> how are you doing a low pass filter in software ?
>
> maybe just point me in the right direction with the "extra 2 or 4 bits"
part

A single pole low pass filter is easy and fast in software.  If NEW is a new
reading and FILT is the filter value to update with the new reading then the
algorithm is:

  FILT <-- FILT * (1 - W) + NEW * W

where W is a 0 to 1 weighting fraction for the new value.  This type of
filter has the same response as an RC low pass filter with NEW as the input
and FILT as the output.  W determines the time constant of the filter if new
values come in at regular intervals.

Another way of looking at this filter is that the output is a weighted sum
of all input values.  The most recent input has a weight of W, the nextmost
recent W**2, the next more recent W**3, etc.  Or you could say that each new
reading starts out with a weight of W in the filtered value, and decreases
by W for each additional new reading.  In other words, the impulse response
is a step followed by an exponential decay.

Computing such a filter is particularly easy if W = 2 ** -N where N is a
positive integer.  In this case, NEW * W is just NEW shifted right N bits.
FILT * (1-W) is FILT shifted right N bits subtracted from FILT.  Using
C-like syntax, the algorithm is:

  FILT = FILT - (FILT >> N) + (NEW >> N)

It is often useful to maintain FILT and N as fixed point values with N
fraction bits to eliminate roundoff error.  That's what I was alluding two
with "extra 2 or 4 bits".


********************************************************************
Olin Lathrop, embedded systems consultant in Littleton Massachusetts
(978) 742-9014, olin@embedinc.com, http://www.embedinc.com

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