Tonality is very important. It is also easy to manipulate when the direct and reverberent fields bear close ties. When this is true, it is possible to nail a target curve and the result will sound just like that curve. Any subjectively desired EQ will be minor, I find typically less than 0.5dB in the critical regions with a little more leeway further down.
On the other hand, there is diffracton. Studies have shown that it is a contributor to level dependent distortions. I believe that clean peaks are necessary to maintain the tonality throughout, and in order to appreciate it you also need to ensure the performers remain as clearly delineated during peaks.
I have a suspicion furthermore, since diffraction artefacts will have a unique timing and directivity structure to any other source which may aid in their localisation, that there may be nothing practical that can be done aside from mopping it up with absorption, other than avoiding it altogether.
On the other hand, there is diffracton. Studies have shown that it is a contributor to level dependent distortions. I believe that clean peaks are necessary to maintain the tonality throughout, and in order to appreciate it you also need to ensure the performers remain as clearly delineated during peaks.
I have a suspicion furthermore, since diffraction artefacts will have a unique timing and directivity structure to any other source which may aid in their localisation, that there may be nothing practical that can be done aside from mopping it up with absorption, other than avoiding it altogether.
While thats true, and most speaker manufacturers offer woefully inadequate measurements for their speakers to the public, (nobody seems to want to provide warts and all measurements, even an un-smoothed on axis response is rare) its not clear what we would do with an exhaustive set of measurement data if it were available when attempting to compare or select commercial speakers.
Even for those knowledgeable in the field there is no consensus of agreement on many of the parameters of speaker design - just look at the threads "What is the ideal directivity pattern for stereo speakers?", and "'Flat' is not correct for a stereo system ?", both of which number in the thousands of posts, and heavily debate the "ideal" directivity and polar response of speakers, (among other factors) with no final conclusion or agreement. Not even close.
(The only conclusion I could draw from those threads was that although directivity does definitely matter, and I personally favour directivity, there is no one particular ideal polar response that is "optimal" or best for all situations)
For each measurable parameter of speakers there are those followers who believe that is the most important one of all, (almost with religious fervor) and that the other parameters are of secondary or minor importance.
To name a few we have:
1) Those who believe that achieving flat on axis frequency response is the primary determining factor in perceived quality.
2) Those who believe that off axis response, polar response, power response and so on are the primary determining factor in perceived quality. Within this group you have two diametrically opposing viewpoints - "as wide as possible" dispersion at all frequencies, vs "controlled/constant directivity".
3) Those who believe that achieving extremely low levels of distortion is the primary determining factor in perceived quality.
4) Those who believe that the time domain response is the primary determining factor in perceived quality - such as cumulative spectral decay, and elimination of time delayed reflections and diffraction from the cabinet.
5) Those who believe that resonances, or more specifically the elimination of them is the primary determining factor of perceived quality. (This ties in with both time domain response and frequency response flatness beliefs as resonances affect both...)
For each category you will find ardent followers, and the designs they devise will typically follow on naturally from their belief of what is the most important parameter, and often focus heavily on that one parameter at the expense (or at least neglect) of others.
Who is right ? Most likely all those factors (and others I haven't listed) are important, but the relative weighting is still up for debate even amongst the experts, otherwise we wouldn't have so many radically different approaches to speaker design.
I agree that tweeter and midrange controls are a bad idea, (especially if implemented passively as switchable L-Pad's) and that overall EQ for room variations should only be applied below about 250Hz, however what exactly are you suggesting ?
"provisions to compensate for variables of different rooms" means absolutely nothing. Give us a concrete example of a room "problem" and what your ideal speaker design would provide that could make compensation of that problem possible without EQ or a fundamental design change to the speaker.
So what solution do you propose then ?
I don't agree that its a battle that can't be won. There is such a thing as a "room tolerant" speaker - a speaker that will sound close to neutral whether its in a dead or a live room.
This is achieved by having the right directivity vs frequency profile, and extremely wide dispersion in the treble (the goal of most small cone+dome designers) is not it.
Its easy to demonstrate. First compare a large full range driver which beams at high frequencies (as narrow as 20 degrees in the high treble) in a well padded fairly dead room vs a typical acoustically live living room.
In one room or the other its possible to get a good balance with EQ, however shift from one room to another without changing the EQ and the balance will radically change.
Start with the dead room and EQ that gives a neutral balance then move the system to the live room. What happens ? There is a perceived lack of treble, and the response sounds dull because there's almost no treble in the reverberant field. Boost the high frequencies several dB with EQ and it may sound a bit more balanced, but then it will be uncomfortably bright if taken back to the dead room.
A large full range driver is an example of a speaker that is not "room tolerant". It's directivity is such that perceived balance shifts too much with room characteristics.
Now try the same test with a small bookshelf 2 way cone+dome design which has been optimized for as wide a dispersion as possible.
Start with the dead room and a perceptually neutral EQ balance. Move it to the live room. What happens ? Now its too bright in the treble (and probably upper midrange) compared to the dead room.
To get a perceptually neutral balance the upper midrange and treble will need to be reduced a bit - not as much as the full range driver needed to be increased, but maybe on the order of 1dB or so.
Although the wide dispersion speaker is more "room tolerant", perceptually it actually errs in the other direction of becoming too bright when the room is made more live. (Who here hasn't noticed that live rooms tend to sound too bright with wide dispersion speakers ?)
The conclusion from this is obvious - attempting to maintain a flat power response at high frequencies or an extremely wide dispersion is wrong, but so is having an extremely narrow dispersion at high frequencies - the optimal lies somewhere in between.
With a degree of controlled directivity at higher frequencies (higher being from middle/upper midrange upwards) you can have a speaker that stays fairly close to perceptually neutral (ignoring bass response changes) with a variety of room conditions.
I don't think there is one ideal directivity vs frequency profile to do this, however there are a certain range of acceptable solutions. Defining what those boundaries are is the challenge, and is an as yet unresolved debate.
I've generally found that the two go hand in hand rather than being mutually exclusive. Once other things like minimising diffraction, (especially at high frequencies) phasing/phase tracking of drivers is correct, and polar response of the speaker is "acceptable", achieving good imaging or not then comes down largely to frequency response balance.
Relatively small errors in frequency response in certain key frequency ranges (particularly the presence region but also elsewhere) can make or break "imaging", and I believe are related to the way we perceive direction of sound sources based on modification of the frequency response. (HRTF etc)
Errors in key frequency regions of the HRTF cause "confusion" in our perception of where a sound is coming from, causing ambiguity in apparent sound source location and a loss of "imaging".
When those errors in critical frequency regions are eliminated good imaging is obtained, and this generally happens very close to a flat neutral response, so tonality need not be sacrificed for imaging.
"however what exactly are you suggesting ? "
That a loudspeaker should have not only flat response in its direct field with respect to the listener at whatever angle he is at but should produce flat reflections as well. There are no speakers designed with this capability. This flaw among many others is why two speakers with the same on axis response sound different in the same location and why the same speaker sounds different in two different locations.
There really is no such thing as flat frequency response unless you not only solve that problem but the fact that the variables of spectral balance of commercial recordings varies all over the lot. The system must be compensated for those variables too or it will sound different on every recording even on the same record label. The term flat response means nothing if it is not flat all the way back to the recording microphone. Yet experience shows that both vinyl phonograph records and cds vary enormously because there is no standard way to make a recording. In this regard, in general cds are even worse than vinyl, one reason so many peole don't like them. So equalization is necessary but not sufficient. BTW, calling equalization "room correction" because it is performed automatic doesn't change what it is, it just disguises it from those whose religion says equalization is bad. My experience is that there isn't a single area of the audio spectrum which isn't critical to hearing musical instruments reproduced accurately. But of all of them, the treble is most critical. If that isn't right, nothing else works.
The experimental systems I design for my own use only incorporate engineering features which allow these problems to be solved or at least mitigated.
That a loudspeaker should have not only flat response in its direct field with respect to the listener at whatever angle he is at but should produce flat reflections as well. There are no speakers designed with this capability. This flaw among many others is why two speakers with the same on axis response sound different in the same location and why the same speaker sounds different in two different locations.
There really is no such thing as flat frequency response unless you not only solve that problem but the fact that the variables of spectral balance of commercial recordings varies all over the lot. The system must be compensated for those variables too or it will sound different on every recording even on the same record label. The term flat response means nothing if it is not flat all the way back to the recording microphone. Yet experience shows that both vinyl phonograph records and cds vary enormously because there is no standard way to make a recording. In this regard, in general cds are even worse than vinyl, one reason so many peole don't like them. So equalization is necessary but not sufficient. BTW, calling equalization "room correction" because it is performed automatic doesn't change what it is, it just disguises it from those whose religion says equalization is bad. My experience is that there isn't a single area of the audio spectrum which isn't critical to hearing musical instruments reproduced accurately. But of all of them, the treble is most critical. If that isn't right, nothing else works.
The experimental systems I design for my own use only incorporate engineering features which allow these problems to be solved or at least mitigated.
Cue speaker dave's response in 5....4....3....2....1 😀
Speakers with different polar responses will sound different in the same location despite having the same on-axis response, but which one is correct ? How about none of them ?
And why wouldn't it sound different ? If you are suggesting that a speaker with the same spectral balance off axis as on axis will somehow sound the same in different rooms, then I think you had better re-think that theory. There is no evidence at all to suggest this is or could ever be the case.
It completely ignores many things about how our hearing mechanism works. Our ears and brain are not a spectrum analyser connected to a microphone...
Yes but lets not wander into the wilderness of the "circle of confusion". Speaker designers can't (and shouldn't) try to "solve" this problem, because they can't. Better to stick to designing the speakers to some sort of standard and let the recording procedures eventually sort themselves out.
I'm not sure if you really mean "flat reflections" as in "the system even compensates for non flat absorption of the room's surfaces" or just that the speaker has flat response at every angle.
I'll assume the later since the former is unlikely to be possible/practical.
Are you advocating an omnidirectional system or just a full range constant directivity system? Either would send the same frequency response in every direction.
If you advocate omnidirectional systems you are out of step with all the modern research that I have read that repeats that flat power response omni systems always sound too bright.
Here is one I have recommended to you many times:
AES E-Library Experiments in Direct/Reverberant Ratio Modification
If you are just looking for constant directivity (same polars for all frequencies, hence same response) you still need to define what directivity you think is ideal.
I'd still suggest that modern research shows that the ear is surprisingly tolerant to a wide range of response in the reflected field and that the direct field dominates in matters of perceived frequency response.
David S.
"If you are suggesting that a speaker with the same spectral balance off axis as on axis will somehow sound the same in different rooms, then I think you had better re-think that theory"
That's not what I said. I said that provisions must be engineered into the system to accomodate different rooms to produce the same end result. No fixed FR will work. It will always be wrong unless you happen to hit it lucky. That never happens.
"Yes but lets not wander into the wilderness of the "circle of confusion". Speaker designers can't..."
No they can't, that's obvious from the results they've produced. In fact the speaker is onlly one element of an entire recording/playback system. When are variables in some elements of the system, the recording elements for example, other elements must compensate or the results will never be right. And experience with commercial recordings and commercially sold equipment demonstrates that to critial experienced listeners of live music, it never is. That is one reason people still attend live concerts. They sound better than recordings.
That's not what I said. I said that provisions must be engineered into the system to accomodate different rooms to produce the same end result. No fixed FR will work. It will always be wrong unless you happen to hit it lucky. That never happens.
"Yes but lets not wander into the wilderness of the "circle of confusion". Speaker designers can't..."
No they can't, that's obvious from the results they've produced. In fact the speaker is onlly one element of an entire recording/playback system. When are variables in some elements of the system, the recording elements for example, other elements must compensate or the results will never be right. And experience with commercial recordings and commercially sold equipment demonstrates that to critial experienced listeners of live music, it never is. That is one reason people still attend live concerts. They sound better than recordings.
Would they ? I don't see how a "full range" constant directivity system could ever be realised, at least in a living room.
That would require maintaining constant directivity control right down to bass frequencies, and I'm not even sure that makes sense in the context of room modes, let alone whether its physically realisable.
Every practical constant directivity system is only constant directivity above some corner frequency, below that it gradually becomes omni-directional, so in fact the response is not spectrally flat off axis if you measure right down to the bass.
Does it matter ? I don't know. If we were able to maintain constant directivity down to the Schroeder frequency, maybe thats enough to be considered truly constant directivity, especially when the frequency response below that frequency is going to be dominated by room modes and need EQ and/or modal smoothing anyway.
Practically speaking an omni design would be the only way to achieve an identical on and off axis response, but you then have the major problem that to achieve a flat on axis response would also require a flat power response, which will sound far to bright. (Not to mention a very poor direct/reflected ratio)
Not saying it is easily realizable, just interpreting what he is saying.
I do think that CD over a range should not be simply refered to as constant directivity.
David S.
I do think that CD over a range should not be simply refered to as constant directivity.
David S.
What "provisions" ? You keep talking in vague and sketchy terms. What exactly would you make adjustable in a speaker design aside from EQ to accomodate different rooms, and are you only trying to address the perceived tonal balance, or also the direct/reflected ratio and so on.
Part of breaking the circle of confusion is to introduce a standard somewhere in the loop. That standard is the playback speaker. If we just design speakers to "sound good" by compensating for the "errors" further up in the chain, then all hope is lost because now engineers will be trying to adjust the music to "sound good" on our speakers which were adjusted to "sound good" on previously recorded music.
Somewhere in the loop a line has to be drawn in the sand that says "playback speakers will measure flat" (both mastering speakers and home listening speakers) so that there is at least an anchor for the circle.
That doesn't mean that overnight all recordings will suddenly sound great, because there is still taste when it comes to mixing. I'm sure that many recordings sound bad simply because the "engineer" has poor taste, or is forced by the producers to go for a certain type of sound. (I think we all know what that is)
To simply throw our hands in the air in despair is silly though. Build speakers to be flat and neutral, and use active tone controls/EQ on recordings that need it.
The funny thing I've noticed is that as my speakers have gotten better over the years I've had less and less need for tone controls.
I always used to find myself turning the bass and treble up or down on different recordings to make them sound ok, but I now very rarely ever deviate from a flat response. Yes recordings vary, some are a little bit bass heavy and some are a bit bass shy etc, but they nearly all fall within a comfortable range.
Last edited:
I used an Acurus L-10 for about 15 years, never missed the tone controls. I always flew with the notion that what was on the disc was what the artist wanted us to hear, even if it felt unbalanced.
It's my belief that the best speaker systems make bad recordings sound really bad and good recordings sound really good. They expose the warts and the beauty. Mediocre speakers tend to make everything sound 'acceptable'. With my system, I have gradually become more discerning in my music purchases to the point of the recording's quality being 1st priority, and performance comes in a close 2nd.
I agree with that, however I was referring to perceived tonal balance variations of recordings rather than sound quality overall.
I think what happens is if you have a speaker with a fairly non flat response with significant peaks and valleys especially at higher frequencies, then specific songs can tend to "ping" those peaks and make them obnoxious - and you then tend to want to turn that frequency range down, for example turning the treble down if its in the treble, or turning the bass and treble up if its in the midrange. (Effectively turning the midrange down)
Other songs may have significant content in the parts of the spectrum where your speakers have holes in their response, and for these songs you may tend to make the opposite "correction". No one EQ setting sounds satisfactory for the majority of your music.
As the speaker becomes flatter throughout its range there are no peaks for the music to excite nor holes for important parts of the song to get lost in, so you tend to leave the tone controls alone even over a wide range of music, aside from perhaps a few outliers which are obviously badly mixed. That's what I find anyway.
Last edited:
My take on that is a little different. Very good speakers (and upstream gear) do show the differences in recordings, but few of these recordings tend to be annoying as they are on lesser gear. In fact better speakers tend to reveal hidden good qualities in bad recordings.
When ever I read someone claiming that "My speakers are so good that they can play only the best recordings, bad recordings are unlistenable" or something similar - I just chuckle to myself and think "Better take a look at your crossover." 😉
A good example of this is speakers that are described in reviews as being very "revealing". On good recordings they will praise the speaker for being so life-like and revealing every tiny detail of the music like no other speaker, but that other "bad" recordings will be unlistenable, with the recording getting the blame.
With almost monotonous regularity when you get to the measurements part of the review it is seen that so-called "revealing" speaker has a sharp several dB peak in the presence region, typically near 3-4Khz.
There are a couple of models of high end and rather expensive B&W speakers (which shall go nameless) that are praised by the reviewers as "ruthlessly revealing" that have just such a 3-4dB spike at 4Khz, and it's by no means uncommon in "high end" speakers. Peaks in the higher treble can have a similar effect.
Closer to home the full range drivers I use in my speakers do also have a modest natural peak of 1-2dB at exactly 4khz, (also my crossover frequency) and although I had largely corrected it in the networks there was still a small bump in the response there, considerably smaller than the peaks I've seen on many stereophile reviews of high end speakers.
Despite that, although "revealing" and enjoyable sounding on nearly all recordings there were a few specific recordings that always bothered me and weren't enjoyable to listen to due to a little bit of a harsh edge to them which I put down to the recording, so I avoided playing them.
More recently I had another go at more accurately correcting the peak in the crossover and touching up the crossover transition region as a whole with the result that a narrow peak of less than 1dB at 4Khz was eliminated.
Well knock me over with a feather - most of the few recordings that had a harsh edge to them now sounded perfectly fine. Even some that were unlistenable before were now almost normal sounding, despite the peak being less than one dB before correction.
Moral of the story - narrowband peaks are not acceptable in the presence region even if it appears to "enhance" the liveliness of some recordings. 1/3 octave averaged measurements are also inadequate, you really do need to examine the frequency response narrow band.
On the other hand there are circumstances where better speakers are more revealing of actual recording flaws - exaggerated top end treble with a peaking filter near 10-15Khz can sound pretty awful on good speakers, but listen to it with a speaker lacking in top octave treble response and it won't be as noticeable or noticeable at all.
Last edited:
- Status
- Not open for further replies.