PSYCHOACOUSTICS AND I

[svart-hvitt]

Psychoacoustics is the subject where the engineer tries and get the engineering just right for the listener. Example: A flat frequency response curve is a great starting point but tests showed that (an average of) people preferred systematic deviations from the «neutral» curve.

There is, however, one thing that strikes me: Is the «preferred» curve of an individual the same all the time, or does it change over time (during the day, after years of experience, aging etc.)? And may the average of preferred curves among many individuals never fit the one individual’s preferred curve?

I guess what I’m trying to ask is if individual deviations from the normal, standard are so large as to make the normal curve of little interest to the individual. In other words: Could the standard curve be right for the average listener only, who may not exist in reality? Is this a possible weakness in existing psychoacoustics science?

Is the correct curve based on psychoacoustics a sort of Ford Edsel* in audio?


*The Ford Edsel story is a bit more complicated than many think, so I suggest we for this purpose employ the popular variant, i.e. thinking about the Edsel as a product that was too heavily based on customer polls.

 

[RayDunzl]

I don't find my preferences changing.

For me, Flat is where it's at.

Though I have the capability, I don't fiddle with the virtual knobs and sliders (once set, except to refine occasionally).

*except for the rare recording of extreme imbalance, like some old cassette transfer of a live gig recorded on a boom-box.

I let the recording determine the frequency response.

The big fault in the bass is a phase anomaly, seemingly not obvious to the ears with the transient nature of music, but a cancellation occurs at the microphone at the listening position with steady or swept tone.

 

[svart-hvitt]

I don't find my preferences changing.

For me, Flat is where it's at.

Though I have the capability, I don't fiddle with the virtual knobs and sliders (once set, except to refine occasionally).

*except for the rare recording of extreme imbalance, like some old cassette transfer of a live gig recorded on a boom-box.

I let the recording determine the frequency response.

View attachment 8899

The big fault in the bass is a phase anomaly, seemingly not obvious to the ears with the transient nature of music, but a cancellation occurs at the microphone at the listening position with steady or swept tone.

What’s the different curves/colours?

 

[RayDunzl]

What’s the different curves/colours?

Gray - the measured swept-sine frequency response at the listening position
Red - Peak sound pressure level history, music playing.
Black - Current sound pressure levels by frequency, music playing

 

[svart-hvitt]

Ray,

if you’re a flat person, it means you are deviant, right? So gear that overweights certain frequencies (for example according to psychoacoustics research) will not be to your liking.

PS1: Why is the great null around 50 Hz?
PS2: Impressive that you have extension way below 20 Hz.

 

[RayDunzl]

if you’re a flat person, it means you are deviant, right?

“Without deviation from the norm, progress is not possible.” ― Frank Zappa

PS1: Why is the great null around 50 Hz?

My room is rectangular, open on the left rear quarter, so, asymmetrical, and that seems to be responsible for causing the standing wave from left and right speakers around 47 hz to arrive at the listening position 180 degrees out of phase with each other, causeing cancellation (deep dip) when both speakers produce the tone, intially, in-phase.

Phase of Left and Right (individually) at listening position

SPL for Left, Right, and Both, at listening position

PS2: Impressive that you have extension way below 20 Hz.

The speaker cones move, pressure and rarefaction occurs, I don't "hear" anything below 25Hz.

The mains are MartinLogan reQuest, with a 12" sealed woofer.

I've added some cheap subs as an experiment, they are ported and claim a port tuning of 23Hz, and roll off below that. I see their purpose as reducing low frequency distortion, allowing me to run them (and the mains woofers) about 9dB lower than without.

 

[Cosmik]

I think that my preference probably varies all the time, and knowing that my preference is varying probably affects my preference, etc. But I don't try to chase it down - if my preference is for more treble or whatever that day, then my preference is probably *wrong*. If I listen for a little while, and try to forget monitoring my own listening, my preference will probably settle down to 'flat'.

But we must be careful what we mean by "flat". We don't mean simplistically flat at the listening position. We mean that the speaker should be flat, but the room will do what it likes.

Quoting Floyd Toole:

It is a bold assertion that a single steady-state measurement in a room—a room curve—can reliably anticipate human response to a complex sound field. Such measurements take no account of the direction or timing of reflections within the sound field. Time-windowing the measurement is useful to separate events in the time domain, but these too ignore the directions from which sounds arrive. Human listeners respond to these cues, in some detail, and they exhibit skills in separating room sound from the timbral identity of loudspeakers, and in adapting to different circumstances. This is, after all, what happens at live, unamplified, musical events. This means that not everything measured is perceptually important, nor can our reaction to such sound fields be constant, we adapt.... The simple measurements therefore cannot be definitive. Assuming that we had a credible target for a room curve, not all irregularities seen in it indicate the presence of a problem. If a problem is thought to exist, the visual cues do not lead unambiguously to a cause, and therefore to the appropriate remedy, as was seen in Fig. 4 and the accompanying discussion. The advent of DSP brought with it many explorations of what could be done when the starting point is a measurement in a room that is dominated by non-minimum phase acoustical interference. Genereux [10] provides a good perspective on early efforts, and Fielder [11] provides a detailed analysis of other approaches. The results are mixed, with substantial limitations being found to what can be corrected. The underlying issue is one that Genereux stated clearly: “[W]e are not interested in removing all the effects of the room. Instead we wish to remove any audible colourations evident to the listener.”

...direct and reflected sounds generally come from different directions, which a microphone cannot differentiate, but binaural hearing can. References [1, 2,12– 14] focus on several of the perceptual consequences of reflections. It turns out that in most small-to-medium sized sound-reproduction spaces human listeners find these multi-directional reflected sounds to be mostly benign, even beneficial if the loudspeaker has relatively constant directivity. A common perception is spaciousness—information about the listening space, not timbre-damaging comb filtering. This is certainly true for recreational listening, but professionals may find that a less reflective space is preferred for mixing but perhaps not for mastering recordings [12]. These measurements are therefore misleading, and even if equalization were capable of removing such reflections, there is the decision of whether it is necessary. Overall, equalizing the spectral fluctuation “errors” found at a few measurement locations, can add new spectral variations to the total sound output of the loudspeaker that is radiated to all locations throughout the room. There is a significant risk of degrading the performance of good loudspeakers

Speakers like the Kii Three don't need 'correcting' for the room - they just sound flat because they are flat at source.

 

[svart-hvitt]

I think that my preference probably varies all the time, and knowing that my preference is varying probably affects my preference, etc. But I don't try to chase it down - if my preference is for more treble or whatever that day, then my preference is probably *wrong*. If I listen for a little while, and try to forget monitoring my own listening, my preference will probably settle down to 'flat'.

But we must be careful what we mean by "flat". We don't mean simplistically flat at the listening position. We mean that the speaker should be flat, but the room will do what it likes.

Quoting Floyd Toole:


Speakers like the Kii Three don't need 'correcting' for the room - they just sound flat because they are flat at source.

I dare say that modern speakers, both «hifi» and «pro», measure pretty flat these days. I don’t have any data to back me up, but I claim that speakers are «flatter» today than they used to be. Yet they sound different. And even if a DSP speaker like Kii measures flat in anechoic chambers, the frequency curve can be chaotic at the listening position because the room enters the equation.

There are different approaches to DSP. Genelec prefer to adjust for the room, but only by cutting peaks, not by trying to fill out dumps and nulls. Off-the-shelf standalone DSP software often try and correct both peaks and troughs (even if troughs can be hard for your amplifiers to rectify, with loss of headroom as a result).

When doing DSP based room correction, I believe you’d better see to it that direct sound is greater than reverberant sound.

My point is this: Toole et al. may come up with general guidance but do they take into account that individual listeners deviate from the neutral, normalized settings?

In other words: Should every audio setup be calibrated to the individual listener instead of the average listener?

FWIW: I like these two blog posts by Genelec’s Aki Mäkivirta and Thomas Lund:

https://www.genelec.com/blog/neutrality-safe-choice

https://www.genelec.com/blog/frequency-response-and-personal-preference

Please feel free to point out weaknesses in the arguments of Mäkivirta and Lund

 

[RayDunzl]

So gear that overweights certain frequencies (for example according to psychoacoustics research) will not be to your liking.

My car pulls to the left a little. I wouldn't be any happier if it pulled to the right.

Speakers like the Kii Three don't need 'correcting' for the room - they just sound flat because they are flat at source.

Well...

Kii III

Looks pretty flat...

JBL LSR 308

Looks flat to me...

JBL LSR 308 in-room here, no EQ:

Has some annoying (to me) low frequency problems.

LSR 308 (left and right) chopped and channeled here at Neverland East:

(at the listening position, because that's where I measure)

 

[svart-hvitt]

My car pulls to the left a little. I wouldn't be any happier if it pulled to the right.



Well...

Kii III

Looks pretty flat...

JBL LSR 308

Looks flat to me...

JBL in-room here, no EQ:

View attachment 8902

Has some annnoying (to me) LF problems.

LSR 308 (left and right) chopped and channeled here at Neverland East:

(at the listening position, because that's where I measure)

View attachment 8903

Just to complicate further:

https://www.princeton.edu/3D3A/Directivity.html

(Which explains my preference for the idea of point-source speakers).

However, let’s try and answer this question:

=> Because my preferences are not your preferences or the group’s preferences we need to calibrate the audio reproduction system to the individual’s preferences, not the average of many persons’ preferences.

 

[March Audio]

The Toole research, IIRC, showed that speakers with certain charcteristics were consistently favoured, so actually preferences are quite similar.

 

[svart-hvitt]

The Toole research, IIRC, showed that speakers with certain charcteristics were consistently favoured, so actually preferences are quite similar.

Yes, correct. And, anecdotally, flat measuring speakers outperform in Stereophile subjective reviews as well.

I just wondered if there is an element of «Ford Edsel» factor in psychoacoustic research. How do we see to it that your new speakers match YOUR (and not everyone else’s) preferences?

DSP is probably key to a solution here. I just wondered how one should go about and map an individual’s true preferences, to be controlled for his time-varying preferences etc.

 

[Cosmik]

...even if a DSP speaker like Kii measures flat in anechoic chambers, the frequency curve can be chaotic at the listening position because the room enters the equation.

I think you may be missing Floyd Toole's main point. He is saying that the frequency curve at the listening position doesn't tell you how it will sound, because the ears/brain are performing more than just a frequency response analysis.

An analogy might be the way the eye focuses light rays to form an image. Suppose we place a TV in the corner of a room which has predominantly red decoration and we display white on it. A photocell somewhere in the room (instead of a microphone) would show an overall red balance to the light being received (a frequency response curve). However, place a lens in the same place, e.g. an eye looking at the monitor, and the observer will see a neutral white monitor - it would take quite an anomaly in the ambient lighting and reflections to upset that perception to any great extent. I think our hearing does something similar: we 'focus' our attention on the source, by 'refracting' the 'rays' of sound that are flying around the room. We use phase and timing to do this. We hear a neutral speaker as neutral, even if the room is not 'neutral'.

Unfortunately, while it is obvious how the eye does its focusing, our hearing doesn't reveal how it works so easily. To a certain extent, you must either observe your own internal responses to signals and measurements and form a judgement of how it may work, or you believe that the ear is nothing but a frequency response analyser. (This is like believing that free markets lead to optimal outcomes - there is a certain logic and simplicity to it, but it is nothing but a belief).

 

[amirm]

THERE is more involved here than psychoacoustics. For example room surfaces easily absorb high frequencies. Lows not so much. In that regard a measurement that includes reflections should not be used to correct that or the direct sound will be brighter.

Another factor us that there is no standard curve that was used in recordings. So if music is the reference as opposed to tones then we may easily like one curve vs another.

If you do a controlled test where the room and music are the sane then our preferences highly converge. See Oliver's test of room eq products.

 

[DonH56]

Also note dispersion is almost impossible to control by applying DSP to a conventional speaker design. Multiple drivers under independent control are usually required for that. Controlling reflections and reverberations with time-varying signals is another challenge. There are limits to what DSP can do.

 

[svart-hvitt]

I think you may be missing Floyd Toole's main point. He is saying that the frequency curve at the listening position doesn't tell you how it will sound, because the ears/brain are performing more than just a frequency response analysis.

An analogy might be the way the eye focuses light rays to form an image. Suppose we place a TV in the corner of a room which has predominantly red decoration and we display white on it. A photocell somewhere in the room (instead of a microphone) would show an overall red balance to the light being received (a frequency response curve). However, place a lens in the same place, e.g. an eye looking at the monitor, and the observer will see a neutral white monitor - it would take quite an anomaly in the ambient lighting and reflections to upset that perception to any great extent. I think our hearing does something similar: we 'focus' our attention on the source, by 'refracting' the 'rays' of sound that are flying around the room. We use phase and timing to do this. We hear a neutral speaker as neutral, even if the room is not 'neutral'.

Unfortunately, while it is obvious how the eye does its focusing, our hearing doesn't reveal how it works so easily. To a certain extent, you must either observe your own internal responses to signals and measurements and form a judgement of how it may work, or you believe that the ear is nothing but a frequency response analyser. (This is like believing that free markets lead to optimal outcomes - there is a certain logic and simplicity to it, but it is nothing but a belief).

Yes, I may be missing Toole's point, or maybe not.

Please note that I made the following qualifier: "When doing DSP based room correction, I believe you’d better see to it that direct sound is greater than reverberant sound".

In other words: When the direct sound overwhelms the reverberant sound, DSP based room correction may have more legitimacy, would you agree?

Everybody agrees that "there are limits to what DSP can do [in terms of room correction]" (that sentence is written in a way that everyone can agree upon and therefore of limited value...). Differences among people seem to stem from how big a part DSP based room correction could or should be. There are obviously different views and schools here. Many experienced, resourceful master engineers use a little DSP based room correction these days, while others do not. Is there evidence that DSP based room correction is "unscientific"?

 

[Cosmik]

Yes, I may be missing Toole's point, or maybe not.

Please note that I made the following qualifier: "When doing DSP based room correction, I believe you’d better see to it that direct sound is greater than reverberant sound".

In other words: When the direct sound overwhelms the reverberant sound, DSP based room correction may have more legitimacy, would you agree?

I think I know what you may be referring to here, but I think I still disagree with the overall aim. As long as, in all cases, a DSP 'anti-pulse' (to kill a reflection as it reaches the listener's ear) is at a higher level than the anti-pulse needed to then kill the reflected results of that anti-pulse, a DSP-based system can perfectly deconvolve the room's impulse response from the signal - at a single point in space, at least. The measured results would be 'perfect' in that the signal would be being projected anechoically to the listener's ear - as though he is wearing headphones. He would have to keep his head very still.

Everybody agrees that "there are limits to what DSP can do [in terms of room correction]" (that sentence is written in a way that everyone can agree upon and therefore of limited value...). Differences among people seem to stem from how big a part DSP based room correction could or should be. There are obviously different views and schools here. Many experienced, resourceful master engineers use a little DSP based room correction these days, while others do not. Is there evidence that DSP based room correction is "unscientific"?

My view is that the desire to achieve 'perfect measurements' at the listener's ear is in direct conflict with the idea of using speakers in a real room. You want to listen to speakers in a room in order to restore genuine, coherent acoustic sound fields to an audio scene that comprises only of a single, static 'sound field' (recorded with stationary microphones, replayed on stationary speakers). The addition of the room ambience gives you the ability to move your head (over a limited range) and locate an audio 'scene' in front of you, as though it is real. Attempting to remove that ambience you have just added by fiddling with the signal in order to get a simplistic 'flat' measurement is an error. Not a compromise, but a straightforward error. There should be no 'correction' of the room above the bass, because it is not correction at all, and will be heard by the listener as an error.

DSP correction should be limited to fine tuning of the errors caused by the speaker's own physical presence - which are not fully correctable, but can be adjusted in a limited way to sound subjectively better.

 

[Fitzcaraldo215]

Is there evidence that DSP based room correction is "unscientific"?

If it is, then vive la "unscientific". I would not be without it.

But, no, I am not aware of anything proving it to be unscientific. DSP developers pretty much all study and use available science, but they might still disagree on what are the proper tradeoffs in applying that science. Room acoustics and perceptual psychoacoustics are so messy and chaotic that it is almost a wonder DSP EQ works at all. But, work it does, though it is not a cure for all problems. Target curves can usually be personalized to your typical preferences, but not usually rapidly or easily.

Toole laments that easy to adjust tone controls have been essentially banished from quality audio gear today. That might be a better answer to what you seek, rather than DSP Room EQ.

Me? I derive more pleasure from listening to the music rather than fiddling with knobs to adjust the sound of that music. I fiddle very carefully during setup with a wide range of recordings, but then I shift gears to listening and not fiddling for months or years. Toole also cites our ability to adapt to the sound that we have, rather than impatiently changing it when we have no standard for what it is supposed to sound like.

Is your momentary seeming "preference" reliable anyway, given mood swings, state of inebriation, and possibly ambient temperature, humidity and atmospheric pressure, etc.? And, let's not forget the circle of total confusion about the quality of the recording itself. They vary all over the place and we know not the sound that was actually recorded. So, indulging in momentary but arbitrary "preference" by changing frequency response seems an unpleasurable waste of time to me.

 

[svart-hvitt]

I think I know what you may be referring to here, but I think I still disagree with the overall aim. As long as, in all cases, a DSP 'anti-pulse' (to kill a reflection as it reaches the listener's ear) is at a higher level than the anti-pulse needed to then kill the reflected results of that anti-pulse, a DSP-based system can perfectly deconvolve the room's impulse response from the signal - at a single point in space, at least. The measured results would be 'perfect' in that the signal would be being projected anechoically to the listener's ear - as though he is wearing headphones. He would have to keep his head very still.

My view is that the desire to achieve 'perfect measurements' at the listener's ear is in direct conflict with the idea of using speakers in a real room. You want to listen to speakers in a room in order to restore genuine, coherent acoustic sound fields to an audio scene that comprises only of a single, static 'sound field' (recorded with stationary microphones, replayed on stationary speakers). The addition of the room ambience gives you the ability to move your head (over a limited range) and locate an audio 'scene' in front of you, as though it is real. Attempting to remove that ambience you have just added by fiddling with the signal in order to get a simplistic 'flat' measurement is an error. Not a compromise, but a straightforward error. There should be no 'correction' of the room above the bass, because it is not correction at all, and will be heard by the listener as an error.

DSP correction should be limited to fine tuning of the errors caused by the speaker's own physical presence - which are not fully correctable, but can be adjusted in a limited way to sound subjectively better.

Interestingly, Genelec has a BYPASS button for you to click on in order to A-B test room correction in real-time. If room correction were flawed, why let their users A-B their audio setup with and without DSP based room correction. I guess DSP based room correction works in many cases.

There are different ways to trim the frequency curve depending on room interaction. Genelec only takes away peaks, not troughs, and mostly in lower frequencies.

Stand-alone DSP software is much more aggressive.

I once heard a very complex system; 8 times 24 inch woofers for lower frequencies, 6 times 15 inch woofers for mid bass. It was all controlled by standalone DSP software for crossover handling and room correction, in a heavily treated, one-purpose room. In sweet spot it was extremely good (like in «most fascinating sound ever»). Outside of sweet spot not good at all.

I guess it depends. The Genelec way with 3-way coaxials and full DSP works very good off and on axis, as well as when I go into another room (the music sounds natural). Last week I had the chance to test them side by side against Kii Three. I think I will write more on that later.

 

[Fitzcaraldo215]

There are different ways to trim the frequency curve depending on room interaction. Genelec only takes away peaks, not troughs, and mostly in lower frequencies.

Stand-alone DSP software is much more aggressive.

Some stand alone DSP is more aggressive, some is not. I once used Audyssey and I now use Dirac. Both had limits of about 8-10 dB of maximum boost they would apply, though there were no limits on the amount of cut. That is fine as long as there is sufficient amp power, which I do. I have never had a problem with that, and I find it a better idea than no boost at all in reducing or eliminating some deep troughs but not overdriving infinite nulls. I have never had a true infinite null, myself.

Some friends are using Anthem ARC in their prepros. My understanding is that it has no limits on the amount of boost or cut, a potentially bad situation. But, they too have not had problems with it.

One friend has a very deep null using his Revel Salon 2s, but he is using Dirac, so no problem.