I snagged the 68HC705xx firmware from a website I don't remember now, the purpose of which was to convert an IBM-PC/AT keyboard into something useful, about 15 years ago, and whittled on it further to reduce the process from converting it to async-serial to parallel with strobe, and then used it with "vintage" hardware for which parallel keyboards were not so readily available any longer.

The doc's that came with that went into considerable detail explaining that the PC's keyboard was encoded on keypress but the result wasn't transmitted until the key was released, or a timeout expired, at which time the result was repeatedly sent. Never having had any reason to doubt those doc's, I've always believed that, no matter how odd it seemed, that's what they did. The rationale was that if multiple keys were pressed on the downstroke, waiting for the release would ensure that the correct key would be chosen.

Maybe that's not what actually happens ... I don't know. It's one of the hazards of believing what's published on the www.

I personally have only encoded on the downstroke, but most of my encoders were for small keypads rather than full keyboards, and they were considerably less likely to have multiple keys struck at once.

Somewhere in the pile of stuff out in the shed, I've got an old PC/AT keyboard with a windowed 8742 on it ... maybe there's a way to cause that thing to give up its code-space content. Then, with a little disassembly, one can find out what they really did.

RE