dsp_plan.txt

- what is dsp
    - digital signal processing
    - signal
        - function conveying information about a system
        - this paper, focusing on audio information although electromagnetic, sonar, image information can also be expressed as signals
    - digital signal
        - essentially a stream of data
    - dsp is the field where modifying signals is studied
        - essentially taking a signal, modifying it in some way, and outputting that modified signal
        - we will focus on taking existing dsp algorithms and talking about how to parallelize them (as opposed to discussing the mathematics of dsp)

- why parallelize dsp?
    - DSP operations are computationally expensive
    - however inherently suited to task partitioning and therefore easy to parallelize
        - hardware
        - software
    - parallel DSP
        - reduce clock speed of hardware for same level 
        - increase clock speed and perform more DSP operations
        - increased data throughput


-> study: software DSP parallelization
- realtime signal processing
    - in its most basic terms, analyzing a signal through a visual medium requires buffering a small piece of the signal, applying a DSP to that piece, and outputting the result of that piece
    - audio frequency visualization -> requires FFT

- FFT - can it be parallelized?
    - what is FFT
        - fast fourier transform
        - samples a signal and divides it into frequency components
        - fourier transform
            - takes a signal, wraps it around a circle
            - function is the "center of mass of the circle"
            - formula

    - performing FFT on a signal (sequential)
        - divide finer and finer and finer
        - eventual lag
    - parallelizing FFT
        - do each frequency in parallel
        - divide finer and finer
        - lag point later

- filtering
    - what is filtering
        - technically, a filter is any medium through which an audio signal passes
        - typically in audio engineering, equalizers and pass filters are used
            - analysis of a waveform done via FFT
        - can be done in hardware using special circuits or via software calculations
        - this demonstration: a software low-pass filter
    - butterworth bandpass bandstop