Reading Rob's info thread I am beginning to get the glimmerings of an idea.

Hugo, are the buttons you are using those minature surface mount, metal diaphragm, positive click things that got popular a few years back? If so, the "sticking button" phenomenon could be that there is a corrosion problem on the switch diaphragm that causes a lot of apparent switch bounce. If this is the case, it could be that the PIC debouncing the key depressions is getting upset. Let me explain.

When I was at GEC, I went through a process development exercise for surface mount components to allow us to produce SMT boards in bulk. These switches were chosen because of solderability and their attractive positive feedback to the user. They had to be sealed prior to passing through a solder wave, but we found that they weren't always sealed properly and the rosin in our solder attacked the metal diaphragms in the switch. They worked OK, it's just that they then went from producing a few spikey transitions to literally hundreds of messy transitions which used to drive our matrix decoder IC's crazy - they were used to the sort of irregular contact bounce of a normal keyswitch. What we started seeing was that just one keypress on the telephone keypad would cause the phone to generate a stream of repeated digits, which we couldn't figure out for ages.

Maybe there is an isolated number of display panels with contaminated keyswitches causing this? Maybe if people with the "bounce" problem reported their serial numbers then you could see if there's a cluster of units with a problem?

Unfortunately, the display boards were all built in one batch so serial numbers won't really help; the switches are through-hole and were flow soldered with a no-clean process I believe. It could be corrosion though, but as the PIC code sends state changes, we should always get a button up however much bounce there is.

Hugo

but as the PIC code sends state changes, we should always get a button up however much bounce there is.

Yeah, but what if, as he suggested, you're getting a string of ups and downs really fast and continuous? Or perhaps the debounce routine in the PIC can't handle that much noise and it's really never getting around to sending the up code?

Of course, you're in a better position to know what's happening. We're just trying to help- you know, throwing out ideas. Maybe it's got nothing to do with the switches...

Tony Fabris
Empeg #144

The way the PIC code works is that it has a current state and a last state: it sends changes between the two. It takes several milliseconds to send a button press/release code, and so if there's bounce inbetween it still maintains a consistent last/current state and sends the delta.

I'm confident that if a button is released (electrically) then the PIC will send the right code. The greyer areas are faulty switches and the code not being recognised for a variety of reasons at the main CPU end.

Hugo