sound processing by the human auditory system and it's relevance to audio

[jkeny]

As suggested by Vacuphile, I started this thread & hope he & others will contribute. As the title says, how the ear (not brain) manipulates & processes audio signals is a relevant area of information & knowledge for the audiophile. It has arisen recently on two threads here how-better-turntable-compared-cd thread here & on the JC Blowtorch preamp II thread here .

vacuphile posted a nice summary of some of the mechanics of the ear which I hope he doesn't mind me reposting here :

1. Do you really think the brain has no control over the functionality and the ear handle everything on its own sending the final signal to the brain??Answer: It is a two way street between the brain and the ear when you look at neuron pathways, with the brain receiving about 10 times more inputs from the ear than the other way around. It is also known, that the inner ear contains tiny muscles, which contract to avoid damage at high SPL. Furthermore, it is known that the inner haircells, tiny band path filters, may have very steep slopes, up till 1000 dB per octave. If you would try to replicate such a high Q filter with man made technology, this could not be done without considering amounts of ringing. Yet, this does not happen inside the ear. So, in short, a lot of processing is done by the ear, but at a low level; to tidy up the signal, so to speak. As to the question how this biological/neurological system works, much remains unanswered. However, we do have a fair understanding of what comes out of the ear subsystem. Unfortunately, they found out by tapping into the auditory nerve in live cats with tiny needles, but thanks to that, we know that what comes out is like a flat cable, with each individual wire connected to an array of bandpath filters (the ear). See my earlier post. And all this is not linear. The frequency response of the ear is like a banana, belly up. And we also know that the banana gets flatter at higher SPL.

So, this is what the ear subsystem basically does, and all further processing takes place in the brain.

2. Are you saying there is more processing in the ear then in the brain when it comes to audio?

Answer:. Although the ear is not a linear transducer, and also performs a Fourier analysis on incoming sound, the main processing takes place in the brain. The auditory nerve weaves a path through the brain. On this path, it is connected to a series of sub processing centres, which all do very different jobs. It may excite a startle response when it passes through one of the oldest parts of the brain, the pons or brainstem. In the superior olives (some brain parts have odd names), intensity and frequency matching between both ears allows us to locate sound in space. When it goes through the thalamus, it may trigger an emotional response. In Wernecke's area, it may be decoded into intelligeable speach.

All this happens more or less in parallel, and in music, it sometimes leads to an integration at all levels.

My definition of audiophile quality is when this integration can take place, without intervening artifacts. I have only my own brain to decide whether this happens to me, or not.

At the same time, much research has taken place to find out what such intervening artifacts could be. Ringing or non-flat frequency reproduction are known issues that lead to audible artifacts, and so is distortion. We also know that dynamic range is a factor. Noise is more complicated; the ear itself is never silent; it even produces its own sounds which can be picked up from the outside!

I personally can live with some noise in a system, so I partly share the earlier comment that the brain can compensate for imperfections inherent to the LP. Alas, my brain is very unforgiving when it comes to scratches or static. Others may have brains that are able to filter those out better than mine, and for them vinyl might still be an option.

There are some useful on-line links that I have found but I would hope others will add some more & also add their comments:

Just posted by FrankWW on JC Blowtorch thread: JJ Johnston's presentations on hearing & Psychoacoustics: PowerPoint Presentations from recent (or not so recent) meetings.

The Phsyiology of hearing : Mechanics of the Mammalian Cochlea

Elements of Psychoacoustics - overview : http://www-dsp.elet.polimi.it/ispg/images/pdf/audio/materiale/elementi_di_psicoacustica.pdf

 

[rsdio]

Some good links that I am familiar with are:

AES PNW Meeting Report - <meeting title/topic here>

and there are many slides here:

PowerPoint Presentations from recent (or not so recent) meetings.

 

[jkeny]

Blind Testing

One thing that was touched on in the JC thread was about Blind Testing & the need for training of individuals along with other criteria which are seldom mentioned by those who champion blind testing as the gold standard for audio evaluation. Here's the extract from the slides:


• If you listen to something differently (for different features or objects)
– You will REMEMBER different things
– This is not an illusion

• If you have reason to assume things may be different
– You will most likely listen differently
– Therefore, you will remember different things

What this all means, in effect, is that any test of auditory stimulii that wants to distinguish only in terms of the auditory stimulii must:

– Have a falsifiable nature (i.e. be able to distinguish between perception and an actual effect)
– Must isolate the subject from changes in other stimulii than audio
– Must be time-proximate
– Must have Controls
– Must have trained, comfortable listeners​

• A control is a test condition that tests the test. There can be many kinds of controls:
– A positive control

• This is a condition that a subject should be able to detect.
• If they don’t, you have a problem.​

– A negative control

• A vs. A is the classical negative control
• If your subject hears a difference, you have a problem​

– Anchoring elements

• Conditions that relate scoring of this test to results in other tests
• These can vary depending on need, and may not be obligatory​