Measurements and Our ears

[The Computer Audiophile]

I just received a link to this video, and I noticed an interesting statement made in it. The guy says our ears don't work like measurement equipment. He is 100% correct in his statement. Our hearing is not linear. 

 

This has me thinking more about measurements and how much stock we put in them. 

 

If for example, our hearing is most sensitive between 2-5 kHz, should we use EQ to bump the other frequencies? Thus, making out ear hear flat frequency response rather than the measurement tools. 

 

All stuff for conversation. I'm not pushing an agenda.

 

 

 

[semente]

20 minutes ago, The Computer Audiophile said:

I just received a link to this video, and I noticed an interesting statement made in it. The guy says our ears don't work like measurement equipment. He is 100% correct in his statement. Our hearing is not linear. 

 

This has me thinking more about measurements and how much stock we put in them. 

 

If for example, our hearing is most sensitive between 2-5 kHz, should we use EQ to bump the other frequencies? Thus, making out ear hear flat frequency response rather than the measurement tools. 

 

You've probably read about the BBC (or Gundry) dip:

 

 

Spendor BC1 as measured by Troels Gravesen

[The Computer Audiophile]

Very interesting.

[elcorso]

1 hour ago, semente said:

 

You've probably read about the BBC (or Gundry) dip:

 

 

Spendor BC1 as measured by Troels Gravesen

 

Excellent !

 

And... speakers measurements from an anechoic chamber. Very far away of reality !

 

Even human listening panels for evaluation, where your ears are not participating...

 

Roch

[wgscott]

1 hour ago, The Computer Audiophile said:

Our hearing is not linear. ... If for example, our hearing is most sensitive between 2-5 kHz, should we use EQ to bump the other frequencies? Thus, making out ear hear flat frequency response rather than the measurement tools. 

 

The problem with this assumption is that the frequency-dependence of the detector (your ears) doesn't mean that your brain will perceive a flat signal as being anything other than flat.  It already does the correction.  

 

We have similar frequency (color) -dependence with our eyes, but it doesn't follow that you should photoshop your images to compensate.

[The Computer Audiophile]

2 minutes ago, wgscott said:

 

The problem with this assumption is that the frequency-dependence of the detector (your ears) doesn't mean that your brain will perceive a flat signal as being anything other than flat.  It already does the correction.  

 

We have similar frequency (color) -dependence with our eyes, but it doesn't follow that you should photoshop your images to compensate.

 

How can one's ears / brain fill in what is unknown? Perhaps I'm not following you. 

[wgscott]

I just had another thought:  Maybe this should be done to correct for individual variations (or hearing impairments).

 

Back to the original point:

 

Let's say your eyes detect blue light with 40% efficiency and yellow light with 90% efficiency (I'm making up these numbers, so don't take the details seriously). A good camera should enable you to produce a neutral representation of an outdoor scene.  But if your eyes see yellow light much better than blue light, the photo shouldn't "look" neutral to you, unless your brain is actually doing some sort of internal compensation or correction.

 

If the information is completely absent (eg UV light), then the brain cannot fill in what is unknown.  

[jabbr]

12 minutes ago, wgscott said:

I just had another thought:  Maybe this should be done to correct for individual variations (or hearing impairments).

 

Back to the original point:

 

Let's say your eyes detect blue light with 40% efficiency and yellow light with 90% efficiency (I'm making up these numbers, so don't take the details seriously). A good camera should enable you to produce a neutral representation of an outdoor scene.  But if your eyes see yellow light much better than blue light, the photo shouldn't "look" neutral to you, unless your brain is actually doing some sort of internal compensation or correction.

 

If the information is completely absent (eg UV light), then the brain cannot fill in what is unknown.  

in imaging this is handled with "color profiles" mostly so editing on monitor translates to prints on different printers. 

Getting an accurate profile is a big deal.

 

But if your eye is more sensitive to yellow in the original image, same for the screen and print so all good. 

 

You might take a photo of a neutral gray card or even better, a color target, so that you can be sure the photo matches regardless of scene lighting.

 

What the guy is saying though is that discrete op amps are better for audio -- that has nothing to do with ear response despite what he is claiming

[wgscott]

Just to make the eye analogy a bit more realistic, here are the absorption profiles for the various cellular detectors (rods and cones):

 

 

Reds, yellows and greens are covered pretty well.  Blue/cyan has the biggest deficit (within the visible spectrum).

 

 

 

Human hearing:

 

[elcorso]

20 minutes ago, wgscott said:

I just had another thought:  Maybe this should be done to correct for individual variations (or hearing impairments).

 

Back to the original point:

 

Let's say your eyes detect blue light with 40% efficiency and yellow light with 90% efficiency (I'm making up these numbers, so don't take the details seriously). A good camera should enable you to produce a neutral representation of an outdoor scene.  But if your eyes see yellow light much better than blue light, the photo shouldn't "look" neutral to you, unless your brain is actually doing some sort of internal compensation or correction.

 

If the information is completely absent (eg UV light), then the brain cannot fill in what is unknown.  

 

Like in the colorblind humans?

 

Anyway, to design an speaker (for example) you have to start from an standard and the standards are not so... then they use measurements equipments to try to create the 'perfect' frequency graph.

 

Roch

[Jud]

1 hour ago, wgscott said:

It already does the correction.

 

If so, then why, in reverberant rooms, do people evaluating speakers almost uniformly perceive a slightly falling frequency response as flat?

[Jud]

1 hour ago, wgscott said:

If the information is completely absent (eg UV light), then the brain cannot fill in what is unknown.

 

For UV light, sure.  But let's beware of extending this principle too far (think of optical or auditory illusions - we plainly see or hear what isn't there).

[Jud]

3 hours ago, The Computer Audiophile said:

If for example, our hearing is most sensitive between 2-5 kHz, should we use EQ to bump the other frequencies? Thus, making out ear hear flat frequency response rather than the measurement tools.

 

Nope.  I think this is what Bill was saying (please correct me if I'm wrong).  Since we go around the world effectively "hearing less" of higher and lower frequencies, if we suddenly heard more as a result of EQ it would seem anything but normal and "flat" to us.

[manisandher]

My previous BD-Design Orelo speakers (http://www.phasure.com/index.php?topic=2923.0) had Munson Fletcher curves built in. I never used anything but the flat curve, because all the others sounded unnatural to my ears.

 

Mani.

[Jud]

6 minutes ago, manisandher said:

My previous BD-Design Orelo speakers (http://www.phasure.com/index.php?topic=2923.0) had Munson Fletcher curves built in. I never used anything but the flat curve, because all the others sounded unnatural to my ears.

 

Mani.

 

No wonder they sounded unnatural - "Munson Fletcher" curves are backwards! 

 

https://en.wikipedia.org/wiki/Fletcher–Munson_curves

[manisandher]

8 minutes ago, Jud said:

 

No wonder they sounded unnatural - "Munson Fletcher" curves are backwards! 

 

https://en.wikipedia.org/wiki/Fletcher–Munson_curves

 

They give the correction that you apply. I.e. a reduction between 1-3kHz.

 

Mani.

[jabbr]

34 minutes ago, Jud said:

 

If so, then why, in reverberant rooms, do people evaluating speakers almost uniformly perceive a slightly falling frequency response as flat?

For the same reason applying haze and coloring toward blue is associated with the visual perception of a scene being farther away. Does falling frequency response extend the sound stage back?

[semente]

23 minutes ago, manisandher said:

My previous BD-Design Orelo speakers (http://www.phasure.com/index.php?topic=2923.0) had Munson Fletcher curves built in. I never used anything but the flat curve, because all the others sounded unnatural to my ears.

 

Mani.

 

The way I see it, whether we listen to live music or recorded music, our perception is always "flat"; that is our reality, how we listen to things.

 

If we choose to deviate from flat (for whatever reason) we will be skewing reality.

 

 

One problem is that microphones and speakers aren't "flat" (though electronics can be reasonably "flat").

Not only that but they are not equally un-"flat" (mics with other mics, speakers with other speakers, mics and speakers), nor the latter compensate for the unevenness of the former...

 

Another problem comes from mic position - closer mic'ing changes the tonal balance by increasing high frequencies - and then there's the mixing/EQ'ing and the mastering that can make a real mess of the whole thing.

 

 

One of the reasons for choosing "flat" in all stages is to avoid skewing as much as possible.

 

But with a "flat" system we will be at the mercy of the engineers and technicians, which is why tonal controls were invented in the first place.

 

R

[semente]

5 minutes ago, jabbr said:

For the same reason applying haze and coloring toward blue is associated with the visual perception of a scene being farther away. Does falling frequency response extend the sound stage back?

 

The "presence region" is so called exactly because it affects perceived depth and immediacy.

 

See here:

 

 

[Jud]

23 minutes ago, manisandher said:

 

They give the correction that you apply. I.e. a reduction between 1-3kHz.

 

Mani.

 

Yup.  Was joking - sure you knew that. 

[Jud]

Our ear-brain systems are funny old things - frequency response affects perception of distance, reverberation affects perception of frequency response....

[semente]

5 minutes ago, Jud said:

Our ear-brain systems are funny old things - frequency response affects perception of distance, reverberation affects perception of frequency response....

 

Although listening room reverberation is time delayed, if I'm not mistaken.

 

But when you measure room response with continuous pink noise you do get an idea of the frequencies in which the room interferes, if you assume that your speakers are anechoic "flat" or you have a free-field measurement of their response.

 

 

Which in turn explains why both anechoic and in-room measurements are important.

 

R

[mmerrill99]

3 hours ago, The Computer Audiophile said:

 

How can one's ears / brain fill in what is unknown? Perhaps I'm not following you. 

 

1 hour ago, Jud said:

 

For UV light, sure.  But let's beware of extending this principle too far (think of optical or auditory illusions - we plainly see or hear what isn't there).

There are many examples of the brain filling in what isn't there - the best known being the missing fundamental where we 'perceive' a fundamental tone when only the harmonic overtones are present in the signal.

 

The focus so far in this thread is all about frequency but let's not forget about time. As Jud has said many times our auditory perception is a pattern matching engine which implies changing frequencies over time. Both auditory & visual perception work on the basis of pattern matching but this is far more complex than it first seems. AFAIK, we don't perceive complete objects (either visual or auditory), the objects we perceive are constructed from the analysis that we have performed on the neurological signals coming down the auditory or cochlear nerve (or the optic nerve). Progress on understanding this analysis is being made but there are many questions still to be answered & much needs researching. For instance, the analysis done on the optic nerve signals seem to be broken down into parallel streams of analysis - one stream isolating out shapes by analysis widths of lines & edges; another just analyses movement; another just color. These 3 analysis streams are later recombined to form what we perceive as a visual object (this is a very simplified description). These streams & combinations of elements of the streams form patterns that are compared to stored abstract patterns to give us the best guess of what is being observed. What is at the heart of this is the stored abstract patterns formed from experience we have built up over time in the course of development & throughout a lifetime's encounter with the world. What is also important is that the 'perceived' object is used to predict it's behaviour, again based on experience. If the next thing we 'observe' does not fit the prediction, we change to the next closest 'model' & continue observing. This is all happening at a level we are not conscious pf & in an appropriate timeframe on a moment to moment basis

 

All this is rather heavy on processing & the signals are not always complete - in most cases there is ambiguity in the mapping of signals to objects hence perception uses other modalities to reduce this ambiguity as much as possible - our natural perceptual mode is to use vision & hearing together with some mix of smell, touch, taste, etc. Illusions (auditory & visual) exploit the ambiguity of our perceptions & help to analyse its functionality but those who try to use the McGurk effect as an example of how sighted listening is flawed are showing how little they know about perception.

 

Auditory perception is thought to share some functionality with visual perception but there are appear to be many, many differences & there is less known about the workings of auditory perception. In audio the whole concept of pattern matching & predictive modelling seem to occupy a more fundamental role - statistical analysis of patterns is actually thought to be important.  Hence when we hear a sound, we quickly determine what family of objects this sound is likely to have emanated from based on the initial attack portion of the sound envelope. If this portion is removed but all other parts of the envelope left intact, decay, sustain, release, we find it very difficult to differentiate between some sound objects. The shape of the full envelope characterises the individuality of this sound - we know for instance the typical sound envelope of a triangle & it's long sustain/decay - as the envelope develops over time we characterise the object, is it damped, etc.

 

Listen here for examples of sound envelopes treated in different ways http://ears2.dmu.ac.uk/learning-object/envelope/

 

As regards measurements & patterns - there are a number of problems with the standard measurements used - the most ubiquitous measurement being FFTs. One obvious example of how this can badly let us down - the FFT of a sound envelope will be exactly the same as the FFT of the same sound envelope played in reverse - yet they will sound completely different.

 

The often stated problem of mostly using tones as signals for measurements is obvious when you realise that pattern matching is an important aspect of auditory perception - tones do not show how a sound envelope will be handled by the replay device - how does it handle multiple dynamically changing tones overlaid on top of one another.

 

I'm of the opinion that the disconnect between measurements & what's perceived lies in this area - measurements which take no cognisance of the pattern recognition used in auditory perception.

 

And, btw, Fletcher-Munson doesn't define the equal loudness curves for all sounds - noise has a whole different set of equal loudness curves & we are more sensitive to noise at a different frequency, not 3KHz instead around 6KHz 

[manisandher]

2 hours ago, Jud said:

Yup.  Was joking - sure you knew that. 

 

That was my immediate thought, but the wikipedia link threw me. Anyway, glad you've still got it... your sense of humour that it.

 

Mani.

[Ralf11]

3 hours ago, Jud said:

 

If so, then why, in reverberant rooms, do people evaluating speakers almost uniformly perceive a slightly falling frequency response as flat?

 

I suspect that is related to the octave to octave nature of hearing.

 

BTW, sensory systems almost always 'fill-in' things that are not there.  hallucinations, and illusions, are common examples of that.