> > Two keys from one keypad with standard routines can be a bit tricky
> > but can be done..   If two keys on different rows AND columns are
> > down you will read all four keys that make a square as down.
>
> It's been my experience that ambiguity only occurs when THREE keys
> that form three corners of a rectangle are pressed, and that TWO keys
> are never a problem to read.
>
> Of course, this applies only to keypads that are scanned by row or
> column; were you talking about another keypad topology?

There are four basic matrix-scanning techniques I've used or seen in signi-
ficant use.

[1] Energize all rows and read columns; then energize all columns and read
    rows.

    This technique is simple to implement in software, and can get results
    in only two steps.  It will not work well for multiple keypresses unless
    all of the keys are in the same row or column, but it's quick and easy
    and it works.  Especially on something like a PIC, this is probably the
    easiest way to go from a keypress to "A number".  Note that the following
    code would probably run too fast on anything other than a 32KHz crystal
    and should be slowed down on faster machines.

        clrf    PORTB
        movlw   $0F
        tris    PORTB
        movf    PORTB,w
        bsf     RP0
        comf    PORTB   ; Really the TRIS register
        bcf     RP0
        iorwf   PORTB,w

    Note that the first three instructions only need to be done once at system
    startup (though doing them more often may allow the system to recover if
    one of the port latches gets glitched).  Interestingly, if the first three
    instructions are omitted from subsequent calls, the behavior of TRISB will
    be different on every other call but the routine will still work fine!

[2] Fully-scanned matrix, no diodes
    This approach energizes the row wires one at a time, reading what signals
    come back on the column wires.  It will read unambiguously any one-or-two
    key press, and will also return presses of more keys with the caveat that
    if three keys forming the corners of a "box" are pressed it will appear as
    though all four are pushed.  If such cases are logically impossible in the
    application, this won't pose a problem.  Note that most PC keyboards are
    handled this way.

[3] Fully-scanned matrix, with diodes
    This approach uses a diode on every key to ensure that the "box" problem
    can't happen.  Many musical keyboards use this approach since it is hard
    to predict what keys people might try to push simultaneously.

[4] Minimalist no-extra-components matrix
    In cases where I/O pins are at something of a premium, it's possible to
    read (n*(n+1))/2 switches with (n) port pins (e.g. 4 port pins can read
    10 switches).  This is more than double the capacity of a conventional
    matrix technique, but does not have the ability to handle even double-
    switch closures well (all such closures are ambiguous; if the aplication
    dictates that some are permissible and others not, they may be supportable
    in such a context).  To place the switches for use with (n) I/O pins, draw
    an NxN matrix, strike out the primary diagonal (top-left to bottom-right),
    and put switches below that.  Then add an additional (n) switches between
    the I/O pins and ground.

    Scanning such a keyboard is more complicated than scanning a square key-
    board with the all-rows then all-columns method, but the savings in I/O
    pins can make it worthwhile.