Bob Stuart from MQA, gave deep-insights in understanding if digital audio is as good as it can be, during an online lecture jointly organised by the AES UK Section Cambridge Group and IET Cambridge Local Network.

His talk explained that by better defining objects we can achieve better sound recording and playback abilities. He started by explaining how the fundamentals of psychoacoustics and neuroscience from auditory sciences help us understand pure tone frequencies/harmonics and contrast them against vocalisation and stimulated sound.
Advances in neurosciences, is now letting us do more, like dealing with more natural signals. When humans encode, for example, the intensity of sound to 'loudness', it is nonlinear, like when we encode frequency to the subjective element of 'pitch'. Our frequency response flattens and the hearing system introduces distortion and thresholds which are the either the upper or lower level sounds the human ear can hear.

During the lecture, he explained how understanding the design of the human hearing system, we are now learning, how to better use spectral analysis of sound frequencies and a human's ability to transfer information data rates, to understand which sound frequency ranges can help digitalise data more efficiently.

For example, when we hear 'noise', it follows the human spectral curve, and from that we can start to understand what sound threshold levels are effected, from a psychoacoustic perspective we can then analyse how the auditory channel information is effected. For example, with this, we can identify the thermal-limit for microphones (the absolute minimum that sound can transmit). Interestingly, the thermal limit on Mars is different to Earth (ET would have known this, don't worry).

These insights, let us understand different types of encoding rates and effects on human hearing, thus, understand how much actual data we need to achieve the type of sound quality required.

He spoke about how the neural paths within the human ear interacts with the brain, so that the different neurons can eventually help you understand what sound you are actually hearing, where short and long-term memory also play a key role, as different neurons are responsible for different sound ranges, for example, some of them run at the medium range, to invoke middle ear reflexes, or some used by the temporal-fine-structure, help coding/decoding closer the brain stem - this happens automatically!

Interestingly, for every inner human cell, we have about 5 outer human cells. The outer human cells, act as motors receiving instructions from the brain - thus, the tuning of our hearing is under continuous monitoring. In the hearing ecosystem, by changing the function of cochlea, we achieve different auditory impulse responses (high/low data rates), that help us understand how fundamental sound can be digitalised, such as when listening to environmental, speech or animal sounds.

Talking through different research spectrums, he explained once digitized, errors can be introduced which require 'dither mechanisms' and new sampling techniques, to help eventually meet the ultimate goal of maintaining and preserving 'sound quality' closer to the 'source', using lower data rates.

A fascinating talk on how 'sounds' are perceived, understood by the human-ear and how best we can take that understanding to efficiently digitalise and preserve sound.

The entire webinar can be seen on the Audio Engineering Audio society YouTube channel here.

Read Bob Stuarts profile here.

Robert Heaton