Hey,

     >For input only lines I'd recommend resistors (100K to 1M) in series
     >with the inputs and placed near the chip.  The internal pull-ups
     >can't be used then, add pullup resistors on the "outside" end of the
     >isolation resistors.

     I am using the internal pull-ups, as I need to keep the amount of
     external components to a minimum. The I/O lines, because they run
     through very long wires (we're talking 120 feet to the switch), are
     going to have some resistance. In fact, because the machine I am
     building uses a specialized standard, the resistance can go up to 100
     ohms when the switch is closed and the machine should still detect the
     change.


     >For real heavy-duty protection split the resistors in two and put a
     >zener, MOV, or similar device between the midpoint and ground.  Also
     >be sure the resistors have adequate voltage capacity so they won't
     >flash over internally.  Stay away from physically small resistors
     >like 1/8W or SMD ones.  Small capacitors on the PIC pins to ground
     >would further attenuate static spikes that may couple capacitively
     >through the resistors.

     The design is not terribly heavy-duty. I am not taking direct hits
     from lightning or anything like that. I can't use the capacitors
     because I am pulsing the I/O lines at a very high rate and I don't
     want to filter all that out.

     >Bidirectional lines are tougher to isolate.  A series resistor may
     >still be tolerable if the loading is light.  For example if its a
     >matrix scan, make the pullup resistors rather large and add some
     >delay so the long RC times can settle.

     I suppose the series resistor is the best technique. About 2K for each
     line should do?

     Data lines look something like this:

     -|
      |----------o|
      |           |--
     P|----------o|
      |         |
     I|         -o|
      |           |--
     C|----------o|
      |
     _|

     This is an example setup and is not quite like the real application.
     The real application has six I/O lines for switches and the controller
     must detect contact between them and in any combination. I send out a
     pulse into a data line and see where it comes back.

     So you see my problem... If I put series resistors on each of the
     lines, I double the resistance. That gives my 4K series resistance for
     TTL logic levels.

     The Microchip application note for static control said that all you
     need is a capacitor from power to ground. I hope this is all I need.

     Thanks for you help, though.

     Later,

     Eric

     P.S. The environment that this machine will be used in is basically a
     regular wooden gym floor with people wearing rubber shoes. One of the
     I/O lines may go to a tremendous metal floor pad with people walking
     all over it.


______________________________ Reply Separator _________________________________
Subject: Re: Static control
Author:  Mike Keitz <mkeitz@JUNO.COM> at INTERNET
Date:    9/24/97 1:13 AM


On Tue, 23 Sep 1997 14:22:29 -0700 ERIC SCHLAEPFER
<eric.schlaepfer@AUTODESK.COM> writes:
>     Hi All,
>
>     Does anyone here know a really good static control method that
>works
>     with the PIC16F84? I have six I/O lines that go outside of the
>     project's case and go through really long wires to switches. Some
>of
>     the wires are exposed and will be fiddled with by people, and I
>don't
>     want the chip damaged. I've checked the Microchip application
>notes,
>     and they say to put at 10000 pf capacitor from power to ground.
>Is
>     this all that is necessary?

For input only lines I'd recommend resistors (100K to 1M) in series with
the inputs and placed near the chip.  The internal pull-ups can't be used
then, add pullup resistors on the "outside" end of the isolation
resistors.

For real heavy-duty protection split the resistors in two and put a
zener, MOV, or similar device between the midpoint and ground.  Also be
sure the resistors have adequate voltage capacity so they won't flash
over internally.  Stay away from physically small resistors like 1/8W or
SMD ones.  Small capacitors on the PIC pins to ground would further
attenuate static spikes that may couple capacitively through the
resistors.

>
>     I thought of putting in a TTL buffer or an optoisolator, but the
>data
>     lines need to bidirectional.

Bidirectional lines are tougher to isolate.  A series resistor may still
be tolerable if the loading is light.  For example if its a matrix scan,
make the pullup resistors rather large and add some delay so the long RC
times can settle.