Hello experts,
In my application, I only used a quadrature encoder as my DC motor feedback which then minus with command position value generating position error. I used this error value to drive a PID unit by which the motor was driven. But after having tuned several PID values, the effect was not good. So I wonder whether it is because I ignored the volecity feedback. Here I want to get some prompt from you experts. Must DC motor position control need volecity control loop? If no, under what circumstance volecity loop can be ignored? If yes, what will happen if I ommitted such loop?

Thank you in advance.

 

There is a historical background to the use of velocity servo loops with position servos. In the early days of applying servos to industrial machines, the goal was to use high-gain position servo loops on industrial machines to attain high performance. This approach was defined by John Dutcher of G.E. as a "hard" servo. Typical position loop gains were 3 or 4 ipm/mil. To obtain the needed accuracy, the position feedback was measured at the machine slide. This meant the machine dynamics were inside the position loop. Since the servos were therefore unstable, the position loop gain had to be reduced to unacceptable levels to attain stability. Servo performance, drive stiffness, and accuracy were not acceptable.

The solution that G.E. used, was to include a high-performance inner velocity servo with the feedback at the servo motor, thus eliminating the machine dynamics from the velocity servo loop. This technique was described
by G.E. as a "soft servo". Position loop gains were typically about 1 ipm/mil. Another feature of the soft servo was the use of extended error with low position loop gains. This meant that position loop gains on all axes of the machine had to be matched. This problem was overcome by the use of a technique called velocity feedforward which in time also included acceleration feedforward control. Thus this meant the machine axes position loop gains no longer had to be matched. Overall performance was greatly improved.

Drive stiffness is proportional to the product of all the servo loop gains. Drive resolution (the ability to maintain smooth feed rates at low feeds) is also related the product of the servo loop gains.

In general, these techniques are still being used for industrial machine servo drives. For very small servo drives it is quite often possible to eliminate the velocity servo loop since the large inertia machine dynamics will not be present. I have available documents that illustrate these principles. In addition there are detailed discussions on these principles in my recent book on industrial servo drives.

Regards
George W. Younkin, PE, IEEE Fellow
Staff Engineer
Industrial Controls Consulting
A division of Bull's Eye Research, Inc.
N7614 HWY 149
Fond du Lac, WI 54935
Ph: 920: 929-6544
Fax: 920: 929-9344
E-mail: [email protected]

 

The "D" (derivative) term of your PID can provide the velocity feedback you need (velocity is the derivative of position), provided that a couple of conditions are met.

First, your position sensor must have enough resolution so that when its digital derivative is taken (deltaP/deltaT), the resulting velocity estimate does not have too much quantization noise.

Second, the position sensor must be tightly enough coupled to your motor that there is no significant resonant mode in this velocity loop. If your sensor is on the load, and there is significant compliance and/or play in the coupling/gearing, you will have problems of the type George describes above.

But the overwhelming majority of positioning servo systems these days do not have a separate velocity sensor; they simply estimate velocity from the position value through derivative action.

Curt Wilson
Delta Tau Data Systems