Lecture 06: Signal Processing

Some of the notes that follow were begun by Owen Green but have been subsequently adapted.

To save you scrolling down, here’s a link to the processing manual you made in class:

masterManual

Introduction

Whilst signal processing per se starts with our recording choices, as workers with sound we are confronted with a slightly bewildering array of processing choices for subsequently shaping our audio. Really, though, processors tend  fall within a few categories.

In keeping with the general thrust of this module, what we are concerned with here are the usages and techniques we can employ as part of our sound designing, ahead of the technicalities of exactly how everything works (although, some appreciation of this is needed to grasp the particular affordances of a piece of equipment). Here is an image showing some possible types of work that we might do with signal processors:

Processing Options: Different types of task, different aspects of sound (Image by Martin Parker)

Processing Options: Different types of task, different aspects of sound (Image by Martin Parker)

So we see four different broad parameters of sound that might be available to work with, and some associated task types. ‘Cleaning’ might entail noise removal, ‘tidying’ might involve the removal of redundant material, etc.

The content that follows will have been partly developed by you, the Sound Design Media students, as you pool knowledge about the following (broad) categories of processor, using the above graphic to help organise your thoughts.

Before we go in to these details, let’s first have a look at some Tonmeisters in action:

Georg Solti – The Golden Ring (Original BBC Mono – 1965).

In the above example, watch Birgit Nilsson move towards the microphone in a quieter bit. That’s dynamic range compression right there.

Preliminaries: Inserts and Auxes; Online and Offline

  • An insert means the process is inserted into the main signal flow, so that we get all or nothing (unless there’s a mix control).
  • Alternatively we can set up a separate bus that runs the effect, and can send a designated amount of our dry signal to that so that it is mixed with the processed.
  • This gives us more control than mix knobs, and we can also send multiple signals to the same effects, for glue or for saving processing.
  • Commonly, things like reverbs are used on auxes.
  • Online means a ‘real time’ effect that we can hear as we playback, and tweak parameters with immediate sonic feedback [we should hear exactly the same sound when bounced].
  • Some processes are too intensive, or need too much knowledge of ‘the future’ to be run like this, so can be run ‘offline, in which case we have to process, audition, tweak, process, audition, etc.
  • The trade-offs are between immediacy and workflow for online stuff, but sometimes higher quality for offline processes.

Today, we’re going to a make a sound processing book

First we’re going to assign people to different tasks, you’re each going to write a chapter of this book, explaining a different sound processing technique or tool.

Download the chapter here: chapters. 

NoiseGate.org
MultibandCompression.org
Limiting.org
Compression.org
BitCrushing.org
Chorusing.org
CombFiltering.org
Convolution.org
DeEsser.org
DelayBySample.org
DelayByTempo.org
DeNoising.org
Distortion.org
Exciter.org
GainChange.org
NotchFilter.org
PhaseVocoder.org
ReverbEmulation.org
ReverbIR.org
RingModulation.org
SpectralFilter.org
SpectralDelay.org
StereoWidth.org
Waveshaping.org
TimeStretching.org
LowPassFilter.org
HiPassFilter.org

Upload your file here: outlook.office.com/owa/?path=/group/sounddesignmedia2018-19-group@uoe.onmicrosoft.com/files

Basic Processes: Filters and Delay

  • Filter Types: Low pass, high pass, band pass, shelf, notch, all pass, bell / cut
    • Cleaning for removing noise, hum, whine, rumble
  • All-pass: Useful for building other filters, making phasors
  • Comb-filters: Just a delay line!
    • Useful for spatial effects (neutral sounding early reflections, extended tails): small delays, low feedback
    • Tidying (can mask grittiness and smooth transients)
    • Slapback (25-50ms) / Echo (> 100ms)
    • Making robot noise (<50ms, high feedback)
    • Precedence effect: Our localisation of a sound depends both on level and time of arrival, so delay can be used to augment pan controls.
    • We can make a more natural sounding resonator with a comb by putting a low pass filter in the feedback path, which will damp the higher frequency resonances.

Bread and Butter: EQ, Reverb, Modulation

  • EQ:
    • Basic controls of a parametric band: filter type, boost / cut, frequency, Q (bandwith)
    • Nominally: Bass ~20-100 Hz; Low-Mids ~100-700Hz; High-Mids: 700-3kHz; Highs >3kHz
    • Cleaning (as above), but also targetted tidying in the bass and low mids is often needed
    • Cuts less noticeable than boosts, by and large; so boosts more suited to creative shaping, cuts to transparent ‘surgery’; can use sharper cuts before becoming noticeable.
    • Spatial perspective (roll of highs and lows, draw back high mids)
    • Presence (boosts to high-mids / lower part of highs)
    • Mass (Boosts to bass and low mids)
    • Air (Targetted boosts to highs)
    • Sculpting: very low Q with very small boosts cuts in the right place on an overall mix can make surprising difference, act as glue, increase coherence

 

  • Reverb:
    • Typical Controls: Wet/dry, revb time / decay time, pre delay, ‘diffusion’, room size.
    • Algorithmic reverbs typically work on a model of sparse early reflections followed by dense, noisier diffuse tail.
    • Creating spaces: putting our sounds in context
    • Glue: Bringing different material together
    • ‘Dizzyiness’ and other psychedelic effects
    • Distance effects (further back, spread across spatial field)
    • Smoothing and masking, as with delays
    • Time ambiguity, by smearing transients we can actually adjust the ‘feel’ of material, appearing to push it back (relative to other material).

There are other spatialisation techniques, such as HRTF filtering for making material ‘binaural’ (for 3d headphone listening), decorrelation for playing with perceived spatial perspective and size.

  • Modulation:
    • Amplitude modulation: <~20 Hz produces tremolo effect; fast produces audible side bands.
    • Ring modulation: Almost a synthesis method! Slow produces an undulating phase oscillation, fast produces very audible side bands. Quite unpredictable on complex sources.
  1. Time Varying I: Modulation of parameters (Phasors, Flangers, etc…)

Modulated delay lines: flangers modulate with sine for virbrato / phasing fx, chorus with noise for thickening

 

  1. Nonlinearities: Distortion

Transfer functions create new spectral components for valve effects, precise new harmonics. Ordering matters.

  1. Time Varying II: Dynamics processing

  2. Offline processes: Resampling, Time stretching…

  3. Spectral processing preview