Which Filter?

In the this guide, important defined terms are highlighted (any item with italics will open in a small popup window) and are linked to a glossary where they are explained.
We have one simple basic suggestion, IF IN DOUBT CONTACT US! We have been doing this for a few years and many of us used to be filter users before we came here, we may have heard your application before.

This section asks the basic questions which apply to any filter requirement. There is no single filter solution to any application: it's always a compromise to find the best filter for your application. We like to address the following questions when recommending a filter:

What does the basic filter need to achieve?

The more difficult the task, the more sophisticated the filter needs to be in its signal response. There are four important criteria for selecting the right filter; do you need to:

  •   Separate wanted frequency components from 'nearby' unwanted ones. Here the difficulty is set by the frequency ratio between wanted and unwanted components. As this number gets closer to 1 (from either side) the job gets harder, and you need a filter with more and more discrimination.
  •   Separate wanted frequency components from unwanted ones which are much 'louder', ie higher in level. The difficulty in this case is determined by how much you need to reduce the level of the unwanted signals. The bigger the reduction needed (expressed in decibels), the more rejection you require.
  •   Preserve the accuracy of waveform reproduction of the signal made up of the wanted frequency components. All filters change the signal passing through them in some way. The more accurate you need your waveform reproduction needs to be, the less vector error can be tolerated from the filter.
  •   Control the disturbance caused to the output signal by a sudden 'step' change in the input signal. All filters take some time to recover after being 'shocked' by a sudden change in the input signal. The more cleanly you want the system to settle down, the lower the overshoot required in the filter response.

We will refer to these four criteria in the rest of this guide.

What are poles and how many are needed?

Briefly, a filter with more poles needs more components, making it larger and more expensive to make. The increased complexity allows the filter response to offer improved performance in one or more of the categories above. Generally, the more 'difficult' your application is, the more poles will be needed in the filter response. 
IMPORTANT NOTE - In many systems the cost of the filter is a small percentage, and in many cases selecting a lower performance filter to save money is a false economy. 
We have many filter responses designed to give a spread of capabilities. We give them 'Response' numbers; some of them also have 'familiar' names. Examples of these families are the Butterworth and Bessel filters, which are available in four and eight pole versions on many of our products. The chief application for Bessel filters is where the fourth criterion, overshoot, is important; for such tasks they excel, although they are really rather poor at the other three.
We believe that, in many signal conditioning systems, the potential for waveform distortion from the filtering is underestimated. To reduce this, we recommend the use of filters which offer good performance in the third criterion, wherever this can be done and still meet the requirements imposed by the other criteria. Using the latest computer techniques, we have designed a range of responses which excel in this respect; our generic name for these filters is Low Vector Error filters.
The eight pole filter range offers several excellent choices. For anti-aliasing applications, the Kemo Response 01 filter (a member of the class of elliptic filters) has become the industry standard response shape. For general purpose filtering the linear phase, Low Vector Error Response 41 provides 82dB rejection at three times the cutoff frequency.
We cannot do justice in this guide to the full range of Kemo filter responses; not only are there many more lowpass responses than the ones highlighted here, but there are a large number of highpass, bandpass and bandstop (notch) responses. Advice on how to assemble the best filtering for your signal conditioning needs is always available, so please do contact us with any questions you may have - we'll do our best to help.

What other capabilities and features?

The Products and Applications section of this site describes key products, separated into various groups according to common themes of usage which have cropped up repeatedly over many years. Products range from benchtop filtering instruments of great versatility, through complete signal conditioning front-ends for data acquisition systems, to basic but state-of-the-art filter modules.