Re: [CAD_CAM_EDM_DRO] Re: Dedicated mb for PID

Since operating frequencies are generally below 100Hz (6Krpm) it is possible
to use a greatly simplified continuous model as an accurate approximation.
It's been a long time since me and LaPlace was drinking buddies but the
Galil filter looks to be a transfer function -

F(s) = P + sD + I/s

Some algorithms for convolution of the FIR filter are given in Numerical
Recipes for Fortran. They don't appear to be compute intensive.

Doug

----- Original Message -----
From: Les Watts <leswatts@...>
>>>>
Fast PID algorithms I have seen use a finite impulse response digital
filter set to implement integration and differentiation. This is simply a
short list of coefficients in memory. Each one is multiplied by a datapoint
in time order then the data is shifted one and it is repeated, forming a
digital approximation of convolution. Simply by choosing proper
coefficients band pass filters can be used to control high gains at high
frequencies for derivative (for noise immunity) and at low frequencies for
integral.

I have written some in c for dos that were blazing fast at least on a pc. I
seem to recall running some at 100 kHz for a project a while back in a
pentium even though I used floats.

Les
>>>>