From the point of view of signal flow, a DAFX is a device that takes an un-processed input signal and outputs a processed signal. In most cases the user can control the signal-processing behaviour by manipulating a number of parameters through a graphical user interface. The aim of the user is to manipulate the signal-processing parameters in order to produce the desired transformation of the input signals. Figure 13.1 shows the standard implementation of a DAFX, where x(n) is the input source and y(n) is the output resulting from the signal processing.
Figure 13.1 Diagram of a DAFX and user.
In an automatic-mixing context we aim to aid or replace the task normally performed by the user. In order to achieve this, some important design objectives should be performed by the AM-DAFX:
1. The system should comply with all the technical constraints of a mixture, such as avoiding distortion and maintaining adequate dynamic range.
2. The design should simplify complex mixing tasks while performing at a standard similar to that of an expert user.
3. For sound-reinforcement applications the system must remain free of undesired acoustic feedback artefacts.
In the case of a system used for live performance mixing, the automatic system must avoid undesired acoustic feedback artefacts at all cost. For this reason several stability solutions have been developed, for example gain shearing [Dug75, Dug89] and self-normalisation techniques [PGR08a, PGR08b]. In most cases these techniques try to prevent acoustic feedback by ensuring a maximum electronic transfer function gain no bigger than unity. This ensures that regardless of the changes in signal-processing parameters the system remains stable.
A diagram depicting a generic AM-DAFX can be seen in Figure 13.2, where xe(n) is an external source. Our aim is to emulate the user's control parameters. An AM-DAFx is formed of two main sections: the signal-processing section and the side-chain processing portion. The signal-processing algorithm is a standard DAFX processing device and can include a user interface if the AM-DAFX is meant to give visual feedback for its actions. The analysis decision section of the automatic-mixing algorithm is what we will refer as the side-chain processing. The analysis decision-making portion of the automatic mixing tool takes audio from one or more channels together with optional external inputs and outputs the derived control data. The controlling data drives the control parameters back to the signal-processing algorithm.
Figure 13.2 Diagram of an AM-DAFX.
The AM-DAFX described herein aims to take objective technical decisions. This is useful for improving the audio engineer's work flow and allowing him to achieve a well-balanced mix in a shorter period of time. The AM-DAFX described in this chapter are not designed to take into account any uncommon mixing practices or to be able to take subjective mixing decisions. In order to optimise the design of AM-DAFX the use of common mixing practices can be used as constraints. Given that the task normally performed by an expert user also involves perceptual considerations, perceptual rules can improve the performance of the algorithms. When combining several basic AM-DAFX tools to emulate the signal flow of a standard mixer, we can achieve a mixture in which the signal-processing flow is comparable to the one performed in a standard mixing situation.
One of the simplest AM-DAFX is the automatic gain control. In a standard digital audio mixer the head amplifier gain control is a simple multiplier which is in charge of ensuring correct analogue to digital conversion. Typically the head amplifier gain is used to scale the signal such that two technical properties are achieved:
1. The input signal should be scaled in order not to have distortion.
2. The signal must make optimal use of the dynamic range available.
This involves setting the gain such that the maximum peak values do not overflow the analogue to digital converter (ADC), while ensuring that its maximum peak value is as close as posible to 0 dB FS. Figure 13.3 shows the implementation of such an automatic-gain AM-DAFX device, where the input signal-processing section of the algorithm consists of a simple gain controller and an ADC. The side-chain processing corresponds to a simple peak feature measurement performed on the input. Every time the peak input voltage is bigger than the maximum peak value held by the ADC, the input head amplifier gain gets diminished by a step. An equivalent implementation can be achieved by using the overflow flag of the ADC instead of performing a direct measurement over the peak input [PGR09a]. In Figure 13.3 the AM-DAFX side chain extracts a feature from the input signal, taking a decision based on the extracted feature and then outputs control data to perform changes to the desired parameter in the signal-processing portion of the automatic-mixing algorithm. This feature extraction and decision-making process is characteristic of adaptive AM-DAFX.
Figure 13.3 Diagram of an automatic gain AM-DAFX.
