If you want to see inside an integrated circuit (IC), you generally have to take the die out of the package, which can be technically challenging and often destroys the device. Looking to improve the situation, [Bunnie] has been working on Infra-Red, In Situ (IRIS) inspection of silicon devices. The technique relies on the fact that newer packages expose the backside of the silicon die and that silicon is invisible to IR light. The IR reflects off the bottom metalization layer and you can get a pretty good idea of what’s going on inside the chip, under the right circumstances.
As you might expect, the resolution isn’t what you’d get from, say, a scanning electron microscope or other techniques. However, using IR is reasonably cheap and doesn’t require removal from the PCB. That means you can image exactly the part that is in the device, without removing it. Of course, you need an IR-sensitive camera, which is about any camera these days if you remove the IR filter from it. You also need an IR source which isn’t very hard to do these days, either.
Do you need the capability to peer inside your ICs? You might not. But if you do and you can live with the limitations of this method, it would be a very inexpensive way to get a glimpse behind the curtain.
If you want to try the old-fashioned way, we can help. Just don’t expect to be as good as [Ken] at doing it right away.
This would be great for detecting fake chips. Package it together to make it portable, and bring it on your shopping trips at the SEG.
Unfortunately this only works for flip-chips without encapsulation or lidding. This won’t help you weed out clone STM32s unless you’re interested in CSP footprint parts.
It’s kind of neat this works with ‘normal’ silicon cameras. It’s operating in a tiny slice of spectrum around 1050nm, where the silicon in the object is transparent enough to see through, but the silicon in the camera can still absorb enough to detect the photons.
If it were the silicon you were interested in seeing then you’d need a different wavelength of light, fortunately the silicon is pretty uninteresting and the bits we do want to see aren’t silicon.
Ditto with an image sensor, the silicon isn’t doing the heavy lifting, it’s the structures foremd, deposited, etched into it that are doing the heavy lifting of light sensing
Sez CJay: an image sensor, the silicon isn’t doing the heavy lifting
Uh, no. The silicon in an image sensor is ABSOLUTELY doing the ‘heavy lifting’. That’s where the photons get absorbed and the requisite electrons get liberated to get detected. The structures created in the silicon just store the electron charge to be read out, and then do the charge-to-voltage conversion that is the image signal.
Make no mistake, it’s the silicon doing the detecting. Which is why a silicon-based camera is blind at wavelengths longer than 1100 nm.
Very impressive hack.
I wonder if stacking or super resolution technic of some sort could use multiple images of low resolution to enfance the resolution ?
Not likely.. the low quality you see is probably due to poor optical performance of the imaging system at the wavelength of 1 micron. Remember that most consumer imaging systems are designed for the visible spectrum (if you’re on a budget at least). If you change the illumination to be outside this range then the optical performance will suffer. Imaging at 1 micron is possible, but to achieve better performance requires a re-design of the optics (different elements, spacing, and materials).
Many “regular” lenses don’t work that well at 1000 nm, but some do work well. The Raspberry Pi HQ 6mm C (or CS?) lens is sharp in the center, the new RPi Camera Module 3 NoIR lens is very good, and I’ve also had good luck with some Rokinon/Samyang mirror less camera lenses.
https://github.com/williamhobbs/PVRW-2023-EL
And isn’t there a fundamental resolution limit of about 1um if using 1um light?
This immediately reminded me of a very old sci-fi movie: “The Man With X Ray Eyes”.
Fantastic film, Ray Milland was brilliant in it.
If your fake IC deceives you, pluck it out!
Is it safe to assume that this IR is a wavelength that can’t affect the gates of the transistors? The potential to affect the operation of an encapsulated device with a strong enough light source sounds like a problem (sort of like that RasPi 2 U16 chip issue from ages ago).