Metamerism (color) - Wikipedia, the free encyclopedia
A well known issue in the printing industry. Don't know if it exists in audio.
You may be reading too far into the analogy.
Interesting. Thanks. Yes, exactly the same as going to window (or full-spectrum bulb) to check colors.
As I said, I'm not proving anything by analogy, only illustrating the general principle known as "equivalent stimuli" and suggesting it occurs in sound too.
As Toole demonstrates, people (yes, even real expert listener sub-samples) are bad at telling the different between a slight broad low-treble EQ boost and a slight high-bass EQ cut (and vice versa). Ears are doing something more like "envelope detection" than freq analysis (not that some arrogant audio fans will ever believe they can be "fooled" that way one bit).
In other words, those two EQs are "equivalent stimuli"... people can't tell them apart, at least with music in Toole's tests.
You are right to question whether it exists in loudness. I can't say; but something similar is happening with loudness with the Fletcher-Munson phenomenon.
Last edited:
Right. The eye and ear don't quite work the same way. We have only three (overlapping) color sensors in the eye, the ear has about 10,000X more sensors. Both ear and eye can be fooled, and that's a good thing. Otherwise our world of electronic audio and video would be very much harder, if not impossible, to obtain. 
Good point on the equivalent EQs. Would like to learn more about that. It's amazing how much one register seems to effect the others. You know, a better tweeter can affect the bass, or vice-versa.
Good point on the equivalent EQs. Would like to learn more about that. It's amazing how much one register seems to effect the others. You know, a better tweeter can affect the bass, or vice-versa.
There would be "equivalent EQs" for hearing, but owing to the ears very high spectral resolution the significance of this effect woulkd be minimal. As you say the eye has three spectral sensors and the ear thousands.
The ear detects differently in different frewquency bands. Below about 500 Hz. it detects syncronously as the cell firings are actually syncronous in time with the signal. But by 1 kHz the cells can no longer fire fast enough to keep up with the signal and do not fire syncronously but more randomly and the "pitch" is detected by its location on the cochlea rather than by the timing of the firings. This is why our hearing has such profound changes in its abilities above and below 500 Hz. Since loudness is detected by the firing rate - number per second - it is natural that as the frequency falls, the firings being syncronous to the signal, that the loudness has to fall as the frequency falls. This explains the shape of the loudness curve below 500 Hz. Above 1 kHz the firings are more random and the spectral relationship is basically flat (phase detection is gone), but shaped by the natural resonances in the ear, hence the peaks at 3 kHz and 7 kHz. Temporal detection also falls below 500 Hz and gets messed up by the resonances of the ear above 3 kHz, which is why our temporal detection rises above 500 Hz to peak at about 2 kHz and then falls again after than. In this frequency range we are very sensitive to temproal aberations in the signal. Below 500 Hz the ear is almost "blind" to temporal aspects - but it has good frequency/pitch resolution.
The ear detects differently in different frewquency bands. Below about 500 Hz. it detects syncronously as the cell firings are actually syncronous in time with the signal. But by 1 kHz the cells can no longer fire fast enough to keep up with the signal and do not fire syncronously but more randomly and the "pitch" is detected by its location on the cochlea rather than by the timing of the firings. This is why our hearing has such profound changes in its abilities above and below 500 Hz. Since loudness is detected by the firing rate - number per second - it is natural that as the frequency falls, the firings being syncronous to the signal, that the loudness has to fall as the frequency falls. This explains the shape of the loudness curve below 500 Hz. Above 1 kHz the firings are more random and the spectral relationship is basically flat (phase detection is gone), but shaped by the natural resonances in the ear, hence the peaks at 3 kHz and 7 kHz. Temporal detection also falls below 500 Hz and gets messed up by the resonances of the ear above 3 kHz, which is why our temporal detection rises above 500 Hz to peak at about 2 kHz and then falls again after than. In this frequency range we are very sensitive to temproal aberations in the signal. Below 500 Hz the ear is almost "blind" to temporal aspects - but it has good frequency/pitch resolution.
Yes and no. For sure the mechanism/operation of each sense is distinct. But it is not always important to query the underlying mechanism to understand the perceptual experience.
Certain principles do seem to hold across the senses: contour sharpening, AC-coupling, negative after-effects, log sensitivity, etc.
"Equivalent stimuli" isn't necessarily a perceptual principle since some families of equivalence are self-evidently physical like the size-distance equivalence in vision.
+1 for Earl
Regarding headphones and flat response:
I dabbled with headphones at PSB. the issue of target response is very different than with speakers. Imagine a listener in an anechoic chamber in front of a perfect (flat) loudspeaker. You can measure at his outer ear and you will get a particular response based on head diffraction and outer ear reflections. If you want to pick a speaker angle you can define a target response curve for that angle (and that particular head).
If you are designing insert drivers (in-ear monitors) then you need to measure within the ear canal and the measurement will peak dramatically at 3kHz or so from the canal dimension. This too will vary from person to person as ear shapes and dimensions vary, although you can design for mean dimensions.
Rather than a free field, it is generally advocated to design for to a target based on a subject in a flat diffuse field, i.e. move the subject into a reverb room. This certainly gets around sorting through the great variation of response vs. angle. The diffuse field ear canal target curve is, I believe, called the Killian curve and is the target for the etymotic units.
So I designed a dual transducer in-ear monitor with a complex external network that was a great fit to the Killian curve and had a listen. Result: way too bright. I believe this is a general consensus for the Etymotic models and in line with Earl's comments.
Perhaps this returns to our speaker-in-room observation that a flat diffuse field is generally too bright and not usually experienced in real acoustic spaces.
One real difference between headphone/earphones and speakers is that we don't have to worry about the spectral response changing with time (early sound vs. late sound), confusing our definition of "what is flat".
Interesting discussion,
David S.
I dabbled with headphones at PSB. the issue of target response is very different than with speakers. Imagine a listener in an anechoic chamber in front of a perfect (flat) loudspeaker. You can measure at his outer ear and you will get a particular response based on head diffraction and outer ear reflections. If you want to pick a speaker angle you can define a target response curve for that angle (and that particular head).
If you are designing insert drivers (in-ear monitors) then you need to measure within the ear canal and the measurement will peak dramatically at 3kHz or so from the canal dimension. This too will vary from person to person as ear shapes and dimensions vary, although you can design for mean dimensions.
Rather than a free field, it is generally advocated to design for to a target based on a subject in a flat diffuse field, i.e. move the subject into a reverb room. This certainly gets around sorting through the great variation of response vs. angle. The diffuse field ear canal target curve is, I believe, called the Killian curve and is the target for the etymotic units.
So I designed a dual transducer in-ear monitor with a complex external network that was a great fit to the Killian curve and had a listen. Result: way too bright. I believe this is a general consensus for the Etymotic models and in line with Earl's comments.
Perhaps this returns to our speaker-in-room observation that a flat diffuse field is generally too bright and not usually experienced in real acoustic spaces.
One real difference between headphone/earphones and speakers is that we don't have to worry about the spectral response changing with time (early sound vs. late sound), confusing our definition of "what is flat".
Interesting discussion,
David S.
David S -
You've clarified a bunch of ideas for me. Thanks.
Right, it isn't that headphones are flat (as opposed to speakers that aren't) but only that headphones are simpler to evaluate. My point is that either way, achieving "flat" in headphones is a whole lot simpler. Either way, we are back to speakers-in-rooms where it isn't simple.
As you say, the spectral shift over the period of decay changes the impression of freq response. So in a pretty heavily damped room (which I rather prefer), there would be greater agreement between mic measurements and hearing. Hence, less tweaking after the measurements are done.
Which suggests that we examine waterfall-like measurements in certain ways (as if anybody was in doubt that would be the conclusion of this thread).
In tuning concert halls, big effort is made to get the reverberation x freq performance to the acoustician's liking. Hard to do at home but truly deserves more attention.
I guess freq coloration would seem different depending on the kind of music - whether notes are sustained by the instrument, etc.
You've clarified a bunch of ideas for me. Thanks.
Right, it isn't that headphones are flat (as opposed to speakers that aren't) but only that headphones are simpler to evaluate. My point is that either way, achieving "flat" in headphones is a whole lot simpler. Either way, we are back to speakers-in-rooms where it isn't simple.
As you say, the spectral shift over the period of decay changes the impression of freq response. So in a pretty heavily damped room (which I rather prefer), there would be greater agreement between mic measurements and hearing. Hence, less tweaking after the measurements are done.
Which suggests that we examine waterfall-like measurements in certain ways (as if anybody was in doubt that would be the conclusion of this thread).
In tuning concert halls, big effort is made to get the reverberation x freq performance to the acoustician's liking. Hard to do at home but truly deserves more attention.
I guess freq coloration would seem different depending on the kind of music - whether notes are sustained by the instrument, etc.
Last edited:
I think it is more the other way around. Shift over the period of decay changes the measurement of response (extending measurement window from direct sound only, through steady state response) but has minimal impact on the perception of response.
So, yes, a deader room will measure more like it sounds.
David S.
Respectfully, there's no ambiguity or confusion about measurements, provided you know what you are doing... which is always possible on a good day. On the other hand, listeners are often unable to distinguish or do some kind of decomposition or analysis of what they hear, even when they think they have such mental powers.
In practice, as you say, faulty measurements lead to tweaking. The real question is what are the right measurements that lead to no tweaking.
In practice, as you say, faulty measurements lead to tweaking. The real question is what are the right measurements that lead to no tweaking.
If I remember right we couldn't all agree on what they should be. Is this it??
Rob
http://www.diyaudio.com/forums/multi-way/166411-measurements-when-what-how-why.html
Rob
http://www.diyaudio.com/forums/multi-way/166411-measurements-when-what-how-why.html
Last edited:
Hello,
"relative loudness across the freq band"
There are numerous models for loudness perception that can be found from the free domain. See for example Zwicker for classical analysis.
Maybe not the best link but gets you going:
http://fridrichacoustical.com/Documents/Zwicker-Method_notes10-MR03-NO08.pdf
Essentially the perceived loudness depends not only on the system the sound is played through but on the properties of the signal itself. This leads to a paradox: Since there are infinite number of signals (temporal and intensity across the freq band) there are infinite number of perceived loudnesses. Which signal from the set of infinite you select to set your system?
- Elias
"relative loudness across the freq band"
There are numerous models for loudness perception that can be found from the free domain. See for example Zwicker for classical analysis.
Maybe not the best link but gets you going:
http://fridrichacoustical.com/Documents/Zwicker-Method_notes10-MR03-NO08.pdf
Essentially the perceived loudness depends not only on the system the sound is played through but on the properties of the signal itself. This leads to a paradox: Since there are infinite number of signals (temporal and intensity across the freq band) there are infinite number of perceived loudnesses. Which signal from the set of infinite you select to set your system?
- Elias
All kinds of things (as in that article) are known to influence the over-all loudness of complex sounds. For example, the S/N level, your SPL exposure immediately before, and don't even ask what influences people in their judgments about how loud airplane noise over their homes is. But we are talking about a much simpler test-setting and yet the factors aren't at all clear.
Yes, a few relevant previous threads about measurement. And that is exactly why I concluded that the evident "bottom line" is that (1) most skilled DIYer tweak after doing mic testing and (2) as in this thread start, they don't know why they have to do so.
Yes, a few relevant previous threads about measurement. And that is exactly why I concluded that the evident "bottom line" is that (1) most skilled DIYer tweak after doing mic testing and (2) as in this thread start, they don't know why they have to do so.
OT posts removed.Interesting discussion of concert halls over at the Curl blowtorch thread, and the two-dozen posts prior.
For a long time, there has been a consensus on certain design goals for concert halls, Stuff like reverberation time x freq.
But more recently, specialists have started to ask what distinguishes great halls from not so great - because they don't want to design ordinary halls which cast a pall on their reputations*. Although I don't have much knowledge of where that discussion is going, I think new and nifty parameters are being debated.
*and they ask for "final cut" rights.
For a long time, there has been a consensus on certain design goals for concert halls, Stuff like reverberation time x freq.
But more recently, specialists have started to ask what distinguishes great halls from not so great - because they don't want to design ordinary halls which cast a pall on their reputations*. Although I don't have much knowledge of where that discussion is going, I think new and nifty parameters are being debated.
*and they ask for "final cut" rights.
Hello,
Are you determined to go beyond the paradox, or remain in the present? If we want to move on it is neccessary to take those things into account in the measurement analysis to be able to know the perceived loudness of a loudspeaker and a room combination. There is no other way?
- Elias
Are you determined to go beyond the paradox, or remain in the present? If we want to move on it is neccessary to take those things into account in the measurement analysis to be able to know the perceived loudness of a loudspeaker and a room combination. There is no other way?
- Elias
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.