Audio Processing and Filter Techniques

A set of flashcards covering key concepts, definitions, and processes related to audio filters and processing techniques.

What is the difference between FIR and IIR filters?

FIR filters have a finite impulse response while IIR filters have an infinite impulse response.

What do FIR and IIR filters have in common?

Both FIR and IIR filters have a delay.

How can a simple FIR lowpass filter be created?

By delaying a signal by one sample, using phase differences to sum waves.It works because when you sum the waves that are 180 degrees out of phase, they sum to 0. At Nyquist frequency, they sum to 0. As you approach 0Hz, phase differences become smaller

How is cutoff frequency defined in lowpass filters?

Cutoff frequency is defined when the amplitude increases by -3dB or reaches 0.707 of peak amplitude.

What is the bandwidth in a bandpass filter?

The bandwidth is the frequency distance between the two cutoff frequencies.

What does the quality factor (Q) of a bandpass filter represent?

Q is the center frequency divided by the bandwidth; a higher Q indicates a narrower bandwidth.

How can the center frequency of a bandpass filter be calculated?

Center frequency is calculated as the square root of the product of the two cutoff frequencies.

What is the attenuation per octave of a one-pole lowpass filter?

-6dB per octave.

What is the attenuation per octave of a two-pole lowpass filter?

-12dB per octave.

What causes a comb filter to output harmonic spectra?

Harmonics reinforce at periods 180 degrees in phase while others cancel out, resulting in a clear fundamental.

What basis is used for the Karplus Strong Plucked String Algorithm?

The comb filter modified by adding an amplitude envelope before it goes into delays

What is the source-filter model in audio processing? How does it emulate the voice?

It consists of a source wave (e.g., sawtooth or square) and a 3 parallel filter bandpass filters to create shapes similar to the vocal tract.

How was flanging created with audio tape?

By using two tape recorders with one slowed down and then sped up to create phase shifts.

How is the flanging effect recreated digitally?

By modulating the delay time of a comb filter using an LFO.

What is needed to emulate the sound of chorusing?

You take several sound sources (voices) and add a random delay to each. This creates slight time and pitch differences.

How is the wah-wah effect created digitally?

By using an envelope follower to control the center frequency of a bandpass filter.

What does RT60 refer to in audio processing?

RT60 is the time taken for the original signal to attenuate by 60dB.

What are the three parts of reverberation?

Direct sound, early reflections, and late reflections.

Why are comb filters placed in parallel in Schroeder’s reverberation?

To emulate the effect of having four walls.

What quality do comb and allpass filters possess that make them suitable for reverberation?

Recursion.

Create a block diagram of Schroeder’s original reverberation that uses comb filters and all-pass filters.

4 Parallel Comb filters that all add together and then feed into an all pass filter, which feeds into a second all pass filter

How does a channel vocoder work?

By taking the amplitude envelope of a modulator (e.g. voice) and imposing it onto the oscillator.

What is FFT size?

The size of each frame in samples.

What is an FFT window? What purpose does it serve?

The window is in a frame, it’s a method to create smooth transitions within a sample to avoid clicking and artifacts.

What is frame overlap and its necessity?

Overlapping frames prevents amplitude modulation by allowing for smoother transitions. (typically 4 overlaps per frame)

What is a bin in audio processing?

A frequency interval representing a specific range of frequencies in spectral analysis.

What is the tension between time resolution and frequency resolution with the FFT?

Larger FFT size improves frequency resolution but worsens time resolution.

How does frame overlap help ease the tension between resolution types?

Frame overlap allows more updates per second, improving time resolution without worsening frequency resolution.

Why are the bins above the Nyquist frequency considered redundant?

They mirror the bins below the Nyquist frequency, thus are symmetrical.

Given an FFT size of 2048 samples with an overlap of 4, and a sample rate of 44,100 Hz, how many bins will be created? What is the frequency width of each bin? How many of those bins would typically be used for audio processing?

a)2048 (half the FFT size)

b)44,100/2048 (21.5, sample rate/FFT size)

c)1024 (half the FFT generated)