stage is ADA4523-1 with feedback from two SMNZ 2k/1k arrays with trim.

Curtis

[branadic]

Here is an update of my ADR1000 after now 5000 h.

-branadic-

[Andreas]

Update on ageing of ADR#01 after T.C. adjustment and with no burn in.

compare:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3995003/#msg3995003

Again ageing drift measurement against 3 instruments (HP34401A, K2000 and ADC27). Now at 1 kHrs
Unfortunately the K2000 shows a lot of popcorn noise so it is the most unstable reading here.

As comparison also from time to time readings on my most stable LTZ#4 reference. (-1 .. -2 ppm/year)
At the end of day 42 it shows that K2000 reading is about 1.7 ppm too high and 34401A reading about 0.5 ppm too high whereas ADC27 is about right.

So the total drift of ADR#01 is -2.2ppm/kHr for the first kHr on the 6.6 V output.

with best regards

Andreas

[Andreas]

Hello,

first measurements on burn in of ADR1000A#03

I decided to do a cycling of 110-130 deg C on this device for minimum 2 weeks.
Up to now (250 hrs) or sqrt(250hrs/1kHr) = 0.5 the average drift is  around -27 ppm/sqrt(kHr).

with best regards

Andreas

[Andreas]

Update on ADR1000A#02

Downcycling finished with no surprise: every 2nd day the output value was still higher as on the 1st day with a -10 deg C step.

I now have set the zero-T.C. of the zener to ~55 deg C to have enough headroom up to 40 deg C environment.
The heater setpoint is also set to ~55 deg C.

Now I have done the first T.C. measurement vs environment temperature with a temperature sensor attached to the top of a metal case.
Average T.C. of the 6.6V buffered output is -30 ppb/K unadjusted on my LTZ1047B PCB.
This is much better than on ADR#01 (-166 ppb) with a different PCB.

On ADR#02 I plan to adjust the T.C. with the (nominal 400K) resistor R9. First guess is 1.5 Meg to achieve ~30ppb/K increase.

with best regards

Andreas

[branadic]

Andreas, read the hints I provided by mail on how to compensate for the remaining t.c., there is no need for this dubious 400 k resistor.

-branadic-

[branadic]

Andreas, I'm a bit confused, here you wrote

6.6V Zener voltage ~     -6 ppb/K
10V Output voltage ~ +20 ppb/K

Now you say:

This is much better than on ADR#01 (-166 ppb) with a different PCB.

So what result is true for your ADR1000A#01? Can you clearify on that?

-branadic-

[Andreas]

Hello branadic,

There are 3 steps in adjustment of the reference:

Step 1: Adjust the zero T.C. of the Ref-Amp of zener and transistor (to 55 deg C in my case). By either changeing current through zener or through collector.
Step 2: Adjust the heater temperature setpoint to the zero T.C. temperature. By influencing the R4/R5 ratio.
Step 3: Adjust the final T.C. of the whole reference cirquit against environment temperature. By either R9 or a copper resistor in series to R1 or shortening the legs.

With shortening the legs I am still unshure. On the A-Version of the LTZ I always got the "right" direction (increasing T.C. by shortening the legs.
 But on the non-A Version of LTZ it went the wrong direction.
The ADR seems to be somewhere in between (at least from thermal resistance).

On ADR#2 I am now after step 2.
So you will have to compare it to the Step 2 results of ADR#1 (and not with the final Step 3 results).

So please compare:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3921830/#msg3921830
with:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4050361/#msg4050361

now you can speculate about the reason.
Accidently (only 2 different samples).
Or result of the different amplification in the heater control cirquit. (one cirquit with transistor used as diode, so transistor amplification is missing).

with best regards

Andreas

[Kleinstein]

Even without the final TC trim in step 3 the TC is quite small. The 30 ppb/K translate to 0.2 µV/K. For the voltage loop OP this would correspond some 40 µV/K at it's input. The extra gain in the voltage loop should have little effect, as the load to the OP/transistor is essentially constant.
There are many other details that can be different between the 2 ref. versions: resistors TC, details in thermal layout, amount of solder at joints, quality of the adjustments and just random variations in the ref chips.

The residual dift after step 2 looks rather linear and they have different sign. The trim with R9 only works in one direction and it is proportional to the heater current and thus the square root of the heater power - so a nonlinear function, especially in the lower power range that is likely more relevent.

[Andreas]

there is no need for this dubious 400 k resistor.

I fear Kleinstein is right. R9 is useful to compensate upwards.
The copper resistor to compensate downwards.
With 30 ppb/K the resulting non-linearity through R9 is low enough.
But with up to 15 deg C temperature difference in my "lab"
between winter and summer I need to have the T.C. as low as possible.

with best regards

Andreas

[branadic]

Andreas, it just depends were you put the copper resistor in

-branadic-

[Andreas]

Hello,

R9 is already planned on my LTZ1047B PCB design.
So more easily populated instead of scratching traces.

First try with 1.5 Meg showed large overcompensation of the negative T.C. So my formula for LT1000A with 1 Meg compensating 40 ppb does not work here.
I would have tried 3.0 Meg but I had only 3.3 Meg in the drawer.
With 3.3 Meg I get -4.5 ppb/K average resulting T.C.
The large hysteresis is due to the fact that the environment temperature sensing NTC is on the top of the aluminium housing so getting the temperature with some time lag.

And a view to the PCB with the 3 trimming resistors across the low T.C. resistors. (R9 is on the right side of the ADR1000A).

with best regards

Andreas

[Kleinstein]

The effectiveness of the R9 compensation depends on the thermal setup.  No cap on the ref chip means more heater power needed and thus a stronger effect of R9. The LTZ  A version needs less heater power and should thus have less effect of R9.

The curve with the 3.3 M resistor does not look so bad: It looks like the slope is larger with low temperature and the range below some 15 C is likely of little interest in later use.

From the picture it looks like there is quite some air space obove the circuit inside the shielding box. This may lead to turbulent convection and resulting low frequency noise. Some filler at the to lid may help.

[Andreas]

Hello Kleinstein,

for the photo I have removed all the thermal isolation stuff.
Of course the ADR is thermally isolated on both sides of the PCB.
The metal housing is only to equalize the environment temperature a little bit.
And of course the measurements are all made with the final thermal isolation installed.

with best regards

Andreas

[Andreas]

Hello,

further measurements on burn in of ADR1000A#03 now at 500 hrs (sqrt(0.5kHrs))

Cycling is unchanged 110-130 deg C
the average drift increased from  around -27 ppm/sqrt(kHr) to -38 ppm/sqrt(kHr)

with best regards

Andreas

[chuckb]

Rev B of ADR Data Sheet on the ADR1000 page at ADI. Has there been a product page before?

[Andreas]

Hello,

update on burn in of ADR1000A#03 now at 1000 hrs (sqrt(1 kHrs))

Cycling is unchanged 110-130 deg C

the average drift from 500-1000 hrs is now -15.5 ppm/sqrt(khr) in blue
against -38 ppm/sqrt(kHr) for the first 500 hrs in green.

so I think the minimum duration for burn in around 120 deg C is ~500 hrs.

with best regards

Andreas

[branadic]

I've passed the 6000 h mark. Some 2700 h ago I've also added the F7000T to the rear terminals of R6581D, this measurement starts directly after the measurement of ADR1000 connected to the front terminals of R6581D.
Unfortunately, I've messed with the setup about 500 h ago, hence why there is a jump of 1.5 µV visible.

-branadic-

[dietert1]

In a previous post you mentioned an exponential relaxation fit to your log.
Can you give an estimate of the time constant? At which temperature are you running the ADR1000?

Regards, Dieter

[branadic]

At some point in time the model stopped fitting the measurement data and I haven't yet played with other models to get a working prediction model again.
The reference is operating at ~55 °C oven temperature, see also the first post of this thread.

-branadic-

[dietert1]

Yes, your measurements seem to exhibit a linear drift recently.
I think your first mention of 55 °C is on page ten of this thread. I missed that. At which temperature are you running your 7 to 10 V amplifier?

[branadic]

You are right, I mentioned it later directly, but had expected that it is clear where I supposed to operate the oven given the t.c. diagram with the zero t.c. visible at ~55°C.

At which temperature are you running your 7 to 10 V amplifier?

Not sure I got this question correct, but the amplifier is at ambient temperature, no additional heating there.

-branadic-

[Andreas]

Hello,

Update on ageing drift on my ADR#01 + ADR#02 after 40 days (~1 kHr) after adjustments.
ADR#01 first firing was in September last year. Mostly all measurements where done at ~50 deg C.
ADR#02 has ~2kHrs of burn in beginning end of November before final adjustments.

Now I am doing weekly measurements relative to my most stable LTZ#4 reference.
#1 drifted ~1.8 ppm/kHr since adjustments
#2 drifted ~1.2 ppm/kHr since burn in + adjustments.
my newest LTZ#9 (non-A) running since 10/2019 is given for comparison

So #2 seems to have a slight advantage in stabilisation. (but this are only 2 different samples so no significant statistics).
But when I compare that with my AD587 references the ageing drift of the ADR is still too high.

https://www.eevblog.com/forum/metrology/ad587lw-10v-precision-travel-standard/msg4147111/#msg4147111

with best regards

Andreas

[Andreas]

Hello,

update on burn in of ADR1000A#03 now at 2000 hrs (sqrt(2 kHrs))

Cycling is unchanged 110-130 deg C

After trendline changed to -15.5 ppm/sqrt(khr) from 500-1000 hrs (in blue)
the ageing reverted back to the initial ageing rate of -38 ppm/sqrt(kHr) from 1000-2000 hrs.

I will now check wether noise has changed (after downcycling to 50 deg C average).

with best regards

Andreas

[branadic]

I've passed the 7000 h mark, so here's an update. I've rearranged the setup a while ago, removed one of the 2182A and added 34420A instead. The remaining 2182A still monitors the difference between F7000-1 and F7000-2 (green curve), while the 34420A now monitors the difference between ADR1000 and F7000-1 (red curve) and ADR1000 and F70002 (black curve). The magenta curve is the calculated and thus expected difference between ADR1000 and F7000-2. Black (measured data) and magenta (calculated data) curves do match quite well.
While the first plot is a measurement with R6581D monitoring the ADR1000 directly, in the second plot (second measurement each day) the R6581D measures the F7000T (averaged output of the 2x F7000). In plot 2 you can see me messing with the setup at ~2200 h leading to a small jump downwards, that recovers at ~2850 h when I added the 34420A to the setup.
Measurement ongoing.

-branadic-

[Andreas]

Hello,

so I still can hope that my ADR1000A will stabilize in some 3-4 kHrs.
At the moment the drift is still  ~1 ppm/month.

with best regards

Andreas

[sahko123]

God I cant wait till I get an ADR1001 put it in a cute little box with a UPS system and seal it away for 50 years

[Andreas]

Hello,

1/f noise comparison before and after burn on ADR1000#03 in shows no significant change.

on 15.02.2022 the noise was
618 nVpp average with Std dev 62 nV over 19 measurements of 100 sec
and 73 nV RMS average with Std dev 3 nV

today on 26.05.2022 I measured
617 nVpp average with Std dev 77 nV 55 nV over 19 measurements of 100 sec
and 55 nV 77 nV RMS average with Std dev 3 nV

typical noise examples before/after burn in attached.

with best regards

Andreas

edit: corrected interchanged std dev / rms value on 26.05.2022

[Kleinstein]

The noise curves do look a little different. The RMS noise of 73 and 55 nV is also aready a bit different.
Somewhat looks like the more higher frequency noise could have gone down, maybe the longer lived states causing popcorn like noise got more stable (longer lifetime). If data are available it may be worth looking at the FFT / noise spectrum.

[Andreas]

The noise curves do look a little different. The RMS noise of 73 and 55 nV is also aready a bit different.

Sorry copy and paste error (RMS noise + std dev interchanged).
Corrected the values above.

In both measurement series I find some "popcorn events (jumps)" of up to 0.3 uV.
So I would not interpret from a single picture (picked to meet about the average peak/peak noise) out of a series too much.

with best regards

Andreas

[Andreas]

Hello,

Noise measurements (statistical) can look rather different.
(you have to be at the right place at the right time).

Just some examples: the "best" and the "worst" noise from one and the same measuerment series.
One measurement series before and one after burn in.

The worst case usually also contains one or more edges of some Popcorn events. (which is filtered by the 0.1 Hz high pass filter).
The popcorn events captured here are not necessarily the worst case popcorn events.

With best regards

Andreas

[branadic]

Thanks Andreas, why are you convienced that what you are looking at is popcorn and not some weird emf event nearby? Maybe you can describe your setup a lil bit more detailed and show some pictures?
I guess you have the reference battery powered in a cookie box together with your low noise amplifier, the Picoscope and your convertable disconnected from mains line too? Every potential source of noise/hum is either turned off, banned from your lab or far far away?

-branadic-

[kleiner Rainer]

I suspect that we have reached levels of sensitivity in our measuring equipment that we have to take events from background radiation and cosmic rays into account. Andreas, do you live in an area with granite rock, think "radon"?

The Wikipedia article about cosmic rays states:
"Cosmic rays have sufficient energy to alter the states of circuit components in electronic integrated circuits, causing transient errors to occur (such as corrupted data in electronic memory devices or incorrect performance of CPUs) often referred to as "soft errors". This has been a problem in electronics at extremely high-altitude, such as in satellites, but with transistors becoming smaller and smaller, this is becoming an increasing concern in ground-level electronics as well.[89] Studies by IBM in the 1990s suggest that computers typically experience about one cosmic-ray-induced error per 256 megabytes of RAM per month.[90] To alleviate this problem, the Intel Corporation has proposed a cosmic ray detector that could be integrated into future high-density microprocessors, allowing the processor to repeat the last command following a cosmic-ray event.[91] ECC memory is used to protect data against data corruption caused by cosmic rays."

If a cosmic ray event can flip a bit in dynamic memory - what could it do to the output signal of a precision reference?

Food for thought.

Greetings,

Rainer

[Dr. Frank]

These spikes extend over longer period of time, instead of a few points only.
Therefore these are highly probable dips coming from the reference, not e.m.f.
2nd argument, other reference does not show such dips. I have observed similar effects on one of my 5 ADR1000 only, scroll upwards, please. These dips did not occur constantly, but there were quiet and noisy phases.
Cosmic rays should not play a role on bipolar, big area semiconductors as such

[branadic]

Let me remind you of Eric's talk at MM2021 and his explaination of surface charges, so it isn't necessarily popcorn.

Also coupled interference can cause all sort of weird effects. For example I saw weird jumps that appeared to be popcorn noise on my ADR1399. They were invisible to me when I used a 7.5 digit meter because of it's larger noise, but came visible when I switched to an 8.5 digit meter. Both meters where running on the same integration time of 20 s though. It turned out that was some sort of common mode noise from my lab power supply powering the ADR when I switched to my ULNLPS (shielded transformer + LT3045) and these jumps immediately vanished. That effect stressed my head for the last days before I finally found the culprit.

So before drawing any conclusion it's good advice to check that this is not a human induced effect or error, like in 99.9% of the cases.

-branadic-

[Andreas]

Thanks Andreas, why are you convienced that what you are looking at is popcorn and not some weird emf event nearby?

Hello,

of course I am using a metal cookies box connected to the ground of the oscilloscope.
The reference itself and the 1/f amplifier are battery powered.
The scope is powered by a laptop which itself is normally mains powered.
(I have checked that this has only minor effect on the output).

The whole setup is placed in a "quiet" corner of my lab.
Hum is checked by FFT before start of the measurement.
The noise measurements are usually done "over night" where I have less influence from EMI sources.

The question is: how does EMI noise (usually influencing as common mode noise over small parasitic capacitances) look like in comparison to a popcorn noise which is a "rectangular jump in > 1 sec range" (in differential mode) at the output of the 0.1-10 Hz 1/f amplifier.

The simulation shows:
Popcorn noise shows up at each edge as "sawtooth" pulse with a time constant of ~1 sec at the output.

Measurement shows:
EMI-noise (e.g. fluorescent lamp switching) shows up as "short pulse" which is much less in duration in the us or ms range.

and also some weird noise (EMI-pulses from a accidently not disconnected USB-connector) showed up as short pulses with 16 sec regularly time interval.

with best regards

Andreas

[Andreas]

For example I saw weird jumps that appeared to be popcorn noise on my ADR1399.

Is there any legend which colour is what?
I see Jumps on the blue line but no real difference between the upper and the lower chart.

with best regards

Andreas

[branadic]

Blue line is the ADR1399 voltage the rest are temperature sensors nearby. The difference is in the scale

-branadic-

[Andreas]

Update on ageing drift of my ADR#01 + #02 now ~2 kHrs after adjustment/burn in.

compare also:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4189336/#msg4189336

Average drift is now ~1.2 ppm/kHr so -2.4 ppm for the last 2 kHrs.

I have added temperature because my lab temperature is now increasing.
(it is internal temperature of my AD587LW references which are nearby the ADR references so ~2-3 deg C above room temperature).

with best regards

Andreas

[branadic]

I passed 8000 h, hence an update on the diagrams. There is a jump at about 7800 h, which is me pulling my Datron 4000A calibrator from the rack for repair and which the R6581D obviously didn't like 
From the results measured I was able to create the plots Drift2 and Drift4, which represents what the R6581D would have measured for each reference. Common mode effects on all curves would then represent the LTZ in the R6581D itself, while the rest would belong to the individual reference.

-branadic-

[Andreas]

Hello,

again update on ageing drift of my ADR#01 + #02 now ~3 kHrs after adjustment/burn in.

On day 98 I added a buffer on the 6.6V previously unbuffered output on ADR#01 added some EMI coils + capacitors and put all into a aluminium enclosure. The additional measurement shows that there is no change on output voltage except measurement stray.

The ageing measurements suffer from high lab temperatures of ~30 deg C leading to nearby measured temperatures ~33 deg C of NTC.
So I cannot really tell wether the ageing rate now slows down to ~1ppm/kHr or if it is only due to high lab temperatures.

with best regards

Andreas

[branadic]

Today, one year ago I've started the measurement of my ADR1000 reference. Here are the results I monitored so far.
New in these series of plots is image Drift3.png, where I've calculated what the R6581 would have measured them, but with the influence of the meter removed (image Drift2.png with common mode effect removed). Note that I've pulled my calibrator from the rack at ~7700h (~4400h in Drift4.png), what influenced the readings of the meter.

Is the reference stable now? Not sure about that, could be that we see this years hot summer as an influence on the ADR, with an overall rise in the lab temperature from 6000h on, while the temperature spikes not being visible. It needs further observation to see how that evolves.

Also interesting to see the difference between both F7000, which is on average 3 µV/year.

-branadic-

[iMo]

..added some EMI coils + capacitors and put all into a aluminium enclosure. The additional measurement shows that there is no change on output voltage except measurement stray..

The most dangerous EMI freq spectra for your measurement is from 50MHz-5GHz today (unless you live close to a powerful long or shortwave transmitter), imho. That filter (on your picture) based on a foil capacitor (good for quite low freq only) and a multi-turn toroid (with rather huge inter-winding capacitance) will work till a couple of MHz effectively.. Better you put 1n-10n ceramics at the binding posts (and at the other side of the choke too) and as the choke a simple 1-3 turn(s) via a small diameter ferrite tube/bead (like the material 52 or 61 for example).

[Andreas]

Hello,

of course you are right in general. Especially with the capacitor value. Best would be a 1 nF feed through capacitor instead of the binding posts.

In my lab WIFI and cell phones are banned.
So most of my EMI are USB cables (@full speed) which are needed to log the voltmeters.

The TN10 ferrites (4A11 material = NiZn) are relative wideband up to some 100 MHz.
https://www.eevblog.com/forum/metrology/emi-measurements-of-a-volt-nut/msg3347342/#msg3347342

with best regards

Andreas

[Andreas]

Hello,

again update on ageing drift of my ADR#01 + #02 now ~4 kHrs after adjustment/burn in (up to 2 kHrs).
Setup uses, like before, a ratio measurement against LTZ#4 (my most stable reference) and LTZ#9 as sanity check.

Still having nearly -1ppm/kHr on the zener voltages.

Only the 10V output on ADR#01 seems to slow down.
But I fear that is some humidity influence (seasonal change) of the statistical output divider.

with best regards

Andreas

[Dr. Frank]

Hello,
my ADR1000 #3, currently 6.668563V, is now running for 1 year.
Its oven is set to 51.4°C, and it has got no initial conditioning / heat treatment.

In the diagram, you see its absolute drift, i.e. relative to the median of my group of 8 references.



Over the years, I determined the average drift of this reference group so I'm able to make absolute drift measurements to  < 0.5ppm uncertainty. This is confirmed by the recent 'baseline' measurement of -0.17ppm in June 22, which is derived from the ring comparison organized by branadic, where two members had calibrated uncertainties of < 0.5ppm each, and I, No. 7, was measuring the travelling LM399 right in the middle between both:

https://www.eevblog.com/forum/metrology/eu-calclub/msg4385281/#msg4385281

Therefore, my ADR1000 shows -4ppm/6months when constantly powered between November 2021 and May 2022.
I didn't power it in October 2021, and in June 2022, to check for hysteresis effects.
This turned out to be +2.8ppm in Oct. 21, and up to +2.2ppm in June 22.
On latter tests, I switched it on 12h prior to taking measurements, but in the end did not investigate on how long it takes to reach the previous "warm" value.

I add a comparative drift measurement on two pre-conditioned LTZ1000A, which show -1ppm/6mo. at most, and seem now to stabilize to much smaller drift rates for the last 7 months.



The FLUKE 7000 shows -1.4ppm/year in 2021, which is quite mediocre. So it is used as a stable 10V transfer standard further on, as I changed its output value twice for my experiments on the 752A and 720A dividers.

So my conclusion is, that the ADR1000 at current is not suitable for serving as a volt-nuts reference like the LTZ1000, due to its high annual drift and big hysteresis. A working burn-in procedure would be required. In the datasheet there's a hint that after 3000h @ 75°C the drift will be reduced to 0.5ppm/Yr., but I don't know if anybody has tested this.

Frank
     

[iMo]

You got the early samples.. Would be better to wait till they fine-tune their production process.. 

[MiDi]

Asking me, how did ADI get to their DS value of 0.5ppm/year, which is at least 10x less than what we see?
For me "typical" means at least 2/3 of significant samples from production line satisfy DS value.
ADI following VPG way with MBS or are all those we see engineering samples with worse specs?

You got the early samples.. Would be better to wait till they fine-tune their production process.. 

Guess or do you have secret information?

[iMo]

In Sept2021 within the Germany remote workshop ADI designers described the pretty recent history of ADR1000 R&D.
While doing such an exercise you certainly would produce XX thousands of chips (early ones). Some quantities get packaged, some stay as dies. There are many subcontractors/customers who evaluate(d) them, not only our voltnuts from eevblog are involved. Thus I can imagine myself there are thousands of ADR1000s floating around I would call "an early sample".. That is with all chips, imho, but for example nobody will spend a year with daily elaborating an LM358 (as you do with the ADR1xxx) whether it fits the strict DS values perfectly, or not..

[dietert1]

One plausible speculation is that early samples are missing some vendor side burn-in and selection. Later they will be able to offer parts including burn-in and selection. Probably the datasheet drift specs were assumptions based on the LTZ1000 datasheet

Regards, Dieter

[iMo]

ADI started to think about the LTZ1000 replacement in 2016. They started R&D in 2016 from scratch. First chips in packages are dated Q34 2018, afaik. As the experts here say you need years to evaluate the parameters of such fine technology all chips you got so far are those I would call "early samples".. 

[branadic]

As far as I've heard, early engineering samples such as date code 1727 haven't seen the bake and burn (168 h @ 150 °C), while all later date codes (1839, 2108) are production samples.

-branadic-

[iMo]

Make a small "heating hat" where you can regulate the temperature and put on it on its case, thus you can burn-in it while it is soldered already in your board..

[iMo]

Make a small "heating hat" where you can regulate the temperature and put on it on its case, thus you can burn-in it while it is soldered already in your board..

Like a micro oven inside another micro oven (414 resistor size for comparison)

I thought a small hat which fits just on the top of the ADR1k package - that is for burn-in only - such you can heat the body of the ADR1k up to those 120-150C, while the ADR1k is already soldered into your board (you do not want to cook your entire pcb with all other components on it in an oven, or do you?

[cellularmitosis]

Can you just use the built-in heater for burn-in?  Temporarily use a pull-down on the temp-control resistor divider?

[iMo]

Can you just use the built-in heater for burn-in?  Temporarily use a pull-down on the temp-control resistor divider?

You have to calculate the current needed. The thermal resistance of the package is 216 C/W (soldered in two layer pcb, DS), thus you would need aprox 49mA through the 240ohm ADR1k heater to get the chip at 150C with 25C ambient (double-check my math plz).
PS: the "annealing recipe" above expects stepping the temperature up and down at various temperatures, thus it seems to me to be easier to have the burn-in hat with easy temp reg put on your head. 

[Kleinstein]

One could use the internal heat for a burn in, at least for the lower temperatures. Depending on ther thermals around the chip the temperature is limited, especcily if there is only a 15 V or so supply.
There is howerver still the question on which temperature profile and maybe zener current would be correct for the burn in. There are different processes going on and hotter is not always better. The glass temperature of epoxies is just within the possible range. Another point is thermal stress that can build up when keeping the ref at a higher temperature for a long time.

[iMo]

You would need 11.8V on the internal 240ohm heater to get 580mW, that is aprox 150C on chip provided the case is 216C/W thermal resistance and ambient is 25C.
PS: the internal heater resistance is specified around 240ohm at 1mA heater current in the DS. It could be pretty different at 49mA (and such high temp), however..

[manganin]

Years ago I built an aging rig for the AD587 references. The total power comsumption of 100 chips was so high, that no separate heater was needed. The temperature control loop simply adjusted the supply voltage.

Thanks to the distributed heat source, the temperature gradients were minimal. So I could trust that all the chips got the same treatment.

Smaller number of chips (less power dissipated) requires of course better insulation to get the same temperature.

[Andreas]

Hello,

update on ageing drift of my ADR1000A#01 + #02 now at 5 kHrs
see also:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4397278/#msg4397278

the 6.6V zener outputs still drift -0.3 ppm/khr while the 10V output is better (the voltage divider drifts the other direction most probably due to still high humidity).
The Jump on ADR#02 on day 180 by +0.7ppm was created by a flat battery for ~3 days (I had forgotten to attach the charger).
At the same time I had also a Jump of 0.4 ppm on the LTZ9 sanity check so not all of the 0.7 ppm are true.
But afterwards I have still a permanent hysteresis of ~0.4 ppm due to the power off phase.

So regarding hysteresis this observation confirms the measurements of Dr. Frank.
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4398226/#msg4398226

with best regards

Andreas

[branadic]

I guess we can call it "All Drift Reference 1000" 
My experiment is still running too, but it is not yet conclusive. At some point (after around 6000 - 7000 h) the drift changed sign and is now drifting upwards instead of downwards.

-branadic-

[branadic]

Let me share an update of my measurement results (10V output, so we don't look at the raw zener voltage).
Any shift of the signals can be explained by either turning on/off instruments, removing equipment from the setup, changing batteries and the like, so what really counts is the overall behavior of the references, not the details.
What can be observed is a downward drift first, a change in sign second, followed by a steady state at the moment, while we have to be patient how that evolves next.

My current suggestion for a future experiment is to have three ADR references running at different oven temperatures:

1. running ADR oven at its zero t.c. temperature (~50 °C, that is a resistor ratio of about 11.5:1 for the datasheet circuit or about 0.52 V for the circuit using the temperature sensing transistor as a diode)
2. running ADR at 75 °C oven temperature (that is a resistor ratio of about 13:1 for the datasheet circuit or about 0.465 V for the circuit using the temperature sensing transistor as a diode)
3. running ADR at 100 °C oven temperature (that is a resistor ratio of about 14.6:1 for the datasheet circuit or about 0.41 V for the circuit using the temperature sensing transistor as a diode)

The comparison of the drift between these three references should indicate what time it takes for the references to stabilize at which temperature, though it's not an abolute measure. Once stabilized the oven of the latter two references can be set to the zero t.c. temperature of their zener. For this experiment I would use the raw zener voltage buffered and preferable nanovoltmeters to measure the differences.

-branadic-

[KT88]

My current suggestion for a future experiment is to have three ADR references running at different oven temperatures:

1. running ADR oven at its zero t.c. temperature (~50 °C, that is a resistor ratio of about 11.5:1 for the datasheet circuit or about 0.52 V for the circuit using the temperature sensing transistor as a diode)
2. running ADR at 75 °C oven temperature (that is a resistor ratio of about 13:1 for the datasheet circuit or about 0.465 V for the circuit using the temperature sensing transistor as a diode)
3. running ADR at 100 °C oven temperature (that is a resistor ratio of about 14.6:1 for the datasheet circuit or about 0.41 V for the circuit using the temperature sensing transistor as a diode)

-branadic-

Ideally a couple of specimens from at least three production lots should be taken to get a more meaningfull result. Of course that would be quite a challenge when it comes to different production lots...

Cheers
Andreas

[Andreas]

Hello,

I also think that you need at least 3 samples per group.
With my LTZ-References one of 9 samples (LTZ#8) showed more than 1 year until it stabilized below the typical 1-2 ppm/year.
During first year I had a -3.4 ppm drift.

For the rising drift: I think it is partly related to the humidity of the output voltage divider.
So I guess it will fall again during upcoming months (winter) when humidity is falling.

with best regards

Andreas

[branadic]

From an academical point of view I agree to the sample size, but it's a question of cost and samples at hand and at least I have 5 ADR samples including the required resistors left with which I can do further investigation. I'm not claiming statistical relevance, but indication only.

Edit: While humidity influence is a valid assumption for possible drift, I don't see any real correleation and hence no causality yet.

-branadic-

[iMo]

Guys, while observing your efforts and cost of material and electricity you spend a single well working 10V reference is worth of a Bugatti Chiron.. 

[openloop]

The ADA4084 suggested in the ADR1000 datasheet may be a bit questionable.

Just realized: Configuration, where grounds for force and signal are separate (as in a diagram in ADR1000 datasheet), appears to be problematic indeed for ADA4084.

Look at the Fig.48 (p18) from ADA4084 datasheet. The left side (right where we are) is not a good place to be. Some minor action of the heater (lifting the "force" ground a bit) might cause offset to go off by up to 0.5mV!
That will translate into, like, 2uV of additional swing of reference, on top of everything else...

I do not like ADA4084 here at all.

[Andreas]

Hello,

update on ageing drift of my ADR1000A#01 + #02 now at 6 kHrs
see also:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4464853/#msg4464853

the 6.6V zener outputs are now at -0.5 ppm/khr while the 10V output is still better at ~ -0.3 pm/kHr.
As it is getting more and more cold outside and now after a rainy phase starting with less air humidity:
I am very curious wether the different 10V output of the statistical divider is correlated with humidity or simple ageing drift.

with best regards

Andreas

[dietert1]

As far as i understand the diagram, the division ratio of the statistical divider drifted about 1 ppm in 250 days. That's an interesting result by itself. Which part are you using for the divider?

Regards, Dieter

[branadic]

Thanks for your update Andreas, 0.5 ppm/khr still isn't the typical 0.5 ppm/a as per datasheet
Also interesting to see the difference between buffered and scaled output.
I'm heading towards 12.000 h with the sample I'm running.

-branadic-

[Andreas]

Hello Dieter,

I am using a TDP16031002AUF.
Effectively 6 out of the 8 resistors in the DIP16 package are used for a *1.5 configuration.

with best regards

Andreas

[ap]

Although this formally seems not to be a NOMC network, it seems to use the same materials (NiCr). So the 1ppm in 250 days seem way more than one would expect.

[Kleinstein]

For the 7 to 10 V step only about 1/3 depends on the resitors. So a 1 ppm drift in the 10 V translates to 3 ppm change for the resistor ratio. The averaging over 2 / 4 resistors each also helps. Leaving out 2 resistors out of the 8 would allow to skip on the outer ones that seem to be ones that can be a bit differente at least for the NOMC.

[Andreas]

Also interesting to see the difference between buffered and scaled output.

Hello,

I made a additional chart with the difference of ppm drift between the 6V6 and 10V output of ADR1000A#1
(so effectively the ratio drift of the statistical divider)
Up to now it is a rather linear drift over time.
But if I compare it with PT1 filtered rH% values of my room data logger it could also correlate to rH%.

From my LT1027 references with DIP8 package I know that the influence from rH% needs several days (more than 5 days) to correlate with voltage output drift. So I used a PT1 filter with 1 week time constant. The values are logged every 15 minutes.

with best regards

Andreas

[Andreas]

Hello,

update on ageing drift of my ADR1000A#01 + #02 now at 7 kHrs
see also:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4547027/#msg4547027

still a relative large drift against a typical LTZ1000.

Humidity is now decreasing (a little) and the 10V output divider drift is slowing down somewhat.
But I fear at the moment that the humidity is not the main factor for the output divider drift.

with best regards

Andreas

[openloop]

Decided to check the "low noise" aspect of my ADR1000 references.
See how they stack-up against my LTZ1000s.
Conditions: Two independently battery powered reference boxes connected to measure voltage difference, HP34420A, 10mV range, 100 NPLC (@60HZ), 4 hours of readings, no filters.

Notice the change in vertical scale.

[branadic]

I passed 13000 hours, attached an update.

-branadic-

[Dr. Frank]

Here's an update of the drift measurement of my ADR1000 #3.
After 1 1/2 years of operation, the drift is slowing down, but still at a rate of about -2ppm/year.
I observed temporary hysteresis of about +2ppm when I shut off the references for some time, in around June 2022.
I shut it off for 8h and then measured the recovery over 17h.
Its voltage returned to its former value within less than 0.2ppm, so I assume that this hysteretic effect might be not so pronounced anymore.

Frank

[miro123]

Marco reps posed youtube video

[iMo]

Btw, - as seen in the above video - burn-in with 140C at 6 o'clock position and then using it at 12 or 3 o'clock position - is that a good idea?
PS: a pity ADI is so slow with the release of the ADR1001, as it would be much simpler and cheaper exercise, imho..

[Andreas]

hmm,

I fear I will not have the time to view the looong video. (i have to watch my references for ageing).
Can someone summarize the quint essence?

BTW I have found some github data:
https://github.com/marcoreps/ADRmu

with best regards

Andreas

[Kleinstein]

I fear I will not have the time to view the looong video. (i have to watch my references for ageing).
Can someone summarize the quint essence?

For a lower TC without the heater active it seems to help to have smaller collector resistors (e.g. 30-50 K range instead of 61 K from the datasheet example circuit) than in the data sheet.
This does no come to a surprise - it was known that the zero TC point is at higher current (for the zener or transistors).
The main point was that the reference boards mostly seem to work well (noise in specs and not much visibe drift, one sample with popcorn noise).
Preaging at 140 seems to cause at least initially more drift - so likely not a good idea.  I think others made a similare experience with aging at a rather high temperature.
The DCDC converter as some positive aspects:  isolation via a simple though relatively large toroidial transformer with relatively large spacing between the windings and the core. No details on the performance shown and I think there may be room for improvements ( at least a partial electrostatic shield, reduction in stray field).

[miro123]

The DCDC converter as some positive aspects:  isolation via a simple though relatively large toroidial transformer with relatively large spacing between the windings and the core. No details on the performance shown and I think there may be room for improvements ( at least a partial electrostatic shield, reduction in stray field).

I question this approach nowadays.
1. The battery technology improved last 50 years.
  - capacity/volume went up
  - reliability went up too - long term stability. number of cycles no memory effect just to name few.
  - BMS made last 50 years huge leap too
  - price went down - e.g. I've bought 100Ah LiFePO4 cells - each has the cost of single LTZ1000
2. Typical use case of Vref - 99,99% Idle/aging  + 0.01% measurements.
3. Cost - BOM and effort to build own dc-dc vs COTS battery backup

[Kleinstein]

Li cells are well performing, but there are regulations on transport, that can make them a hassel. NiMH cells are not that great an alternative and a bit tricky when it comes to charging.

There is one possible alternative: use 2 sets of batteries and relays to switch over. One set is used to power the device and the 2nd set can be charged with an automatic (low glitch) switch over if needed.  So no more transformer in a DC/DC, but 4 relay contracts to set the isolation. The battery to case part can still add quite some capacity - so look out for a low capacity battery  .
Another point could be the variable heat production during a charge cycle. So it is not that ideal too.

Those references are usually made to be run 24/7. So a battery only solution is not really an option. It is nice to have the battery option though. So use mains power most of the time and battery operation for the really critical measurments and on transit if one wants to keep it hot all time.
A good DCDC would allow to keep it mains powered for more of the less critical tests (e.g. 24/7 difference data to a 2nd ref).

[miro123]

Humidity is now decreasing (a little) and the 10V output divider drift is slowing down somewhat.

you can apply correlation function to fanout any dependency, delay times and trends. That said does statistical methods works better with higher amount of sample datapoints.

[Dr. Frank]

Anyhow, it's a pity that Marco does not join this community to share his evaluations, although he evidently has used knowledge especially from this thread. 

[DavidKo]

Andreas: I have watched it in the train to work (2 trips required) 

The PCB should be universal for LTZ and ADR (ready for critical divider from normal resistors, resistor array or Vishay), it should have output current protection and lost power indicator. Both old and new refences are evaluated. What I have found interesting is T.C. measurement and tuning of resistors. During the resistor tuning the whole setup is in external oven to include the T.C. of the whole setup, but at the end he uses the internal heater. In video is said, that the github data should be updated to see further development of values.

Design is on 6 layer PCB. The cap for reference is glued - I expect that screwing will be better due to the different temperature expansion of FR4 and plastic.

[Kleinstein]

I don't know the plastics used for the cap, but most types are relatively compliant compared to the FR4 board. Ideally the cap should not be massive plasitcs, more with small voids, as sometimes used with 3D printed parts to save material. Not sure how it is really made.  The force from the cap should be relative moderate. In addition the temperature is expected to be relatively constant, partially regulated by the reference.  I would have preferred to glue the cap only at 3 local spots to allow even more compliance. Not sure that srews would be much better.

Tunig the ref. TC for the whole circuit in one step is a bit odd:  much of the reference TC will in the final configuration be suppressed by the oven part. The final gain part would still be 100 % effective.
So it does not help to do the test for the full circuit.

It should still be OK to do the trim this way: the final gain stage needs a very good TC (e.g.  < 2 ppm/K preferrably better). This way it would not really interfere with the trim of the unheated reference that does not need to be that good. Even the initial about 20 ppm/K range is OK and 5 ppm/K well good enough.  One could use an alternative way by keeping the environment temperature constant and just add a disturbance to the set point divider. So change the oven temperature and choose R2 (e.g. 30 K to 100 K range) for low sensitivty to the set point divider.  I would consider a jumper from the set point divider to a resistor (e.g. 100 K range) and trimmer (or external control voltage) for that test to shift the set point. This way no extra temperature controled box is needed.
The test in the variable temperature box with the heater enabled would mainly test the final gain stage.

[branadic]

First a quick comment about the history of the ADR1000. What we've learned from Eric Modica at MM2021 was:

"... I had just taken over a project to second source the LTZ1000, which you all know. And this was obviously before the merger with Linear Technology. The primary reason ADI started to look at second sourcing the LTZ1000 was because, one of our VP saw an ADI AppNote that used the LTZ1000 as a reference for a 20 bit DAC and that was the AD5791. So by the time I took over the project, there had been a couple of test chips to look for various candidates for internal processes to use at ADI to match such a part..."

So in my understanding the goal was not to create an update or improved successor, but something similar to the LTZ1000 and by "...We got lucky?..." the final reference showed lower noise.

Given the charts from this thread in the video showing measurements after 8000 hours it must have taken Marco quite some time to create his video, almost a year.

A few technical comments. What I could extract from the video is that Marcos references show quite large noise compare to other measurements done before, which makes me wonder why. A possible source could be the DC/DC converter and to verify that powering the reference circuit from batteries could help. Also the measurement seems to be performed without protection from ambient noise, here is where your lovely cookie box comes in handy.

Furthermore, he build a reference with an early engineering sample with datecode 1727. As far as I understand this is a sample without the later introduced PMO which is a 150 °C, 168h both bake and burn, which Eric explained to be the cause for jumps that appear like popcorn noise.

"... what I’ve observed is that bake and burn is helpful primarily for low frequency noise reduction in this context. We have had tier 1 manufacturers tell us the same thing when we send them ADR1000s. Prior to instituting a bake and burn most of the parts look okay from a low frequency noise perspective, but if you look at it enough you find one that is extremly noisy and actually remedies with the bake and burn. So only being a process hobbiest myself, the only explaination I can give there is that you have these dangling hydrongene bonds at the surface of the SI and sometimes that amounts to surface charge and if that ends up in the vacinity of the zener, if you are unlucky enough, then this is the kind of thing that you get. And it’s really the only thing that would move at such low temperatures where you are running a bake and burn, say like a 150 °C..."

Nice to see Marco adopted some of the concepts from W/F7000, that were proven to work with results shown here on eevblog too, such as zero t.c. adjustment. However, some of the t.c. curves make me wonder as I have never observed any kink in my curves when sweeping through the temperatures.
It would be nice if he would present his results not in some special Python or Grafana format, but as charts with proper image file extension. Finally, it remains interesting to see more longterm results from more specimen in general, so I'm curious to see how his samples will do in the future.

-branadic-

[miro123]

A few technical comments. What I could extract from the video is that Marcos references show quite large noise compare to other measurements done before, which makes me wonder why. A possible source could be the DC/DC converter and to verify that powering the reference circuit from batteries could help. Also the measurement seems to be performed without protection from ambient noise, here is where your lovely cookie box comes in handy.

I'm going to experiment with ZVS/ ZCS half resonant converters.  I'll post the result in separate tread. they will be together with transformer bode100 plots - just to avoid the lucky shots in inductors winidings

It would be nice if he would present his results not in some special Python or Grafana format, but as charts with proper image file extension. Finally, it remains interesting to see more longterm results from more specimen in general, so I'm curious to see how his samples will do in the future.

I tend to disagree here. My reasons are listed below
1. Giving access to raw data allows other engineers to analyzed it in make own conslusions.
2. Files are in jupyther notebook format. De facto standard in EE world. I think this is essential  EE skill nowadays - like handling soldering iron, Ltspice matlab and CAD product.
e.g. jpyther screenshots shows few wrong assumptions made by Marco. Source https://github.com/marcoreps/ADRmu/blob/main/results/ADRmu3_TC.ipynb

[branadic]

I agree that access to the raw data is nice for sanity check by other, I disagree though giving the raw data only in a very speciallized format. It is unscientific to provide data and just say: "analyze them on your own". It would be way better to provide them in a way so that everyone understands what you want to present and that is a plot/chart/image. You can feel free to provide the raw data additionally.
I don't care whether or not jupyther notebook format is a de facto standard, you simply don't put such files into a review paper and tell the reviewers to plot whatever they like to see themself, do you?
And what if at some point the de facto standard changes? Hence I prefer raw data in plain text files, csv format or similar, which is platform independent and readable "For All Times, For All Peoples".

-branadic-

[Andreas]

Hello,

from the diagram you can clearly see that 16 deg/C per hour is much too fast.
The delay between reference output and Temperature is around half an hour.

with best regards

Andreas

[iMo]

Hello,
from the diagram you can clearly see that 16 deg/C per hour is much too fast.
The delay between reference output and Temperature is around half an hour..

That would require 10x longer sweep time, imho. Also 0.024ppm/C with the components used seem to be too good to be true..

[Andreas]

Hello,

update on ageing drift of my ADR1000A#01 + #02 now at 8 kHrs
see also:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4623766/#msg4623766

Now at -6 ppm reference drift over 8 kHrs or -0.75 ppm/kHr in average.

Humidity is now decreasing but the 10V output divider drift is still rising.
Obviously the drift over time (ageing) is the main factor.

with best regards

Andreas

[branadic]

I passed 14000 hours.

-branadic-

[iMo]

While looking at your above pictures - that reminds me on the first ever measurement I did with myADR1001#1 - see below (from the thread on ADR1001). I ran the chip with temperature set to something like 115-125C (it was not my intention, btw). The measurement started perhaps 1 hour after first ever power up and took aprox 13 hours.
Almost the same data and shape as your long term measurement above. Interesting..
Could it be you may speed up the burn-in with higher set temperatures?

[branadic]

That is unlikely. Correlation doesn't necessarily mean causation. You can't speed up multiple thousands of hours to just a few tens of hours by higher temperature.

-branadic-

[Kleinstein]

In some cases one can indeed speed up things with just a higher temperature.  Many prodesses speed up by a factor or 2 every 5-10 K. So it would need some 50-100 K more to get a factor of 1000 speed up.
With some thermal activated processes the exponential law holds over a really large range. The limits may be more with the measurement (both at the fast and slow end).

The shape of the curve loops like 2 processes, a slighlty faster settling one going negative and a slower one going positive.
Given the same type of sample and no additional turn around visible it is even quite likely that this are the same 2 processes.

The rather high temperature however also comes with a downside: the residual aging processes also get faster and this result in more drift after the initial settling.

[branadic]

I disagree, you can't compare ADR1000 with a totally different schematic and components involved with ADR1001 and correlate aging curve shapes. You could though for same devices and designs, such as multiple ADR1001 running at different temperatures.

-branadic-

[Kleinstein]

I have not realized that imo' curve was for a different, though still similar chip.  So hard to tell if the same porcesses are involved.

The point of the parts around the actual reference is also a good one.  There is drift from the actual reference chip, but possibly also from the resistors around them.  The ADR1000 should have a realtively good attenuation of drift - from the DS a factor of ~ 200 for the set temperature and ~ 230 for the "60 K" resistor R2. It is a bit unclear, but I would not expect much effect from the resistors.  I tend to see R2 as the more critical. 1 ppm overall drif would need some 200 ppm of drift from the resistor. Over 14000 hours is not that likely to get for a good resistor.
So chances are most of the drift seen for the ADR1000 ciruit is actually from the reference chip and not from the resistors.
What resistor types are used for R2 in brandic's reference ? ( the info may be somewhere at the start of the thread, but hard to find - the picture shows 2 metal cans, some SMD and a DIP16 array (likely for the set temperature)).

[iMo]

The initial drift at the high temperature with my ADR1001 has been measured directly at the 10V output of the ADR1001. That involved basically the same "schematics" as the most 10V refs with ADR/LTZ1000 with 10V output. Sure - the resistors are different technology. My guess would be the initial drift comes from the ADR1000 structure we most probably have the same technology (as I doubt ADI has developed a new kind of reference for the ADR1001).

[branadic]

That is just speculation unless someone shows a bare die image of what's inside an ADR1001 package. Until then I wouldn't even think about similar behavior especially if some 13 hours on a single specimen were measured with a 6.5 digit meter.

-branadic-

[Noopy]

That is just speculation unless someone shows a bare die image of what's inside an ADR1001 package.

I would love to!!! :'(

[balun]

That is just speculation unless someone shows a bare die image of what's inside an ADR1001 package.

I would love to!!! :'(

I would donate one to you if you tell me where to get it.

[Noopy]

That is just speculation unless someone shows a bare die image of what's inside an ADR1001 package.

I would love to!!! :'(

I would donate one to you if you tell me where to get it.

That's very kind but the problem is you can't buy one and you (or at least I) can't sample one. 

[Andreas]

Hello,

update on ageing drift of my ADR1000A#01 + #02 now at 9 kHrs
see also:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4709861/#msg4709861

Humidity is still low but the 10V output divider drift is still rising.
So now it is clear that the 6V6 to 10V divider has its own ageing drift.

If I take the X-Axis not in days but in sqrt(kHrs) getting 3 kHrs out of
the sqrt(9 kHrs) the ageing drift of both
6V6 outputs is nearly linear with -2 ppm/sqrt(kHrs).
Except for the last kHr where ageing drift seems to get slower.

with best regards

Andreas

[Dr. Frank]

Here's my ADR1000 #3 drift report, fresh from the 3458A as of today.
It's the buffered, raw V_Z of the ADR1000.
For the last 6 months, the voltage drift seems to have stabilized to reasonable -1ppm/a.



As a reference, here are the overall drifts of my Volt Group. One can see the improved Transfer Accuracy of about 0.1ppm.



Yesterday, TiN finally uncovered his revolutionary 10 Digit Cryogenic Standard HP3458A.
As a premiere, I'd like to present my mediocre 9 digit, 0.02ppm Standard Deviation, hand written protocol from today.
I can keep up with TiN a little bit, as my transcript was done with a PTB pencil, so it must be extremely precise.



Frank

[branadic]

So this is ADR1000#3, what about ADR1000#1 and #2?

-branadic-

[Dr. Frank]

I already had 10 references running continuosly,  2 further intermittently, so I assembled 1 PCB only. Sorry.

[branadic]

I meanwhile switched from hours to days on the horizontal axis, so here is an update after 634 days (15228 hours) of operation. In general the ADR1000 based reference seems to be stable, but shows some variation that I can't really correlate with temperature nor humidity. Due to missing access of the buffered raw zener voltage it is unknown whether this is a result of the resistor network of the boost stage or the reference and its associated components.

-branadic-

[Dr. Frank]

Here are the stability data for the ADR #3 as of today.
The reference now seems to have stabilized, as its drift was < 0.5ppm over the last seven months.
In comparison there are my two LTZ1000A, which had been thermally pre conditioned, with -1.0 and -1.5 ppm/ 2 years.
My Fluke 7000, which is the worst drifter from the batch, still has a higher drift of about -1.5ppm/year, but that's still inside its specification.
I assume, that its 7V => 10V stage is the culprit.

I never want to suggest anything by such comparisons, but maybe the single ratio divider IC is also the root cause for branadics ADR1000's big 10V drift.
It wouldn't cost you anything but an additional jack at the lower point of this divider, to measure the direct reference voltage.

Frank

[branadic]

@ Frank
From what I see the initial drift is not that much different to your buffered zener voltage. However, on my reference at some point a drift with reverse direction is taking over, which could potentially be the resistor network, I fully agree. Nevertheless, this drift meanwhile seems to have stabilized too and I only see some variations, that can't directly be correlated with temperature or humidity, which is a bit odd.

Connecting an additional jack to the inverting input of the gain stage is a little tricky. I had quite some bad experience with that in the past, since this node is of high impedance, prone to introduce all sorts of noise/hum/disturbance.

On the other hand my reference board was designed to connect both Z- and Z+ points to add a second and different gain stage as a piggyback solution, hence why the regulated voltage is accessable on the board. I had hoped that a different and working solution (PWM-based gain stage or PWM-supoorted resistor divider gain stage) would show up here on the forum, but unfortunately that hasn't happened.
Custom made precision resistor networks at least where not in my scope, but could still be an option, including waiting forever for them to arrive.

-branadic-

[DH7DN]

@Dr. Frank

I don't know what "baseline 3458A" means and at this point I'm afraid to ask.    Could you explain it briefly, please? 

[Dr. Frank]

@Dr. Frank

I don't know what "baseline 3458A" means and at this point I'm afraid to ask.    Could you explain it briefly, please? 

That's my expression for a new calibration point, derived from a voltage reference with low uncertainty.

Over the past 10 years, or so, me and other volt-nuts exchanged voltage references, either on Metrology Meetings @ branadic, from the calibrated Fluke 7000 ensemble, which was calibrated by TiN to <0.5ppm uncertainty, by sending a known voltage reference from one participant to another, by personal meetings, or by sending one of my own DIY references to 'ap' for calibration.

This way, I was able to determine the drift rate of my ensemble of 8 references, and the averaged "Volt" of this whole ensemble. This predicted mean value fitted within 0.5ppm to the latest traveling standard which branadic had sent around last year. I meanwhile monitor the drift of 4 additional references, and the LTZ1000A inside my HP3458A.

So it's time now to get a new baseline point.

Frank

PS: I got my first baseline point in December 2009, when I acquired my Fluke 5442A (with unknown uncertainty) and compared it to my first two LTZ1000 prototype references, built in around 2004 and running since then.

[DH7DN]

Doing "inter-voltnut" comparisons is a good (and also very important/necessary) metrological practice. Thank you very much for your explanation, Dr. Frank. 

[chris_11]

So long story short, within 40 years there is only a very small improvement on heated buried zener references. Todays much cleaner and more precise processes make popcorn noise less likely, but in essence the ADR1000 winds up with comparable drift to the LTZ1000A.
Or did I miss something?

[Andreas]

Hello,

update on ageing drift of my ADR1000A#01 + #02 now at 10 kHrs
see also:
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4789418/#msg4789418

At the moment the zener drift is around -0.3 ppm for the last kHr on the 6.6V outputs.

Humidity is still low but the 10V output divider drift is still rising.
It does not fully compensate the falling drift of the zener so the resulting drift is ~-0.1 ppm for the 10V output.

with best regards

Andreas

[branadic]

I passed 16550 hours, so here is an update.

-branadic-

[Andreas]

Hello,

I now passed 11 kHrs:

the 10V output on ADR#01 seems to stabilize.
the 6V6 output ageing rates are slowing down.
And this even with SQRT on the X-Axis.

Divider drift is still rising.
Since spring has passed the humidity rH% is now rising.

with best regards

Andreas

[branadic]

Divider drift is still rising.

Could you perform a fit to tell what formular it follows? Looks like a ld(x) function.

-branadic-

[Andreas]

Hello branadic,

Good  idea!
if I use the time axis as sqrt(kHr) I get a near linear behaviour above 1kHr

with best regards

Andreas

[branadic]

Two years of data, time for an update of my graph.

-branadic-

[Andreas]

Hello,

I now passed 12 kHrs:

the 10V output on ADR#01 seems to further stabilize.
the 6V6 output ageing rates are slowing down/seem slightly goint into drift up phase
And this even with SQRT on the X-Axis.

Divider drift is still rising. (mostly linear with SQRT on X-Axis)
Since spring has passed the humidity rH% is now rising.

with best regards

Andreas

[Andreas]

Hello,

I now passed 13 kHrs:

nothing new against previous observations.

with best regards

Andreas

[alligatorblues]

The stats of the raw regulator IC is irrelevant. Once you get the output to a practical, usable value, like 10VDC, then take your stats. I made 2 LTZ1000 references, used all low-drift components after the regulator IC, and got 1K/C. So, you really have to ovenize everything but the psu.

I have never seen anyone build their own DC Standard that was outstanding, after the output was adjusted to something practical and useful. The only reason Fluke can do it is because it is a ultra-precision resiator manufacturer, but it doesn't sell them.

732As, 732Bs, and 732Cs are all slightly different within the model. No one can obtain Fluke's resistors, so it is impossible to even come close.

[Kleinstein]

With the ADR1000 reference and a 6.6 V raw reference votlage it is actually possible to get a quite stable scaling to something like 10 V as the ratio is very close to 3/2 which can be realized based on a off the shelf resistor array (e.g. LT5400).  It is true that it is difficult for the more common 7 V to start with, but even this is possible, e.g. with PWM based scaling. Even if not super linear it usually is pretty stable over time.

[iMo]

..
I have never seen anyone build their own DC Standard that was outstanding, after the output was adjusted to something practical and useful. The only reason Fluke can do it is because it is a ultra-precision resiator manufacturer, but it doesn't sell them..

I do not know whether my DC standard is outstanding (most probably not with the engineering samples), but the ADR1001 has got all the resistors with useful outputs (7V/5V/10V) ovenized

[branadic]

I've passed 800 days, here is an update of my diagrams.

Edit: I've plotted the differences, both from calculated (cyan) and measure (blue) data taken by the nanovolt meters. From here we can extract that ADR stopped its initial drift roughly after 250 days, before some other drift mechanism took over, which is linear between two references just yet.

-branadic-

[Andreas]

Hello,

I now passed 14 kHrs:

the 10V/7V divider drift is still nearly linear in sqrt(kHrs) display

with best regards

Andreas

Edit: (explanation of the diagrams for those who are not familiar with the history)

I am monitoring 2 ADR1000 samples for ageing drift.

ADR#1 sample built on a PCB from branadic with 6.6V (Vz) and 10V output based on a statistical TDP16031002AUF resistor array.
ADR#1 is operated in a zero TC temperature setpoint of ~55 deg C with -6ppb/K for the 6.6V output and +20ppb/K for the 10 V output
after T.C. adjustment with a copper resistor in series to the 100 Ohm current setpoint resistor.

Ageing measurement started after some run-in phase.

https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3921830/#msg3921830
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg3931196/#msg3931196


ADR#2 sample is built on one of my LTZ1000 PCBs after 2kHrs burn in (110-130 deg C) of the reference. There is only one 6.6V (Vz) output.
T.C. adjustment is done via variation of the 400kOhms resistor. Resulting in -4.5 ppb/K for a 3.3 Meg resistor.
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg4054810/#msg4054810

20231027_ADR1000A_01_02_ageing_drift.PNG
shows both 6.6V output drifts of ADR#1 in green and ADR#2 in pink.
The 10V output drift is in blue.
LTZ#9 (my 2nd best reference with <1ppm/year) in red is only as sanity check (so can be ignored).

All (corr) readings are corrected with the reference voltage of my LTZ#4 (my most stable reference well below 1 ppm/year)
Otherwise I could not make sub ppm observations with my 6.5 digit multimeters and 24 Bit ADCs.

20231027_ADR1000A_01_02_ageing_drift_sqrt.PNG is the same diagram only with the x-axis in sqrt(kHrs) instead of days.

20231027_ADR1000A_01_divider_drift.PNG shows the calculated 6.6 to 10V divider drift over time with humidity.

 20231027_ADR1000A_01_divider_drift_sqrt.PNG is the same diagram (as suggested by branadic) with x-axis in sqrt(kHrs).

[branadic]

Andreas, would you mind sharing a plot with all of your ADR1000s and briefly summarize what your legend means? I find it hard to understand what the meaning of "LTZ#9 corr", "ADR#01 Vz corr", "ADR#01 10V corr" or "ADR#02 corr" is and everyone has to go through the thread to see what treatment which reference has seen. Would be neat to have it summarized in one post. Thanks.

-branadic-

[Andreas]

Hello branadic,

I updated some information in my last post.

with best regards

Andreas

[branadic]

Thanks Andreas, now the diagrams make way more sense to me. 
So from here we sort of can conclude that the 2kHrs burn-in (110-130 deg C) procedure on ADR#2 didn't change the drift behavior compared to ADR#1. Furthermore, both references seem to have changed drift behavior after ~300 days (from log or square root to something more linear, by the looks of it).
However, the divider drift over time seems to stabilize too. And we can conclude, that the network is not directly affected by humidity, which is also very conviencing.

-branadic-

[Andreas]

Hello,

made a power off state of 124 hours followed by a 4 hours warm up time for my both samples.

Result (see day 800 in measurement) a "dent" with ~ 0.3 ppm after power off phase on both references (and 10V output of #1).
So somewhat better than the "dent" on day 180 with ~ 0.5 ppm on ADR#2 which was accidently running flat for some days (forgot to attach the charger).

The question is: is the "power off" hysteresis getting better when ageing the references?

with best regards

Andreas

[alm]

The question is: is the "power off" hysteresis getting better when ageing the references?

Based on my understanding of "power off" hysteresis being caused by the current taking a different path through the silicon, I'd expect it to be random with a spread that covers zero shift. So I don't think you can draw any conclusions from 2 experiments.

[iMo]

The hysteresis is so small compared to other factors surrounding the chip and the measurement itself, that it would be hard to finger-point at the ref chip, imho. If the "dent" was 3ppm, I would point at the reference..

[Kleinstein]

The question is: is the "power off" hysteresis getting better when ageing the references?

Based on my understanding of "power off" hysteresis being caused by the current taking a different path through the silicon, I'd expect it to be random with a spread that covers zero shift. So I don't think you can draw any conclusions from 2 experiments.

AFAIK the picture of different paths is a thing for simple surface zeners and old stabilzer tubes, not for burried zener diodes.

There are several effect that could contribute to the hysteresis: With a heated reference part of the hysteresis can be from mechanical stress from differences in the thermal expansion. This stress may in parts (specially at stress concentrations) relax in both the cold and hot state. Another possible factor could be water vapor moving between parts/surfaces inside the sealed metal can. Also electrical transients on power on / off may be an issue trapping charge somewhere on or inside the chip. With part of the circuit still separete also the external parts like resistors may contribute.

Still the effect seen is rather small and it is hard to tell if the difference is real or just noise.

[Andreas]

Hello,

I now passed 15 kHrs:

the 10V/7V divider on ADR#01 drift seems to stabilize (so no longer linear in sqrt(kHrs) x-Axis)

The 7V outputs of ADR#01 and ADR#02 drift ~1 ppm in 5 kHr now.

Detailed explanation of the diagrams see previous post (14 kHrs)
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg5138967/#msg5138967

with best regards

Andreas

[Dr. Frank]

Well,

I haven't reported about my ADR #3 for a while, because I switched it off after June 18th.

In between, I got a new baseline comparison @ ab precision, and after several iterations at his lab, the mean value of my 8 references agreed within 0.25ppm with his 732B reference, which itself has 0.3 uncertainty, confirmed by a fresh calibration of the 732B afterwards. You'll see the baseline point in the diagram, which also is another confirmation for the recent ring comparison, branadic had organized around June 2022, with similar uncertainties by each Philipp and ab precision.

At first, the ADR #3 had stabilized before the shut-off. I powered it on again on 30th November 10 p.m., letting it warm up with the 3458A and the 5442A over night, for about 12 hours. So the transfer measurements showed a +5ppm shift, very huge compared to the other shifts, when I switched the ADR1000 off for several days or weeks.

Anyhow, a measurement on December 16th showed, that the ADR 1000 evidently returned to his recent voltage within a few tenth ppm.

All of my LTZ1000 and LTZ1000A, especially in my 3458A which is only switched on for measurements, never ever showed such a shift after being powered down.
They all return within 2h at most to their nominal value.
This effect - if it turns out to be inherent to the ADR1000's technology - is not convenient for usage as a intermittently used reference, like in DMMs or cold travelling standards.

I switched on my equipment on Thursday, 14th, and the next morning, I found my Fluke 5442A got a defect, the processor PCB obviously freaked out, so I have something to do Between The Years. Strangely, the transfer measurement of my other baseline references revealed, that my oldest prototype references Ref_1 and Ref_2, which were still not optimized and not in a tuner box enclosure, like all the others, both show an unexpected +0.4 ppm shift.
Something has happened during the night, damaging the 5442A, and influencing both the references.

By chance, the Solar outbreak just has begun on 14th December, but that influence is quite improbable.
Maybe there was another E.M.C. impact, or a short circuit occurred inside the 5442A, like a burnt tantalum, which impacted the references. We'll see.

Frank     

[iMo]

You had several highly correlated shifts with the REF1 and REF2 in past..

[Dr. Frank]

You had several highly correlated shifts with the REF1 and REF2 in past..

That only seems to be the case from view of this diagram, but had different reasons.
E.g. I preliminarily have adapted a new normalization factor for the whole group, that gave a discontinuity.. I have to find a better algorithm for averaging the reference group, smoothly integrate new baseline measurements, and a method to remove/ add members. I have another 3 or 4 references, which might replace the defect 5442A.

 This shift really is unique.
Frank

[Andreas]

By chance, the Solar outbreak just has begun on 14th December, but that influence is quite improbable.

Hello,

When exactly did you recognize the shift? In the morning of 15th?
I also had a strange reading on one of my AD586LQ based (unshielded ADCs):
In the morning of 14th and 15th the reading of ADC13 was about 2 ppm too high when reading PCB#2 of my ADR1399.
The reading healed on 16th

Normally I would think of a bad contact on the negative power supply of ADC13 which gives together with PSRR a too high reading.
But the idea with a solar outbreak is also tempting ...

with best regards

Andreas

[Dr. Frank]

Andreas, I actually measured on 16th, so I cannot assign this event to the fault of the 5442A. Normally, I don't report such singular observations, but this time, I could not resist. 
Frank

[branadic]

This is rather interesting, as I observed increased standard deviations and some sort of jumps in my 1h measurements of ADR1399 I take every day automagically within the last days, which I couldn't correlate with similar changes in ambient conditions.

-branadic-

[Dr. Frank]

Here's my New Years datalogging.
The ADR1000 definitely recovered to the previous value, after it's been powered off for several months:



My 20y/o Ref_1 and Ref_2 as well seem to recover from their shift, but still are not back to normal:



The environmental data logging (UNI-T UT330C) shows the usual rise during warm up and measuring session on 15th and 16th December, when my 5442A failed, and both references drifted upwards,



but what's really interesting is the quite fast and huge pressure rise exactly during that time period. I never have seen such a big rise before.
Maybe I'll ask the DWD (German Weather Service), what has happened.



The other references, LTZ1000As #6 and #7, as well my F7000 had also not been affected.



@ branadic: The F7000 meanwhile has stabilized during the last 8 months, or so.
   

[branadic]

but what's really interesting is the quite fast and huge pressure rise exactly during that time period. I never have seen such a big rise before.
Maybe I'll ask the DWD (German Weather Service), what has happened.

Indeed, quite some change I observed too:

but found a similar change about 300 days ago in my data 

@ branadic: The F7000 meanwhile has stabilized during the last 8 months, or so.

Nice to hear that 
It seems like that for no obvious reason references can start random excess drift that settles after quite a while for again no obvious reason, hence why they show up on ebay. Illya reported a settling of a drifting F732B he got after some years. Although we monitor ambient conditions it's still hard to explain the mechanism behind it 

-branadic-

[maat]

Maybe I'll ask the DWD (German Weather Service), what has happened.

This was the high pressure area FIONA. See for here: https://www.dwd.de/DE/wetter/thema_des_tages/2023/12/11.html (unfortunately only in German).

Attached is a data dump I got from from someone working at DWD It's their data from Frankfurt/Main.

[ramon]

... surface pressure around 1045 mbar ...

Marko Korosec (16/12/2023)  -  A Massive Heat Dome Develops over Europe
https://www.severe-weather.eu/global-weather/powerful-heat-dome-warm-wave-forecast-europe-christmas-december-2023-mk/

[Dr. Frank]

I managed to repair my Fluke 5442A.
It was quite a profane root cause: A current sensing resistor, carbon composite, inside the LM723 based +5V regulator circuit, drifted towards 3 times its nominal value. 
That's reasonable after 40 years, I guess.
Anyhow, the 5442A is now warming up, and I will redo my comparison measurements, and then send my LTZ #7 to branadic, for Volt comparison.
Frank

Well, that 0.15 Ohm resistor now measured 0.418 Ohm, so the ~ 1.7A, 5V supply was clamped down, and the processor board failed with under-voltage..
I wonder, how long the board was running on lower voltage, hopefully the calibration last year was not affected, i.e. the storage of the calibration constants.

I show the original test tips of my 121GW, from the early adopters, as these inhibited a direct identification of the fault.
The red tip is defect, i.e. it's making bad contact inside the tip, so I have varying Ohm readings on the order of several 100mOhms.
So I could not directly identify this increase in resistance. 

[The Soulman]

I managed to repair my Fluke 5442A.
It was quite a profane root cause: A current sensing resistor, carbon composite, inside the LM723 based +5V regulator circuit, drifted towards 3 times its nominal value. 
That's reasonable after 40 years, I guess.
Anyhow, the 5442A is now warming up, and I will redo my comparison measurements, and then send my LTZ #7 to branadic, for Volt comparison.
Frank

Good news it was a "easy" fix, I've seen drifting carbon resistors before on different equipment, lm723's (v-ref zener) do seem very stable over the years.

[Andreas]

Hello,

I now passed 16 kHrs:

the 10V/7V divider on ADR#01 drift has stabilized

The 7V outputs of ADR#01 and ADR#02 now drift ~1.3 ppm over the last year.

Detailed explanation of the diagrams see post of 14 kHrs
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg5138967/#msg5138967

with best regards

Andreas

[Dr. Frank]

The ADR 1000 #3 has recovered to its previous value, after the 6 months of been powered down.
It's the raw reference output, 6.668V, and it's still not yet properly assembled into a tuner box, unlike the other references.



I plan to put it in my 3458A first, to check how the noise of the 3458A would improve.

Frank

[Andreas]

Hmm,

so I fear you may want to operate your HP3458A 24/7 to maintain stability in future instead of saving power.
How about a backup supply in case of power outage?

with best regards

Andreas

[Dr. Frank]

Hmm,

so I fear you may want to operate your HP3458A 24/7 to maintain stability in future instead of saving power.
How about a backup supply in case of power outage?

with best regards

Andreas

Hello Andreas,
definitely, never would I let my 3458A run constantly.
That would reduce its life time, due to deterioration of the fast comparators EL2018, and it would bust my energy bill 

Anyhow, the 3458A just serves as a stable and ultra linear comparator and transfer instrument, but not as a voltage reference.

No,  I just want to assemble the ADR1000 PCB temporarily into the 3458A, just for repetition of the noise figure measurements on different NPLC numbers, as we did during the big DMM noise campaign, which TiN once had initiated.

The ADR1000 is very probably not suitable at all for permanent use in DMMs.
Therefore, it will not replace the LTZ1000, nor the LTFLU in these applications, unless somebody (in England? @ CERN?) finally finds a stable bake-in method.

I still think, that a 50% less noisy reference might give some noise improvement on long scale DMMs.
So maybe Philipp can test that on his Fluke 8588A as well? 

Frank

[Andreas]

I still think, that a 50% less noisy reference might give some noise improvement on long scale DMMs.

Hello,

I think the effect will be more at a input voltage near full scale.
At zero volts I think that mainly the input cirquit and comparator are the main noise source.
But we will see ...

with best regards

Andreas

[Kleinstein]

The reference noise and ADC noise are in principle different parts. Ideally the reference noise only matter proportional to the input voltage.  Many of the multislope and similar ADCs react to some reference noise even with zero input: they probaly missed that higher frequency noise in the range of the feedback modulation (e.g. 10-100 kHz) can also contribute, even with zero input. This part is easy to filter and does not need a lower noise reference. Often there is at least partial filtering - probably for a different reason.

It is the low frequency ref. noise that may want a lower noise reference.  A good ADC may reach the 100-150 nV/sqrt(Hz) range for a 10 V range. At low frequencies (slow conversions and possible additional averaging) the references will be higher than this. The relevant frequencies are often lower than the standard 0.1 - 10 Hz band.  A lower noise ref. could help a 8 digit DMM, not just near full scale, but mainly there. The longer the integration the more relevant the ref. noise also with less than full scale.  For > 1 Hz even the LTZ1000 is not that bad.

Beside the ref. in the DMM there are also only few external sources to measure that are lower noise - the easier part is to use the new ref. as an external reference / source.

For the 3458 ADC the largest noise source at the ADC are the resistors at the integrator. After that there are several similar sized contributions (e.g. OP-amps, jitter, other resistors, maybe the higher frequency ref noise). The comparators should not contribute much, even for fast conversions and even less at 1 PLC and above.

The strong effect of powering down the reference makes the ADR1000 less practical for a normal DMM.  A new DMM may get away with a lot less power than the 3458 and may not run too hot to stay on 24/7.

[branadic]

Given the results shown so far, how do you guys interprete the numbers and figures given in the datasheet?

Long-term drift
   200 h (early life drift) 8.9 ppm
1,000 h 7.7 ppm
2,000 h 6.6 ppm
3,000 h 6.2 ppm

Are these numbers really long-term drift or is it drift rate instead and if one was about to cumulate the numbers and plot a graph we are looking at a more realistic inverse exponential chart?

   200 h (early life drift) 8.9 ppm
1,000 h 16.6 ppm
2,000 h 23.2 ppm
3,000 h 29.4 ppm

-branadic-

[Andreas]

Hello,

Of course we are all hoping the 0.5 ppm/year after 3000 hours.
But obviously we are all not clever enough to follow the recommended cirquit and layout ?

with best regards

Andreas

[The Soulman]

Second option is how I read it. The first wouldn't make much sense.

Edit: the first option does fit the graph Andreas shared. Makes you wonder if those "typical" numbers in the datasheet are based on just these four samples?

[The Soulman]

And of course these are "typical" numbers only, actual performance could be much worse and you wouldn't have right to complain.

[KT88]

measured specs would mean that the drift would have been tested for example for 1k hours - that's not viable. The only way to know for sure is a burn-in. And even then the further LTD would follow the random walk scheme.

[iMo]

I remember the video workshop (I think organized by branadic) with the ADR1000 creators where it was said they did the LTD measurements in an oil bath, with not good results as the oil contained water residuals. They then replaced the oil with a better oil. Double-check that in that video. Perhaps it is why they get such nice LTD.
PS: ..and perhaps they do it with a naked package - the zener only..

[miro123]

Hello,
I fully understand why they use oil bath for LTD measurements.
However I don't understand how do they do it at ovenized references.
My experience is that there is no single manufacture who specify the test conditions.
Datasheet numbers itself  does not tel a lot if you don't know the test conditions.
In this case I the customer starts guessing. e.g. Branadic just guess something. I can guess something else too. Lets see who will win the lottery. :-)
My guess - Maybe the oil bath is heated to T=45..50C. The ADR oven is switched off on such way they eliminated the influence of ADR precision temperature control. It tests only the performance of the zenner.

[miro123]

it was said they did the LTD measurements in an oil bath, with not good results as the oil contained water residuals. They then replaced the oil with a better oil.

I use silicone oil for my resitors references. It performs better is terms of mix with humidity

[iMo]

Hello,
I fully understand why they use oil bath for LTD measurements.
However I don't understand how do they do it at ovenized references..

Imagine you are the R&D engineer at ADI. How would you test say 10 randomly chosen ADR1000 samples for 1000-3000hrs?
For example you may create a 10x10cm large pcb, with 10 pads for the ADR1k packages, solder the chips in, and wire the zeners only out through a flat cable (like 40 wires). Put the pcb into the oil bath, the oil temperature at say 70C. Leave it there during the measurements. You may even sweep from -40 to 125 when the oil allows.
You would definitely not build 10 Vrefs boards with all the resistors and opamps on it, as the ADI is not going to guarantee a performance on someone's fully populated pcb with "random components and design" either..

[Kleinstein]

For a ref. amplifier like the ADR1000 or LTZ1000, there is no good way to only use the zener. One could get away without the temperature regulation, but would still need the external loop for the voltage to get at least a reasonable low TC.   So they may have to build really complete ref. PCBs of some kind.  So the drift should indeed include a set of resistors. To be on the safe side, chances are they used quite good resistors, at least for a few sets. For tests with more parts (e.g. 50 pieces) they may also have lower cost versions.

A problem with the drift tests is that the treatment (e.g. soldering and maybe a bake before soldering, storrage before use) can also effect the drift. With the ADR1000 another parameter is the set temperature. This can make quite some difference in aging. The example cuves are for a rather high set temperature (75 C).

[iMo]

This is what I ment with the test jig (the pcb with packages only) for ie. 10 samples and 40 wires kept in the XXdegC oil bath.
On the other side of the flat cable (off the oil bath) is one opamp with two (70k and 120ohm) resistors and one diode, but only 1x, and there are 10 (x4 contacts) relays or switches or whatever which switch the particular sample in during the measurement..

[branadic]

Hello,

Of course we are all hoping the 0.5 ppm/year after 3000 hours.
But obviously we are all not clever enough to follow the recommended cirquit and layout ?

with best regards

Andreas

Andreas,

if we assume the values given in the datasheet are drift rates instead of long-term drift and cumulate them, we get the blue curve. If we also assume the elapsed time has to be cumulated we get the red curve, which looks way more like an exponential shape and thus more realistic to our observations of time needed for the reference to stabilize, don't you agree?

-branadic-

[Andreas]

Hello branadic,

usually it is more complicated.
My best guess:
- the 0.5 ppm/Year is that what they promised to their first customer who wanted a improved LTZ1000A. -> datasheet (target) spec.
- In reality they have forgotten (the LT-engineers left due to retirement) how to produce a reference with LTZ1000 specs.

has anyone seen a "errata sheet"?

By the way there is a PCN available: (they are changeing the die attach - of course without any changes on electrical specs)
https://www.analog.com/media/en/PCN/ADI_PCN_23_0197_Rev_-_Form.pdf

with best regards

Andreas

[Dr. Frank]

Hello branadic,

usually it is more complicated.
My best guess:
- the 0.5 ppm/Year is that what they promised to their first customer who wanted a improved LTZ1000A. -> datasheet (target) spec.
- In reality they have forgotten (the LT-engineers left due to retirement) how to produce a reference with LTZ1000 specs.

has anyone seen a "errata sheet"?

By the way there is a PCN available: (they are changeing the die attach - of course without any changes on electrical specs)
https://www.analog.com/media/en/PCN/ADI_PCN_23_0197_Rev_-_Form.pdf

with best regards

Andreas

Anyhow, the drift diagram diverges from all of our observations, that the (initial) drift is negative in reality in most/all cases.

After my observations about the hysteresis after months of power down, I'm even more convinced, that most of the initial drift, and all of the hysteresis is related to the adhesive material.

That PCN shows, that they use a special epoxy material, for electrical isolation. It's  LOCTITE ABLESTIK QMI536.
For most semiconductors, you want to use epoxy or a joint with high electrical and thermal conductivity.
As the Ablestik stuff has quite a high thermal conductivity, I assume that they use a similar method for thermal isolation (> 200K/W) like in the LTZ1000A.

Btw.: You'll find an Automotive classification of this change @ ZVEI.org, search for "PCN", then 'More information and Downloads', and then download the ZVEI Delta Qualification Matrix.
This is case SEM-PA-07, Change of die attach material, i.e. a C-level change, which usually requires no further electrical tests.
'No change in form, fit, function' is a buzzword you'll often find in PCNs, but that is the pre-requisite for any PCN, which always implies no change of the specified and non-specified parameters of the component.
The electrical characteristics will not be changed, that is evidently true, but those drift and hysteresis parameters, which belong to the non-specified (or not fully specified) characteristics, might be affected.

Unfortunately, the attached test reporton the ADI product page does not reveal further details, but I guess that Henkel takes care that the thermal and elastic parameters of their epoxy does not change much.

Would be interesting to know the replacement material and download its datasheet, or Henkel will only change the recipe w/o changing the product designator.

In the LTZ they might have used a similar epoxy, maybe there is another PCN on this component, where we can learn more.
Indeed, there's a PCN about the glass beads, they are using for the A version.
 
The LTZ1000A might behave differently concerning hysteresis, when transported cold at lower temperatures, as now.
Maybe I should send my LTZ #7 to branadic right now?

Frank   

[branadic]

Maybe I should send my LTZ #7 to branadic right now?

That would be great.

-branadic-

[Dr. Frank]

Maybe I should send my LTZ #7 to branadic right now?

That would be great.

-branadic-

I will send you the stuff on Monday, as in Stuttgart there is night frost predicted for the weekend, but none for next week.

Congratulations for your successful defense, Dr. branadic!     🥳🍾

[aronake]

https://www.analog.com/en/products/adr1000.html

pricing now avilable on analog webpage.

1.000 pieces for 57.93 USD each
500 pieces for 66.63 USD each

Selling option only "Distribution Only" what that now means, but looks like things are moving forward.

[Noopy]

I didn't find the ADR1001 in myAnalog so I asked the customer service if it is possible to get a small quantity. The answer was promising but I didn't hear more since then... 

[Andreas]

Hello,

I now passed 17 kHrs:

the 10V/7V divider on ADR#01 drift has stabilized perhaps with a slight upward drift.

The 7V outputs of ADR#01 and ADR#02 trend to further settling.

Detailed explanation of the diagrams see post of 14 kHrs
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg5138967/#msg5138967

with best regards

Andreas

[branadic]

Thanks Andreas. I haven't posted recent results for a while as we were waiting for our paper A 10 V Transfer Standard Based on Low-Noise Solid-State Zener Voltage Reference ADR1000 to be published.

Now that this is done I will post some new diagrams soon.

-branadic-

[branadic]

While the above mentioned paper covers an observation time of 850 days, I'm now approaching 1000 days.
For the paper noise measurements of the 10 V output (after 850 days) had to be performed and it was found that the output was slightly oscillating. A bit more capacitance between opamp output and inverting input was added, which resulted in a shift of the output voltage.

Furthermore, the output wasn't as stable as before, powering the reference from my Ultra Low Noise Lab Power Supply Extender for your existing lab supply. Hence why after 920 days I finally switched to a copycat of an LT1533 based power supply, from which it was running since then, but again slightly shifted the output. It will now be interesting to see, if things are stable. Attached is an updated diagram.

-branadic-

[ivo]

While the above mentioned paper covers an observation time of 850 days, I'm now approaching 1000 days.
For the paper noise measurements of the 10 V output (after 850 days) had to be performed and it was found that the output was slightly oscillating. A bit more capacitance between opamp output and inverting input was added, which resulted in a shift of the output voltage.

Furthermore, the output wasn't as stable as before, powering the reference from my Ultra Low Noise Lab Power Supply Extender for your existing lab supply. Hence why after 920 days I finally switched to a copycat of an LT1533 based power supply, from which it was running since then, but again slightly shifted the output. It will now be interesting to see, if things are stable. Attached is an updated diagram.

-branadic-

Is such a shift from change in PSU expected? It's within line voltage regulation? I agree with your paper's conclusion that you should have separated analysis of the standalone zener voltage from the amplification to 10V. As far as I can see - in the modern world although 10V is nice, the only true arbitrary number to choose, if implementing such a choice is not sacrificial to other parameters, is one that will be most easily used and useful to relevant ADCs (e.g near the top of their range[ s ]). I guess the only other competing factor is history - if other references are set to X, also aiming for X potentially allows an tighter comparison by using one as a null point (relative 0).

[dietert1]

If a setup has problems (in this case oscillation and power supply EMI) one needs to solve that. I think the apparent shifts of 3 and then 0.6 ppm are less remarkable than the fact that the first problem went unnoticed more than two years.
The first 920 days of the log may be or may not be typical for a ADR1000AHZ. That depends on whether that oscillation is typical..

Regards, Dieter

[iMo]

..A bit more capacitance between opamp output and inverting input was added, which resulted in a shift of the output voltage..
-branadic-

I reported here aprox +2ppm after adding the 4n7 foil into the feedback of the OPA189 in my myADR1001#1 (10V->10V buffer)..

[Andreas]

Hello,

I now passed 18 kHrs:

the 10V/7V divider on ADR#01 drift has stabilized but has still a slight upward drift.

The 7V outputs of ADR#01 and ADR#02 trend to further settling.

Detailed explanation of the diagrams see post of 14 kHrs
https://www.eevblog.com/forum/metrology/lowest-drift-lowest-noise-voltage-reference/msg5138967/#msg5138967

with best regards

Andreas