My problem with PID is people don't wrap it around a model and use it to correct the errors in the model. Often their tuning ends up "carpet-bumping" between m

My problem with PID is people don't wrap it around a model and use it to correct the errors in the model. Often their tuning ends up "carpet-bumping" between multiple critical points in the system. But as others have mentioned, it gets you ok, predictable performance.