Automated PCB reverse engineering idea

[Rjc987]

Hey guys,
One piece of test equipment I've never seen is an automated PCB reverse engineering solution - something that could take a PCB and either through imagery or direct measurement, output a general schematic of what connects to what. there's been times when I've been very busy and get in a piece of equipment that has a convoluted board which I don't have any schematics or information about and which something like that would be a huge timesaver and a massive help when reverse engineering or diagnosing.

So I was wondering why and thinking about ways it could be done - the first idea I had was using simple pcb imagery and machine learning but the fact that traces are hidden underneath ic's and inner layers, plus perspective and lighting means bigger components are going to be blocking things, and glossy black or white solder masks make it almost impossible to get a clear image of traces etc. I feel like that might not work very well.

So how about this- just a simple, very dense bed of nails style fixture completely filled top and bottom, Connected in some matrix array allowing a continuity signal to be applied sequentially from pin to pin and record every other pin which has continuity or a resistance value. This could then be processed into a boardview type file using images of the boards with each node selectable to display all related connections. It obviously won't work with bga's etc, and it would probably work best for low to medium density PCBs but am I missing anything here? It doesn't seem like it would be that difficult of a project to me, I imagine the hardest part would be the coding. Any thoughts? Suggestions?

[pcprogrammer]

I see problems with the bed of nails and component height, and not just BGA's that pose a problem. QFN packages also have the pads underneath, and there are other components with the connections underneath the component, like some coils, LED's, etc.

Even with pogo pins there can be components in the way for reaching in deep enough to make contact. Flying probes would be a better solution, but it would need a robot arm with good articulation and a long needle to reach in to every nook and granny, and that still leaves BGA's

An X-ray image of the board can be of help, but with multi layer boards it still is a challenge.

Further more it would need automated recognition of components, and that can be difficult too, with the same SMD marking on different components, so knowing the package is also needed, and then still.

Distinguishing a SMD ferrite bead from a SMD capacitor can also be a challenge, and components with very low resistance can mislead the continuity test and make it think there is a direct connection.

Distilling a proper schematic is also not that easy, even with a full node to node list.

So all in al not an easy feat to accomplish.