What Destructive Chip Inspection Can Tell You

Opening a chip gives you access to its internals. It can also destroy the best opportunity to understand what was wrong with the intact device. If there's only one suspect part, the order of the tests matters as much as the equipment in the lab.

That tradeoff gets lost in discussions of hardware implants. An image of an exposed die looks conclusive, but someone still has to explain what they're looking at, how it differs from a legitimate part, and whether the difference accounts for the behavior that prompted the inspection.

Before opening the package

Consider a replacement component whose markings differ from the rest of an approved lot. There are several possible explanations: a legitimate manufacturing change, a substituted part, a remarked device, or something that needs a closer look. The marking alone doesn't distinguish them.

I'd preserve the packaging, supplier records, photographs, and handling history first. Then I'd ask the manufacturer about the marking and establish whether the part behaves as expected electrically. If it fails only under particular conditions, record those conditions while the failure can still be reproduced. Removing the package may change that behavior.

Imaging can answer some questions without opening the part. X-ray inspection can reveal internal assembly differences; acoustic microscopy can help locate package defects. Texas Instruments puts these techniques before destructive preparation in its failure-analysis process. The sequence preserves information that later steps may remove.

For the replacement-part example, the first useful result might be quite ordinary: the internal assembly differs from a known reference, or the electrical response doesn't match the specified device. Either result gives the lab a more focused reason to continue.

What decapsulation exposes

Decapsulation removes package material so the die or internal assembly can be inspected. The preparation depends on the package construction and the feature the analyst needs to preserve. Removing material can damage bond wires, alter a failure site, or leave an area inaccessible. Before proceeding, the lab needs to know which measurements still depend on an intact device.

An optical image may reveal a die marking or an assembly anomaly. Examining a particular circuit feature can require higher-resolution imaging, electrical probing, or removal of additional layers. Each step narrows the area under examination and changes what remains available for later analysis.

This is why I would be skeptical of a quote for a generic “implant check.” Looking for the wrong die in a package is a different assignment from locating a small change within an otherwise plausible design. The latter may require design information and a comparison sample that the customer doesn't have. Preparation and microscope time are only part of the cost.

A difference needs an explanation

Suppose the exposed die has a different marking from the reference. That is a concrete observation worth documenting. The next question is whether the reference represents the same revision and manufacturing source. Comparing the suspect device with an unrelated revision could turn an ordinary design change into an apparent anomaly.

The same problem applies to an unexpected connection or layout feature. A lab needs enough context to distinguish a production variation, damage, repair, substitution, and deliberate modification. Even after a modification is established, attributing intent requires more than its appearance.

I want the report to let another analyst follow that reasoning. It should identify the sample and reference, show the relevant images at usable resolution, and describe what was removed or damaged during preparation. “The die marking differs from the supplied part description” tells me something I can act on. A broad assurance about the component's integrity would need much wider support.

A clean image has a similarly limited reach. It describes the structures the lab could inspect at that resolution. Buried layers, firmware behavior, and circuits activated only under particular conditions may need other methods. One inspected component also says little about the rest of a shipment unless the sampling plan accounts for how the parts could differ.

Spending the sample carefully

Before shipping a scarce part, I'd agree with the lab on the initial measurements, the reference material, and a stopping point before destructive work. I'd also want to know whether remnants and intermediate images will be retained, and what happens if preparation damages the area of interest.

A second representative sample can leave more options open, but it doesn't solve a missing-reference problem. If nobody can explain which differences would be legitimate, a more detailed image may still leave the original question unanswered. Resolve that problem before sacrificing the only device that reproduces the failure.