I posted this on an other forum, but thought I would raise the same questions here.
Comments made in the "Accurate Speakers" thread, e.g. the difference between 'accurate' vs 'natural' have got me thinking (which is not always a good thing). Feel free to correct any terms that I may be misusing here (as I'm pretty much a lay person when it comes to the technical side of audio).
From what I understand, Toole, Olive, and others suggest that listeners will tend to prefer more accurate speakers over less accurate ones. But what is meant by 'accurate' here? Are Toole and Olive using using 'accurate' to mean 'smooth response' (e.g. low rates of distortion, resonance, and so on), and if so is that the same as 'accuracy' in the sense of 'hi-fidelity.' For the purposes of discussion, let me propose an idealized notion of high-fidelity (and remember, this is an ideal).
A speaker is hi-fidelity if it is able to reproduce or represent (i.e. re-present) the auditory features of a performance environment (e.g. a studio performance or concert) in a spatially and/or temporally distinct listening environment (e.g. one's living room or basement). Obviously the faithful reproduction of a 'performance environment' presumes a sonically neutral 'listening environment' (which I realize is unrealistic, but remember this is an idealized account. It does, however, stress the importance of the listening environment in assessing the accuracy of a speaker or system). Basically, the ideal is that an accurate or hi-fidelity speaker should be able to reproduce the auditory features of the performance environment, e.g. the sounds of Wynton Marsalis playing the trumpet, within the listening environment in such a way that one would literally believe that the performance was taking place within the listening environment, e.g. if one closed one's eyes it would sound like Marsalis was literally in the room playing the trumpet or, conversely, it would sound like one was actually present in the original presentation environment (e.g. a studio or concert). Borrowing a term that actually has its origins in medieval philosophy/theology, where it meant that something had the same effect as the real thing without the real thing actually being present, we might say that an accurate speaker would make it so that Marsalis's trumpet playing was present or real virtually within the listening environment. Now, taking this as a kind of idealized account of hi-fidelity accuracy, I want to raise the following questions:
1. What is the relationship between a flat (smooth) frequency response and hi-fidelity accuracy?
2. Is flat or smooth frequency response a necessary condition of hi-fidelity accuracy (i.e. no flat frequency response equals no possibility of hi-fidelity accuracy)?
3. Is flat or smooth frequency response a sufficient condition of hi-fidelity accuracy (i.e. whenever you have flat frequency response you also get hi-fidelity accuracy)?
4. If flat or smooth frequency response is not a sufficient condition of hi-fidelity accuracy, then what are the sufficient conditions of hi-fidelity accuracy (e.g. what role do other design elements such as cone shape, cone materials, magnetic properties, cabinet shape, cabinet materials, and so on, play in a speaker's hi-fidelity accuracy)?
5. If flat or smooth frequency response is not a sufficient condition of hi-fidelity accuracy, then is it more or less important than things like cabinet design, cone shape, and so on?
Please, in discussing this, let's try not to bring in things like the tired old subjectivist/objectivist distinction (which I actually think is a red herring) and focus on the weight that might be assigned to all the factors that go into the production of a hi-fidelity speaker.
Thanks,
Comments made in the "Accurate Speakers" thread, e.g. the difference between 'accurate' vs 'natural' have got me thinking (which is not always a good thing). Feel free to correct any terms that I may be misusing here (as I'm pretty much a lay person when it comes to the technical side of audio).
From what I understand, Toole, Olive, and others suggest that listeners will tend to prefer more accurate speakers over less accurate ones. But what is meant by 'accurate' here? Are Toole and Olive using using 'accurate' to mean 'smooth response' (e.g. low rates of distortion, resonance, and so on), and if so is that the same as 'accuracy' in the sense of 'hi-fidelity.' For the purposes of discussion, let me propose an idealized notion of high-fidelity (and remember, this is an ideal).
A speaker is hi-fidelity if it is able to reproduce or represent (i.e. re-present) the auditory features of a performance environment (e.g. a studio performance or concert) in a spatially and/or temporally distinct listening environment (e.g. one's living room or basement). Obviously the faithful reproduction of a 'performance environment' presumes a sonically neutral 'listening environment' (which I realize is unrealistic, but remember this is an idealized account. It does, however, stress the importance of the listening environment in assessing the accuracy of a speaker or system). Basically, the ideal is that an accurate or hi-fidelity speaker should be able to reproduce the auditory features of the performance environment, e.g. the sounds of Wynton Marsalis playing the trumpet, within the listening environment in such a way that one would literally believe that the performance was taking place within the listening environment, e.g. if one closed one's eyes it would sound like Marsalis was literally in the room playing the trumpet or, conversely, it would sound like one was actually present in the original presentation environment (e.g. a studio or concert). Borrowing a term that actually has its origins in medieval philosophy/theology, where it meant that something had the same effect as the real thing without the real thing actually being present, we might say that an accurate speaker would make it so that Marsalis's trumpet playing was present or real virtually within the listening environment. Now, taking this as a kind of idealized account of hi-fidelity accuracy, I want to raise the following questions:
1. What is the relationship between a flat (smooth) frequency response and hi-fidelity accuracy?
2. Is flat or smooth frequency response a necessary condition of hi-fidelity accuracy (i.e. no flat frequency response equals no possibility of hi-fidelity accuracy)?
3. Is flat or smooth frequency response a sufficient condition of hi-fidelity accuracy (i.e. whenever you have flat frequency response you also get hi-fidelity accuracy)?
4. If flat or smooth frequency response is not a sufficient condition of hi-fidelity accuracy, then what are the sufficient conditions of hi-fidelity accuracy (e.g. what role do other design elements such as cone shape, cone materials, magnetic properties, cabinet shape, cabinet materials, and so on, play in a speaker's hi-fidelity accuracy)?
5. If flat or smooth frequency response is not a sufficient condition of hi-fidelity accuracy, then is it more or less important than things like cabinet design, cone shape, and so on?
Please, in discussing this, let's try not to bring in things like the tired old subjectivist/objectivist distinction (which I actually think is a red herring) and focus on the weight that might be assigned to all the factors that go into the production of a hi-fidelity speaker.
Thanks,
A flat frequency response is ideal but worthless if the driver has high levels of distortion. A low distortion driver is ideal but far less so without a flat response curve.
They are separate parameters, but one can affect/degrade the other.
So you want both; you cannot change a driver's distortion characteristics, but you can tailor its frequency response through filtering.
They are separate parameters, but one can affect/degrade the other.
So you want both; you cannot change a driver's distortion characteristics, but you can tailor its frequency response through filtering.
You also want low resonance. Resonance and driver breakup may or may not show in a frequency response chart. Cabinet resonance can cause a "boxy" sound, driver resonance can cause a "harsh", "tizzy" or other sound, even when the "measured" response is flat. Room contribution can change response, too. "Controlled dispersion" is another factor you need consider.
There are many things to consider and compromize. There's no easy measurement to fully describe how a speaker sounds.
There are many things to consider and compromize. There's no easy measurement to fully describe how a speaker sounds.
Good reproduction starts with a smooth and flat on-axis response. Next comes the off-axis response. This has to a replica of the on-axis response, except some high frequency narrowing. The criterion of smoothness still applies to the off-axis response.
All other measurements are derivatives of the on- and off-axis response. Distortion is important, but it is assumed that in well-designed speakers it is sufficiently low to not cause a problem.
Olive created a weighting scheme and an equation that predicted listener preference based on on- and off-axis measurements. The following are not exact numbers but it went something like this: 30% weight to on-axis smoothness and flatness, 30% weight to off-axis smoothness, 40% weight to bass extension and flatness. There is no dependence on non-linear distortion because most the speakers that were measured and the SPL they were measured at did not produce gross levels of distortion. If you plan to listen at disco club SPLs then you should consider getting speakers that can produce those levels.
There is something to be said about dynamic range also. Basically, the important thing to check is that the frequency response doesn't change with SPL. If at 85 db it is flat, but at 100 db it is wonky, it is no good.
All other measurements are derivatives of the on- and off-axis response. Distortion is important, but it is assumed that in well-designed speakers it is sufficiently low to not cause a problem.
Olive created a weighting scheme and an equation that predicted listener preference based on on- and off-axis measurements. The following are not exact numbers but it went something like this: 30% weight to on-axis smoothness and flatness, 30% weight to off-axis smoothness, 40% weight to bass extension and flatness. There is no dependence on non-linear distortion because most the speakers that were measured and the SPL they were measured at did not produce gross levels of distortion. If you plan to listen at disco club SPLs then you should consider getting speakers that can produce those levels.
There is something to be said about dynamic range also. Basically, the important thing to check is that the frequency response doesn't change with SPL. If at 85 db it is flat, but at 100 db it is wonky, it is no good.
1. A flat frequency response maintains the proportions of loudness vs frequency of the original sound. No speaker ever made has a perfect flat response. Most manufacturers and reviewers will apply heavy smoothing to remove the rough jagged peaks and dips that look almost like fuzz. The result looks like a more or less consistent black line. If the response is a little fuzzy and the line looks a few dB thick at higher frequencies, this is a much better indication of the actual, real, genuine response.
2. If by high fidelity you are implying accuracy, then yes.
However, some recordings are so poor that one may not enjoy it being reproduced with high accuracy. I can play 30 piano recordings and every one of them is equalized different at the studio. Adding to this, people with more sensitive hearing will prefer attenuated treble and accentuated bass in real world long term listening. Younger people are a good example of this. Also consider the number of people constantly swapping gear and seeking "warm midrange" and "deeper solid bass". I suspect what they seek is a different response. These changes are best carried out at the digital or line level though equalization.
3. Absolutely positively not. Bloggers, hobby testers, and even some big name researchers have spent their time investigating and attempting to perfect the frequency response. Some also aim for a specific harmonic distortion. However, they largely ignore power response (directivity), impulse and frequency versus time (CSD). Two speakers with a flat response can sound completely different, as there are other well documented characteristics. If we measure parameters other than frequency response we can see stark differences. If you digitally equalized a PSB speaker it will NOT sound like a Quad ESL-989 because they behave differently. Frequency response can be measured with chirps "or" continuous tones and both give a different response on some speakers. Impulse response, harmonic distortion, power compression, and frequency versus time as some parameters. In terms of frequency response; power response and room interaction, frequency versus time (build up, settling time, and energy storage as additional frequencies) can alter the frequency response with dynamic signals. It's a whole different all game when we put those players onto the field.
4. See Post #45 in your other thread. Beyond a hobbyist conjecture, you'd need to talk with an experienced engineer that's worked in that area. In the end, it would be a lot of math and terminology, and there will be 20 years of catchup for you to play.
5. Things like shape, materials, color, brand name and so on have no direct connection to hi-fidelity accuracy themselves - those properties and design considerations are a means to an end. Materials, magnet design and shapes affect the behavior of the speaker, and it's the behavior which affects its accuracy.
Last edited:
By accurate it is meant reasonably flat frequency response.
This doesn't mean it has to be within 1dB. It is known and
Toole mentions it that eyes get more offended by ragged
frequency response than ears do.
High fidelity and accuracy in my mind are synonyms. The
loudspeaker should not be altering recording created by the
sound engineer.
Good drivers come with flat response and low enough distortion
figures to be worthy of implementing. All kinds of resonances
tied to cone(dome) material, cabinet, room, boundaries have no
place in an accurately performing system.
A big part of the work in designing a good listening experience
is based on eliminating resonances (smoothing).
Recreating the sound field the audience at the concert was enveloped
in, is an art form depending on the recording staff skills. Mikes hear
differently than people do. I am hoping they won't assume people own
crappy equipment
.
Last edited:
The nature of the materials used in the diaphragms of the drivers and the driver combination will also play a part in the overall presentation of the sound you will hear.
With all that mentioned in the above posts in the mix and the tonal differences between drivers it is really a bit of a porridge as to what one gets.
C.M
With all that mentioned in the above posts in the mix and the tonal differences between drivers it is really a bit of a porridge as to what one gets.
C.M
So from what I've garnered to date there doesn't seem to be any widespread agreement regarding the most important determinant of speaker accuracy. Some seem to think that frequency response is the most important determinant, others think it's low distortion, and others again think its good transient response (at least with acoustic instruments).
So which is the more important determinant of speaker accuracy, and why?
1. Frequency Response?
2. Distortion Levels
3. Transient Response
So which is the more important determinant of speaker accuracy, and why?
1. Frequency Response?
2. Distortion Levels
3. Transient Response
Last edited:
You may be able to put weightings on them, but if any one of the three is too bad then the others won't matter any more. The other factor to bear in mind is the room response to the speakers, particularly off-axis. This may be why some speakers which are described as sounding accurate have a slightly declining response into HF (the frequency curve looks tilted down a little).
Robert Greene's blog
REGONAUDIO
and of course Linkwitz, both have more to say on this.
One UK manufacturer uses accurate headphones as a reference to get his (commercial) speakers accurate. (In this case, going back a step, using headphones recommended by the recording engineer who made the reference recordings used for the tests.)
Robert Greene's blog
REGONAUDIO
and of course Linkwitz, both have more to say on this.
One UK manufacturer uses accurate headphones as a reference to get his (commercial) speakers accurate. (In this case, going back a step, using headphones recommended by the recording engineer who made the reference recordings used for the tests.)
Last edited:
Speaker accuracy is mostly dependent on frequency response on axis
and off axis as well. Why is that? People have been put to test to evaluate
sound quality(timbre) and it was done in a scientific manner (blind test).
The overall conclusion was that wide band uniform response is what
primarily matters. It was perceived as good quality.
You can read about it in Toole's "Sound Reproduction (Loudspeakers and Rooms)".
Distortion levels are low enough with any decent speaker unit that it should
not matter much. Modern motors have built in shorting rings, fans of lower
distortion figures can rejoice.
What about transient response? Usually novice DIYers get wrong impression about
it by observing impulse (step) response of the drivers. The higher the frequency
a speaker unit can reproduce the better it looks. It doesn't say anything about how
it will perform except it has the potential of reproducing HF. Low pass that driver
and the pretty response is not so pretty anymore. Translating it to FR graph gives
an easier way to look at useful data.
What should matter is how well the unit performs off axis. That is the reason why
simple one on axis plot is not enough information to have a picture of how the
product will perform altogether. A very high FR variation (6 dB) is not a sign of high
quality.
My understanding is that Toole and Olive equate accuracy with low distortion. If that's mistaken (and it's something like frequency response instead) then I'm open to being corrected.
All I'm doing here is asking whether low distortion (or frequency response) is the primary determinant of what most people would refer to as accuracy in a speaker, or is there more to accuracy than Toole and Olive's account suggests.
I have no particular agenda here and am asking this as an honest question. Since I don't have the expertise to address it myself, I thereby defer to others who have more knowledge in this field and can offer their learned views on the matter.
Let me see if I can make my question clearer.
In their tests of listener preference, Toole and Olive test listener preference against a measurable aspect of a speaker's performance which they call speaker accuracy. As I understand it, the indicator of speaker accuracy that Toole and Olive focus upon is measurable distortion (but I am open to being corrected here). Thus, in their tests, Toole and Olive expose listeners to speakers with various levels of measurable distortion, with the results indicating that listeners tend to prefer speakers with lower levels of distortion. From this Toole and Olive then conclude that listeners tend to prefer more accurate speakers over less accurate speakers.
From my informal queries, however, it seems that not everyone agrees that distortion (or linear response, if that's what they test for) is the primary determinant of speaker accuracy. Some claim, for example, that transient response if a more important determinant of speaker accuracy.
Now as I understand it, it's possible for one speaker 'A' to have lower levels of distortion than another speaker 'B,' but for speaker 'B' to have better transient response than speaker 'A.' Similarly, it's possible for one speaker 'C' to have a smoother linear response than another speaker 'B,' but for speaker 'B' to have a better transient response than speaker 'C.' Assuming all other things to be equal, which of these three speakers is more accurate? To the best of my knowledge, Toole's and Olive's studies don't test for this? They have not tested, for example, listener preference with respect to transient response vs. linear response, or transient response vs. levels of distortion.
I'm simply asking which, if any, of these should be considered the primary determinant of speaker accuracy (and, possibly, listener preference). After all, it's possible that listeners may tend to prefer speakers with a high transient response over speakers with a smooth linear response, or they may prefer speakers with high transient response over ones with low levels of distortion (all other things being equal). I don't think Toole and Olive test for this, and I'm simply wondering what other, more technically learned people here might have to say about the topic.
Thanks,
In their tests of listener preference, Toole and Olive test listener preference against a measurable aspect of a speaker's performance which they call speaker accuracy. As I understand it, the indicator of speaker accuracy that Toole and Olive focus upon is measurable distortion (but I am open to being corrected here). Thus, in their tests, Toole and Olive expose listeners to speakers with various levels of measurable distortion, with the results indicating that listeners tend to prefer speakers with lower levels of distortion. From this Toole and Olive then conclude that listeners tend to prefer more accurate speakers over less accurate speakers.
From my informal queries, however, it seems that not everyone agrees that distortion (or linear response, if that's what they test for) is the primary determinant of speaker accuracy. Some claim, for example, that transient response if a more important determinant of speaker accuracy.
Now as I understand it, it's possible for one speaker 'A' to have lower levels of distortion than another speaker 'B,' but for speaker 'B' to have better transient response than speaker 'A.' Similarly, it's possible for one speaker 'C' to have a smoother linear response than another speaker 'B,' but for speaker 'B' to have a better transient response than speaker 'C.' Assuming all other things to be equal, which of these three speakers is more accurate? To the best of my knowledge, Toole's and Olive's studies don't test for this? They have not tested, for example, listener preference with respect to transient response vs. linear response, or transient response vs. levels of distortion.
I'm simply asking which, if any, of these should be considered the primary determinant of speaker accuracy (and, possibly, listener preference). After all, it's possible that listeners may tend to prefer speakers with a high transient response over speakers with a smooth linear response, or they may prefer speakers with high transient response over ones with low levels of distortion (all other things being equal). I don't think Toole and Olive test for this, and I'm simply wondering what other, more technically learned people here might have to say about the topic.
Thanks,
I'd say an "accurate" speaker reproduces what's on the recording. It's up to the recording engineers to decide if the perspective is to bring the performers into your room, or to bring the listener into a different environment, or even to create a perspective that is totally unreal. It's not the speaker's job to impose a particular perspective.
Hi Phil,
In Toole's "Sound Reproduction" the author mentions staging a listening
test in 1966 that went on for several days to yield enough subjective data
to warrant rudimentary statistical analysis. There was a good agreement
about the products that were preferred and those that were not. The
winning loudspeaker had the best looking set of measured data (smooth
FR on and off axis).
Among tested speakers was Quad Mark 1 ESL which surely qualifies
for a good transient response product. Its on axis performance was
exemplary and many people found great pleasure listening to it in
places where direct sound field was prominent. Because electrostatic
speakers have large diaphragms, their directional performance is
problematic. In a reflective room this loudspeaker did not sound
as good as on axis curve suggested. Off axis curves were highly
attenuated at frequencies beyond 2 kHz.
So the conclusion of the tests performed is that smooth on axis and
well behaved off axis response ( low linear distortion) is an important
part of an accurately sounding speaker.
I have come to exact same conclusion conducting mine experiments.
When in doubt, turn any of the drivers off to experience the extreme
lack of accuracy.
I reviewed Toole's book and Olive's 2004 AES convention preprints regarding his work to correlate objective measurements with listener preference. Both found distortion was not a major contributor to listener preference.
I also agree with wintermute's comment in post #10. Olive did the research and stat analysis and put the weightings on the variables.
I also agree with wintermute's comment in post #10. Olive did the research and stat analysis and put the weightings on the variables.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.