HACKACALC

I partially assembled the board - only 7 CD4051 and one 74HC138 were populated. Then I removed transparent keyboard circuit from calculator remains and soldered 15 wires to his "brain":

Added 2xAAA power source (3V) and magic - it worked :)

I even changed code a little on JP2 inputs and it printed something ;)

So it looks like it is right direction for further experiments...

P.S. Just uploaded Eagle files (with Gerbers) to the project

Thank you, OSHPark!

Now I need to perform 1st experiment - keypress simulation :)

I was thinking for so long how I should make my 1st prototype - on breadboard, as wire-wrap assembly or dead-bug style soldering, but finally I've decided to go with PCB (will order through OSHPark as usual) designed in Eagle v5 (version that I have standard license for). So main part of the prototype is 7 analog switches 8:1 (CD4051) and one decoder 3->8 (74HC138):

Switches implement keyboard matrix 7x8. Also you can see here 2K ROM that will store "program". Additional circuitry is address counter (I'm not so sure that it will work right away):

And this is manually routed PCB 18x10cm that I will order tomorrow:

Circuit doesn't have clock generator on board because I didn't decide yet how to do that - most likely here it will be manual step button (with RS-trigger) on prototyping area at the bottom. Will see how it goes...

UPDATE: Ordered on Nov-16 through OSHPark:

Ok, I've got CASIO fx-350MS:

it looks exactly the same as all other calcs:

but inside it's more like fx-82ms:

it means for now we will use "A&W" clone with flexible keyboard circuit exclusively (at least for some time).

OK, lets take EXACTLY THE SAME model of clone - A&W 08207:

and it's... DIFFERENT from other A&W 08207 that I use here!

PCB looks different, number of signal that go to keyboard is different (18 vs 16):

So still no reproducibility yet, but PCB has name on it and it's 350ms (even though flexible circuit says 82MS-9R) - so it might be clone of CASIO fx-350ms (that should have the same face as well)? Stay tuned to find out...

P.S. It looks like 2 additional contacts on the left is RESET (in the center on the back), but in the same time other 16 contacts connected to keyboard matrix very differently!

Lets check other calculators with the same face - just in case somebody will need to independently reproduce this project - CASIO fx-82ms, CASIO fx-300ms and "NONAME" from Amazon for $7.99:

Without packaging they look exactly the same (except for solar cell on fx-300ms):

1st one is fx-82ms that looks DIFFERENT from inside:

2nd one is fx-300ms that looks similar to fx-82ms, but with 1 button battery instead of 1 AA and solar cell:

3rd one is "NONAME" that is DIFFERENT from originals and from previously opened clone:

Interesting observation: clones use 2 batteries while originals use only 1 (AA or button).

So no reproducibility yet...

OK, now we can take closer look at transparent keyboard circuitry and trace it by index finger :)

Green PCB connected to keyboard by 16-signal bus - we can enumerate signals from 1 (left) to 16 (right) and last one is not connected anywhere. After one night of insomnia I've got this matrix (what should be connected to what in order to simulate keypressing):

It could be directly used as a base for our instruction set - just rename 15 to 0 and we can take columns as higher nibble of opcode and rows as lower nibble of opcode, so ON would be 0x67; A=1 would be 0x48 (1), 0x6E (SHIFT), 0x4E (STO), 0x3E (A) and so on. Rest of the codes could be used for future branching, looping, comparison etc.

P.S. CASIO fx-82ms manual can be downloaded from here

Chosen NON-programmable scientific calculator is clone of CASIO fx-82ms:

While original CASIO fx-82ms costs $12 (see here), "noname" clone is available for $7.99 (see here).

Inside of the calculator we see a small green PCB with on-board chip under black compound:

In this particular model a keyboard circuit is located on transparent flexible "board":