LM399 - die analysis

[Noopy]

Hi all,


I have took a closer look on a LM399 shunt reference.



Actually it´s a HP-part...


You can find the whole story here:
https://www.richis-lab.de/REF02.htm
(german)




A nice die with big structures! 


Reverse engineering is not a big deal:




And here is the buried zener:




I found a circuit that´s not in the datasheet:



It seems they had to do some fancy compensation. 


And here the heater-circuit:




I wrote a lot about the circuit but it´s too late to translate all that words to english. Please use the google translator and ask me whatever you want. 


BTW:
I also have a PMI REF01:
https://www.richis-lab.de/REF01.htm



Interesting point: The new ones are ADR01 named REF01.


Have fun! 


…does anybody want to donate me a (dead) LTZ1000? 
Update: Hey, it seem´s that I´ll get one out of the german mikrocontroller.net-Forum! 

[iMo]

Bei der Z-Diode D1 handelt es sich anscheinend um die gleiche Subsurface-Diode wie im Shuntregler.

Yes, the Dobkin's paper confirms that.
2 burried zeners there.

PS: I can send you a MAC199 (TESLA's highest grade LM399 clone) for de-capping if interested.

PPS: TESLA's schematics show the missing transistor and resistors, there is even one additional transistor more

[rhb]

What is a 1991 die doing in a 1985 date code package?  Where did the device come from?

[Squiddaddy]

Looks like 199 I to me, not 1991.

[chickenHeadKnob]

What is a 1991 die doing in a 1985 date code package?  Where did the device come from?

The original part was LM 199

[Noopy]

PS: I can send you a MAC199 (TESLA's highest grade LM399 clone) for de-capping if interested.

PPS: TESLA's schematics show the missing transistor and resistors, there is even one additional transistor more

A MAC199 would be very interesting! I´ll write you a PN.
And thanks for the hint to the schematic of the MAC199 and now I see the Double-Transistor. Will correct that on my homepage! 


And regarding the "datecode-problem":
Thanks for the hint. Didn´t realize that! 

[Gyro]

You can find die photos of the MAC199 in this thread:  https://www.eevblog.com/forum/metrology/tesla-mac199mae299mab399-and-mac01-teardown-and-tests/

[Noopy]

PPS: TESLA's schematics show the missing transistor and resistors, there is even one additional transistor more

I´m not sure whether that´s really a transistor:



The bigger thing above the resistor-lines.

In my view that is still a resistor.
Well I´m not sure but I see nothing that indicates a transistor.
Otherwise it would interact with the other resistors placed in the same well.

I´ll try to take a better shot of the area… 

[Cerebus]

To save scrabbling about, here is the die photo of the Tesla chip (a MAB399) from the thread mentioned above:



Note that the image links to a higher resolution version of the same image; so click on it if you want more detail.

[Noopy]

It´s high resolution but not as high as I am able to get it. 
It seems to have some different structures but without polarised light it´s difficult to clear that up.

[Noopy]

I´ll try to take a better shot of the area… 

No, I don´t think that´s a Transistor...

[Wimberleytech]

I´ll try to take a better shot of the area… 

No, I don´t think that´s a Transistor...

Looks like a resistor to me

[Noopy]

Hi all!


Thanks to imo I was able to take some pictures of a MAC199:

https://www.richis-lab.de/REF02a.htm




The structures are very similar but not exactly the same.




Tesla had to do quite some mask revisions.




In my view like in the LM399 there is no second additional transistor as described in the tesla datasheet.

The schematic looks like this:



The grey line is disconnected in the MAC199. So whatever the lower transistor compensated it´s not functional in the MAC199.


Another interesting point is the die attach. The die is not directly bonded to the package. It sits on top of something which looks like a thinner die:



...the LM399 has the same "double-die"...

 

[Cerebus]

I can't find anything specific to the LM199/299/399 but it's not uncommon for voltage reference chips to have unconventional die attachments. Differential temperature related movement of the case and die, owing to the different temperate coefficients of expansion of the disparate materials, induces varying strains in the substrate causing drift. So some kind of 'floating' or compliant attachment makes sense for high precision references.

[Noopy]

Do you remember me saying that in the LM399 there is a additional circuit similar that shown in the datasheet of the MAC199?
I claimed one of the two additional transistors isn´t actually there. I was pretty sure that it´s a resistor:




Well some days later: It could be a transistor!




Probably the brown "resistors" are p-doped. If there is an additional n-doped area under the upper contact than we have a pn-structure. With the n-doped well you get an npn-transistor.
Fine! 
If you don´t want that every structure in the well interacts with the half transistor the well has to be more positive than everything else. The most positive potential is the red one coming from left into the picture. But it seem´s that the well is connected to the second yellow contact of the resistor. Since the role of the transistor is a shorting of the resistor it´s ok to connect the well to the second highest potential, the second contact of the resistor.
Fine! 
It´s more a parasitic transistor and possibly not a very good one but it should work.

 

[David Hess]

Another interesting point is the die attach. The die is not directly bonded to the package. It sits on top of something which looks like a thinner die:

...

...the LM399 has the same "double-die"...

I suspect that was added for strain relief.  All precision analog ICs have extra packaging requirements to prevent straining the semiconductor die.

I remember an interview where I think it was Bob Dobkin who mentioned a National executive who kept complaining about the extra packaging cost of analog ICs and why couldn't they use the less expensive digital IC packaging which lacked an extra encapsulation which lowered strain.  Then later someone actually ordered that change and suddenly all of their linear ICs were failing testing and they had nothing to sell.

I found a circuit that´s not in the datasheet:

...

It seems they had to do some fancy compensation.

Schematics often leave out circuits for things like transconductance reduction like common LM324 and LM358 schematics do.

[Noopy]

Another interesting point is the die attach. The die is not directly bonded to the package. It sits on top of something which looks like a thinner die

I suspect that was added for strain relief.  All precision analog ICs have extra packaging requirements to prevent straining the semiconductor die.

Often heard about this.
It´s interesting that the "better" LTZ1000 doesn´t have a "double-die":

Richis-Lab - LTZ1000

[Cerebus]

Another interesting point is the die attach. The die is not directly bonded to the package. It sits on top of something which looks like a thinner die

I suspect that was added for strain relief.  All precision analog ICs have extra packaging requirements to prevent straining the semiconductor die.

Often heard about this.
It´s interesting that the "better" LTZ1000 doesn´t have a "double-die":

Richis-Lab - LTZ1000

The LTZ1000A has its own proprietary die attach. This differs from the non-A version, the A version being engineered for lower thermal conductivity. It's mentioned on the first page of the data sheet.

[Noopy]

The LTZ1000A has its own proprietary die attach. This differs from the non-A version, the A version being engineered for lower thermal conductivity. It's mentioned on the first page of the data sheet.

I know the LTZ1000A has a special die attach but the datasheet descripes only special thermal properties.
Of course it´s possible that it has additional special features...