The cochlear amplifier

Last Monday, August 21, I attended the first lecture of an audio seminar by the Multimedia interdisciplinary working group here at the University. This one and the next lecture scheduled for Aug. 28, are devoted to the human ear and auditory system, and I thought it may be of interest to share, given the fact the level is both accessible and up to date, and of course of central concern for those involved with audio technologies.

The most striking observations for me at least, can be summarized as follows:

The cochlear mechanical response (it is an elongated, coiled organ which works as a spectrum analyzer), is substantially nonlinear. This implies significant production of harmonic and intermodulation products.

Related with above, there is an actual distributed acoustic amplifier, formed by specialized cells similar to the actual sensorial ones, but dedicated to sense vibrations and regenerate them. This feature allows for both about 50 dB increase in sensitivity, and for a sharpening of frequency discrimination well above what the Basilar Membrane should in a purely passive mechanical way. This amplification, which requires energy supply readily available, obviously only works on living tissue and the protein involved (prestin) was characterized as recently as year 2000.

All this system in turn, apart from generating stimuli upstream to the auditory cortex (afferent pathway), is in turn heavily controlled and modulated by downstream (efferent pathway) signals responsible for compression among other features.

If there is interest on this issue, I will be posting noteworthy facts from this seminar, as well as keep this thread obviously open to welcome contributions.

Rodolfo

P.S. Moderators may choose to move this thread to whatever forum they deem better suited.
 
Further material

The support slides of the first lecture (in spanish) can be found here. There is also relevant material for review i found in a cursory search.

From Physiological Reviews by Robles et. al. Excellent compilation of different facts and research findings.

From Stanford U. a course including fine diagrams and fairly up to date (2002) facts. (go to syllabus for lectures download).

Among other things, the Robles review mentions the measured -in vivo- mechanical intermodulation distortion diminishes with increasing SPL. I wonder whether this has something to do with the fact (at least for me) that low level music tends to be muddier, and upong bringing up volume, definition and transparency seem to pop out much better.

Rodolfo
 
The thing that people always claiming "ohhh but the ear has 10% THD!!!!!11" keep forgetting is that obviously, a 1kHz sinewave does NOT sound like it has any 2kHz, 3kHz, 4kHz, 5kHz, ... elements. A pure tone is clearly distiguishable from a distorted tone. NOT ONLY that, but if 10% were true, nobody would notice 1%, maybe even 2% THD. Which isn't true (as I recall...I don't have a THDometer to find out myself). Thus, as it is an established fact that the cochlea distorts, the brain MUST counter it in a forward fashion, passing linear sound to the conciousness.

Hmm, to the conciousness eh? That's where you freaks oughta start... tap right into the brain. -50 to +200dB with no pain or compression. Or into the memory center and you can enjoy a flawless performance at any time!

Tim
 
There was a very good, very long article on this topic linked to IIRC in a thread on this site. Aside from the increasing amplification with lower SPL, there was an interesting side note about how the first researchers used corpses to test hearing response... the sounds got so loud that the bodies were literally being moved by the SPL before electrical signals they could measure were produced...

kind of gory, in retrospect...:dead:
 

MBK

Member
2002-04-03 3:40 pm
Singapore
Actually the first hint that the ear works as an active amplifier apparently came from Thomas Gold in the 60's. It was his thesis and it wasn't well received... Thomas Gold had a lifetime production of theories well ahead of his time, sometimes confirmed, sometimes (as of this time) refuted, in fields as diverse as biology, astrophysics, cosmology, and geology, to name a few...

(oops, just saw he's cited in the review article, and it was... 1948, gasp)
 
MBK said:
....
(oops, just saw he's cited in the review article, and it was... 1948, gasp)


Yes, in fact the lecturer last Monday mentioned there is a Nobel prize photograph where if I don't remember wrong, Von Beckesy was the one awarded on the basis of his extensive research but on the premises of a passive sensory mechanism, and in the same frame was an unhappy Gould looking at it. At last justice was made.

Gould - and engineer by profession - had no basis to support his hypotheses, but from an engineering standpoint that there was no other way except some active operation for the actual hearing performance to be achieved. He was right.

Rodolfo
 
Sch3mat1c said:
The thing that people always claiming "ohhh but the ear has 10% THD!!!!!" keep forgetting is that obviously, a 1kHz sinewave does NOT sound like it has any 2kHz, 3kHz, 4kHz, 5kHz, ... elements. A pure tone is clearly distiguishable from a distorted tone.

Tim

sorry, but you're just plain wrong ...

Actually, the non-linear mechanics and electrophysiology of the inner ear organ indeed does make a loud sine sound like a distorted sine.

I have read in a book about the history of acient european music, that harmonics and scales were invented by monikken who very carefully listen to their own ears IMD and HD.


This non-linear mechanics and electrophysiology of the inner ear is being used every day in countless hospitals to check the hearing of childs suspected of having hearing problems: no non-linear products = something is wrong.
 
I think it is not relevant if our ears cause distorsion, or not.

There is sound in nature and our ears and brain together lead
us to certain perception of that particular sound.
We can easily notice small changes in the sound, even if the
change of the sound is less than our nonlinearity of the ears.

Let's look at our ear+brain as another signal amp in the chain.
The input signal for this amp is the sound in the air. The output of this amp is our perception.
Now let's estimate that our ear-brain-gain has 10%THD, - or worse.
If a pure sine wave fed to it's input, then output will show the sine wave + the amp specific harmonics.
If you change the input, even if you just add 0.5% 2nd harmonic, then output will change. ... and this means also our perception will change.
From this, I would still stick to the technical goal that our electronics and speakers should not add to much colour to the signal.

Nevertheless the input of ingrast could help to put some light to the question, which sort of distorsion a percepted more or less inconvinient.
 
ChocoHolic said:
I think it is not relevant if our ears cause distorsion, or not.

.....
Now let's estimate that our ear-brain-gain has 10%THD, - or worse.
If a pure sine wave fed to it's input, then output will show the sine wave + the amp specific harmonics.
If you change the input, even if you just add 0.5% 2nd harmonic, then output will change. ... and this means also our perception will change.

I guess this is in the right direction.

From this, I would still stick to the technical goal that our electronics and speakers should not add to much colour to the signal.

Agree fully. Our reproduction systems should be capable of just that, reproducing the original sound with arctifacts below perceptive threshold.

Nevertheless the input of ingrast could help to put some light to the question, which sort of distorsion a percepted more or less inconvinient.


Tomorrow I will be asking some questions at the lecture. Will post results.

Rodolfo
 
The key is this:

What is the perceptive threshold?

That's the crux of the matter. This is where all discussions turn muddy.

Somewhere else in the forum I remember seeing a graph with the HD / IMD of the human ear. It seems it can reach tens of percent... Still the threshold for *some* distortions can be low, while it can be high for other distortions. It's surprising enought hat we can hear <0.5% distortion of an amp when not just the ear has much more than that, (you could say, ok, the brain knows that and compensates) but more importantly, when the *speaker* itself likely has >1% HD!! Clearly, the ear discriminates between various distortions and can pick them out *below the overall* noise and distortion floor.

This is why distortion spectrum matters, not just absolute magnitude. This is also the subject of GedLee and of the Cheever thesis, as incomplete or faulty as they may otherwise be - the subject is crucial.

Rodolfo, question on the side - some time ago you asked in the forum on speaker driver choice for testing amplifiers... What did you finally settle on?
 
MBK said:
The key is this:

What is the perceptive threshold? -----

This can readily be asessed objectively measuring the inner hair cells firing rate as a function of SPL, will be back today with some data in this respect.



....Rodolfo, question on the side - some time ago you asked in the forum on speaker driver choice for testing amplifiers... What did you finally settle on?

I am building four boxes with the RS225 / RS28 reference series from Dayton. Will use biamplification with a special design. I will post photos and results as soon as possible (it will take some time still, I am experimenting on looks and finishings).

Rodolfo
 
Hi Rodolfo,

interesting that you chose the Daytons over the Seas you were considering. It must be tough to build speakers for amp testing - you'll likely need a monitor with narrow dispersion to get maximum direct sound and minimum room interaction, and best transparency. In other words it'll likely be unpleasant on a lot of music materiual to listen to ;-) ... I am sure you're aware of the Linkwitz site that has a host of good considerations about what is important in a speaker...

About hearing threshold, I meant not the absolute threshold, but the threshold of what distortions we can hear. I'm afraid physiological data can at best give orders of magnitudes - there is alot of brain acticity to hearing, not to mention possible physiological unknowns as happened in Gold's case. For instance it seems from nerve cell refractory phase time scal you'd expect only miulliseconds of resolution,l when in reality it seems it is in the microseconds - dues to parallel processing. I think this will have to be resolved by banks of psychoacoustic, subjective tests with large sample sizes and good stats. And, it is a bit circularm because what constitutes a "transparent" system, presumably neutral, as the test machinery? It will always be debatable, because I see no good agreement of what constitutes a "good" amp or speaker in the first place. Maybe headphones used with fast op amp circuits can be a key, with DUT in the chain of ampluification...

Tschrama,

interesting about the monks inventing musical scales according to their hearing observations. Do you have source for that (books of www)?
 
MBK said:
Hi Rodolfo,

interesting that you chose the Daytons over the Seas you were considering. .....

In fact I was more looking for the best affordable speakers for a general listening application, more from a complete sytems perspective, and I am betting on the Daytons because of reported performance (including distortion) comparing well with far higher priced brands.



About hearing threshold, I meant not the absolute threshold, but the threshold of what distortions we can hear....

Yes, I blundered on my first interpretation :mad:, only it was getting late to correct once I noticed (I was rushing to the Faculty for the second lecture).

I asked this same question, and got no definite answer though it is a well known fact we can discern distortion levels far below what is actually generated within the cochlea itself. The clue is in the extensive suppression mechanism similar to other senses and I will be extending on this - which was central to today's lecture - in a forthcoming post.

Rodolfo
 
...getting more and more curious...
And definitely I must admit: The data and articles are impressive
for me, but up to now I am not able extract what sort of distorsions, delays, compressions and other errors are inconvinient for us.

My simplest approach would be to guess that errors, which are similar to the natural errors of our human ear-brain-gain are probably more comfortable to our perception than other errors.
But even with this approach I am not able to transform the results
of the biological examinations to a technical specification of 'good' or 'bad' properties for an audio amplifier. :bawling:

...difficult...