Noise Reduction by Filtering

We have defined noise as something that masks wanted signals and does not have a strong correlation to the signal. To maximize the signal to noise ratio it is advisable to understand the nature of the wanted signal and the noise. Filtering is a real science, but like so many things, it follows the 80:20 rule. Simple filtering theory requires only 20% of the effort but can provide 80% of the results! An effective starting point is to qualitatively compare signal and noise power spectrums.

Power Spectrum

The power spectrum of a signal or noise waveform is a measure of the power density over a range of frequencies. It provides a powerful tool for understanding filtering issues.

picture NF1

In the power spectra plot, the area under the curve is related to the power of the signal or noise. So, by selectively rejecting some frequency bands, it is possible to favor the wanted signal over the noise. This is what filtering is designed to do and is how the signal to noise ratio is improved.

If the power spectrum of the signal and noise are very different, then filtering is usually easy to accomplish. Unfortunately the converse is also true - if the power spectrums are very similar then filtering can be extremely difficult or impossible.

Prevention is Better than Cure

Eliminating noise at its source always pays dividends. This is particularly true with low frequency (like 50/60 Hz) inductively coupled noise that can be difficult to filter. The shielding of cables can also minimize capacitively coupled noise.

Filtering Strategies

   Filtering can be done at various points in a measuring system. There are a number of "rules of thumb" that can be followed:

• filter as close to the source as is practical

• identify the signal requirements

• keep the bandwidth as narrow as possible

• pay special attention to line hum (50/60Hz)

• know how to identify"bad" readings

In practice, filtering can be done at various points in a system. Frequently a combination of filtering points and methods is employed. As a user of the measuring device you may not have control or even be aware of internal filtering.

Hardware Filtering

   There are numerous types of filters that are implemented with electronic components. These generally have a fixed characteristic although sometimes they can be modified under software control.

RC Filter

LC Filter

Synchronous Filter

Switch capacitor filter blocks

Integrating Analog to Digital Conversion

By using the analog to digital conversion (ADC) process as a filter,

Software Filtering

By incorporating microprocessors in measuring systems, it is possible to manipulate a series of readings mathematically. This is digital signal processing (DSP) and can be done by any microprocessor although special "DSP chips" have been developed to do the task quickly.

Far point rejection

Simple Averaging

Simple DSP

Complex DSP