Home Page Forums MAME MAME Writing a driver from scratch - help, please

Then latch it from the other direction. Look into PORT_CHANGED_MEMBER.

Assuming a similarity to the PC/XT (commented disassembly at https://github.com/Halicery/8042/blob/main/8048_XT_INTERN.TEXT) the microcontroller tells the external circuitry when to latch the bits. On the PC/XT there are 2 latches, Q (latches the selection bits) and RESET (latches the value read from the keyboard matrix).

This is typical late 70s/early 80s IBM overengineering. Most keyboard microcontrollers do what you'd expect: set the selection bits and read live.

I have been able to stabilise the values of the counter and the select signals until T1 is read. I can see them changing through the logs.

The next step is to assign the real key to the emulated one. I have been trying to assign "Reset Error" (6e) to F5 as it is the position it has in the original keyboard ("Reset error" became F5 in the transition from the Datamaster to the IBM PC) but I haven't got any reaction to the key being pressed. What would be the best way to test this?

Thank you in advance!

Are those inputs read as active low or active high? Wrong input pressed polarity can cause a stuck case where the host CPU believes everything is pressed and do nothing (or at best cause havoc).

No, I don't think this is the issue. If there was a key pressed test 35h would fail and display the offending key's scan code.

It's like the key input at T1 was not recognised.

It's somewhat common for microcontrollers to not properly detect stuck keys if that's the initial state. But add some printf()s or whatever and see if the T1 read callback is returning what you expect it to.

I have successfully read keys, but the scancodes don't seem to match... I imagine I have to revise the matrix again.

EDIT:
By the way, I would like to assign the unassigned ROM space to a known value (for example 0xff). What would be the best way to so? With uninitialized memory I get some sporadic ROM errors between tests 1A and 29...

Replace with

This time I think I do require some revision by someone else. I have to revise the keyboard matrix again, but there's something that is troubling me: the scancodes generated do not match the characters I get to the screen. There's a catch, that is the verification scan codes are working fine, but when it comes to the regular ones, I even get some keys with multiple key codes...

For those of you who have cloned jlopezm's github and are following along, to get past the startup tests you need to unset all the DIP switches except for the ones that deal with memory.

It's great to see some keystrokes working.

I haven't got time today but I'll try to figure out why the matrix isn't what we expect. (Hopefully it will stop snowing here...)