I didn't read any other answers, but here is mine:
Directivity (DI) also matters, a lot, but it is still not as important as the listening axis frequency response.
Cabinet materials matter, but only to the extent that cabinet vibrations are not well dealt with - any material works if used properly. No material is a panacea.
Cone angle affects directivity which can matter depending on how well the system polar response in constant.
I think that your last question was answered above. These are simple answers to complex questions and there can be multiple ideals depending on specific situations. I like music fairly loud and so power handling is an important aspect as is diffraction. If all you listen to is low level stuff then these things don't matter so much. There are lots of different conditions that affect the answers.
Basically to be accurate the response must be smooth and "near flat. The directivity index (DI) will play into how "flat" it should be as well as the room itself. A higher DI will want more HF roll-off, but it must always be smooth. The DI should always be near flat as possible over as wide a range as possible.
Simple answer here, "Yes" if it is not smooth and near flat then it is not accurate and hence not high fidelity.
Almost, but not quite. Too much distortion and frequency response will not matter. But once the distortion is low enough then frequency response is paramount. Most loudspeakers have low enough distortion unless they are "broken".
Directivity (DI) also matters, a lot, but it is still not as important as the listening axis frequency response.
These would mostly be secondary effects, except that diffraction is a particularly interesting form of diffraction and is strongly influenced by the cabinet design. Studies have shown that diffraction is most audible at higher SPL levels so its effect can be very situational dependent - best not to have any.
Cabinet materials matter, but only to the extent that cabinet vibrations are not well dealt with - any material works if used properly. No material is a panacea.
Cone angle affects directivity which can matter depending on how well the system polar response in constant.
I think that your last question was answered above. These are simple answers to complex questions and there can be multiple ideals depending on specific situations. I like music fairly loud and so power handling is an important aspect as is diffraction. If all you listen to is low level stuff then these things don't matter so much. There are lots of different conditions that affect the answers.
You have to understand the terminology. "Distortion" means anything that changes the waveform, so frequency response is a form of distortion, a linear distortion. There is also nonlinear distortion and nonlinear perceived distortion of a linear effect.
Toole and Olive will come right out and say that "nonlinear distortion" is irrelevant, that is why they do not measure it. I would agree with this.
They do not, however, deal at all with Nonlinear Perceived Distortion. They have never tested with it, for it or anything else. They do no tests that have SPL level as a variable - so they cannot make any statements about it.
Given that even the best speakers are so bad in comparison to where we can get to, it is possible to have 2 very different loudspeakers that are equally accurate.
With so many design trade offs with current art, what is "best" can be a very different set of tradeoffs, and hence many different views.
dave
Thanks for all the very informative and helpful comments. It's very much appreciated.
So my take from this so far is that most agree that linear response, both on and off-axis, is a vital indicator of speaker accuracy (and listener preference), but that things like transient response, diffraction, and so on can also play an important role in speaker accuracy and listener preference. Does that sound fair?
If this is the case, then why do many people report not liking some speakers that have a smooth frequency response? I've seen a lot of comments from people on this and other forums complaining that some speakers that seem to measure well in terms of frequency response can nevertheless sound flat or uninteresting. I recall one person, who typically puts a lot of faith in measurements as an indicator of speaker quality, admitting that he actually EQ's his speakers a little off of flat to make certain frequencies more pronounced. Could many of these complaints about the 'flatness' of some speakers with a smooth linear response be the result of those same speakers having a poor transient response (or some other consideration), or is there something else going on there?
So my take from this so far is that most agree that linear response, both on and off-axis, is a vital indicator of speaker accuracy (and listener preference), but that things like transient response, diffraction, and so on can also play an important role in speaker accuracy and listener preference. Does that sound fair?
If this is the case, then why do many people report not liking some speakers that have a smooth frequency response? I've seen a lot of comments from people on this and other forums complaining that some speakers that seem to measure well in terms of frequency response can nevertheless sound flat or uninteresting. I recall one person, who typically puts a lot of faith in measurements as an indicator of speaker quality, admitting that he actually EQ's his speakers a little off of flat to make certain frequencies more pronounced. Could many of these complaints about the 'flatness' of some speakers with a smooth linear response be the result of those same speakers having a poor transient response (or some other consideration), or is there something else going on there?
Transient response is more than likely going to be complained about in the bass region. You hear the term "tight bass" often. I would postulate that an otherwise excellent speaker may get a poor rating if the bass is regarded as "sloppy" or "boomy".
Also please note that Toole's listening tests (as far as I am aware) were conducted with trained listeners. If you did the same tests with people off the st, you might get completely different results.
What people are used to can have a huge influence on what they think sounds good. If they are used to hearing a glaringly inaccurate speaker, then an accurate one may well sound lifeless or dull, or lacking whatever superlative is used to describe their favourite sound.
It doesn't mean that the accurate speaker is bad, it just means it is not to their taste. Let them listen to it long enough, and then ask them to listen to their normal favourite, and they may just change their mind.
I do think that there are other factors such as phase tracking that make a difference (some will agree some wont). experiments with different crossovers which result in very similar on axis FR can sound quite different.
I also have my own theory, that different people are more or less sensitive to certain problems with speakers. As Planet10 has already said there are so many compromises, the compromises one person can live with, another may not be able to.
Tony.
Also please note that Toole's listening tests (as far as I am aware) were conducted with trained listeners. If you did the same tests with people off the st, you might get completely different results.
What people are used to can have a huge influence on what they think sounds good. If they are used to hearing a glaringly inaccurate speaker, then an accurate one may well sound lifeless or dull, or lacking whatever superlative is used to describe their favourite sound.
It doesn't mean that the accurate speaker is bad, it just means it is not to their taste. Let them listen to it long enough, and then ask them to listen to their normal favourite, and they may just change their mind.
I do think that there are other factors such as phase tracking that make a difference (some will agree some wont). experiments with different crossovers which result in very similar on axis FR can sound quite different.
I also have my own theory, that different people are more or less sensitive to certain problems with speakers. As Planet10 has already said there are so many compromises, the compromises one person can live with, another may not be able to.
Tony.
I would completely discount "transient response" as this is widely misunderstood. It exists and matters, but it is completely accounted for in the frequency response. Transient response and frequency response are just two different ways of looking at the same thing. Granted, I do look at impulse responses, but for the most part they don't have any new information. Mostly they will show diffractions if they are pronounced.
What any one person hears as "preference" is pretty much beside the point. Contrary to Tony's comment, Olive has shown that trained and untrained listeners "prefer" the same things on average. The trained listeners show less variation and come to this mean faster - less trials - that's all. We also tend to "like" what we know. If someone only hears a poor speaker then they will come to "prefer" that sound even though it is not "accurate". Listen to an accurate speaker for an extended period of time and "un-accurate" or "colored" speakers will begin to all sound "unnatural". Training by listening to a known accurate speaker is the only way to advance your perception of what is "good".
Edit: I see my post appeared just after Earl's. We've said pretty much the same things 
Have you ever asked the question: what do you mean by transient response? Accurate transient response and flat frequency response are one and the same thing. If a speaker has a sharp peak in the response, the impulse will show ringing. The time domain (transient) and frequency domain (amplitude) and inextricably linked. You cannot separate one from the other.
Studies show that we cannot hear ringing, but we do hear the frequency abberation. Bass notes that linger too long can be fixed by removing the frequency peak. Flatten the response and the ringing is gone. It is important to note that in the bass region, what you measure in the room is what you hear. So, you can put the M2 Reference speaker in your room and it can sound like crap because the room dominates the bass region. Yes, the speaker needs to have deep response but to get ultimate quality, EQ is probably necessary at the listening position. Dr. Geddes' multi-sub approach works well too.
Diffraction is also entirely captured in the frequency response.
- There is no such thing as "fast bass" or "tight bass." There is only "deep" bass and "even bass." To some extent, you want the bass to boom because low-frequency musical instruments have a certain boom. But again, to recreate this, all you need perfectly flat response.
- IIRC, what Toole said was off the street listeners also preferred the same loudspeakers, just that they took longer to settle to a statistically significant answer. This is because they are not trained and so, the variation in their responses from one test to another was large.
- Phase tracking is important, but again it's effect is reflected in the on- and off-axis frequency response near the crossover, isn't it?
- Cannot agree more on your last paragraph! People do have preferences. Most of the time, we don't know what we want.
Edit: Like Earl says in his post, I followed the suggestions about flat response and deep, even bass and now I can easily spot resonances and problems with other speakers.
Have you ever asked the question: what do you mean by transient response? Accurate transient response and flat frequency response are one and the same thing. If a speaker has a sharp peak in the response, the impulse will show ringing. The time domain (transient) and frequency domain (amplitude) and inextricably linked. You cannot separate one from the other.
Studies show that we cannot hear ringing, but we do hear the frequency abberation. Bass notes that linger too long can be fixed by removing the frequency peak. Flatten the response and the ringing is gone. It is important to note that in the bass region, what you measure in the room is what you hear. So, you can put the M2 Reference speaker in your room and it can sound like crap because the room dominates the bass region. Yes, the speaker needs to have deep response but to get ultimate quality, EQ is probably necessary at the listening position. Dr. Geddes' multi-sub approach works well too.
Diffraction is also entirely captured in the frequency response.
- There is no such thing as "fast bass" or "tight bass." There is only "deep" bass and "even bass." To some extent, you want the bass to boom because low-frequency musical instruments have a certain boom. But again, to recreate this, all you need perfectly flat response.
- IIRC, what Toole said was off the street listeners also preferred the same loudspeakers, just that they took longer to settle to a statistically significant answer. This is because they are not trained and so, the variation in their responses from one test to another was large.
- Phase tracking is important, but again it's effect is reflected in the on- and off-axis frequency response near the crossover, isn't it?
- Cannot agree more on your last paragraph! People do have preferences. Most of the time, we don't know what we want.
Edit: Like Earl says in his post, I followed the suggestions about flat response and deep, even bass and now I can easily spot resonances and problems with other speakers.
Last edited:
Dull sounding speakers are anything but flat, mostly rolled off response.
Let me paraphrase what Toole wrote in his "Sound Reproduction".
Why is it that opinions of reviewers are held in so much esteem, why do
people trust them so much? The listening tests they make are not free of bias,
they neither offer proof of performance nor a proof of their hearing
capability. They are simply talented to put to words what they believe they
heard.
Fortunately, in the right circumstances most of the reviewers and ordinary
audio enthusiasts do possess the abbility to form useful opinions about
sound and express them in the meaningful ways. The prerequisite is an
unbiased mind.
This is exactly the experiment I was hoping to try with myself (hence the point of this and another thread that I started on affordable, 'technically accurate' speakers). While I have some speakers right now that I really like, I would like to try listening to a set of speakers that are known to be accurate to see if my tastes change over the long term. This is an idea that has really piqued my curiosity. As I said in my other thread, it's quite possible that some of the speakers I already have are reasonably accurate, but I personally don't have the means nor the expertise to measure/determinate whether that's the case. As a result, I'm curious to try a set of speakers that are known to be reasonably accurate and give them a listen over an extended period (e.g. a year or so, maybe more), to see if my tastes (and my assessment of my current speakers) will change as a result.
Just curious, that's all (but sometimes that's enough).
Edit: p.s. Thanks for clarifying the relation between linear and transient response. I didn't know they were that intimately bound up with one another, so that helps as well.
Hi Philosophil,
Wrt to frequency response, one would think the most accurate would be ruler flat. However, as Olive found out in his, “The Subjective and Objective Evaluation of Room Correction Products” Audio Musings by Sean Olive: The Subjective and Objective Evaluation of Room Correction Products , “flat in-room response is not the preferred target”. So what then is the preferred frequency response target?
In the Olive paper, the preferred target frequency response is a straight line starting at 0 dB @ 20 Hz to -10 dB at 20 kHz.See slide 24:
In 1974 B&K wrote an interesting paper http://www.bksv.com/doc/17-197.pdf called, “Relevant loudspeaker tests in studios in Hi-Fi dealers' demo rooms in the home etc. using 1/3 octave, pink-weighted, random noise” From their research, the most preferred in-room frequency response at the listening position is flat to 200 Hz, -3db at 2 kHz and -6db at 20 kHz.See Figure 5 in the paper:
As an ex 10 year recording and mixing engineer, the above graph in Figure 5 became known widely as the B&K curve.Most pro recording facilities from that point on had at least one set of control room monitors that were eq’d to the B&K curve. Given all things equal, the curve produced a perceptually flat frequency response at the mixer’s listening position.Meaning no-one frequency stood out over the other, and from our ear/brain perception, it sounds like it is a flat frequency response.
In 1998, the European Broadcast Union produced a Tech note (EBU-Tech 3276) called, “Listening conditions for the assessment of sound programme material: monophonic and two–channel stereophonic”. https://tech.ebu.ch/docs/tech/tech3276.pdf Again from a frequency response perspective, the recommendation was to have a flat frequency response out to 2 kHz with flat 1 db per octave rolloff.See Fig 2 on Page 6:
Again, very similar to the B&K curve.Fast forward to today and if one spends anytime on forums such as REW, Audiolense or Acourate, one finds out fairly quickly that indeed the target frequency response that provides the most perceptually flat (accurate?) listening experience at the listening position is still very similar to the original B&K house curve.
In my case, I use Acourate DSP and can shape the frequency response at the listening position any way I want with incredible precision (64 bit resolution). After trying and listening to dozens of frequency response targets, I have ended up with something very similar: flat to 1 kHz, and using 1 kHz as a hinge point, a straight line to -6 dB at 20 kHz. Many on the forums I have listed above use the same or very similar frequency response target and really is not much of a variant compared to the B&K curve some 40 years ago:
In conclusion, I would suggest after one does the research and experimentation, it is fairly well known what the in-room frequency response measured at the listening position is that produces a perceptually flat frequency response to the listeners ears. In other words, the tonal balance or timbre is the most neutral. Does that mean the most accurate?
What is not as well-known is the other side of the same coin, which is the time coherence of a loudspeaker system. Some call it transient response and most folks will point to the impulse response graph, but as one can read in this article: http://www.stereophile.com/content/measuring-loudspeakers-part-two-page-2 (the entire article is worth reading) that the issue is when one is looking at the impulse response, the amplitude is predominantly weighted to high frequencies.From the stereophile article: “An impulse response is extremely hard to interpret, not least because, with a loudspeaker, it is visually dominated by the tweeter's output. Its shape doesn't really tell you much in itself about woofer and midrange-unit behavior. It is also important to note that the impulse response and any information derived from it are only relevant to the loudspeaker's output at the specific microphone position used to capture it.A plot of the step response appears to give more-or-less equal visual weighting to the outputs of all of a speaker's drive-units. You can now glean a lot more information about the lower-frequency drivers, as well as getting a good idea of how time-coherent the speaker is.”
I also noted that when I was looking at the impulse responses separately for my tweeter, mid, and woofer, using the same input level, that the midrange impulse peak was 1/10 of the amplitude compared to the tweeter's peak amplitude and the woofers impulse peak was 1/100 of the tweeters peak amplitude. That can be seen in the Advanced Acourate article on CA. http://www.computeraudiophile.com/content/556-advanced-acourate-digital-xo-time-alignment-driver-linearization-walkthrough/
What is the most accurate step response then? A perfect electrical step response looks like this with the transformation to idealized pressure step response in air:
From the stereophile article:
“The acoustic pressure in a speaker's step response cannot rise instantaneously because the risetime is limited by the tweeter's restricted ultrasonic response. But as long as the speaker has a bandwidth greater than the upper audio limit of 20kHz, the difference in risetime is trivial. More importantly, because a loudspeaker cannot reproduce the acoustic equivalent of a DC voltage—all loudspeakers can be modeled as a high-pass filter of some kind—the sound decays back to the time axis in an exponential manner. It always crosses that axis but eventually returns to it so that the areas on both sides of the time axis are equal, ensuring that there is no DC component present. An ideal step response should therefore resemble a slightly concave right triangle.”
A good speaker’s step response:
Not so good step responses:
As noted in the stereophile article: “Many loudspeakers are claimed by their manufacturers' marketing departments to be time-coherent. There are also a number of speakers that have sloped front baffles, implying that they are time-coherent. However, its step response immediately gives you an indication of whether or not a loudspeaker is time-coherent (on the chosen measurement axis). And almost all loudspeakers are not. Along with false claims of high sensitivity, false claims of time coherence are among the commonest lies in high-end audio.”
With the computational power of modern computers and sophisticated DSP software, one can use digital XO’s which opens the door to time align each driver in a speaker system to approach the idealized step response.With a bit of patience and know-how, this can be applied to virtually any speaker system.The Computer Audiophile article I linked above shows one way to do that.
Other factors not mentioned are the speaker’s distortion characteristics, which I view as a non-issue, and a speaker’s polar pattern or directivity index. I believe the latter to be a personal preference. For me, constant directivity speakers I enjoy the most. Room acoustics, early reflections, decay times, etc., is another long topic, but suffice to say, there are also well known specifications around that as well, some can be found in the articles mentioned above.
Hope that helps.
Cheers, Mitch
Wrt to frequency response, one would think the most accurate would be ruler flat. However, as Olive found out in his, “The Subjective and Objective Evaluation of Room Correction Products” Audio Musings by Sean Olive: The Subjective and Objective Evaluation of Room Correction Products , “flat in-room response is not the preferred target”. So what then is the preferred frequency response target?
In the Olive paper, the preferred target frequency response is a straight line starting at 0 dB @ 20 Hz to -10 dB at 20 kHz.See slide 24:
In 1974 B&K wrote an interesting paper http://www.bksv.com/doc/17-197.pdf called, “Relevant loudspeaker tests in studios in Hi-Fi dealers' demo rooms in the home etc. using 1/3 octave, pink-weighted, random noise” From their research, the most preferred in-room frequency response at the listening position is flat to 200 Hz, -3db at 2 kHz and -6db at 20 kHz.See Figure 5 in the paper:
As an ex 10 year recording and mixing engineer, the above graph in Figure 5 became known widely as the B&K curve.Most pro recording facilities from that point on had at least one set of control room monitors that were eq’d to the B&K curve. Given all things equal, the curve produced a perceptually flat frequency response at the mixer’s listening position.Meaning no-one frequency stood out over the other, and from our ear/brain perception, it sounds like it is a flat frequency response.
In 1998, the European Broadcast Union produced a Tech note (EBU-Tech 3276) called, “Listening conditions for the assessment of sound programme material: monophonic and two–channel stereophonic”. https://tech.ebu.ch/docs/tech/tech3276.pdf Again from a frequency response perspective, the recommendation was to have a flat frequency response out to 2 kHz with flat 1 db per octave rolloff.See Fig 2 on Page 6:
Again, very similar to the B&K curve.Fast forward to today and if one spends anytime on forums such as REW, Audiolense or Acourate, one finds out fairly quickly that indeed the target frequency response that provides the most perceptually flat (accurate?) listening experience at the listening position is still very similar to the original B&K house curve.
In my case, I use Acourate DSP and can shape the frequency response at the listening position any way I want with incredible precision (64 bit resolution). After trying and listening to dozens of frequency response targets, I have ended up with something very similar: flat to 1 kHz, and using 1 kHz as a hinge point, a straight line to -6 dB at 20 kHz. Many on the forums I have listed above use the same or very similar frequency response target and really is not much of a variant compared to the B&K curve some 40 years ago:
In conclusion, I would suggest after one does the research and experimentation, it is fairly well known what the in-room frequency response measured at the listening position is that produces a perceptually flat frequency response to the listeners ears. In other words, the tonal balance or timbre is the most neutral. Does that mean the most accurate?
What is not as well-known is the other side of the same coin, which is the time coherence of a loudspeaker system. Some call it transient response and most folks will point to the impulse response graph, but as one can read in this article: http://www.stereophile.com/content/measuring-loudspeakers-part-two-page-2 (the entire article is worth reading) that the issue is when one is looking at the impulse response, the amplitude is predominantly weighted to high frequencies.From the stereophile article: “An impulse response is extremely hard to interpret, not least because, with a loudspeaker, it is visually dominated by the tweeter's output. Its shape doesn't really tell you much in itself about woofer and midrange-unit behavior. It is also important to note that the impulse response and any information derived from it are only relevant to the loudspeaker's output at the specific microphone position used to capture it.A plot of the step response appears to give more-or-less equal visual weighting to the outputs of all of a speaker's drive-units. You can now glean a lot more information about the lower-frequency drivers, as well as getting a good idea of how time-coherent the speaker is.”
I also noted that when I was looking at the impulse responses separately for my tweeter, mid, and woofer, using the same input level, that the midrange impulse peak was 1/10 of the amplitude compared to the tweeter's peak amplitude and the woofers impulse peak was 1/100 of the tweeters peak amplitude. That can be seen in the Advanced Acourate article on CA. http://www.computeraudiophile.com/content/556-advanced-acourate-digital-xo-time-alignment-driver-linearization-walkthrough/
What is the most accurate step response then? A perfect electrical step response looks like this with the transformation to idealized pressure step response in air:
From the stereophile article:
“The acoustic pressure in a speaker's step response cannot rise instantaneously because the risetime is limited by the tweeter's restricted ultrasonic response. But as long as the speaker has a bandwidth greater than the upper audio limit of 20kHz, the difference in risetime is trivial. More importantly, because a loudspeaker cannot reproduce the acoustic equivalent of a DC voltage—all loudspeakers can be modeled as a high-pass filter of some kind—the sound decays back to the time axis in an exponential manner. It always crosses that axis but eventually returns to it so that the areas on both sides of the time axis are equal, ensuring that there is no DC component present. An ideal step response should therefore resemble a slightly concave right triangle.”
A good speaker’s step response:
Not so good step responses:
As noted in the stereophile article: “Many loudspeakers are claimed by their manufacturers' marketing departments to be time-coherent. There are also a number of speakers that have sloped front baffles, implying that they are time-coherent. However, its step response immediately gives you an indication of whether or not a loudspeaker is time-coherent (on the chosen measurement axis). And almost all loudspeakers are not. Along with false claims of high sensitivity, false claims of time coherence are among the commonest lies in high-end audio.”
With the computational power of modern computers and sophisticated DSP software, one can use digital XO’s which opens the door to time align each driver in a speaker system to approach the idealized step response.With a bit of patience and know-how, this can be applied to virtually any speaker system.The Computer Audiophile article I linked above shows one way to do that.
Other factors not mentioned are the speaker’s distortion characteristics, which I view as a non-issue, and a speaker’s polar pattern or directivity index. I believe the latter to be a personal preference. For me, constant directivity speakers I enjoy the most. Room acoustics, early reflections, decay times, etc., is another long topic, but suffice to say, there are also well known specifications around that as well, some can be found in the articles mentioned above.
Hope that helps.
Cheers, Mitch
Audio Musings by Sean Olive: Part 3 - Relationship between Loudspeaker Measurements and Listener Preferences
Audio Musings by Sean Olive: Part 1- Do Untrained Listeners Prefer the Same Loudspeakers as Trained Listeners?
Well measuring speakers:
http://www.stereophile.com/content/dynaudio-evidence-master-loudspeaker-measurements
Audio Musings by Sean Olive: Part 1- Do Untrained Listeners Prefer the Same Loudspeakers as Trained Listeners?
Well measuring speakers:
http://www.stereophile.com/content/dynaudio-evidence-master-loudspeaker-measurements
Last edited:
Mitch gave a very good explanation of what is required to give accurate and undistorted audio reproduction. From elementary filter theory and transmission line theory it is known that there are just two criteria that ensures a signal passing through a system is undistorted, and that is;
1) Signal must be passed equally well at all frequencies, in other words a flat frequency response overall.
2) Signal phase should not be altered but, if we do not have a choice, then a linear phase response would also result in a perfectly flat group delay in the time domain.
In most audio systems the crossover is the weakest link when it comes to non-linear distortion but offset drivers can also contribute to group delay aberrations. Adding even a single capacitor in series with the tweeter introduces phase shift.
Say we use 3 drivers, it would be virtually impossible to design a 3 way passive crossover which has the same linear phase shift performance in each band. Using continuous time (active) filters or digital filters it would be possible, but required 3 power amplifiers per channel.
To recap, if we stand a small chance of designing a low distortion audio system our system must have a flat frequency response and all the crossovers in the system has to have the same linear phase tracking response. That will ensure no frequency will be delayed in time, relative to any other frequency, i.e. a perfectly flat group delay in the time domain.
This is a very theoretical outlook on the issues that plague audio designers so a linear response (flat frequency response) is only one aspect of a good audio design, but the challenges of Ultra-Fi goes well beyond that single criteria and a flat response in the time domain is now considered just as important in good audio design.
Cheers, Peter.
1) Signal must be passed equally well at all frequencies, in other words a flat frequency response overall.
2) Signal phase should not be altered but, if we do not have a choice, then a linear phase response would also result in a perfectly flat group delay in the time domain.
In most audio systems the crossover is the weakest link when it comes to non-linear distortion but offset drivers can also contribute to group delay aberrations. Adding even a single capacitor in series with the tweeter introduces phase shift.
Say we use 3 drivers, it would be virtually impossible to design a 3 way passive crossover which has the same linear phase shift performance in each band. Using continuous time (active) filters or digital filters it would be possible, but required 3 power amplifiers per channel.
To recap, if we stand a small chance of designing a low distortion audio system our system must have a flat frequency response and all the crossovers in the system has to have the same linear phase tracking response. That will ensure no frequency will be delayed in time, relative to any other frequency, i.e. a perfectly flat group delay in the time domain.
This is a very theoretical outlook on the issues that plague audio designers so a linear response (flat frequency response) is only one aspect of a good audio design, but the challenges of Ultra-Fi goes well beyond that single criteria and a flat response in the time domain is now considered just as important in good audio design.
Cheers, Peter.
The pictured presentations of a single good step response vs. not so good
of the B&W 801(and others) is nonsense. If that were true then the best
loudspeakers would have been full range drive unit based.
In this review :
B&W Nautilus 801 loudspeaker Measurements part 3 | Stereophile.com
the comments about step response isn't doing justice to good speaker
performance.
Bill Waslo has made his Xsim software available for free, so anyone can easily
test what the step response would look like with or without a filter.
The time allignment of the drivers has never been a must have for good
performance. For the sake of fun, someone try to tilt the whole speaker
backwards and see if then time alligned will sound better.
of the B&W 801(and others) is nonsense. If that were true then the best
loudspeakers would have been full range drive unit based.
In this review :
B&W Nautilus 801 loudspeaker Measurements part 3 | Stereophile.com
the comments about step response isn't doing justice to good speaker
performance.
Bill Waslo has made his Xsim software available for free, so anyone can easily
test what the step response would look like with or without a filter.
The time allignment of the drivers has never been a must have for good
performance. For the sake of fun, someone try to tilt the whole speaker
backwards and see if then time alligned will sound better.
I quite possibly have a wrong concept of transient response. I generally think of it as being how quickly the speaker comes back to stillness after an excitation signal. The step response being a good indicator of whether a speaker has good or bad transient response. My question is, can you tell good transient repsonse from bad just by looking at a frequency response curve? I wouldn't have thought so.
When playing around with different alignments for vented enclosures some have very good step response and others can be terrible. Two different speakers may show a theoretical similar response FR wise but quite different step response. Maybe the difference would be obvious with real world measurements?
I don't get this. Is the ringing just a resonance? I thought it was a result of being under-damped??? or am I off on a different tangent?
Just a subjective term which I have always thought could probably be linked to how good the transient response was. No technical basis for this just my own thoughts on what it might mean when people say it
Yes Earl said the same. I was mistaken
Well that's the interesting thing! It doesn't necessarily show in the on axis response, but will almost certainly show in the off axis
Being relatively new to crossover design, I made quite a few mistakes along the way to getting what I was finally happy with. I found it was possible to make flat on axis measurements that actually had quite bad phase tracking (reverse nulls showed the nastiness) I never did off axis measurements of them but pretty sure that they would have shown problems, as would the power response. I have too. Earl pulled me up on it maybe four to five years ago when I was saying I thought it was better to Taylor the speaker to what I liked the sound of. I ended up trying what he said (ie don't do that) and I did once I got used to it, end up preferring it over what I originally had.
Tony.
Last edited:
Mitch's post #31 was quite interesting. My take on it was the implied ON-AXIS decline in FR was the favored FR by listeners.
It would seem to me that a loudspeaker with flat FR from LF roll off to 20 kHz would naturally have OFF-AXIS HF roll off as we have all seen with polar measurements. So, listeners can easily obtain that desirable sloping down FR by simply listening off-axis or, am I missing something?
It would seem to me that a loudspeaker with flat FR from LF roll off to 20 kHz would naturally have OFF-AXIS HF roll off as we have all seen with polar measurements. So, listeners can easily obtain that desirable sloping down FR by simply listening off-axis or, am I missing something?
There is little to no evidence that phase linearity matters. Toole and Olive never even look at it, Griesinger makes a case for its importance in the band from about 800 Hz - 4000Hz, but outside of that there is no data to say that phase linearity matters at all.
Most people believe as you do, but the fact is that good frequency response does mean that you will have good transient response. There are a few theoretical situations where this is not true, i.e. non-minimum phase systems, but most speakers are nearly minimum phase and so frequency response does tell the whole story.
Most people believe as you do, but the fact is that good frequency response does mean that you will have good transient response. There are a few theoretical situations where this is not true, i.e. non-minimum phase systems, but most speakers are nearly minimum phase and so frequency response does tell the whole story.
You are missing the fact that most speakers are not very well behaved off-axis - very few are. So just going off axis does more than just a small falling of the HF response - the whole frequency response gets all messed up. You can see this very clearly in my PolarMap database of real speakers. It is the rare exception where the response off axis is smooth and near flat.
Mitch's post fails to mention an important difference in the measurements he is talking about and the ones I (and probably Earl and Tony) are talking about. The measurement Mitch is talking about is in the primary listening position. This is not an anechoic or quasi-anechoic measurement. The same goes for the B&K curve he mentions.
The measured response at the listening seat is a combination of the speaker response (on- and off- axis) and the room size, room absorption and listening distance. So, you can have a conventional cone and dome two-way with a flat on-axis response (measured anechoically or quasi-anechoically), with a uniform off-axis response that is mildly narrowing in the high frequencies in a room with moderate furnishings and get a falling response like the B&K curve in Mitch's post. Or, you can have a harsh sounding speaker with a tilted up on axis response with a very narrow off-axis response in a small dead room (acoustically dead) and get the same measurement at the listening spot. Will these two speakers sound the same? Highly doubt it.
Using EQ, you can get a response like the B&K curve at the listening position with any loudspeaker in any room. But that is not the way to go. What you would be measuring at the listening position will be the power response of the speaker combined with the absorption of the room. Now, it is a well-known fact that rising power response is not liked by anyone. It sounds bright and harsh.
So, what do you do? In my experience, the ideal way to get to the B&K room curve is by starting with a flat on-axis response, then a uniform off-axis response, i.e., get your speaker right first, and then depending upon your room, add furnishings or absorbers, diffusers and so on, to get the B&K curve at the listening position. If you start with a flat on-axis response and uniform off-axis response, but have a very lively room, and then start tilting down the listening position measured response of the speaker to get to the target curve, it will sound dull.
The measured response at the listening seat is a combination of the speaker response (on- and off- axis) and the room size, room absorption and listening distance. So, you can have a conventional cone and dome two-way with a flat on-axis response (measured anechoically or quasi-anechoically), with a uniform off-axis response that is mildly narrowing in the high frequencies in a room with moderate furnishings and get a falling response like the B&K curve in Mitch's post. Or, you can have a harsh sounding speaker with a tilted up on axis response with a very narrow off-axis response in a small dead room (acoustically dead) and get the same measurement at the listening spot. Will these two speakers sound the same? Highly doubt it.
Using EQ, you can get a response like the B&K curve at the listening position with any loudspeaker in any room. But that is not the way to go. What you would be measuring at the listening position will be the power response of the speaker combined with the absorption of the room. Now, it is a well-known fact that rising power response is not liked by anyone. It sounds bright and harsh.
So, what do you do? In my experience, the ideal way to get to the B&K room curve is by starting with a flat on-axis response, then a uniform off-axis response, i.e., get your speaker right first, and then depending upon your room, add furnishings or absorbers, diffusers and so on, to get the B&K curve at the listening position. If you start with a flat on-axis response and uniform off-axis response, but have a very lively room, and then start tilting down the listening position measured response of the speaker to get to the target curve, it will sound dull.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.