why not use direct measurement system?


Thread Starter


In motion control, indirect measurement system is usually used in the position control loop. But the indirect measurement system cannot provide a good accuracy due to the elastic connection between the motor and the load, or due to the backlash. So why don't people use the direct measurement system which can provide a better accuracy?

Davis Gentry

That depends on the application and the
accuracy/repeatability requirements. If you are using feedback on the load only and have any backlash or elasticity in the drive train you will burn your motor up dithering around a small deviation in load position which can be addressed only with a large deviation in motor position. At least you will if you have tight requirements.

Many of the applications with our controllers utilize dual feedback - both on the motor and on the load, with separate inputs for the velocity and position PID loops. Sometimes we also see a third feedback of some sort brought in as a modifier to the PID loop. One common example of this is a force feedback for applications where a force setpoint is used in addition to a position trajectory.

Davis Gentry
Delta Tau Data Systems
I agree with What Davis Gentry said;
May I add am additional prospective; (hopefully a simple view)

>From a control point of view, feedback closest to the controlled device is best

>From the end result point of view, feedback from the direct measurement is best

The difference is the mechanical coupling.
Is it stiff and stable?
Can it be modeled?

Ah, now that IS the challenge.

Hope this helps.

Dave Kane
Thank you,Davis and Dave!
I have set up the model in the Simulink and try to find the problem.
The system is a simple mass-spring-mass (motor-elastic connection-load).
The control block has a P controller for the position control loop,PI controller for velocity control loop, and a time lag 'Tequi' to model the current control loop.
And I tried to using the indirect measurement and direct measurement as the position feedback. The velocity feedback is the from motor speed(so the speed feedback is indirect). The result shows that I must lower the proportional control gain in directment system in oder to stabilize the same system. So, the dynamic response is worse than in the indirect measurement system.

Now,i bring up the question, is there any possibility to improve the dynamic response in the direct measurement system?

Davis Gentry

That depends on your controller. Our controllers
allow you to state an amount of backlash and the
takeup rate for that backlash. This can help, but
only is used when you change direction of motion, so will not help with highly dynamic flexure in your system. We can also use a deadband range - when position is within a specified distance of the setpoint P can be either multiplied or divided, depending on how you set it up - in this case it would be divided if P is showing better response at lower values. Last, you can use derivative in your algorithm to help dampen response to some extent.

Davis Gentry
Delta Tau Data Systems
I think it depends on the particulary system and needs. I am in fact right now programming a four axis welding system, on three axis we only are interested in velocity, so we are using ball screw mounted encoders, but on the fourth we want absolute postion, without backlash, so we are using a direct measurment method, linear encoder, from Danaher on that axis. Thankfully, the fourth axis is a short axis. Long linears can get very expensive.