Dear Colleagues

In the old days – we couldn’t survive without decent filters on our camera to take a good photo. However, today much can be done digitally after the photo has been taken and the filtering and photo enhancement done later. As long as you have grabbed all the critical data with quality high resolution images together with good lighting.

A Noisy Old World
When you are working with real equipment and instrumentation, you will often see noisy signals. I am thinking particularly of the Process Variable (PV) – the signal from a temperature, pressure or level sensor that you are happily measuring and using to control a valve or process. In many cases, much to your chagrin, you will notice noise on the lovely clean temperature signal. And you will immediately consider the use of filtering to clean out the noise and to give you a nice smooth signal which you can then control with.

Higher frequency noise can derive from electrical interference, jitter and sampling problems. Process noise is generally lower frequency and can arrive from bubbles or splashing or even temperature variations from mixing activity. The first port of call is always to try and fix the problem at the source - the process or instrument. Filters aren’t the optimum solution as they tend (?) to hide potential problems.

Having noise on your clean PV signal can upset your PID controller with the control output bumping around a bit unnecessarily because of the noise with Derivative Control. The greater the filter constant, the smoother the signal.

Please Do Not Use Filtering
As most experts will warn you – do not use filtering as this will add further lags to your control action which will move you to a more unstable regions of your control regime. Generally filters for instrumentation operate as low pass filters. They pass the low frequency signals (slowly moving signals) and block (or attenuate) the higher frequency short spiky signals.

As the filtering action is increased, the original signal starts to lose its real process characteristics.  So large filtering action should be avoided.

A way of looking at the problem is to consider that increasing the filter time constant tends to add to the overall lag of the system. As you may dimly recall from your control theory– more lag in a system tends to push the overall system towards the unstable region and this can result for example - in the control action overshooting.

The tragedy of course is that the plant operators tend to like filters as they clean up noisy unpleasant signals and give one a more friendly view of what is happening in the process. But naturally, the filters could be disguising a real problem.

A Toolbox of Suggestions
Some suggestions on use of filters when you are designing your next control system are:

  • Avoid use of filters in your control or PLC system if you can, and try and fix the problem at the instrument itself i.e eliminate noise at the source.
  • Examine the control action (output of controller and valve) and see the process noise is creating a real problem or not.
  • Train operators and techs on what filters are and why they are not good to have.
  • Filter time constants should be less than 10* (the process dead time + scan time of PLC)

Before I go - what on Earth is Anti Aliasing?
I have considered anti aliasing filters in an earlier blog and these should be used if your have unwanted noise for example from high frequency variable speed drive harmonics intruding on your process signal. If the sampling rate of your PLC for example is too low compared to the PV signal; you may get an aliasing problem. Effectively meaning that the signal your PLC thinks it is dealing with – is actually not the correct representation of the real world signal (because of sampling). This means that you are not controlling with correct information but something which is a badly constructed digital signal which has no connection to the reality of the process that you are measuring. Filters are required in this case to eliminate these higher frequency harmonics.

My grateful thanks to Michael Brown ( and a few other process controls savants for articles on filters which I have adapted.

Yours in engineering learning