How would you connect a million+ analog pins? (Seriously)

[shakalnokturn]

Just for clarification, this is for analysis of fully assembled PCB's or unpopulated boards?

[ballsystemlord]

Just for clarification, this is for analysis of fully assembled PCB's or unpopulated boards?

I intend to use it on both populated (unpowered) and unpopulated boards.
Just to be clear, I know my tester will not work on populated BGA pads.

[tszaboo]

Just for clarification, this is for analysis of fully assembled PCB's or unpopulated boards?

I intend to use it on both populated (unpowered) and unpopulated boards.
Just to be clear, I know my tester will not work on populated BGA pads.

Your tester wouldn't work. Even if you can design it, you wouldn't be able to pay for it.

[shakalnokturn]

The project sounds totally unrealistic, I'm curious to hear how it works out if OP is stubborn enough...

Thermal camera's can tell you quite a bit for diagnosis, that and this topic got me thinking that it would be nice to have a tool that allows non-contact imaging of voltage levels on a powered PCB

[ballsystemlord]

The project sounds totally unrealistic, I'm curious to hear how it works out if OP is stubborn enough...

Thermal camera's can tell you quite a bit for diagnosis, that and this topic got me thinking that it would be nice to have a tool that allows non-contact imaging of voltage levels on a powered PCB

Maybe if I'm both stubborn and insane enough, I'll try it. 

[donlisms]

"Arrogance is the goa'uld's greatest weakness."

[ejeffrey]

Try making a mock up that is a  3x3 or 5x5 array of probes of the type and pitch you want.  You don't need to worry about the scanning yet, just add individual wires to each pin.  You can use some Teflon insulated blue wire.  See if you can get them to make contact they way you expect.  See how much force you need to apply and multiply by 100,000.

[inse]

How is this supposed to work anyhow?
Am I too stupid to imagine how a fixed array of contact needles is supposed to make any sensible contacting?
If a single contact only cost you 10ct…?
Look at the flying probe video linked earlier and then explain to me, please.
Furthermore how do you think to achieve the programming and control of the measurement equipment?
Why not start with something easier for the beginning like a computer made from TTL chips with dedicated programming language and OS?
I feel trolled

[tooki]

You can also have a look at the custom hardware that CERN has built. They have a many channel setup that produces bursts of Peta bytes per second and have FPGA based systems to discard nearly all of the generated data and filter out the interesting bits to be saved.

Link! Link! A few years ago, I tried to learn about what CERN was doing, computer HW wise, but got nothing more than the most basic of overviews! Ugh!

CERN projects are huge collaborative efforts by numerous collaborating universities and institutions. So a lot of the documentation is similarly dispersed. A lot is found in research papers. (CERN is famously the source of a research paper with a record-breaking thousands of listed authors.) A lot isn’t publicly available.

The CERN website has documentation which has links to more detailed documentation, e.g. https://cms.cern/detector/triggering-and-data-acquisition . Most of the info is out there, if you hone your research skills and have access to academic journals.

(My old boss is working on PCBs for the next-generation CMS detector for the LHC. That’s a heck of a thing, since it’s extreme data rates, in very small spaces where you can’t use thick wiring, in radiation. So both circuit design and materials selections are interesting challenges.)

Fun fact I can gloat about: There are actually boards in use at CERN that I laid out (and many of which I assembled).

[coppercone2]

this sounds like what replicators do on a ha-tak

[tinfever]

If the goal is to have an automated system to reverse engineer PCBs, that's certainly an interesting and challenging problem.

I think trying to probe the entire thing at once is a non-starter. You'd have to built a custom jig with thousands of extremely precisely placed pogo pins for every unique board you want to probe. This alone would probably make it not cost effective in money or labor.

That brings you to a flying probe machine, that can probe selected points on a board very quickly. How many pads are on a motherboard? I'm guessing at least 5000 easily. If you want to check resistance between each pad and the other 4999 pads, that's something like 5000*5000 = 25 million measurements. Let's cut that in half since there will be many duplicates from ground pins and others, and you don't need to duplicate measurements on already measured nets. So 12.5 million measurements. Let's say a flying probe machine can do 10 measurements per second. This means it will take 14.5 days to probe out an entire board. That's a long time but at least it isn't years. Although if you double the number of pads on the board, the time will quadruple.

If you applied some intelligence to the probing strategy, you could probably reduce the measurement count somewhat. (CPU socket usually has many traces going to DRAM channels and PCIe channels. Check GND and power nets first using known ATX power connector pinouts.)

You'd also need some very fancy computer vision to let a flying probe machine know where to probe just from a picture of a board. Also, the board would probably need to be stripped of all components for probing. A mechanical process like this will always be more error prone than something without moving parts.

It might be easier to strip all the components off the board, get a high-resolution CT scan of it, and then design some very sophisticated software to analyze the copper connections and generate the PCB net list from that. As a bonus, you could probably take it a step further and have your fancy software recreate the PCB in gerber format or other PCB software.

If you want to get on the AI hype-train, then you could design and train an AI model to convert the net lists into schematics. The output wouldn't be perfect, but it would be much faster than manually laying out a schematic from a netlist of thousands of components.

Identifying the components on the board is another problem as others mentioned.

If you haven't done any of this before, I would consider that it might take years to accomplish this. Is it worth it?

[CatalinaWOW]

An approach which I am pretty sure has been used by those who do such things is the mechanically tomograph the boards.  That is glue, clamp, or vacuum the board down to a flat surface.  Photograph.  Grind a small amount off the surface.  Return to step one and repeat as required.  Then follow the traces through the photographs.  Variations to speed up might include not flat grinding but adjusting to follow the observed surfaces of the board layer.  In some cases I believe they even apply this technique to the semiconductors, maybe substituting etching for some of the grinding.

I am sure that the actual process is much more difficult than the simple description above.

As with all of these techniques you have to really, really want to reverse engineer the board to make it worth the effort.

[tszaboo]

An approach which I am pretty sure has been used by those who do such things is the mechanically tomograph the boards.  That is glue, clamp, or vacuum the board down to a flat surface.  Photograph.  Grind a small amount off the surface.  Return to step one and repeat as required.  Then follow the traces through the photographs.  Variations to speed up might include not flat grinding but adjusting to follow the observed surfaces of the board layer.  In some cases I believe they even apply this technique to the semiconductors, maybe substituting etching for some of the grinding.

I am sure that the actual process is much more difficult than the simple description above.

As with all of these techniques you have to really, really want to reverse engineer the board to make it worth the effort.

Why? You can just xray boards to get the layout.

[coppercone2]

I bet it gets confusing on a 64 layer board to try to x-ray without ultra premiuim equipment. Instead of using a carbide fly cutter or something at 0.0005 increment, if fly cutters work that low


Actually that might be a really simple and cheap operation on a good mill with the board glued down and leveled (thin superglue in a press)


Or perhaps just a silicon carbide or diamond wheel surface grinder.

Put a camera on the wheel and photo every pass


THe main problem is that you need a very sharp wheel, so you don't smear epoxy. So possibly there is 2 things that you can do
1) frequently dress the wheel *use SiC
2) freeze the PCB in nitrogen to discourage smearing
3) heavy flood water (cooled to 2C)
4) maybe a CBN or diamond wheel (need to try to determine what happens obviously, this is based on theory. I am not leaning towards surface grinders because I fuking hate dust)
5) strong suspicious that the ways and mechanisms of a surface grinder machine might be trashed by fiberglass dust. If you talk to a manual machinist about FR4 material, its like asking for weapons grade uranium. Toxicish materials (i.e. bakelite)? No problem. Circuit (glass) board? OMGWTFBBQ (I am suspicious that this means they have to take apart the machine afterwards). I have a FR4 based part I use, I Feel like the bad man when I had to get that made. I bet if he had his way there would be a duplicate machine shop outside under a tree, for use with fiberglass parts, replaced every quarter (don't bother making a roof it goes strait to the smelt recycler)


I would recommend investing in the x-ray if you can, because its the lung healthy choice. I bet you absolutely nothing good will come from grinding circuit boards, and you might be cursing yourself in a few decades about why you did not leave the stone age. Nibblers are so nice.



BTW the trick for simple boards (maybe 4 layers) to help you out but not guaranteed at low budget
1) get a  bright flashlight that can focus good (LED)
2) shine bright flashlight on PCB bottom
3) look at the traces show up

[CatalinaWOW]

Any method is confusing on a complex board.  I know I get confused on a 2 layer board that has relatively few covered traces.  The bigger boards require a great deal of patience, use of various methods and correlating the results and head scratching about intended purpose and so on.  You can imagine that at least some of the makers of these boards are trying to hide something and will take steps to up the complexity.  Things like aligning paths on different layers for example will complicate X-ray examination forcing either multi-angle exposures (along with trying to keep track of information across exposures).  I am sure things like this are child's play for the advanced players in these games, and have really no idea how the big boys play the game.

Grinding circuit cards is messy and potentially hazardous, but remember that automobiles grind rubber and hydrocarbons together mixed with a heady topping of trash, fluid leaks, animal droppings and Lord knows what else and sends the results into the air in quantities several orders of magnitude above any board grinding any of us will do.  Put an exhaust fan on your shop, with a filter to make nice with the neighbors, drink a beer and chill.

[coppercone2]

bad circuit designers do this because hiding crap generally reduces reliability, repair ability (for the advanced people to fix internal traces) and also has poor EM and thermal behavior, IMO decreasing the quality of the product. Even things like copper concentrations causing board stress during reflow and shit like that

Interesting how one fried internal trace can take out 20 more if someone is trying to hide something instead of having it burn out in a isolated area where it can be fixed. hello landfill

and the most likely reality is that the reverse engineering shop charged a slightly bigger NRE (say 15%) because of all the obsfuration efforts making the design take longer to map, and the reputation of the original product has a chance to go down because they started playing camouflage. Then its entirely feasible for the Chinese clone to be better then your design because they fixed the problem caused by your counter engineering efforts lol!

[tooki]

Grinding circuit cards is messy and potentially hazardous, but remember that automobiles grind rubber and hydrocarbons together mixed with a heady topping of trash, fluid leaks, animal droppings and Lord knows what else and sends the results into the air in quantities several orders of magnitude above any board grinding any of us will do.  Put an exhaust fan on your shop, with a filter to make nice with the neighbors, drink a beer and chill.

Cars aren’t fine-grinding glass. PCBs are fiberglass and epoxy, and inhaling ground glass leads to silicosis, a serious condition. This has to be taken seriously, especially when talking about prolonged grinding.

[tooki]

bad circuit designers do this because hiding crap generally reduces reliability, repair ability (for the advanced people to fix internal traces) and also has poor EM and thermal behavior, IMO decreasing the quality of the product. Even things like copper concentrations causing board stress during reflow and shit like that

Interesting how one fried internal trace can take out 20 more if someone is trying to hide something instead of having it burn out in a isolated area where it can be fixed. hello landfill

and the most likely reality is that the reverse engineering shop charged a slightly bigger NRE (say 15%) because of all the obsfuration efforts making the design take longer to map, and the reputation of the original product has a chance to go down because they started playing camouflage. Then its entirely feasible for the Chinese clone to be better then your design because they fixed the problem caused by your counter engineering efforts lol!

I very, very much doubt that obfuscation is even considered when laying out PCBs. Multilayer boards offer more than enough inherent complexity, especially when laid out quickly.

[coppercone2]

Grinding circuit cards is messy and potentially hazardous, but remember that automobiles grind rubber and hydrocarbons together mixed with a heady topping of trash, fluid leaks, animal droppings and Lord knows what else and sends the results into the air in quantities several orders of magnitude above any board grinding any of us will do.  Put an exhaust fan on your shop, with a filter to make nice with the neighbors, drink a beer and chill.

Cars aren’t fine-grinding glass. PCBs are fiberglass and epoxy, and inhaling ground glass leads to silicosis, a serious condition. This has to be taken seriously, especially when talking about prolonged grinding.

and whatever cars make is done outside in very small amounts so it gets diluted by alot. The best way to see is to get a powerful flashlight (or better yet a lazer, if you don't have OSHA) and see the air quality when you are working. It surprisingly hard to keep clean. Enough that experienced people in industry do not like doing it. Unless its some psychos at a boat shop lol

[CatalinaWOW]

Grinding circuit cards is messy and potentially hazardous, but remember that automobiles grind rubber and hydrocarbons together mixed with a heady topping of trash, fluid leaks, animal droppings and Lord knows what else and sends the results into the air in quantities several orders of magnitude above any board grinding any of us will do.  Put an exhaust fan on your shop, with a filter to make nice with the neighbors, drink a beer and chill.

Cars aren’t fine-grinding glass. PCBs are fiberglass and epoxy, and inhaling ground glass leads to silicosis, a serious condition. This has to be taken seriously, especially when talking about prolonged grinding.

and whatever cars make is done outside in very small amounts so it gets diluted by alot. The best way to see is to get a powerful flashlight (or better yet a lazer, if you don't have OSHA) and see the air quality when you are working. It surprisingly hard to keep clean. Enough that experienced people in industry do not like doing it. Unless its some psychos at a boat shop lol

Actually cars are.  More in some areas than others.  Recycled glass is widely used in asphalt.  Concrete dust is also nasty to breath.

I do agree about dilution.  Which is why I am not too concerned about grinding an occasional board.  Silicosis is lifetime dose, so an occasional board, particularly if done with at least moderate precautions isn't likely to put you on oxygen tanks.  Different answer if you try doing this as a profession.

[tszaboo]

I bet it gets confusing on a 64 layer board to try to x-ray without ultra premiuim equipment. Instead of using a carbide fly cutter or something at 0.0005 increment, if fly cutters work that low

Nobody who uses 64 layer boards will be reverse engineering someone else's work. Even at 8-12 layers the software will be orders of magnitude more complicated than the PCB and it wouldn't be worth the effort. Besides for most PCBs any competent engineer can draw a block diagram just by looking at a board. And then what, copy the exact layout of DDR memories or PCI-e lanes? A parallel backplane? Why?