< Previous Page Return to Title Page Next Page >

Classical Signal Processing 101

  • Classical signal processing (studied by every second year Electrical Engineering student) generalizes Fourier's Theorem:
    • Any signal can be represented by a vector in an abstract space, with an infinite number of dimensions. The "dimensions" consist of a family of signals called an "orthogonal basis set"
    • The signal can therefore be decomposed into a unique linear combination of the signals in the basis set
    • The set of all sine waves (at all possible frequencies, phases, and amplitudes) qualifies as an orthogonal basis set
    • Therefore, every signal has a unique Fourier transform which represents the combination of sine waves necessary to reproduce it

  • A radio receiver (or any device which receives signals from a shared communications medium) works by:
    • Starting with everything it "hears"
    • Rejecting portions which fall outside the desired subset of the basis set (noise and interference)
    • Decoding what remains (the signal)

  • The Shannon-Hartley Theorem (AKA Shannon's Law) dictates the maximum amount of information that can be transmitted without error if noise falls within the desired subset of the basis set or if filtering is imperfect

  • The instructor usually mentions -- but only in passing -- that the set of all sine waves is only one possible orthogonal basis set. There are an infinite number of others, including:
    • Square waves (Walsh Transform)
    • "Sampling" functions (e.g. sinc(t))
    • Wavelets and wave packets

...all of which leads to two very vitally important but generally overlooked revelations....