On 2/18/2017 7:37 PM, Chuck Guzis wrote: Yes, of course. I meant running through the combinations of the inputs on the PAL and obtaining the outputs for each set of inputs, then attempting to minimize the logic (for combinatorial). REgistered is a more complex kettle of fish. I am working on creating the schematic now, but it's a tedious process. Jim

As a function of "give a man a fish, you feed him for a day; teach a man to fish, you feed him for all time", I dloaded Logic Friday, and figured out a quick way to read the two 16L8 (combinatorial only) PALs on the board with my EPROM reader (cue duct tape and baling wire snickers). After getting some good reads, and minimizing the binaries (I read 8kB of data, and then split the files into 1kB chunks to handle the one unit's 10 input lines, and split the other file into 4 2048 byte chunks due to the 11 inputs), I then wrote a really crappy C program to convert the data to CSV format for Logic Friday. After dealing with some nuances in the format, I was able to minimize the equations, and I thank you Chuck (I probably am still going to hit you up on the 16R4 and the 16R6, as I assume they are not so easy to reverse engineer) for the suggestion. Now, my new question: Is there a way to get Logic Friday to perform "negative" logic? Specifically, by looking at the CSV file, it is obvious that one of the outputs is low only when A15:A10 and R/W are all low. So, basically: O3 = (A15 + A14 + A13 + A12 + A11 + A10 + R/W) But, what I get is: O3 = A10 + A11 + A15 A13' A12' + A14' A13' A12 + A14 R/W' + A13 R/W' + R/W I assume this is the "wetware" process you mention, but I thought I would check if I need to tweak something in Logic Friday to achieve better results. Jim

Jim Brain wrote: Don't forget that 16L8 means 8 logic outputs and 16 inputs even though at first glance there only seems to be 8 inputs. That is because every output is also an input and can possibly be just an input by forcing the three state option. Both options complicate things and in the first case you can have stuff like: O1 = /I1*I2 + I1*O1 If you count up the possible values of I1 and I2 you will get: I2, I1 => O1 0, 0 => 0 0, 1 => 0 1, 0 => 1 1, 1 => 1 but if you count down you get: I2, I1 => O1 1, 1 => 0 1, 0 => 1 0, 1 => 1 0, 0 => 0 which is different, and if you repeat this a bunch of times you will sometimes get 1 output in the first line and you might get 0 output in the third line once in a while. The IBM PC AT had two PALs. One was a simple address decoder while the second included an interface to the FPU and had a tiny state machine implemented like this. I helped some friends who were doing an AT clone in 1986 figure this out in a few minutes, but most of their competitors in Brazil took nearly two years to reverse engineer this. -- Jecel

I have seen someone making a small device to do this. http://dreamjam.co.uk/emuviews/elec/pal.php Unfortunately the software doesn't seem to be available any longer. Anyone that has it? It would definetely save me some time. /Mattis How about: http://dreamjam.co.uk/emuviews/pal/pd-070908.zip

2017-02-21 22:01 GMT+01:00 jim stephens <jwsmail at jwsss.com>: I did try to contact him but got no reply. But now it solved because Kelly Leavitt somehow did find the correct link, which is not the one linked to from the page. Thanks a lot for this! Unfortunately there were no source code in there unless they are self extracting binaries. No PC nor DOSEMU running so I cannot check right now. If there are no source I need to try to figure out the the ordering of the pins that the dump uses from the schematic. It seems not to readable by the version of Eagle I have here though. Maybe it will be easier to DIY anyway... Anyone else that done this kind of PAL analysis so I can avoid reinventing the wheel? /Mattis

On 2/21/2017 2:06 PM, Mattis Lind wrote: There are PD executables which run from the dos command line for both versions of the hardware. Manufacturing info for the boards are there. I'll targz the source and email it to you offline if that would help. It is compiled with mingw, so should hopefully be simple to make it run on gcc and run the analyze component. thanks Jim