Ultra Precision Reference LTZ1000

[dietert1]

Isn't the broken one an invitation to break others in order to confirm they are hollow? I mean the device is already ruined anyway.

[MiDi]

Those are Soda Lime Solid Glass Microspheres - SLGMS-2.5 90-106 for $133.37.
Thermal conductivity: ~1 W/mK
Specific heat: ~0.9 kJ/kgK

[KT88]

The density of the cospheric sphered is stated as 2.5g/cc which is the density of (solid) glass.

[Noopy]

Isn't the broken one an invitation to break others in order to confirm they are hollow? I mean the device is already ruined anyway.

Scratching over the material didn't really break bubbles. But I was not sure if I had put enough force into the material.

Those are Soda Lime Solid Glass Microspheres - SLGMS-2.5 90-106 for $133.37.
Thermal conductivity: ~1 W/mK
Specific heat: ~0.9 kJ/kgK

The density of the cospheric sphered is stated as 2.5g/cc which is the density of (solid) glass.

Thank you, now that is clear. 

[floobydust]

Does anyone know the history of die-attach compounds with the glass/ceramic balls/spheres? It looks like a Linear Technology Secret Sauce?
Not seen on any older heated-die references like LM399 I think.

Perusing patents on this for more information, I found a little:
"Die structure using microspheres as a stress buffer for integrated circuit prototypes" Clear Logic {1998} US 6,020,648
Linear Technology Patent US 4,888,634 {1989} (expired)  "... improving the uniformity of operating temperature of packaged integrated circuit chips." Carl Nelson not mentioned in a relevant (die-attach) patent.

I think any voids are not a good sign. "Voids are the small areas in the Die Attach area that are filled with a mixture of gases and other environmental variables, such as, air, moisture and dust. After the die placement, there could be a significant amount of waiting time before the subsequent curing process. It can result in a high level of moisture absorption in the Die Attach pastes. While the organic substrates may absorb moisture, moisture may also be present on the metal leadframe surface.  As the temperature increases during curing, absorbed moisture will evolve as steam to cause a void."
https://oricus-semicon.com/common-issues-of-die-attach

What happens when you buy or copy high technology or outsource it- the Secret Sauce gets bungled. I've seen it happen 1000x.
I fully expect ADI to have problems transplanting fab and packaging. They probably haven't even noticed, laid off all the important LT staff that would.

[MiDi]

ADR1001 seems to use glass microspheres, ADR1000 not.

from Richi´s Lab

[Noopy]

I have taken a closer look at the LTZ1000A die attach and some other similar die attach materials:

LTZ1000: https://www.richis-lab.de/REF03.htm
LTZ1000A: https://www.richis-lab.de/REF44.htm
ADR1000: https://www.richis-lab.de/REF19.htm
ADR1001 Engineering Sample: https://www.richis-lab.de/REF29.htm
ADR1001: https://www.richis-lab.de/REF45.htm
LT1088: https://www.richis-lab.de/LT1088.htm




As we have seen the beads in the LTZ1000A have quite different diameters.




The PCN shows that the new beads are solid glass. We don´t know much about the old material we see in this LTZ1000A.




Now let´s take a look at the specifications and the different die attach materials.

The LTZ1000 is specified with a thermal resistance of 80K/W. In the LTZ1000A, it was possible to increase the thermal resistance to 400K/W. The mass used here as the attachment must therefore have an additional thermal resistance of 320K/W. With the known dimensions and the estimated thickness of the die attach material, it can be calculated that the thermal conductivity of the die attach must not be higher than 0,08W/Km.

The table below shows the thermal conductivities of various materials. Epoxy is in the range of 0,14W/Km. There are epoxy mixtures that offer a higher value, but it is not easy to reduce the thermal conductivity. This explains why the ADR1000 has a slightly lower thermal resistance. The newer versions of the LTZ1000A contain solid glass beads from Cospheric. This type of glass has a thermal conductivity of 1,46W/Km. Even though these are spheres and not a solid block, the conductivity is initially far too high to achieve the thermal resistance of 400K/W.

Air conducts heat very poorly. This explains how the high thermal resistance of the LT1088 could be achieved. Assuming that the thermal resistance of the package is similar to that of the LTZ1000, the additional measure must have a thermal conductivity of 0,08W/Km to 0,22W/Km. The foam shown in the LT1088 can represent this value. Hollow glass spheres, such as those produced by 3M, are another way of achieving such low thermal conductivity.




As we have seen at the edge there is a hole in the die attach material.




Removing the die reveals that the special material was only applied to the outer areas. This means that the necessary thermal resistance can also be achieved with solid glass spheres. A large part of the area is insulated by the air cushion.

The cavity probably also explains the opening on the left edge. It enables gas exchange, which could be necessary during production. In addition, a sealed gas volume under the die could be problematic in the event of temperature changes. Mechanical stresses caused by pressure differences can have a noticeable effect on such an accurate reference stress source.




A few glass spheres are broken open and expose a cavity. In this picture there is such a sphere at the top left.




There is a glass sphere under the die, where it is clearly visible that hollow spheres were used. This is surprising. The aforementioned PCN describes a conversion to solid glass spheres, but does not indicate that the previous material contained hollow spheres. Even if the material is only applied to the edge area, it can be assumed that the thermal conductivity does change somewhat.

When using hollow spheres, it should be possible to apply the material over the entire area. This would reduce the risk of undefined contact surfaces leading to mechanical stresses in the die. Such stresses, which may be temperature-dependent, can have a problematic effect on a high-precision reference. In general, the question must be asked whether the advantage of the higher thermal resistance compensates for the potential disadvantages of a somewhat undefined mounting of the die. Just my two cents...


https://www.richis-lab.de/REF44.htm

 

[dietert1]

For me these pictures look like the die attach is handmade and an ad-hoc solution. Wouldn't an expensive reference chip deserve something better, e.g. a clean, machined support made of teflon or a metal clamp with three fingers?
Also i am wondering about the ADR1399, as it runs pretty hot. Won't the epoxy outgas into the hermetic enclosure each time the heater turns on?

Regards, Dieter

[iMo]

The "opening on the left edge" is definitely a random stuff. Nothing to do with intentionally made hole for escaping the gases, imho.
Or, perhaps, an insect (bug) made it while trying to escape from his cave during the initial burn-in process 

[Noopy]

The "opening on the left edge" is definitely a random stuff. Nothing to do with intentionally made hole for escaping the gases, imho.
Or, perhaps, an insect (bug) made it while trying to escape from his cave during the initial burn-in process 

I´m not sure about that. We have seen such openings twice. Just the engineering sample of the ADR1001 doesn´t have it. In my view such an opening would make sense. A closed volume under the die doesn´t sound very good.

Perhaps the opening emerges as a byproduct in a final process step before closing the package. Perhaps there is some vacuum before filling the package with dry nitrogen or something like that. If the polymer is not cured completely in this process step the opening can occur by itself.
Of course I´m just speculating...

[iMo]

Frankly, I really doubt that is something of an intentional nature. They would do it in a different way if planned. It could be after they apply the caviar, put the die on it and cure it somehow with an UV blink, then after bonding they solder the upper lid on, the high temperature creates pressure below the die where the residual gases may escape through the polymer where it is the weakest with a "puff"  .
The process of applying the caviar paste looks rather random to me, indeed..

[Noopy]

Perhaps you are right. Either way I don´t like the idea of an undefined gas volume under a voltage reference... 

[floobydust]

How do they attach the bonding wires without breaking or straining the die if it's lightly supported?
I thought you'd need to press down with force for wire bonding (ultrasonic weld?).

I've also read common die attach could be a B-phase epoxy, which has a second thermal curing possible.

I think automated dispensing would require a dedicated machine and nozzle for the special caviar. For an ISO9001 company, it's looking pretty hokey.

You could figure out what compound they're using.
Shin-Etsu Die Attach Material
Henkel Die attach adhesives

[Noopy]

I´m no expert for bonding but in my view looking at the whole die the force of the bond process is quite small. You have to be careful not to damage the bondpad but the crystal itself is quite strong.

Dispensing for sure isn´t easy especially this mixture of different diameters we see in this LTZ1000A.

Would be hard to find the exact die attach material.
It is interesting that they mix the compound by themself...

[iMo]

..Either way I don´t like the idea of an undefined gas volume under a voltage reference... 

Absolutely. That would require a process where the escape tunnel(s) will be well defined. While looking at the pictures I doubt it is. You would need a "spacer" which gets removed after curing the polymer in the caviar. And a subsequent test in a pressure chamber.