Author Topic: ADR1001 - Ovenized Voltage Reference System  (Read 110399 times)

wry and 1 Guest are viewing this topic.

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14391
  • Country: de
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #375 on: May 29, 2024, 09:32:42 am »
The have the disclaimer to allow for changes. However for a precision part the users expect a not on any change and they usually get the PCNs.
Especially with a ref. chips one may have in house tests on properties the DS does not specify.

Ti even included logic chips in a not about a change of the case material, though these are usually less critical (could still be an issue with fire proofing or similar).

For the ADR1001 I would not expect a large change - more details in the materials / fab used. So nothing one expects to be visible on the photo.
 

Offline Noopy

  • Super Contributor
  • ***
  • Posts: 1809
  • Country: de
    • Richis-Lab
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #376 on: May 29, 2024, 10:52:58 am »
For the ADR1001 I would not expect a large change - more details in the materials / fab used. So nothing one expects to be visible on the photo.

I agree with you. ...but we can never be sure until we have taken a look under the lid...  ;D

Offline MegaVolt

  • Frequent Contributor
  • **
  • Posts: 926
  • Country: by
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #377 on: May 29, 2024, 10:55:51 am »
As an engineer we learn to work to interfaces. In the case of a semiconductor part the interface is the datasheet maybe plus other documents like errata sheet. Everything behind that interface remains a black box. AD can change whatever they want whenever they want as long as the part complies to its specs. I like the images but would not want to depend on your findings inside the black box.
This is all true when we create a regular device. If we are trying to get 110% out of a detail, any additional information is helpful. Including reverse circuits, photographs of the chip, etc....
 

Offline dietert1

  • Super Contributor
  • ***
  • Posts: 2193
  • Country: br
    • CADT Homepage
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #378 on: May 29, 2024, 12:22:30 pm »
If noopy can't characterize the device in a meaningful way, he could give it to somebody who can, instead of destroying it or maybe after soldering the lid back on..
I'm not interested though and can't follow the obsession. The only thing missing now is AD find an exclusive buyer, like it happened before with LT and Fluke on the LTFLU part.

Regards, Dieter
 

Offline iMoTopic starter

  • Super Contributor
  • ***
  • Posts: 4882
  • Country: vc
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #379 on: May 30, 2024, 07:05:07 am »
As I wrote in past there are two things missing in my prototype - the heater limiting resistor and the Cfb (100n in the DS) in the output buffer. Interestingly - the simulation shows the best stability with none Cfb.
« Last Edit: May 30, 2024, 07:10:23 am by iMo »
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 3272
  • Country: de
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #380 on: May 30, 2024, 07:38:32 am »
Hello,

no wonder. CFB has to be on the other side of R1 to improve stability.

with best regards

Andreas
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14391
  • Country: de
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #381 on: May 30, 2024, 07:59:25 am »
The other side of R1 is not accessible externally. The effect on stability is indeed limited if the output capacitor is with very low ESR - add some series resistance of a few ohms and the FB capacitor can again help.
The capacitor in the feedback is not just for stability, it also provides a little filtering with little effort.

It is hard to judge the stability just on the closed loop output signal.
 

Offline iMoTopic starter

  • Super Contributor
  • ***
  • Posts: 4882
  • Country: vc
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #382 on: May 30, 2024, 09:17:21 am »
Yep, the internal divider wiring does not allow to provide ie. the FRA analysis therefore.
Btw the 1R7 acts as the series resistor to the Cfb (in the first two sims).
Sure, it is just a simulation with a "model"..
 

Offline EC8010

  • Regular Contributor
  • *
  • Posts: 106
  • Country: gb
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #383 on: May 31, 2024, 09:55:02 pm »
Well, what do you know? Looks like the Aku Aku was right and Kon Tiki has quietly docked at the US of A! I've got a tracking number that says my ADR1001 evaluation board is at Grand Prairie, Texas. I think someone must have cheated and fitted an outboard to that raft. My board now has merely to make it across the Atlantic ocean and enter the British postal system...
 

Offline picburner

  • Frequent Contributor
  • **
  • Posts: 509
  • Country: it
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #384 on: June 01, 2024, 03:42:19 am »
Quote
....I've got a tracking number that says my ADR1001 evaluation board is at Grand Prairie, Texas.
Lucky you, my order was postponed again to June 25th instead.
 

Offline Noopy

  • Super Contributor
  • ***
  • Posts: 1809
  • Country: de
    • Richis-Lab
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #385 on: June 01, 2024, 01:56:33 pm »
Do you remember my preview of the ADR1001?
I have no picture of the new ADR1001 here. You will have to wait some more days because I'm very busy right now. But I have to tell you that they have changed something!  So it was a good idea to open one more... ^-^
The output resistance we have discussed has been lowered by connecting the output driver with the bondpad through a "large" metal area. This is now less an issue but you still have some resistance in the output. The "ADR1002" would be the better architecture.  ;D

Coming soon...  :-/O
 
The following users thanked this post: Andreas, picburner

Offline dietert1

  • Super Contributor
  • ***
  • Posts: 2193
  • Country: br
    • CADT Homepage
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #386 on: June 01, 2024, 03:56:52 pm »
The label ADR1002 has been reserved on December 13th, 2022 by imo for a change in pinout - in order to have separate pins for output drive and sense (divider input). As of now the part cannot drive a meter without additional output stage.
I think the only way around the instabilities with capacitive load without changing the pinout is an internal capacitor parallel to the upper divider resistor. Then we want to know the capacitance and not its size.

Regards, Dieter

Edit: noopy proposed to label the device he is sampling ADR1002, but edited his post later. Don't know how he did that without the Last Edit: ... attribute appended. Certain contributors need quoting.
« Last Edit: June 02, 2024, 08:21:45 am by dietert1 »
 

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7140
  • Country: ca
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #387 on: June 01, 2024, 06:52:35 pm »
Nobody stuffs a huge capacitor load on an op-amp's output. It's bad practice.
COUT datasheet 1-100uF is "lipstick on a pig" or a rookie-level mistake or both  :palm:
What are ADI trying to cover by doing that? There's an RC-filter CNF  to the buf amp's input, to filter out noise coming in. Either the buf amp itself is noisy, or picking up some artifacts from other sections, or it's just unstable. I don't see a on-die comp cap, there are no specs such as gain/BW spec for the op-amp either.

LT engineers, we miss you.  ADI egg on face continues.
 

Offline DimitriP

  • Super Contributor
  • ***
  • Posts: 1360
  • Country: us
  • "Best practices" are best not practiced.© Dimitri
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #388 on: June 01, 2024, 07:13:23 pm »
Quote
LT engineers, we miss you.  ADI egg on face continues.

Nice summary of a long thread!  :-+
   If three 100  Ohm resistors are connected in parallel, and in series with a 200 Ohm resistor, how many resistors do you have? 
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14391
  • Country: de
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #389 on: June 01, 2024, 08:30:43 pm »
It may be possible to have an amplifier with a current type output. In this case the capacitance at the output could do the jop of setting the BW and doing the compensation. So it depends on the details of the buffer amplifier, if the capacitive load is OK and possibly even needed.

Some of the lower noise SD and SAR type ADCs need a very stiff reference all the way to high frequencies (e.g. 1-100 MHz) and thus essentially the capacitance in the 10 µF range. So it makes some sense to have a driver that can drive such a load. With this capacitance the ESR can however be important. 10 or 100 µF X7R are different from foil or tantalum or polymer or a mix. I would have expected more information on what ESR / loss is wanted / needed.

p.s. : For comparison the Ti Ref6025 as a ref. with ADC driver wants 10-47 µF capacitance at the output.
« Last Edit: June 01, 2024, 08:33:16 pm by Kleinstein »
 

Offline dietert1

  • Super Contributor
  • ***
  • Posts: 2193
  • Country: br
    • CADT Homepage
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #390 on: June 01, 2024, 09:05:25 pm »
Actually i meant an output stage as in the schematic. It can be stable under resistive and capacitive load without change in pinout. Except i have no idea what it takes to have a low leakage 1nF cap on chip.
Of course a "ADR1002" solution with separate force and sense pins would be superior.

Regards, Dieter
« Last Edit: June 02, 2024, 08:22:34 am by dietert1 »
 

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7140
  • Country: ca
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #391 on: June 02, 2024, 03:11:55 am »
I thought current-output amplifiers require different compensation (configuration) so the CF 100nF cap would be in the wrong place if that is the case?
You could be right - "The ADR1001 output buffer requires a minimum 0.2μF of load capacitance placed near BUF_F for stability."
We don't know what macrocells are used for the op-amps. I don't speak "die-shot" but the number of (buf amp) transistors looks sparse compared to say half an LT1013.
Datasheet unit-step response graphs Fig. 21-23 are +/-500uA and it's fairly slow and ringy.

Unless we can free up a pin, or find some more real estate for a comp cap... ADR1002 could be a challenge  :)
« Last Edit: June 02, 2024, 03:32:13 am by floobydust »
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14391
  • Country: de
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #392 on: June 02, 2024, 06:36:50 am »
The response curves look quite good. It is ringing only with a really small capacitor. Otherwise it is a single pulse and settling. So the buffer really looks like it is made to drive a capacitive load.

A point could be that it could be hard to add an in loop buffer to keep the heat away from the ref. chip.  Still one is not forced to use the ref. internal buffer and use a complete exernal buffer.
 

Offline iMoTopic starter

  • Super Contributor
  • ***
  • Posts: 4882
  • Country: vc
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #393 on: June 02, 2024, 07:56:57 am »
The label ADR1002 has been reserved on December 13th, 2022 by imo for a change in pinout - in order to have separate pins for output drive and sense (divider input). As of now the part cannot drive a meter without additional output stage..

Yep, my "ADR1002" (C) (TM) - You may save a package pin when you connect the output buffer with the first 7V->5V divider directly (on chip). Then do use the saved pin for the 2/1 output divider decoupling..
« Last Edit: June 02, 2024, 08:12:11 am by iMo »
 

Offline Noopy

  • Super Contributor
  • ***
  • Posts: 1809
  • Country: de
    • Richis-Lab
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #394 on: June 12, 2024, 07:50:39 pm »
Now let´s finally look into the series production ADR1001!

If you haven´t already seen my LTZ1000A update you should read that before going on:
https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg5540135/#msg5540135







This ADR1001 was produced in calenderweek 7 in 2024.




The electrical connection of the package lid has not changed compared to the engineering sample. It is still connected to the inner housing base, but it is otherwise insulated.






The internal structures are very similar to the ADR1001 engineering sample.




Like in the ADR1001 we see this glass bead polymer mixture.




The diameter of the glass beads is approximately 0,1 mm. In contrast to the material used in the LTZ1000A, only very similar diameters and no significantly smaller particles can be found here. There is one layer of spheres under the die. The gap is therefore approximately 0,1mm. In the ADR1001 engineering sample the layer appears to be somewhat thicker.




Although some of the glass spheres contain air bubbles, they appear to be basically solid. They are probably the same spheres that are described as new material in PCN 23_0011 of the LTZ1000A. These would then be the glass spheres from COSPHERIC with diameters between 90µm and 106µm.




As with the LTZ1000A, there is also a small area that has been left open.




Removing the die reveals that the die attach material has just been applied to the outer areas of the die. A cavity in the middle provides better thermal insulation. As shown with the LTZ1000A, the cavity is absolutely necessary as the glass beads still conduct heat relatively well. In contrast to the LTZ1000A, however, the cavity here is relatively small.

In this context it is interesting that the engineering sample of the ADR1001 has no opening at the edges of the die. Either the material has been applied over the entire surface or there is a closed volume under the die. It is at least conceivable that a sealed volume could have a negative effect on the specification of the reference. On the one hand, substances could remain in this volume that are normally flushed out at the end of production. In addition, temperature changes result in pressure differences in relation to the rest of the package, which in turn generate mechanical stress in the die.




It is now clearly visible that there is exactly one layer of glass beads under the die.




Except for one point, the die has the same structures as the die in the engineering sample. This image is also available in higher resolution: https://www.richis-lab.de/images/REF01/50x10XL.jpg (44MB)






We have talked about the parasitic resistor in the output of the output amplififer. The voltage drop across this resistor changes with load changes. As the voltage drop is outside the control loop, it is not compensated.




Obviously, the output resistance has also been recognized at Analog Devices. It would have been ideal to resolve the node at the output of the opamp so that the feedback resistors could be contacted externally. However, this would have meant changing the pinout. Instead, the output stage is now connected to the bondpad via a relatively large metal surface. The connection to the feedback resistors is made via a long contact between the two metal layers. This contact is located as close as possible to the bondpad.

The long contact of the feedback resistors would not have been necessary, as not much current flows through this path. In fact, it would have been better to choose a small contact and bring it even closer to the bondpad. However, this path is no longer critical. Its resistance should be 15-30mΩ. The bondwire alone adds a resistance of 140mΩ. In addition, there are contact resistances at the bondpad and at the contact pad of the housing.


https://www.richis-lab.de/REF45.htm

 :-/O

Offline iMoTopic starter

  • Super Contributor
  • ***
  • Posts: 4882
  • Country: vc
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #395 on: June 12, 2024, 09:42:52 pm »
Thus the EEVBLOG contributed to the ADR1001 improvement  :D  :D

The "1R7" fix will improve the voltage drop with 10V output (say 6-7 4-5x smaller against the engineering sample). When using an external opamp buffer it does not matter which version is used. Let us wait on my ADR1002 with the "INV0" output divider fix :)

@Noopy: have you made a picture diff to see all changes?
« Last Edit: June 13, 2024, 07:18:27 am by iMo »
 
The following users thanked this post: SeanB

Offline Noopy

  • Super Contributor
  • ***
  • Posts: 1809
  • Country: de
    • Richis-Lab
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #396 on: June 13, 2024, 04:00:53 am »
Would really be interesting to ask one of the designers if they are reading EEVblog.  ;D

To check for changes I switched the pictures back and forth. That works quite well if the pictures are good enough and you take your time to do it.

Offline iMoTopic starter

  • Super Contributor
  • ***
  • Posts: 4882
  • Country: vc
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #397 on: June 13, 2024, 06:22:23 am »
They do, be sure..  ;)

BTW: TiN is starting his experiments as well..

https://xdevs.com/pow/adr1k1_pow/
« Last Edit: June 13, 2024, 06:56:20 am by iMo »
 
The following users thanked this post: Dr. Frank, Noopy

Offline magic

  • Super Contributor
  • ***
  • Posts: 6877
  • Country: pl
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #398 on: June 13, 2024, 09:41:28 am »
Obviously, the output resistance has also been recognized at Analog Devices. It would have been ideal to resolve the node at the output of the opamp so that the feedback resistors could be contacted externally. However, this would have meant changing the pinout. Instead, the output stage is now connected to the bondpad via a relatively large metal surface. The connection to the feedback resistors is made via a long contact between the two metal layers. This contact is located as close as possible to the bondpad.
There is still one square worth of resistance outside the loop in addition to the whole bondwire.

Most references have no force/sense pins, but what they do is run separate force/sense bondwires to the output pin. Several types like LT1027 or MAX6250 are able to guarantee <10mΩ output resistance. AD's own AD587 uses this technique and guarantees <100mΩ.

The ADR1001 doesn't even have a load regulation spec in 10V mode...
 

Offline Noopy

  • Super Contributor
  • ***
  • Posts: 1809
  • Country: de
    • Richis-Lab
Re: ADR1001 - Ovenized Voltage Reference System
« Reply #399 on: June 13, 2024, 09:56:39 am »
Obviously, the output resistance has also been recognized at Analog Devices. It would have been ideal to resolve the node at the output of the opamp so that the feedback resistors could be contacted externally. However, this would have meant changing the pinout. Instead, the output stage is now connected to the bondpad via a relatively large metal surface. The connection to the feedback resistors is made via a long contact between the two metal layers. This contact is located as close as possible to the bondpad.
There is still one square worth of resistance outside the loop in addition to the whole bondwire.

Most references have no force/sense pins, but what they do is run separate force/sense bondwires to the output pin. Several types like LT1027 or MAX6250 are able to guarantee <10mΩ output resistance. AD's own AD587 uses this technique and guarantees <100mΩ.

The ADR1001 doesn't even have a load regulation spec in 10V mode...

You are right, two bondpads and bondwires to the output pin/pad would have been a small impact solution for the resistance problem.


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf