Keypad design and the behavior of output_float()

[kd5uzz]

Hello everyone.

I'm designing a PCB that will incorporate a 4x4 keypad. I would like to be able to detect and decode multiple key-presses.

In an effort to save space on the board I would like to use the minimum number of components. Each input will need a pull-up or pull-down resistor (depends on the rest of the keypad design).

I see some designs that use resistors on the ROW lines to limit the current sourced by the PIC on those lines. I also see some designs that use diodes, the stated reason is to allow the decoding of multiple key-presses.

The code below uses output_high and output_low. My question is:

Can I instead use output_float on the ROWs I'm not checking, and use output_low on the 'active' row, allowing the pull-ups to pull the 'FLOAT' pins high? Why wouldn't this detect (and decode) multiple key presses (without the use of diodes)?

I feel like I'm missing something.

For the record I intend on using the RTOS to run a task similar to the code below every 100ms or so - any thoughts as to if that is too often/not often enough?

Example:

Assume a 4x4 keypad, user is holding (A,1), (B,1) and (B,2).

[PCM programmer]

The CCS kbd.c keypad driver floats the row pins, with a pull-up on each
pin. It then sets each column pin low, one at a time, and reads the row
pins to detect a keypress. That's similar to what you want to do, in
terms of scanning the keypad.

[kd5uzz]

PCM,
Ah, exactly what I was thinking except swap ROWs for COLs, thank you.

Can someone explain what would happen if, using a basic matrix keypad with this code, would happen during multiple key presses? I don't see why I wouldn't be able to detect and decode multiple key presses.

[PCM programmer]

Here are some articles on keypad rollover issues:
http://www.lintech.org/comp-per/04KB.pdf
http://www.national.com/an/AN/AN-734.pdf
http://www.dribin.org/dave/keyboard/one_html

[FvM]

[kd5uzz]

Thanks everyone. The links were very useful.