Almost OT: Pushbutton switch latching

[Sridhar Ayengar]

der Mouse wrote:
>> Picture this - a number of switches, say, eight.  Momentary contact,
>> normally open switches.  And then eight outputs, one for each switch,
>> TTL level.  I want to be able to push one switch and bring [its]
>> corresponding output high, and have it stay high even when the switch
>> is released.  But, making this more complicated, I only want to be
>> able to allow one switch to be active at a time.  So, say switch four
>> is pressed, output four is high.  Then, switch one is pressed, and
>> output four goes low, and output 1 goes high.  Timing is not
>> critical.
> 
> Do you care what happens if two switches are pressed at once?  In
> particular, do you care what the outputs are like during the time when
> both switches are pressed?  Do you care what the state is like on
> powerup, before any switches are pressed?

Couldn't one just hypothetically feed all of the switch outputs through 
a bunch of XORs and then AND the output of the XOR array with each of 
the outputs of the flip-flops you suggest below?  Or am I 
overcomplicating the problem?

> My first cut would be to set up a bunch of R-S flipflops (each one
> being two cross-coupled NAND gates, probably).  A switch would then
> drive low the S input for its flipflop and the R input for all other
> flipflops.  (How this is done is more or less irrelevant.  Personally,
> I'd probably put pullups on the inputs (omittable if they're true TTL,
> as opposed to something like HCT which is TTL-level but not real TTL)
> and then have each switch ground the cathode side of a diode matrix
> which then pulls the relevant inputs low.  If you want to "do it right"
> you'd need a bunch of 8-input AND/NAND gates - you might even make the
> flipflops out of those, with one input of each flipflop dedicated to
> each switch.)
> 
> If you care about power-up state I'd probably add some kind of
> discrete-component circuit to fake pressing a switch for a brief time
> during powerup.  I can see a possible way to do it with a diode, a
> transistor, a resistor, and a capacitor....
> 
> Another possible solution is to feed the raw switch bits into a
> priority encoder chip, latch the resulting encoded bits, and feed the
> latch outputs into an N-to-one demux chip.

This is the method I would normally choose.

Peace...  Sridhar