> Say we multiply by 10 to get a faster response for the high level
> If we sample in 10kHz we would reach 20 000/10kHz*10=2 sec
> And we would reach 0 level in 20 secs: 20 000/10kHz*1=20 sec
> The counter will _never_ be reset since its an endless process.

I have now implemented and tested in production the solution I talk
about here. It works outstanding !!! It follows the tank level and gives
a nice smooth reading of the level on the LEDs. No false triggers
what so ever, and pulse response is close to perfect.

By changing two constants in the algorithm I can easily change the
responsetime for either increasing or decreasing level. Currently I
have 2.5 times faster response for increasing level. Seems to be
a good value in combination with the mashine.

Must say this was a more edjucational project than expected !
/PJ

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.