Essential Tweaks
Part One:
An Overview of Parametric Control of Audio Signal Processing Devices
Audio Signal Processing (ASP) is an activity that, generally speaking, invokes irrevocable changes in a time-varying signal. These processes may involve significant audible transformations, such as those produced by harmonizes and frequency control units, while others may produce effects that are not readily audible but may be perceived using TDR and FDR methods and tools, such as noise reduction and data compression. Dick Moore, one of the leading pioneers of the art-science of digital audio, has remarked on occasion, ...everything is a filter! This seemingly simple statement is full of implications and holds a remarkable degree of truth which becomes more evident as one grows toward mastery of digital audio science. It so happens that filters are important for time-domain modification of signals, for reverberation and many other processes. Signal processing is the Swiss Army Knife of audio production technology - the synthesis engine known as the Motorola 56001 integrated circuit, can be programmed to perform a wide variety of signal processing tasks from gating to harmonization. Thanks to VLSI (very large scale integration) the price of signal processing has dropped while power and sophistication has increased such that sound processors which cost over $10,000 ten years ago are now available for under $400.00!
ASP today is really a subset of Digital Signal Processing, - a vast and esoteric science which attempts to solve a wide range of problems in a vast array of fields - and it is reasonable and expected to discuss musical/sonic transformations effected by digital means as Digital Signal Processing (DSP) rather than as ASP. It is the digital part of DSP that is of such importance to us. The implication is that DSP engines can be arbitrarily programmed to perform certain kinds of tasks on signals. This means that innovation is focussed clearly on the end result and the process to achieve it rather than on the hardware technology alone. In this way, DSP can be seen as an organic science which allows for the development of software solutions to audio problems.
signal processing - a term that applies to all devices used to modify signals during recording and transfer applications.
Signal Processing Devices/Applications |
Two Basic Types (examples) |
sound modifying (discernable) musical applications |
compressors equalizers reverberation |
signal modifying (not discernable) intended to improve transmission quality |
HF limiters (high frequency) noise reduction distortion correction devices |
Understanding these two contrasting classifications of ASP technology is important. Yet, as can be discerned from the broadness of the classifications, the vast array of tools and techniques for ASP are only alluded to here. We may profitably expand this table into the following:
Sub-type Classification |
Audio Application |
Dynamic Range |
Reshaping the amplitude profile of a sound: compression, expansion, gating, ducking, companding |
Mixing |
Combining multiple tracks of audio: crossfading, cut and paste, addition |
Filters/Equalizers |
Modifying the frequency spectrum of a sound: parametric, graphic, paragraphic low-pass, high-pass, band-pass, band-reject may be time-variant |
Time Delay Effects |
Creation of echoes, chorus effect, flanging, phasing Sometimes using a multieffects unit - but often a stand-alone unit which a recirculating memory buffer |
Convolution |
Simultaneous time-domain and frequency domain transformations: often used to create reverberations effects by convolving an impulse response with a sound |
Spatial Projection |
Early refection generation, reverberation, sound path vector design - all serving the elaborated concept of an acoustic space: |
Noise Reduction |
de-hissing, de-crackling, etc. Applications of the FFT for sound analysis and subsequent inverse filtering to remove noise |
Sample Rate Conversion |
Changing speed of the sound, with or without pitch shift. Using either decimation or interpolation. |
Time-Compression |
Changing pitch and/or duration irrespective of the other. |
Each of the techniques outlined above has its own specialized tools and associated techniques. Working with the concepts and finding musical applications for using the techniques will speed their mastery. An engineer moving into the 21st Century needs to be as conversant with these concepts as possible.
Dynamic Range:
In the audio field we are concerned with the equalization of the overall dynamic range. We are also concerned with the individual signal and control of its dynamic range. In order to shape the dynamic range we use tools that are generally referred to as compressors. These units, as we know, can achieve a number of effects through a group of similar processes:
compression - reducing peak amplitudes above user-defined threshold
expansion - the inverse of compression, increasing peak amplitudes above a threshold
companding - combining both compression and expansion
gate - closing off signal flow according to a user-defined threshold
keyed effects - all processing can be keyed by the amplitude of another signal
In digital audio we are also concerned with some manipulations that extend beyond the normal context of the dynamic processor. Chief among these is normalization.
Normalization is the process whereby a signal is first analyzed to find its highest amplitude peak. Then, either gain is added or subtracted to the entire signal in order to adjust the overall amplitude. In a sense, this is just a fancy way of adjusting volume. The important distinction is knowing precisely what the signal level is before normalization and what it will be afterwards. The reason this is important for digital audio is that there is no headroom above the peak amplitude value of 1.
Digital audio (16 - bit words) 0 - 65536 - actual numbers possible. The number 65536 will equal 1 in this way of thinking. And 0 would be the lowest value expressed, and would be the theoretical noise floor of the system.
If we want to take full advantage of the clarity and sonic precision of digital audio we need to be generating amplitudes which will be encoded as larger 16-bit numbers. Another style of digital processing which attempts to do this is dymanic peak-limiting with re-dithering and noise-shaping. Peak limiting is a type of dynamic processing which reduces only instantaneous peaks in a signal and not a larger duration of signal as a compressor may. In fact, this is the type of dynamics processing that the L1 plug-in on Sound Forge does.
end of part one