(no subject)

[randomdreams]

Apparently my style of decapsulating integrated circuits is so different we don't know how to take pictures of the results that show what we want to see. We have damaged chips. When we etch the top off with nitric acid there are spots that won't etch because they're a mixture of metals, epoxy, and silicon. The way I remove the top doesn't leave those spots, so we have to compare the pictures to pictures of a good chip and even then the evidence for damage is subtle: lines that aren't quite straight, for instance. I may have to come up with some way of producing contrast. But, generally, it's extremely successful save for our lack of ability to electrically connect to the die anymore, and we can even manage that with our probe station.

Comments

[twoeleven]

I've got some ideas. How paranoid are you about suggestions and proprietary information?

[randomdreams]

Not at all: this is entirely internal analysis, not even (in this case) of a competitor's product.

[twoeleven]

How about optical effects when photons interact with objects their own scale, like diffraction or interference? Two lines that ought to be parallel ought to be a dandy slit for diffraction of the right wavelengths. But since I'm not sure what feature sizes you're looking at -- you're examining the chips under a light microscope? -- I'm not sure if that's reasonable.

[sistawendy]

Yay, right?

[acelightning]

How about removing the top your way, then chemically etching it down to within a few atoms of the circuitry?

[randomdreams]

We talked about that. Coworkers are pretty much all "dude that stuff is NASTY". They don't get that I have a degree in this kinda shenanigans. (I think hot acetone might be enough to remove 0.05mm of epoxy.)

[acelightning]

As long as you have the equipment for handling hot acetone, I don't see a problem...

[secretagentmoof]

I've always been vaguely curious as to why, when doing chip dev, they didn't have an easily-dissolvable packaging or clear resin - although I suppose that might be orders of magnitude more expensive than getting some lucky person to figure out how to do clever things with a vise.

[randomdreams]

We used to: a hand-encapsulation technique, colloquially called scoop-and-goop. We used it mostly because we could get it two weeks earlier than the first packaged parts, but what we ran into was that it didn't get us much information we couldn't get from taking a wafer and putting probes on the die pads and doing tests that way, and the information it did give us was questionable because the packaging material was different. We're generally good enough at design that what we run into are subtle problems involving how our packaging, leftover chemical residue, and the bond wires interact at elevated temperature/pressure/humidity involved with thermal cycling tests and rapid aging simulations.

[darkoshi]

Is the epoxy the substance that looks like black plastic on most chips?

Are there specialized x-ray machines for chips that can show what is under the epoxy, without having to remove it?

On a related note, found this recent article:
http://spectrum.ieee.org/nanoclast/semiconductors/processors/xray-ic-imaging