Bob Dobbs EC42 wrote:
A technique called successive overrelaxation is the most successful in
searching for a null.

Are you suggesting some 'play' in the null seek process?

No, it's simple geometry.

Draw a long narrow ridge running SW to NE on a piece of paper. A line will do.

Now imagine that the middle of your line segment is the actual null sought.

Suppose gain moves you N-S, and phase moves you E-W.

Start S of the SW end of your line. Move N to the best null, then E to the best
null, then N to the best null, the E to the best null, etc.

You observe you follow a zigzag path making no progress up the ridge to the true
null to speak of.

Now do the same thing going _past_ the best null. Your progress is swift towards
the true null.

Successive overrelaxation goes a fixed fraction past the best null each time.

It is unnecessary where the ridge is instead a nice circle, for two steps take you
to the true null. But a long narrow ridge is very common, and SOR is the way to
find the null then.

--
Ron Hardin


On the internet, nobody knows you're a jerk.