Sun, I think the foremost thing to remember is that very little in the world responds instantaneously. Everything takes a finite amount of time to change this is the phase component. Take my little chinese generator I bought the other day. It has a 63cc 2 stroke engine coupled to a 750w alternator. Its control system consists of a mechanical governor coupled to the throttle. When I place a 750w load on the generator, the motor slows and as the governor opens the throttle fully, the motor coughs then slowly picks up speed again until it overshoots and speeds up/slows down a little until it settles again. I remove the load and the motor overspeeds - the governor shuts the throttle and the motor regains the correct speed eventually. What we see here is a classic control problem. The motor has only limited power and can only accelerate at a given rate - this is phase lag. The governor is a proportional control - too slow so it opens the throttle a proportional amount. What this generator needs is some integration to slow down the opening of the throttle - opening it fully when under load causes the motor to cough. If the governor opened it more gently, the motor would not cough and would most likely accelerate faster. Also, if the governor could sense that the motor is accelerating faster than it can respond then it could estimate when to back off the throttle before it overshoots the target motor speed. As we can see, a purely proportional controller will most likely overshoot, then oscillate a little before settling (it may continue to oscillate). So by adding some integration and differentiation we can better control the feedback loop. The net aim of a control system is to reach the control setpoint as fast as possible without oscillation or overshoot and to maintain the setpoint value within the prescribed tolerance. It all comes down to phase and magnitude - thus the Bode plots. By doing some measurements on the response time of my generator, I could model the control system and simulate a better control system and derive the required P,I and D coefficients for a pid controller for the generator that hopefully would work better than the simple governor setup it has at the moment.

Think about what you do when driving a car - you're travelling at 60km/h (laju) and then you start to travel up a hill. You gradually press the accelerator to regain the lost speed. Then you go over the hill and start to gain too much speed (laju besar) so you must take pressure off the accelerator and maybe even use the brake. This is your human control system at work.