EEVblog Electronics Community Forum

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-

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.

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.

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-

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.

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.   

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-

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

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-

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.

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

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.

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

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

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

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

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

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

 

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

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"

Quote

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?

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.

...
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

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-

Quote from: branadic on October 02, 2021, 07:03:30 am

Quote

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.

Quote from: Noopy on October 02, 2021, 05:07:25 am

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?

@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?

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-

Test whole night and datas:

Quote from: Noopy on October 02, 2021, 07:45:59 pm

Quote from: Dr. Frank on October 02, 2021, 12:53:49 pm

Quote from: Noopy on October 02, 2021, 05:07:25 am

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!

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.

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

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

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

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 ?

I found a ADR1000 with datecode 2108. 

Wanna see what´s inside of this batch? 

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

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

 

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... 

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.

 

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.   

 

Quote from: Noopy on October 07, 2021, 04:28:42 pm

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".

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.

What has eight legs and an eerie green glow?

Looks dangerous! 

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

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.

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..

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.

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..

On 1 o'clock..

..

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

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.

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.

see picture as below:

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..

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?

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 ?

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-

Quote from: Noopy on October 10, 2021, 09:26:25 am

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.

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.

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-

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.

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-

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

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

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