Open
Access
Exploration of audio at digital resolutions higher than CD began in the late 1980s, arising from a wealth of interdependent sources including listening experiences, rapid technical advances, an appreciation of psychoacoustics, acoustic measurement, and the ethos that music recording should capture the full range of audible sound. High-resolution formats were developed and incorporated into successive generations of distribution media from DVD, SACD, and Blu-ray, to internet downloads and now to streaming. A continuing debate throughout has been whether, and especially why, higher resolutions should be audibly superior. This review covers the history of the period, focusing on the main drivers of experimentation and development up to the present, and then seeks to explain the current view that, beyond dynamic range, the most likely technical sources differentiating the sound of digital formats are the filtering chains that are ubiquitous in traditional digital sampling and reconstruction of analog music sources.
Download: PDF (HIGH Res) (447KB)
Download: PDF (LOW Res) (124KB)
Open
Access
Recent interest in high-resolution digital audio has been accompanied by a trend toward higher sampling rates and expanded bit depths, yet the sound quality improvements show diminishing returns, which can be explained by a failure to reconcile human auditory capability with the information capacity of the audio channel. The authors propose an audio capture, archiving, and distribution framework based on sampling kernels having finite length, unlike the “ideal” sinc kernel that has an infinite duration. With these new kernels, the original transient events need not become significantly extended in time when reproduced. This new approach runs contrary to some conventional audio wisdom, such as the complete elimination of aliasing. This paper reviews advances in neuroscience and recent evidence on the statistics of real signals, from which it can be concluded that the conventional criteria may not be helpful. This proposed approach can result in improved time/frequency balance in a high-performance chain whose errors, from the perspective of the human listener, are equivalent to those introduced when sound travels a short distance through air.
Download: PDF (HIGH Res) (3.4MB)
Download: PDF (LOW Res) (1.6MB)
Open
Access
There is still disagreement over the ways in which sound quality might benefit from higher sampling rates or wider bit depths in a digital path. The authors show that if a digital pathway includes any unintended or undithered quantizations, then several types of errors will be created whose nature will change with increased sampling rate and word size. Although dither methods for ameliorating quantization error have been well understood in the literature for some time, these insights are not always applied in practice. It is rare for an audio performance to be captured, produced, and played back with a flawless chain. The paper includes a tutorial overview of digital sampling and quantization with additive, subtractive, and noise-shaped dither. The discussions also include more advanced topics, such as cascaded quantizers, fixed and floating-point arithmetic, and time-domain aspects of quantization errors. Guidelines and recommendations are presented, including for the design of listening tests.
Download: PDF (HIGH Res) (5.2MB)
Download: PDF (LOW Res) (2.4MB)
Open
Access
High-resolution audio has started to use modern sampling principles to further enhance the quality of digital audio reproduction. This tutorial seeks to explain how these new methods are an improvement over traditional Shannon sampling. Although perfect in theory, a Shannon-sampled signal will always have some aliasing errors when reconstructed because the infinite-duration filters required cannot be realized in practice. The tutorial first reviews the relevant parts of traditional sampling, and then goes on to introduce the modern method based on splines, which are one of many possible approaches. As an alternative, splines offers the possibility of a lower level of error for a given filter length as well as reconstruction filters that are very compact in time. The challenge for the future will be to develop circuits that allow these benefits to be achieved in practice.
Download: PDF (HIGH Res) (6.2MB)
Download: PDF (LOW Res) (519KB)
Open
Access
Intermodulation distortion (IMD) arises when a nonlinearity causes two or more signals to interact. This paper investigated this distortion mechanism by measurement and listening tests using three models of high-quality loudspeaker. The authors investigated IMD arising in combinations of amplifiers and loudspeakers, concentrating particularly on circumstances where very high frequency signals might induce audible distortions. Using selected ultrasonic signals (i.e., above 20 kHz) higher than 80 dB-SPL, IMD could be measured in all three loudspeaker systems tested, and it was just audible in the absence of any signals below 20 kHz. The aim was to discover whether IMD of ultrasonic signal elements could lead to their detectability and thereby confound listening tests or otherwise modify the listening experience. The results show that while such distortion can be found and must be accounted for in some psychoacoustic threshold experiments, it is not pertinent to playback of current high-resolution recordings, since the level of ultrasonic signals tends to be significantly lower.
Download: PDF (HIGH Res) (6.1MB)
Download: PDF (LOW Res) (1.0MB)
Discuss this paper (1 comment)
A significant revolution that has taken place in the recent past involves the potential created by combining computer modeling, digital signal processing, and low-cost or compact transducers. Transducers don't always perform exactly to the manufacturers' average specification, and there may be opportunities to correct for batch variations using signal processing. We also find that horn-loaded bass loudspeakers suffer from the same acoustic problems in small rooms as conventional ones. Papers from the 146th Convention are summarized.
Download: PDF (663KB)
