Gregory's ServerI'm doing a more thorough reverse engineering of the power PCB from the Thinkpad 700 series. lots of work so far and there are still more parts to remove on the other side of the board!
 52 comments
@tubetime Are you sure you're not just doing this to show off your mad desoldering skills? 😀 that's both sides. phew, lots of work! the next step is to photograph the top and bottom sides of the PCB. I really ought to get a scanner. 😅 @maehw it's a mix. there are some that I had to assign myself (the ones starting from 300)  I could just use a regular scanner, but this is a good excuse to try out this prototype Labsmore.com microscope that they loaned to me for free. (thanks guys!) i'm using Fiji/ImageJ to do the stitching. basically you go plugins->stitching->grid/collection stitching. then you pick "Filename defined by position" and then use c{xxx}_r{yyy}.jpg" for the file names parameter. and here we are in Kicad 7 using the image overlay feature. it was pretty easy to get it going but the program slows down quite a bit for large images. you'll want to scale them down. pro tip: add a scale factor in GIMP before exporting so Kicad gets the dimensions right.  got all the parts placed and many of the traces. i will need to create custom footprints; the originals are much more compact than the Kicad defaults i've been using. down to the last few traces. i'm using a multimeter to ohm out and discover what they are connected to. you can also see the faint outline of traces on layer 2 (the board has four layers). i'll need to route those next. it's actually going pretty well despite me not being able to see where the traces go exactly. since i have net connectivity it helps guide me. layer 2 is basically done. i've even got some copper pours! all this from staring *through* the top copper layer. layer 3 was sorta hard and easy. it is mostly a ground plane, but there are a few other copper pours. i've got 100% connected, but there are some mystery pours that are just wired in parallel with pours on the top or bottom layer.  not satisfied with layer 3, so i am going back over it with a new technique: on a user defined layer, i'm drawing lines ONLY where i see a visible plane cut in that layer. then i'm adding "S" on top of every via that has spoked connections to the copper on that layer. ahh that makes more sense. there are a few higher current connections, but more importantly, there are cuts in the ground plane! this creates a star ground at the large highlighted pad near the middle. this is the negative battery terminal (but after the current sense resistor) anyway, i could fix up a few footprints but this reverse engineering job is essentially completed. you can find the design files here: https://github.com/schlae/Thinkpad700CPower i haven't reverse engineered many 4-layer boards, but i'm kinda amazed it worked out this well. @tubetime Amazing job, would love to see a video on how exactly you go about it. reassembled like a magic trick. the layout CAD was quite helpful for finding footprint locations.  @tubetime That's amazing. What's the quote? Ah, yes. "Sometimes, magic is just someone spending more time on something than anyone else might reasonably expect." -- Teller @tubetime Wow. How in the world do you figure out the middle layers? It’s a good thing this one was only four - I think some modern PCBs are 12+ layers. The aim is not to be able to re-make a PCB, but just to be able to debug one? @gogobonobo the thread explains how i did it. higher layer count boards require more invasive methods. but yes i wanted to be able to debug boards corroded by a leaking capacitor. @tubetime How did you get to these middle layers? I mean, you can measure some connections, but not get the layout exactly, unless you open it or have some 3d x-rays or something. @http i have a schematic (that i ohmed out) already, and i can see *part* of the traces on that layer. that and a process of logical deduction led me to an approximation of the original layer. @tubetime does kicad have any way to see multiple layers side by side? Sometimes when I have components on both sides that overlap it gets difficult even when I’m changing layer visibility.  @tubetime Ohming it out on that scale is impressive. I just did the same on a board with just 20 dip ics and even that was a long task. Well, I guess it didn't help that the ohming was done remotely by a friend who was a few hundred miles away and had borrowed the card to clone while i was doing the KiCad part :) @tubetime It can't produce a schematic from the traced PCB, right? That's the kind of thing that could really cut down on the time needed to RE a board... @phaseseeker i'm starting with a schematic i already ohmed out the hard way. i'm not sure about the workflow going the other way around, i'd have to think about that. probably i'd place all the footprints first and then figure out connectivity.   @yusufm yes, i'm very interested to see how the focus stacking works out once i get to that point. |
@tubetime That’s dedication! Hard to do without damage. Good work!