Signal Processing

Signal Processing 

 

Modern signal processing encompasses the manipulation of information in both the analog and digital domains. Because of the availability of high speed computing devices (including digital signal processors) a great deal of signal processing work today is being done in the digital domain.  In some applications, however,  it is simply cheaper and more expeditious to do relatively modest processing while the signal is still in the analog domain.  Xecon has considerable experience in both analog and digital signal processing.

 

When digital signal processing is indicated, it is usually essential to do some limited conditioning of the the analog signal prior to its A/D conversion.  A good example of such pre-processing  is the low pass filtering of the input signal in order to prevent aliasing after conversion.  Another critical consideration represents the proper loading of the analog signal into the A/D converter so as to optimize conversion efficiency and minimize noise injection.  Finally, it is necessary to correctly design the front end sample and hold circuitry so that the converted signal represents a true replica of the input value.

 

 A/D converters have many parameters that may be of extreme criticality to the proper functioning of the overall system. Examples are conversion rates and types, linearity, missing codes, and the basic consideration of quantization noise.  All must be carefully examined prior to choosing the proper device for A/D conversion.  Non-linearities and certain missing code problems can be a real disaster in feedback control systems.  Quantization noise is always present in sampled data systems, but can mitigated and minimized through proper design analysis and hardware selection.

 

Once the information has been properly converted into the digital domain it is then possible to apply one of the numerous techniques of modern digital signal processing to the problem.   These are sophisticated mathematical methods whereby one can manipulate the data ensemble to obtain the desired results.  Common DSP examples are filtering (smoothing, emphasis, de-emphasis), the use of transform methods (FFT, DFT, Hadamard, Haar, etc.) to perform spectral analysis and manipulation, and the utilization of convolution techniques.

 

The two-dimensional (spatial) aspect of images make them particularly amenable to digital manipulation. By using the appropriate DSP technique one is able to mechanize processes that can produce remarkably noticeable effects on the image such as: contrast improvement, edge crispening, noise cleaning, and color enhancement.  Contrast enhancement is usually produced by histogram re-scaling and/or the use of statistical differencing.  Edge crispening can be achieved via the use of unsharp masking or a high-pass convolutional filter.  Noise cleaning is usually performed via either median filtering, non-linear smoothing, or the use of a low pass convolutional mask.  A valuable technique that is particularly useful when displaying features developed from digitally processed images is the use of false and/or psuedo color.

 

 Finally the use of DSP  is essential in preparing signals for either compact storage or optimal data transmission through the use of coding and/or data compression techniques.