"Ronald Leenes" wrote in message
...
Hi,

for a project we have a very large key-matrix (32x24). We have a problem
with phantom keys (when three keys in a rectangle are pressed, also the
fourth corner is seen as pressed).

In any matrix system that is not properly isolated, it will be impossible to
avoid this problem. That is because pressing the corner key creates a path
between the two diagonal keys, and that is why your phantom key shows up.

How do we solve this problem? I know something can be done with diodes,
but we can't get it to work. Can somebody give some hints.

Typically, diodes are a good solution, but they are also slightly leaky.
This means that on small keypads, you should not have a problem, since there are
not many parallel diode on any given leg of the matrix. However, when you have
many keys in a matrix, the leakage currents quickly become quite significant.
The surest way to isolate this would be with parallel loaded shift
registers. One 8 bit register can be clocked to take a "snapshot" of 8 switches
on a single column with no leakage, phantoms, or unpredictable behaviors.
However, the circuitry can't be used with most matrix keypads- they are already
hard wired and not easily broken out to be used this way.

An additional problem is that the 'keyboard' consists of panels that can
only be opened with great difficulty.

Yes, that is a problem- meaning that you cannot open a common trace and run
it to a specific column outside of the keyboard. Well, you might try diode and
resistive solutions where each key has a fixed resistor in series, but once
again, this can be pretty difficult to do with most matrix keyboards.
Look inside a typical PC keyboard and you can see how they put over a
hundred keys together with a controller chip. Be prepared to program a
controller for your keyboard if you want key "rollover" and best isolation.

What is the most clever way to interface say 500 switches to a computer
(serial or usb)?

I would personally choose to completely forego matrix arrangements, when
given the choice. However, that is from trying to eliminate spurious signals in
robotics- a flaky sensory input can cause the machine to do crazy things. My
choice is parallel to serial shift registers, no more than 32 keys per column
(for speed), and go for 16 columns. That gives you 512 switches, read 16 bits
at a time. You end up reading 32 sixteen bit words and that can move pretty
quickly on USB.

Best regards,

Ronald

Cheers!

Chip Shults
My robotics, space and CGI web page - http://home.cfl.rr.com/aichip

Ronald Leenes wrote:

Hi,

for a project we have a very large key-matrix (32x24). We have a problem
with phantom keys (when three keys in a rectangle are pressed, also the
fourth corner is seen as pressed).
How do we solve this problem? I know something can be done with diodes,
but we can't get it to work. Can somebody give some hints.

An additional problem is that the 'keyboard' consists of panels that can
only be opened with great difficulty.

What is the most clever way to interface say 500 switches to a computer
(serial or usb)?

Best regards,

Ronald

If you add a diode in series with each contact point (either in series
with each row trace and the contact point, or in series with each
column trace and the contact point) you eliminate the phantom keys.
This makes the flexible circuits lumpy, though.

--
John Popelish

Ronald Leenes wrote:

Hi,

for a project we have a very large key-matrix (32x24). We have a problem
with phantom keys (when three keys in a rectangle are pressed, also the
fourth corner is seen as pressed).
How do we solve this problem? I know something can be done with diodes,
but we can't get it to work. Can somebody give some hints.

An additional problem is that the 'keyboard' consists of panels that can
only be opened with great difficulty.

What is the most clever way to interface say 500 switches to a computer
(serial or usb)?

Don't. use fewer keys and context switch, or use a virtual keyboard, it
would be cheaper than the hardware.

Al