Computerised Enhancement of Spectral Signals

In modern instrumentation spectral signals are collected and stored in digital form within appropriate channels of a microprocessor or a dedicated computer. The speed and power of commercial devices are now so great that these signals may be analysed in real time and the results of the analysis used to conftol the behaviour of the entire system. General computer algorithms suitable for these purposes are now widely available but for optimal performance these need to be tailored to the properties and behaviour of particular instruments. In this section we consider specifically the collection and processing of data derived from FM MMW spectrometers. 

In the first case convergence may be expected to be slow, however, and the model parameters will be contaminated by large errors if the signal to noise ratio is poor. In the second case some prior assumptions about the profiles to be extracted must be made. 

The simplest smoothing procedure is a moving window average, in which each of 2N data points is replaced by the mean of itself and its nearest neighbours  

This averaging process reduces random fluctuations by a factor (2A/ +l)i whilst leaving any pedestal or linear background unaffected, though it does distort a quadratic background. More seriously, it broadens and decreases the height of any spectral feature present despite leaving its area unaffected. 

If integrals over spectral profiles are to be determined experimentally, the most 

---