It has come to my attention that there is a definite lack of understanding on
how this simple procedure is achieved. Especially for low  numbers of  switches
in non-microprocessor controlled devices. If the number of buttons you  require
on your front panel numbers  less than  16 then  this is  how you  get a  clean
button press.

  You  can use  CMOS or  TTL but  the component  values change.  I'll show  the
component values for CMOS but it should be noted that the values should also be
sweetened to taste. That is to say that you can Taylor the reaction time of the
switch depending on your application's requirement and the feel  you'd like  to
give it.

  Figure 1 shows the basic arrangement. A resistor pulls  a capacitor  up to  5
volts (or what ever your positive supply voltage is if you're using CMOS. Up to
15 volts) Because of the time coefficient of the capacitor  and resistor,  this
takes time to occur. The bigger the capacitor or resistor, the longer it takes.
Placed at the junction of the resitor and capacitor, the momentary press button
pulls the capacitor to ground. It doesn't short out the power supply because of
the resistor. When the switch is made, the voltage on the capacitor falls  away
very  rapidly to  nothing. or  near nothing  depending on  the switch.  (That's
another story and I'll get to that) When the switch is  released the  potential
across  the  capacitor  is  charged  up  again  slowly  by  the  resistor.  The
charge/discharge rate is a smooth logarithmic curve. One faster than the other.

  However CMOS and TTL  don't like  smooth transitions.  Even TTL  has a  small
linear region and will tend to get into undecided  states. CMOS  is even  worse
because it can easily run linear and because of it's high impedance nature  can
flollop  round  like an  old mattress.  A device  called a  Schmitt Trigger  is
needed. A Schmitt trigger is a device that will not change state till a certain
threshold voltage has been reached. Called a hystereesis level. Actually  there
are two. An upper hysteresis and a lower. Assume the  output is  at 'zero'.  In
order to change this state to a one, the input  voltage must  attain or  exceed
the upper hysteresis level. Likewise if the output is a 'one' the input voltage
must fall  bellow the  lower hysteresis  level before  it will  switch back  to
'zero'. Shown bellow.

  There are a number of parts in both CMOS and TTL. The most common of which is
the 7414 hex inverting schmitt trigger. This is  basically an  invertor with  a
preset hysteresis function. It is also available in CMOS called  a 74C14.  Also
in CMOS is the 4093 which is a NAND  gate schmitt  trigger in  a quad  package.
These are commonly used to clean up  nasties in  data streams.  Referred to  as
"squaring up" since it can take a rough waveform and put square  edges back  on
it. Also useful for resetting microprocessors which  is similar  to the  switch
debounce. But we're getting ahead of ourselves.

  Above  is  the full  debounce circuit  using one  sixth of  a 74C14  package.
Remember the output is inverted. If the input  is high  the output  is low  and
visa versa. So If the resistor has pulled the capacitor high then the output of
the  schmitt trigger  will be  low. When  the switch  is engaged  it pulls  the
voltage on  the capacitor  to near  zero. As  it crosses  the lower  hysteresis
threshold the  output of  the schmitt  trigger flys  high until  the switch  is
released. At which time the capacitor begins charging again and when it charges
enough to cross the  upper hysteresis  level, the  output flys  low again.  The
speed at which one can press the switch is determined by  the time  coefficient
of  the  resitor  and  capacitor. If  the button  is pressed  again before  the
capacitor has charged up sufficiently, the output will remain high. It is  also
possible to give the switch on a time coefficient as well by placing a resistor
in series with the switch to ground. However it should be noted that this  must
be significantly smaller than the value of  the pull-up  resistor otherwise  it
will merely equalize  somewhere in  the middle.  No switching  will take  place
since nether threshold was obtained. 

  Which  brings me  neatly to  a word  of caution  about momentary  pressbutton
switches. Most switches have a moment that pressed two bits  of metal  together
shorting the two switch terminals together. This kind of switch has very little
resistance. So little as to make no odds  as far  as our  switch is  concerned.
However it is common these days  to find  switches made  from rubber  actuators
with conductive surfaces. These typically have a resistance of 50  ohms or  so.
But can be as high as 200 ohms. This means that when the switch makes, it  only
pulls down by the resistance across the pad. As if you had placed a resistor in
series with the switch. In  which case  it will  have a  slight switch-on  time
depending on the value of the other resistor. This is particularly important to
realize with TTL. Since a typical pull-up resistor of 4K7 with a  pull down  of
around 200 ohms probably wont reach the lower threshold. Or worse. It will some
times but not when you really need it.

  The reason for pulling up with a resistor and not down is also typical of TTL
Circuits.  Though  it is  possible to  turn the  whole scheme  upside down.  As
mentioned, TTL inherently pulls it self up.  To pull  it down  may require  200
Ohms or less and this may be difficult to work with.  Note that  CMOS and  it's
high impedance nature neither pulls  up nor  down. and  it can  be easily  used
upside down. But it is more general to stick with  the conventions  of TTL.  If
you wished to have an output that is stable high and goes low  when the  button
is pressed, run the output through a second invertor instead.  Either a  normal
invertor from another package or a spare one from the Schmitt trigger package.

  It should be noted that upon power-up  a pulse  will be  generated from  this
debounce  circuit.  This  is  because  the capacitor  needs time  to charge  up
initially. The Schmitt trigger's output will rise very quickly on power-up  and
then fall when the capacitor charges. Ironically this  is the  simplest way  to
reset a microprocessor or logic system. If  the Microprocessor's  reset pin  is
active high (Meaning it's reset when the pin is high)  then the  micro will  be
reset for a time period of  the capacitor/resistor.  When the  cap charges  the
schmit trigger will go low and the micro will initialize and  fetch it's  first
instruction. Likewise if this initial switching is a problem for further  logic
that it controls then a reset circuit could be used  to inhibit  the action  of
the   logic   until   everything   stabilizes.   Just   make   sure  that   the
capacitor/resistor  of  the  reset  circuit  is  something   larger  than   the
capacitor/resistor combos of the debounce circuit.

  Now I mentioned that this is useful for switches up to 16.  After that  there
are a couple of ICs from Nationa Semiconductor in their CMOS range.  Known as  
MM74C922 and MM74C923. 16 key and 20  key encoders  respectively. These  single
chips solutions can debounce up to 20 switches in a matrix with  only a  couple
of external components. It outputs a 4 or 5  bit code  representing the  switch
pressed  and also  a signal  to notify  further logic  that a  switch has  been
pressed. It can be configured  to work  in almost  any situation.  The DATA  on
these can be found at National's most  excellent Web site.  All data sheets are
in Adobe's acrobat format.

  If a very large array of switches is required then it's time  to learn  about
microprocessors and that's a whole other topic of discussion.