Reverse engineered a circuit board

[laggeddag]

This is a BTR / RS232 to tape reader converter board out of a Mori Seiki SL1 lathe w/ Yasnac 2000GII controller.

It was made by an out-of-business company in the 90's.

I decided to see if I could make a copy of it given that they are somewhat rare and I know several people who would definitely like one for their old mori.

Its 2 layer and through hole. There are about 50 unique components on it. No firmware or processors. All standard logic parts, with a baud rate generator and a UART. No surprises.

It took me about 16 hours to make a fully digital copy in professional CAD software, ready to turn into a PCB. I havent made the layout yet but it will be trivial for this design.

The process was straightforward but pretty boring. I have an improved process in the works which will greatly reduce the time it takes and make it almost completely automated as far as generating the netlist directly from the PCB traces. The only work left will be to generate the component library from visible markings and online datasheets. 2 layer is definitely doable. 4+ layer would be much more difficult but not impossible depending on the complexity and whats on the inner layers.

The BOM cost for this PCB looks to be about $20 in low quantity. The PCB itself could be made for about $10 each in 5 qty from pcbway.

Given how this board works, I think nearly the entire circuit could be replaced with a microcontroller and a few high voltage translators. I would imagine thats probably true for alot of the electronics in old machines.

Im looking for another PCB challenge now, something machining related, that hasnt been done before, and would be useful to alot of people. So probably another RS232 to tape adapter. But I'd be open to other stuff for sure.

 

[Pete Deal]

Fun stuff. I used to do a lot of reverse engineering pcb's back in the 80's and 90's. It has gotten much less practical with modern pcb's with custom ic's and programmable logic being so common.

 

[laggeddag]

bump!

No ideas? Theres gotta be something more juicy than a BTR to reverse engineer and copy. If its something I really like I'll post detailed instructions on the process, especially the graphics enhancement and trace to netlist steps.

Likely candidates:

-physically available for me to disassemble..cant copy/reverse it if I cant get my hands on one of them.

-will help alot of people (I guess a BTR would fit the bill technically especially if its for a machine with no or very expensive options to get one)

-2 layer

-no black box logic chips that cant be either directly or inductively reverse engineered...i.e. read with a programmer, or simple enough to determine their function by the rest of the circuit. so a pcb with a single xylinx 9000pin fpga and a connector is probably not a good candidate

-special bonus points if the reverse engineering aspect is just as or more interesting than the copying aspect..i.e. knowing how it works is somehow very interesting (unlike a BTR which is pretty obvious)

 

[laggeddag]

Heres an example of a device where I wanted to know how it worked, but didnt need to make a copy. I got this Federal Millatron 830 (832?) electronic test indicator setup for practically free. The meter and transducer worked but the analog needle would stick if driven past maximum, I tried to fix the meter movement but I just couldnt get it perfect. This meter is designed to measure down to 10 microinch resolution. I was interested in what kind of circuitry such an important and precise device might have so I reversed it.

Simultaneously eye-opening and also somewhat disappointing, its a textbook chain of precision opamps with basically no tricks or special techniques. The signal even passes through gain resistors which are directly switched by the dial switches and the layout is care-free through hole with traces going wherever they feel like for the most part. There are some guard traces around the oscillator circuitry but otherwise there is really nothing at all special going on here, which is interesting in itself and a testament to the engineering packed into even decades old precision amps by burr brown and analog. I dont see any reason why a gain resistor couldn't be changed slightly and instead of the resolution being 10 microinch, being dropped down to 1 microinch or whatever you want really. I doubt linearity would be affected by such a small increase in gain and it could easily be verified with a few gage blocks (that have already been measured on a much more expensive machine).

Really this does make sense, since the signal is DC and nulled manually, pretty much any noise source is trivial to suppress since there is such an enormous difference in time constants. The limitation on resolution will almost certainly be the transducer itself followed by the reality of physically "measuring" displacements that are in the same range as light wavelengths. I did actually try that and it seemed to be stable enough to make measurements with 1 uinch resolution with a fairly sloppy setup on a granite surface plate.

Quickly looking over the schematic which I drew awhile ago, it looks like the overall DC gain is around 1000.

Note that this schematic is not complete. I didnt bother drawing up the sine wave generator, which is an off the shelf chip, or the second channel circuitry. This was done without taking the parts off the board.

 

[Perry Harrington]

Seems to me you might have problems going to 1uin without guard traces around the inputs of the op-amps.

 

[atomarc]

Seems to me you might have problems going to 1uin without guard traces around the inputs of the op-amps.

Man....I was thinking the very same thing! I make it a point to guard all my op-amps, they're just too valuable not to.

Stuart

 

[andrewmawson]

You've reminded me that I've STILL not got the BTR RS232 interface on my Fanuc 6 driven Fanuc Tapecut Model M wire EDM machine working.

.... damn you sir please send more Round Tuits

 

[kb1cjp]

Curious if you would expand your process in greater detail for laying out a PCB of that size. You even state a 4 layer board would have been doable. Maybe you just have more patience than I. I certainly have not been that ambitious to do a whole board of that size, but have done much smaller, or just small sections of a board for troubleshooting proposes.

Brian

 

[Georgineer]

Just out of interest, how do you disassemble the board? I had a 1980s PCB to reverse engineer but we had several available, so I just had the components snipped off flush. The board was scanned and reproduced photographically and the new version looked wonderful, but there were two very sneaky, almost invisible links between pads in a densely populated area. That took some finding.

Is component obsolescence an issue with what you're doing? I had to identify, specify and where necessary arrange procurement of components to cover a planned thirty-year future life. There are some very grey areas in that part of the component market.

George

 

[kb1cjp]

George, he is not duplicating the board, he is tracing out the circuits and creating a new layout. So for any discontinued componets will need to be cross refferrenced and applied to the new layout.

 

[Bruce Griffing]

Memory expansion cards would be popular - especially for machines for which commercial options (eg. memex) do not exist.

 

[laggeddag]

Curious if you would expand your process in greater detail for laying out a PCB of that size. You even state a 4 layer board would have been doable. Maybe you just have more patience than I. I certainly have not been that ambitious to do a whole board of that size, but have done much smaller, or just small sections of a board for troubleshooting proposes.

Brian

Id be happy to share the entire process I use in detail I just think it would be more fun if there was an example to work on that others wanted to see reversed or copied as well..I dont have pictures of the boards I've already done during the process.

The 4 layer would be alot more work than 2 layer. But if there isnt alot going on in the inner layers or its inductively determinable whats going on in it by the outer layers it might not be that bad. For instance if its just power planes then it could be inductively determined which vias connect to it and not elsewhere, by capturing the out layers, vectorizing and netlisting them, and then seeing which power pins clearly do not bridge to each other via the outer layers, and therefore must bridge on the inner layers, and so the nets could simply be joined. Its also possible to remove the copper from the outer layers once they are captured so its easier to "see" the inner layers and possibly get real vectors from them.

By laying out do you mean reversing? Or do you mean making a new PCB layout for the copy? For making a new layout I use professional PCB software but anything that imports vectors like DXF and has high quality netlisting and free copper detection would probably work. For reversing the layout from an existing boards there are alot more programs and steps but its all pretty straightforward. If we can find a good example board to reverse I'll "lay it all out" in detail..get it hee hee

 

[laggeddag]

Just out of interest, how do you disassemble the board? I had a 1980s PCB to reverse engineer but we had several available, so I just had the components snipped off flush. The board was scanned and reproduced photographically and the new version looked wonderful, but there were two very sneaky, almost invisible links between pads in a densely populated area. That took some finding.

Is component obsolescence an issue with what you're doing? I had to identify, specify and where necessary arrange procurement of components to cover a planned thirty-year future life. There are some very grey areas in that part of the component market.

George

On the BTR (first board above), I tried using a reflow oven but it was pretty messy and slow, and darkened the board and seemed to generally be rather sloppy. Although I think it could be refined to where it was a quick and clean process. Maybe next time. So I just clipped all the chips out and pulled the pins out individually with a soldering iron and tweezers. Its pretty quick and fast. I then ran over it with braid to get off stray solder. The problem with the oven is that you have to get the board very hot for quite awhile to allow all the solder to liquify, then when you pull the board out of the oven you have to pull the chips out before it cools down too much, and they dont always come out nicely. If you tap the board to make them drop out you're likely to make solder splatter on the board which is a big pain to get off. But I'm working on the above process and will probably have some improvements. There are alot of people out there removing components from PCB's like this and some of them use heat guns. If there aren't alot of components, like the above BTR, clipping leads and pulling things out with a hand iron and tweezers if clean and pretty fast.

As far as obsolescence, getting physical parts isn't required unless you are trying to make an exact copy. Once the board is reverse engineered, you have alot more options as far as making a copy that functions the same, without necessarily using the same parts, because now you know how it works. So in that case it becomes more important to be able to identify the components, and pretty much every datasheet for every component ever made is online. And even if the datasheet isn't available, you can probably figure out what a part does by context once you see the rest of the circuit.

In the case where you do actually want to physically get identical components, you are still likely to find what you want on ebay or aliexpress. Its very rare that I cant find a chip for sale that I need, even if very old and obscure.

 

[laggeddag]

George, he is not duplicating the board, he is tracing out the circuits and creating a new layout. So for any discontinued componets will need to be cross refferrenced and applied to the new layout.

For the BTR, making a copy actually is the goal, all those components are still available so it would be possible to make an identical copy if desired. I'll end up making a new PCB layout, but it would be possible to directly use the vectorized layout I captured from the board and skip that step. But I'd rather have a live PCB layout of that board. Since its so simple it isnt much work, especially since I can just place the chips in the same locations on the board and route it exactly how the original is.

For the federal gage meter board, the goal was just to reverse engineer it. So I dont need to make a layout.

 

[laggeddag]

Memory expansion cards would be popular - especially for machines for which commercial options (eg. memex) do not exist.

noted, and a few ideas:

candidates for reversing with full process details posted up here:

1. memory expansion cards for memex systems
2. high power/high voltage power supplies (early pulse-controlled EDM?)
3. high power spindle/servo drives

 

[Bruce Griffing]

noted, and a few ideas:

candidates for reversing with full process details posted up here:

1. memory expansion cards for memex systems
2. high power/high voltage power supplies (early pulse-controlled EDM?)
3. high power spindle/servo drives

It may not be clear from my post, but I meant memory expansion cards for various cnc controllers, eg. Okuma or Mori Seki. Memex is a supplier of after market memory expansion for some (eg Fanuc) but not all cnc controls. The idea is to fill in the gaps left by Memex and its like.

 

[laggeddag]

It may not be clear from my post, but I meant memory expansion cards for various cnc controllers, eg. Okuma or Mori Seki. Memex is a supplier of after market memory expansion for some (eg Fanuc) but not all cnc controls. The idea is to fill in the gaps left by Memex and its like.

ahh okay in that case a realistic pathway for that to happen would probably to find which machine has alot of interest for a relatively low cost memory card, then get a broken/cheap example of a memory card that works on that machine, reverse it to see how it functions, then extrapolate to make a new one from scratch that is as big as possible, etc..

 

[dcsipo]

ahh okay in that case a realistic pathway for that to happen would probably to find which machine has alot of interest for a relatively low cost memory card, then get a broken/cheap example of a memory card that works on that machine, reverse it to see how it functions, then extrapolate to make a new one from scratch that is as big as possible, etc..

I wonder if it is really cost effective to engineer a new board for an obsolete system , Unless it is a cookbook design and the components can be had real cheap and there is demand for more than a dozen, i doubt you can recoup the time invested. Say you spend a week on it what is worth? about 5000 in engineering it? Can you sell more than 13 at 500 bucks, and produce them under 100 a piece? Remember even after you made them you have to test each, package and ship....so there is work involved post design. The retrofit market is a tricky place.

dee
;-D

 

[laggeddag]

I wonder if it is really cost effective to engineer a new board for an obsolete system , Unless it is a cookbook design and the components can be had real cheap and there is demand for more than a dozen, i doubt you can recoup the time invested. Say you spend a week on it what is worth? about 5000 in engineering it? Can you sell more than 13 at 500 bucks, and produce them under 100 a piece? Remember even after you made them you have to test each, package and ship....so.prx.webanonymizer.org there is work involved post design. The retrofit market is a tricky place.

dee
;-D

I try to make business plans after there is a specific project in mind..otherwise all the numbers are just made up.

 

[dcsipo]

I try to make business plans after there is a specific project in mind..otherwise all the numbers are just made up.

I buy that....also did a quick search on ebay, the CNC add on memory boards do not seem to be that expensive, or at least i did not run into any that is.

dee
;-D