In essence, this approach is a single-slope integrator. Because the
comparator (the PIC port input, which ought to be a schmitt trigger
input, for maximum noise rejection) always trips on a positive-going
ramp, there is potential for some jitter due to noise. Capacitors also
suffer from various non-linearities which can affect the reading.

More sophisticated A/D approaches use dual-slope integration. In this
approach, the capacitor is first charged until the port reads a logic 1.
It is then full-charged to the supply voltage and then discharged until
the comparator reads 0. The two times are averaged.



> ----------
> From:         DREITEK[SMTP:DREITEK@AOL.COM]
> Reply To:     pic microcontroller discussion list
> Sent:         Thursday, January 08, 1998 8:14 AM
> To:   PICLIST@MITVMA.MIT.EDU
> Subject:      Re: Best way to read a POT without the A/D?
>
> In a message dated 98-01-07 11:48:01 EST, you write:
>
> <<
>  What is the best way to read a POT without using an A/D converter?  I
> need
>  to generate values between 0 and 128 from the reading.  It also needs
> to be
>  repeatable i.e. when the POT is not moved, the same number is
> generated.
>  The time limit for reading is not too critical, maybe 250 msec or
> less.
>  Idealy it should not need "calibration" so that multiple units would
> behave
>  the same.
>
>  What I am currently doing is the method shown at:
>  http://www.geocities.com/SiliconValley/2499/answers.html#PIC00065
>  and measuring the charge time.  The times seem to vary quite a bit
> when the
>  pot is in the same position.
>
>  Is there a better way to get more stable readings?
>
>  BTW I am using a 16C84 for the project.  If I can't get it to work
> right,
>  I'll switch to the 16C71 and use the A/D converter.
>
>  Thanks for any help
>
>  Norm >>
>
> Hello Norm,
> I haven't looked at the URL you mentioned but I have used this method
> quite a
> bit and it works very well.  You need to select good quality caps to
> prevent
> drift.  Use a mylar cap.  Also experiment with the cap values.  I used
> a 16
> bit counter instead of the usuall 8.  This steadies things up a bit.
> Then you
> only use the MSB of the counter for your value.  All of the drift will
> show up
> in the LSB which you ignore.
> I use this method when I need simple numeric input.  End users
> understand
> knobs.  I made a bottled water pump for refridgerator ice makers.  The
> timer
> on the pump was set with a simple knob.  It works great.
>
> Dave Duley
>