Measurement Noise in PID Control Loop

Snapshot of a part of the front panel of the simulator:



Description of the system to be simulated

In this simulator a process represented by a transfer function (a seried combination of two first order systems and a time delay). You can add random (white) measurement noise. The measurement signal can be filtered by a first order filter or a dead band filter.


Aim

The aims of this simulator is to observe the unfortunate consequences of meadurement noise in a control loop, and to experience ways to reduce these conseqences.


Motivation

Measurement noise is a problen in most practical control loops. The noise gives false or erroneous process information which, via the controller, may give unfortunate excitation of the actuator and the process to be controlled. Below are a few examples of measuremen noise:

  • Ultra sound based level measurement in liquid tank. The level is measured based on the reflection time from the liquid surface. Waves on this surface will produce measurement noise.
  • Noise from electric components.
  • Noise induced by the mains.

If the noise source can not be eliminated, the noise can be attenuated by using a proper signal filter.


Tasks

If you use the default process parameters you can use the following PID controller parameters (found using the Ziegler-Nichols' closed loop method:

Kp=2.8; Ti=1.2; Td=0.3

The controller is assumed to be in auto-mode unless otherwise stated in the tasks below.

Start the simulator. Introduce random measurement noise.

  1. Measurement filtering using a first order lowpass filter:
    1. How does the bandwidth of the lowpass filter influence the ability of the filter to attenuate the measurement noise? (Observe the signal ymf at a few different bandwidth values. The noise attenuation can be quantified using the variance ymf.)
    2. How does the variance of the control signal depend on the filter? (Observe the variance with and without filter.)
    3. What happens to the stability of the control loop if the filter bandwidth is set to a very small value? What can be done to avoid the stability problems observed?
    4. What happend to the correspondence between the unfiltered process measurement and the filtered measurement if the bandwidth is set to a very small value?

     

  2. Measurement filtering using a deadband filter:
    1. How does the size of the deadband filter influence the ability of the deadband filter to attenuate the measurement noise?
    2. What happens to the stability of the control loop when the deadband filter is used in the loop?

[SimView] [TechTeach]


Updated 18. January 2008. Developed by Finn Haugen. E-mail: finn@techteach.no.