Hi Richard,

As you think about the PID implementation, be sure to use "enough" arithmetic resolution in the sensor measurement and the algorithm. Without "enough" resolution, oscillation will appear in the countrol output. The oscillation can look like an unstable system as found in the theory of linear systems. It's really a limit cycle created by the lack of sufficient resolution. Beware also that when various control elements such as amplifiers and motors have high gain, the need for more bits of resolution increases.

You might write a quick a dirty VB program to simulate your speed control if you don't have access to a mor powerful tool like Matlab. Simulate quantization effects using a dead zone of appropriate width, while making use of normal floating point arithmetic libraries in VB or in whatever language you wish to write the simulation. Apply the quantization to the model an A/D process and as much of the algorithm as you wish.

Here's a couple of good reads on the Zeigler-Nichols procedure -

http://www.eng.adelaide.edu.au/subjects/.../chap4.pdf

http://www.shu.ac.uk/schools/e...torial.htm

Bruce