Abstract
Spectrum analysis of an input signal is one of the most common processing tasks of a Synthetic Instrument (SI). Instruments that perform spectrum analysis have evolved over the years from parallel filter bank implementations through sequential swept frequency implementations to modern windowed Fast Fourier Transform (FFT) implementations. Frequency range, desired bandwidths, dynamic range, and other considerations dictate which of the three techniques or combination of techniques are best suited to a particular application. The parameters of a spectral analysis for acceptance testing are usually specified and at minimum include resolution bandwidth, spectral span, and dynamic range. A system designer can select various combinations of anti-alias filters, A-to-D converter sample clocks, and firmware-based FFT size to precisely meet these requirements. More often than not, strange design decisions or compromises are invoked. Examples include selecting sample rates which are multiples of a power of 2 (4.096 MHz for instance) to achieve a specified spectral resolution in a 4096 point FFT. An engineering compromise may be that 1024 Hz is close enough to 1000 Hz that it does not matter because no one will notice the difference. This is incorrect. A more cost effective and versatile option, free from capricious design numbers and questionable engineering compromises, is based on the flexibility and capability of embedded DSP engines. These engines are applied to the task of performing arbitrary sample rate changes in the DSP domain, thus obtaining precise matching of system parameters to specified parameters.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have