a presentation by: John Beerends (TNO)
The Netherlands Section of the AES met on 13 December 2011 at the Netherlands Institute for Applied Physics Research, to attend a presentation by John Beerends of the same Institute. The topic of Dr. Beerends’ presentation was: “HiFi philosophy how do you measure ‘sound quality’ and does ‘the truth’ exist”? Some 40 members turned up to participate in the discussions.
The first hurdle is the stated goal of the audio reproduction. For instance, do you want to transport yourself, as a listener, to the original event so that you have the feeling you are There and Then? Or, conversely, do you want to transport the performers to your ‘Here and Now’? Obviously this has a great impact on the requirements for, for instance, your loudspeakers: the performers in your living room do not need any reproduction of the original venue, while your being ‘There and Then’ requires that the venue acoustics are faithfully reproduced but not your listening room attributes.
The quality of sound reproduction can be defined as ‘the distance’ between the ideal and the actual reproduction. In the past several attempts have been made to model the entire listening process, and Dr. Beerends presented the latest model he worked on (model POLQA 2011 which is standardized as ITU-T Recommendation P.863). This is a very detailed model including perception aspects like masking (intensity and time-frequency) and time-frequency decomposition, as well as cognitive aspects like time-pitch averaging and difference interpretation. In measuring reproduction quality, the model would compare an input signal with an output signal and produce a metric of the reproduction quality, i.e. ‘the distance’ between ideal reproduction and the actual one. The model has been used in determining speech reproduction quality in voice communications, where the target is not necessarily the most faithful reproduction but to get as close as possible to an ‘ideal’. Obviously, in cost- and commercial-sensitive voice communications, this ‘ideal’ could be a less than a 100% accurate reproduction.
The model could also be used to measure the reproduction quality in HiFi systems. In that case, the ‘ideal’ would not be separate input to the model but just the original source material. Whether this type of measurements will be done depends on the availability of funding, but the participants agreed that it would be an interesting step towards a possibly more objective metric for sound reproduction quality.
And oh yes, ‘the truth’. Indeed, it does exist, but it is context dependent and subject to personal interpretation…