Authors:Giannoulis, Dimitrios; Massberg, Michael; Reiss, Joshua D.
Affiliation:Queen Mary University of London, London, UK
Because dynamic range compression is a nonlinear process with memory, parameter settings involve a complex trade-off even if there is advanced knowledge of the input and desired output. By studying the behavior of real human sound engineers, the authors created an independent set of algorithms to automate the values of the parameters using features extracted from the side-chain. The user need only set the threshold for the preferred amount of compression. Subjective evaluation was performed with two groups of subjects: nine expert mixing engineers and seven amateurs who had experience with dynamic range compression. The preferred human choice for each setting was compared to that created by the automation method. A small correction to Eq. 3 on page 718 of this paper was posted in 2014 February.
Download: PDF (HIGH Res) (1.9MB)
Download: PDF (LOW Res) (439KB)
Affiliation:Rochester Institute of Technology, Rochester, NY, USA
Two-port and multiport models, which are used in electrical engineering, optics, and photonics, can be also be used to characterize acoustics elements as linear time-invariant models. This paper presents an overview of such acoustic elements in signal processing applications. For example, a cascade of two-port elements can be used to model an expansion chamber, exponential horn, or audio filter. Similarly, waveguides (with and without loss), terminations, and multiport junctions can be used to model state space filters and feedback delay networks. The ability to see a system as a network of waveguides and junctions can provide some insight that may not be evident in equations and code. Audio signal processing algorithms can be constructed using two-port and multiport building blocks. Both frequency domain and discrete time domain analyses are included along with notes for MATLAB implementations
Download: PDF (HIGH Res) (2.0MB)
Download: PDF (LOW Res) (309KB)
Authors:Bilbao, Stefan; Webb, Craig J.
Affiliation:Acoustics and Audio Group, University of Edinburgh, Edinburgh, UK
With the rapid advances in the processing power of parallel special purpose hardware, it is now possible to create models of instruments embedded in a full 3D environment. This paper explores a test case of a timpani drum using finite difference time-domain methods, which are particularly suitable for parallelized hardware. Processing limitations usually required a simplification of the timpani model as a linear membrane and cavity under low striking amplitudes. Using the Nvidia Tesla architecture, the model of the timpani drum can the include high-impact nonlinearity of the membrane and the air, as well as the nonuniform tension of the membrane. In addition, 3D synthesis of the entire acoustic space is possible. Matlab code was ported to Nvidia’s native language without significant optimization, which still results in high-speed processing.
Download: PDF (HIGH Res) (3.2MB)
Download: PDF (LOW Res) (506KB)
Authors:Portnuff, Cory D. F.; Fligor, Brian J.; Arehart, Kathryn H.
Affiliation:University of Colorado, Denver, CO, USA; Boston Children’s Hospital, Boston, MA, USA; University of Colorado, Boulder, CO, USA
As a public health concern, portable listening devices are a significant contributor to music-induced hearing loss because they are able to generate high levels over long periods of time. Rather than depend on expensive measuring equipment to evaluate the risk, this research explores two techniques using a dosimetry system and a clinical probe-microphone system. The microphone technique provides calibrated data in the laboratory, while the dosimetry technique allows for the evaluation of actual exposure in a real-world setting, which includes the behavior of the listener over time.
Download: PDF (HIGH Res) (1.5MB)
Download: PDF (LOW Res) (141KB)
Discuss this paper (1 comment)
Authors:Berg, Jan; Bustad, Christofer; Jonsson, Lars; Mossberg, Lars; Nyberg, Dan
Affiliation:Luleå University of Technology, Piteå, Sweden; Swedish Radio, Stockholm, Sweden; Swedish National Laboratory of Forensic Science - SKL, Linköping, Sweden
The perceived audio quality of a digital broadcasting system (such as DAB+) is dependent on the type of coding and bit rates selected. Because of bandwidth constraints, the required number of channels, and conflicting auxiliary services, audio quality is sometimes degraded. In designing a broadcast system, it is necessary to have well-defined criteria for minimally acceptable quality. Two studies explored quality criteria and how quality degrades for various bit rates. For DAB+ the subchannel rate should not be less than the currently available maximum of 192 kbits/s for a stereo signal, which would be comparable to the quality of a modern FM system. Rates below 160 kbit/s can significantly degrade certain types of program material. To be truly perceptually transparent, bits rates of close to 300 kbits/s may be needed when using the current generation of coders.
Download: PDF (HIGH Res) (2.4MB)
Download: PDF (LOW Res) (623KB)
Authors:Thorborg, Knud; Futtrup, Claus
Affiliation:Scan-Speak A/S, Videbaek, Denmark
An earlier report by the authors described a lumped parameter model for a dynamic loudspeaker that incorporated a mechanical resistance proportional to frequency. That model neglected frequency dependence of compliance as having no significant influence at audio frequencies. This paper refines that model by including the frequency dependence of compliance according to the so called “LOG model.” However, loudspeakers with a low degree of viscoelastic properties still have frequency dependence of damping not explained by the original “LOG model.” The new model accurately fits the measured loudspeaker magnitude and phase impedance curves. This improved accuracy is achieved with the addition of only one parameter.
Download: PDF (HIGH Res) (960KB)
Download: PDF (LOW Res) (778KB)
Mixing engineers are not yet made redundant by new technology, but there are a number of common trends in the ways people mix. These can enable automatic systems to emulate the behavior of sound mixers, with the aim of bringing a mix somewhere close to what a human engineer might achieve. There is also a role for novel controllers and technology-assisted mixing tools to make the human engineer’s job more straightforward in the age of multiple channels and 3-D mixing.
Download: PDF (277KB)