Jez:
I am really surprised that you felt that the only way to go was to debounce switches with the external switch debouncer. That seems to me to add unecessary extra cost to a system. I have been using software filtering for many many projects on many platforms. The processor overhead of really rather quite small in the percent or two category.

Whenever one builds a key matrix to scan a ton of buttons software debounce has worked for me quite well in these cases too.

So guess what, I have never used switch debouncer chips in systems where software had control. Goodness even in all the industrial embedded PC based applications I have done I used C code to debounce all the opto coupled discrete inputs to the computer, whether they be switches or other types of inputs. The last time I made a hardware debounce circuit was probably 15 years ago when circuits were 'all hardware' and then the infamous 74279 came to the rescue at the expense of having to use double throw switches!!!


Are you sure you had your code written correctly in your pre-Maxim switch debouncer chip days?

Michael Karas

PS
Using a scheme similar to the post I made earlier in this thread it is possible to use simple arrays for the filters and previous/current status I assign the individual switchs to bits in the variables. Thus one can filter 8, 16, or even 32 inputs depending on whether you use unsigned char, int or long variables. Such scheme was used in a C program I did once on a 386/SX-25 computer to filter 32 inputs at 150 Hz (sample rate was 450 Hz). It consumed about 2->3 % of processor bandwidth.
MJK