The missing chapters of audio science
Thought-provoking and (hopefully) original ideas about loudspeakers, headphones and related principal fields (room acoustics, hearing, measurements, etc.). Dedicated to the appreciation of real scientific methodology and model thinking. Can be consumed by anyone, anywhere, anytime.
"First principles, Clarice. Simplicity."
Many areas in life have their own "ten commandments": a list of principles presented as commandments. It would be fun to have a ten commandments list for audio science.
Experimenting with YouTube
I like to experiment with YouTube... One result is a YouTube collection (300+ videos) and browser-based video player in one. In this version, videos are randomly selected, but categories can be selected by the user.
Different 'schools' in audio
Some thoughts on the different 'schools' in audio, methods, audio measurements.
Although the audio community seems to be divided into "objectivist" and "subjectivist", this kind of portrayal is too simplistic and misleading. Starting with subjectivism, subjectivism is a faith-based system that comes from a misinterpretation of measurements, misinterpretation of the principles of sound reproduction or sometimes the scientific method. Or to put it another way, subjectivism is using subjective test methods without an objective understanding of the world. On the other hand, "objectivism" is more versatile, more complex and offers more sophisticated flaws for us (it is only right when it is done right). Although there may be several paths to the truth, there are certainly very few...
Subjective testing (subjective evaluation) and "subjectivism" have different meanings, which can lead to a lot of confusion. Subjective testing (where subjective testing means traditional song based listening tests) is just a method, and like all methods, it can be used correctly or incorrectly, though its reliability is quite limited when the audible differences are minor. Subjective testing is acceptable when it is supported by measurements. However, there is one important thing that neither subjective tests nor "objective" blind tests can provide: understanding how amplifiers, speakers, audio formats or human hearing work. Furthermore, there is no "objectivist" who don't use subjective testing at all.
The objective approach (based on models and measurements, supplemented with tests) also has its pitfalls. Typical flaws:
- Mindless application of data analysis tools and test methods (pseudo-intellectualism, data worshipping in academic fields). In science experiments are the first-class methods and tests (blind or not) are secondary, safety tools only.
- Measuring without understanding hearing, without a valid hearing model (it can lead to the "listening to music with the eyes" phenomenon).
- Measuring without understanding basic concepts, or models, such as the concept of bandpass system and "bandlimiting", since all components in the signal chain are bandlimited (e.g. maximum signal rate of change depends on cut-off frequency and maximum level).
- Ignoring the properties of real-life audio signals (this is the least bad).
Blind test is a solution to only one problem: eliminates the possibility of self-deception arising from perception (unreliable auditory memory). Blind test doesn't provide a solution for how to avoid false positive results and if the test is free of errors, correlation still doesn't imply causation. The exclusive use of tests can lead not only to misconceptions but also to superficial knowledge.
The goal is to remove human fallibility from the test procedure by any means...
Don't be fooled. Big Data can't save us. Our salvation lies within ourselves.
The most important criterion for an audio measurement is the correlation with perception and fidelity. How can we measure the "right thing" in the right way and interpret the result in the right way? Any problems related to audio fidelity requires a three-level analysis:
- an accurate model of the system being measured (system behavior, causes and diagnosis of signal changes),
- a hearing model (how we hear, what we can hear, differences between age groups),
- characteristics of audio signals, sound soruces (temporal and harmonic structure of musical instrument sounds and human voice, maximum third-octave band levels in music).
This three-level analysis is the most important principle in audio science: just as a compass helps us find the right direction, it helps us find the right questions. (A similar three-level analysis can be applied to display technology and vision.)
In addition to a different attitude to measurements, blind tests are often portrayed as a dividing line between subjectivism and objectivism. Now we can see that this is a huge misconception, mainly because our knowledge doesn't come from blind tests. Blind tests with music can be used as a secondary, demonstrative tool, but to find a causal relationship we have to perform an experiment that doesn't depend on blind testing ("If your experiment needs statistics, you ought to have done a better experiment." - Ernest Rutherford). The point is that understanding technology and hearing is more important than constant blind testing.
Our auditory memory evolved for communication and surviving, not for testing amplifiers or cables. If we compare hearing and vision then we can see that the ear's analytical ability is much worse than the analytical capability of vision. We can completely separate one part of the image from the rest, but we can't completely separate one part of the music from the rest. Analytical and focusing ability of hearing is very limited. If we also add that humans are visual thinkers, which means that we need a visual representation of a problem to understand the problem, it's no surprise that hifi has become so subjective. (Although we perform better in visual analysis, visual perception is also fragile when the differences are small. See the crazy 4k and 8k resolutions and framerates.)
Csaba Horváth