A Broad Discussion of Speakers with Major Audio Luminaries

[NTK]

I guess trained listeners means that they got trained until they could, with great consistency, discern tonality differences. In general, that should mean that most of these trained listeners got better and better with more time training. So, how many hours did they get trained compared to, say, a mastering engineer who has worked 10 hours a day, 5 days a week, for the last 10 or 20 years?

I can't see why a guy who got a few months or maybe a year(?) of listening training at Harman would be better at hearing tonality problems than a guy who has already been doing it professionally for 10 or 20 years. More training should, in general, mean you get better at it, don't you think?

I would definitely want to see a German "gedankenexperiment" on this.

From the Introduction section of this book:

Audio Production and Critical Listening | Technical Ear Training | Jas

Audio Production and Critical Listening: Technical Ear Training, Second Edition develops your critical and expert listening skills, enabling you to listen to

www.taylorfrancis.com

 

[rdenney]

I don't believe that is a defensible position any longer.

That is. However, the argument that bothers me is the one that says that moving from a good stereo system to a multichannel one of the same cost is inevitably a compromise. It presumes the incremental expenditures on a stereo system are as rewarding as applying those expenditures to multichannel expansion

Kal, not really. Example:

I found a used set of Revel F12's for $800. That is coupled with a Buckeye stereo NC502MP, which is now about $700.

To go with 5.2, and sticking with the same brand, vintage, and quality level, I'd need three M12's and a couple of, say, B10 subs. The M12's seem to get about $300 each on the used market, and the B10's about $500. So, that adds up to $2700--not a lot to spend on very good stuff, of course, but 3.5 times what I paid for speakers.

And the Buckeye amp that would drive six channels would be a lot more than the stereo model I bought--about a thousand bucks more.

And then there's the preamp. My vintage stereo preamps cost a few hundred. Finding a vintage multichannel preamp with preamp outputs isn't that easy, to be honest. But I see an Adcom GTP750 5.2 preamp that went for about $700. That's $400 more than I paid for any of my preamps. I bought several because that's where my hobby went, but frankly the first one I bought was fine (an SAE) and I only paid a hundred bucks for it. But I'll compare the Adcom 750 to the Adcom 565, for which I paid $300.

So, what I paid $1800 for would have cost $5100 for a 5.2 system--same brands, same vintages, same used stuff, same quality. Yes, I would expect a more immersive experience, especially for multiple listeners. But my current system sounds superb, especially considering what I paid for it, and 99.8% of my library is stereo anyway.

What might I find for for $1800? Probably pretty close to what I have for watching TV: a Yamaha AVR that I paid $600 for used, but it lacks preamp outputs, so I use the built-in amps. I have a pair of Linn Index Plus speaker for L and R, but those are so old as to be worthless. So, let me suggest: Infinity Reference speakers which might fetch $1200 for a preowned 5.2 set including two towers.

I'm sorry, but even though I think that would make a competent system, I think that's a cost-driven compromise when listening to stereo stuff, which, as I say, is 99.8% of my collection, compared to what I have.

Rick "everything is a compromise" Denney

 

[Heinrich]

From the Introduction section of this book:

Audio Production and Critical Listening | Technical Ear Training | Jas

Audio Production and Critical Listening: Technical Ear Training, Second Edition develops your critical and expert listening skills, enabling you to listen to

www.taylorfrancis.com

On training methodology: compare to a reference, is there a difference? If there's a difference, what quality has it, possible parameters are predefined 'loudness', 'sharpness (Q factor)', 'frequency', ... ? Again, at least a reference is given, naturally as they might say.

Regarding the practicability: now, listening to the recording some 40 years later, I have no reference. And so I'm lost entirely.

Something totally different: I would love to see a definition of 'equivalence' between speakers in regard to their purpose.

It is understood that people can be trained to most everything that is imaginable. And you could, may I put it so, use people to detect tiny remaining flaws in speakers. But what is the practical use, if even the producers won't make it in that challenge?

As an anecdote on bad recordings, I sometimes think that a mix is peculiar, but then I say, it's the artistic freedom. I have to take it as it is. And finally I come to realistically appreciate deviations from an imaginary ideal as utterly intended. Otherwise it would be too boring.

 

[NTK]

On training methodology: compare to a reference, is there a difference? If there's a difference, what quality has it, possible parameters are predefined 'loudness', 'sharpness (Q factor)', 'frequency', ... ? Again, at least a reference is given, naturally as they might say.

Regarding the practicability: now, listening to the recording some 40 years later, I have no reference. And so I'm lost entirely.

Something totally different: I would love to see a definition of 'equivalence' between speakers in regard to their purpose.

It is understood that people can be trained to most everything that is imaginable. And you could, may I put it so, use people to detect tiny remaining flaws in speakers. But what is the practical use, if even the producers won't make it in that challenge?

As an anecdote on bad recordings, I sometimes think that a mix is peculiar, but then I say, it's the artistic freedom. I have to take it as it is. And finally I come to realistically appreciate deviations from an imaginary ideal as utterly intended. Otherwise it would be too boring.

Perhaps you should be asking the people running the music schools this question instead. They are the one who decided to make learning critical listening skills a requirement of a sound engineer's university education.

 

[Floyd Toole]

Something totally different: I would love to see a definition of 'equivalence' between speakers in regard to their purpose

All loudspeakers, professional monitors or consumer products, should aim to be as neutral as possible. If they play loud enough and long enough for the intended purpose, they qualify. If everyone followed this principle, the circle of confusion would be less of an issue. In our loudspeaker evaluations there have been pro monitors and high-end audiophile loudspeakers that end up in statistical ties in terms of sound quality, not sound quantity. A good loudspeaker is a good loudspeaker. All electronics are fundamentally neutral and always have been. The best studio microphones are almost as accurate as measurement microphones (there are other types that are used as equalizers of a special kind). The entire sound record/reproduction path should have neutrality as its baseline. Then recording engineers and musicians can create whatever art they like, realistic or abstract, and consumers might have a chance of hearing it as it was created. That is my kind of "equivalence". Those in the recording business call it "translation", but as things are the sound translations are rarely "literal". Fortunately music itself translates very well, and. if there is enough bass to get the juices flowing most of us can find pleasure, if not always complete gratification. And I haven't mentioned soundstage and imaging . . .

 

[amirm]

Everybody has an access to for example AES eLib to search the studies I have mentioned about timing. Lib is behind payment barrier also for AES members, but I can show a short summary what I have found and downloaded from there:

Hardly anyone here has access to AES library for the reason you mention. You have to have a membership or pay per paper. So you need to properly reference the paper and explain what it is about and the quotations to back your postion.

I'm know that this is not all, but until more information is shown, audibility was found when GD >= 1.6 ms. Just 3/7 was done with speakers, and probably just one with signal suitable for evaluating possible drop in dynamics due to delay at L.F.

Again, you need to properly quote the paper and explain what it was about. First, they used an impulse function for the primary test. That showed that bass group delay threshold is quite high:

"It is seen that the group delay can exceed 10 ms below
200 Hz when compared to the group delay at
high frequencies without the difference being audible.
However, an interesting frequency range arises when
Fig. 11(a) is compared to the two other subfigures,
since the major differences between the three groups of
signals are seen between 300–1000 Hz. In Fig. 11(a),
the values are below 1.0 ms in this range and the differences
are inaudible. In Fig. 11(b), the group delay
values are between 1–2 ms, which resulted in the difference
being sometimes audible. Finally, in Fig. 11(c),
the values are above 2.0 ms and the differences were
mostly audible."

To raise audibility of group delay in bass region, they had to modify the already corner case signal of an impulse by convolving it with pink noise. Only then the threshold of hearing for bass frequencies lowered. As a result, their key finding was that of the threshold for 300 to 1000 Hz.

But before you go celebrate, you need to pay attention to the last paragraph in the paper:

"The results of this study should not be interpreted
as saying that the group-delay responses of the loudspeakers
for which differences were heard are defective
and would lead to poor sound quality. The results
merely suggest that the group-delay deviations are large
enough to create audible differences in an exaggerated
test such as the pairwise comparison of a test signal auditioned
in the forward and backward directions. However,
the results could be interpreted to say that the
loudspeaker impulse responses for which differences
were not heard in this exaggerated arrangement are sufficiently
short and that the group delay probably is not a
factor in creating audible effects in sound reproduction."

So sure, get everything else right and you can work on group delay then. But don't commit the usual sins we see in loudspeaker design in the name of "correct timing." And know that even if you got there doing everything perfect, there is no assurance that someone likes your speaker better while playing music.

Other studies have reached similar conclusions that while in specialized tests we could tease out differences, it simply is not a concern.

 

[rdenney]

They don’t read, or they read and misapply the research beyond what it was intended. It’s routine. In “their” defense, me included, there are interrelated but distinct concepts that are not always apparent that lead to this. One example: sighted listening bias; adaptation effect and need to use 3 or more pairs of speakers for DB testing.

So @Floyd Toole Dr. Toole, what should a consumer do to make the best decision on a purchase of floor standing speakers for their particular listening area?

Review Spinorama data
Make a short list
Get 2,3 or 4 pairs in to A/B blind?

I can answer that in my case. I reviewed anechoic spinorama data prepared in the Canadian NRC anechoic chamber--the only such tests published about the Revel F12. I reviewed further measurements that John Atkinson conducted. I read Kal's comparison of the Revels with his other speakers, but in Kal's case he's had enough of a grounding in measurement that I was looking for validation only, given the only approximate correlation between measured performance and preference. One of the things I looked for was how good was the on-axis response and directivity when the speakers were pushed hard, and these Revels were one of only two or three pairs of speakers in the Soundstage database (that were tested at the NRC chamber) that were pushed to higher levels because of good performance at lower levels. They performed well with an average output of 100 dB SPL at two meters in the anechoic chamber--equivalent to 106 dB at one meter and that for one speaker only. After all that, I was prepared to find a preowned pair to buy and try. Note that I did not read the subjective review in the Soundstage website as the basis for a decision, and frankly Kal is the only reviewer at Stereophile whose opinion I would use to validate spending actual money. And buying them preowned I recognized that if they didn't work out, I could turn them over, so the risk was not huge.

My short list was short indeed--that was the only speaker that 1.) I could afford on the secondary market, 2.) that received spousal approval, 3.) that could make a big sound from a big amplifier to meet the needs of my use case of playing the tuba along with a recording at realistic levels, and 4.) that were generally of a type approved by those who have had more access to speaker measurements and listening than I have. While I did spend a bit of time being offended by the crappy sounding speakers at a local Best Buy--due to terrible placement and nothing resembling room integration--I did not perform direct comparisons. Thus, I acknowledge there may be better speakers out there, and maybe even for the $800 I paid for the pair.

I did not have access to the resources that Dr. Toole has had, of course, and so had to scrounge for what was available, so take this as a data point.

Rick "trying to catch up with 30 pages of posts over one freaking weekend" Denney

 

[Heinrich]

All loudspeakers, professional monitors or consumer products, should aim to be as neutral as possible.

Neutral (in terms of frequency response etc) is the acclaimed standard, which became your achievement, once you were supported by Harman as a big player. Thanks, no criticism in that regard. I only wonder what a 'good enough' would be. Is it +/- 1dB or even +/-2dB, so that people could rest assured that the art is conveyed true to the intentions of the record's design? I hope that composers, musicians and engineers work together in order to create a something that's worth all of the effort.

If I look at the recent releases it comes to me, that the actual quality of speakers (nearly?) fulfills the promise, eventually. Do we really need trained listeners still, once we have the NFS and well defined technical parameters?

 

[Timcognito]

First I would like to thank Dr Toole for his time on this forum and Amir for his usual calm approach those naysayers and those of us learning. I will reward Dr Toole when his 4th addition comes out and Amir with another annual donation.

I have found this list on another site and wonder how comprehensive it is, if some of it represents the work of Dr Toole and if there are differences from the measurement techniques used on ASR, in other words apple to apples for comparison when shopping or helping those that are.

SoundStageNetwork.com | SoundStage.com - Loudspeakers

www.soundstagenetwork.com

 

[Floyd Toole]

First I would like to thank Dr Toole for his time on this forum and Amir for his usual calm approach those naysayers and those of us learning. I will reward Dr Toole when his 4th addition comes out and Amir with another annual donation.

I have found this list on another site and wonder how comprehensive it is, if some of it represents the work of Dr Toole and if there are differences from the measurement techniques used on ASR, in other words apple to apples for comparison when shopping or helping those that are.

SoundStageNetwork.com | SoundStage.com - Loudspeakers

www.soundstagenetwork.com

Thank you Timcognito. The Soundstage measurements use some of the same measuring equipment that was at the NRCC when I left in 1991! The data are presented in the pre-spinorama format I used at that time, which is still accurate and useful, but the magazine does not buy the full set of curves. It is a trustworthy database to help zero in on basically neutral loudspeakers.

 

[MattHooper]

Thank you Timcognito. The Soundstage measurements use some of the same measuring equipment that was at the NRCC when I left in 1991! The data are presented in the pre-spinorama format I used at that time, which is still accurate and useful, but the magazine does not buy the full set of curves. It is a trustworthy database to help zero in on basically neutral loudspeakers.

I’m not sure if you’ve commented on this before, but I’m wondering:

Do you have an opinion on active speakers?

There are some of these days who wouldn’t even consider a passive speaker, believing active speakers to have that much advantage.

Do you see the future as belonging to active speakers? Or do you find there are enough passive speakers that are neutral enough that the gains from active speakers aren’t necessarily game changing?

 

[amirm]

I only wonder what a 'good enough' would be. Is it +/- 1dB or even +/-2dB, so that people could rest assured that the art is conveyed true to the intentions of the record's design?

Answer is in Dr. Toole's book and papers. Audibility doesn't just depend on level but also Q (how broad it is). The wider the deviation, the more chances that musical notes hit it and show the deficiency. At the extreme, eve 0.5 dB deviation with low Q can be audible. This is why I build EQs and experiment to detect level of audibility of frequency response errors.

 

[Floyd Toole]

I’m not sure if you’ve commented on this before, but I’m wondering:

Do you have an opinion on active speakers?

There are some of these days who wouldn’t even consider a passive speaker, believing active speakers to have that much advantage.

Do you see the future as belonging to active speakers? Or do you find there are enough passive speakers that are neutral enough that the gains from active speakers aren’t necessarily game changing?

It is extremely challenging to design passive loudspeakers that do not have "any" audible resonances. It can be done, but efficiency/sensitivity frequently suffers. It is much easier if there are dedicated electronics with an equalizer - built-in or external. A purely active loudspeaker can be tweaked at the factory using anechoic data on that product to guide the process. This comes at a price, but the results can be exemplary.

Thanks to Amir, and other sources, anechoic spinorama data are available on many passive loudspeakers. Looking at the published data there are some that are remarkably neutral out of the box. But there are many more that are "almost" neutral, needing just a little equalization to achieve a higher level of performance. Because transducers are minimum-phase devices, as are strong low-frequency room resonances, parametric equalization is a powerful tool to attenuate resonances and finesse the performance of a system. For bass frequencies steady-state room curves are needed, just don't trust them above about 400-500 Hz. For the technically inclined this has to be a very attractive option, making accessible many more loudspeaker designs, and addressing the small room resonances for the prime listening location. It may be well to select a reputable manufacturer who might care about production variations - they happen.

As for the future, I see quite a few passive loudspeakers that are probably good enough to satisfy fairly critical listeners, and a few that are as neutral as they need to be (I have passive loudspeakers). Small rooms cannot be completely eliminated from the experience but multiple subs and EQ go a long way. For those with the money, the best active systems are seriously good, but the darn room is still there - can't fix that at the factory.

 

[Kal Rubinson]

Kal, not really. Example:

I found a used set of Revel F12's for $800. That is coupled with a Buckeye stereo NC502MP, which is now about $700.

To go with 5.2, and sticking with the same brand, vintage, and quality level, I'd need three M12's and a couple of, say, B10 subs. The M12's seem to get about $300 each on the used market, and the B10's about $500. So, that adds up to $2700--not a lot to spend on very good stuff, of course, but 3.5 times what I paid for speakers.

And the Buckeye amp that would drive six channels would be a lot more than the stereo model I bought--about a thousand bucks more.

And then there's the preamp. My vintage stereo preamps cost a few hundred. Finding a vintage multichannel preamp with preamp outputs isn't that easy, to be honest. But I see an Adcom GTP750 5.2 preamp that went for about $700. That's $400 more than I paid for any of my preamps. I bought several because that's where my hobby went, but frankly the first one I bought was fine (an SAE) and I only paid a hundred bucks for it. But I'll compare the Adcom 750 to the Adcom 565, for which I paid $300.

So, what I paid $1800 for would have cost $5100 for a 5.2 system--same brands, same vintages, same used stuff, same quality. Yes, I would expect a more immersive experience, especially for multiple listeners. But my current system sounds superb, especially considering what I paid for it, and 99.8% of my library is stereo anyway.

What might I find for for $1800? Probably pretty close to what I have for watching TV: a Yamaha AVR that I paid $600 for used, but it lacks preamp outputs, so I use the built-in amps. I have a pair of Linn Index Plus speaker for L and R, but those are so old as to be worthless. So, let me suggest: Infinity Reference speakers which might fetch $1200 for a preowned 5.2 set including two towers.

I'm sorry, but even though I think that would make a competent system, I think that's a cost-driven compromise when listening to stereo stuff, which, as I say, is 99.8% of my collection, compared to what I have.

Rick "everything is a compromise" Denney

I must apologize for saying that without adding any qualifications. This a principle that I would apply in proportion to the initial expenditure and, thus, with increasing confidence, to much more expensive stereo systems. At the lower end, a single decent speaker may be optimal.

 

[mglobe]

I’ve been thinking about this for quite a while. In my opinion, Dr Toole and his team actually proved that listeners are capable of discerning the quality of speakers by listening without the benefit of measurement. However there is the caveat that it is VERY difficult to do so if they know what speakers they are listening to due to their preconceived notions about how brand, size, appearance,… impact their expectations.

It’s not that they can’t hear the differences. It’s that they can’t judge them due to those expectations. I would think/hope that to some extent, with enough expertience/training, it MIGHT be possible to overcome that, but I also have no doubt that that is VERY difficult. I’m confident that I can’t do it, and I’m also confident that some folks will disagree with my premise.

 

[Curvature]

This thread proves, over and over, that reading is one of the best

I’ve been thinking about this for quite a while. In my opinion, Dr Toole and his team actually proved that listeners are capable of discerning the quality of speakers by listening without the benefit of measurement. However there is the caveat that it is VERY difficult to do so if they know what speakers they are listening to due to their preconceived notions about how brand, size, appearance,… impact their expectations.

It’s not that they can’t hear the differences. It’s that they can’t judge them due to those expectations. I would think/hope that to some extent, with enough experience/training, it MIGHT be possible to overcome that, but I also have no doubt that that is VERY difficult. I’m confident that I can’t do it, and I’m also confident that some folks will disagree with my premise.

Entirely the right distinction.

 

[Arindal]

Many more measurements since then, and Sean Olive's correlation study confirmed the importance of the direct sound in sound quality assessments.

Fully agree for situations in which the direct sound is the main factor of perceived tonality, for example when focussing on vocals without with little or no meaningful reverb. But listeners preferring neutral direct sound does not automatically mean they accept colorated reverb or do not perceive it as colorated, in my understanding.

Particularly when taking modern home listening environment into account with larger distance speaker<>listener, and more reflective surfaces, we could expect to have mostly setups beyond the critical distance at medium frequencies. Which in my understanding means the indirect sound generated in the listening room, is dominating over the direct sound from the speakers (no reverb from the recording taken into account).

The off-axis sounds are subject to longer propagation path lengths (-6 dB/double distance) and to correspondingly increased air attenuation at frequencies above about 1 kHz.

I agree to the fact that absorption for frequencies above 8K, or lack thereof for frequencies under 400Hz, are in most of cases causing some different plateaus on the chart and a somehow decreasing tendency. But this does not necessarily affect the frequency range in betwee,n which is the important one for reverb and its tonality. Attenuation due to air, which recording engineers know as dissipation, is for example expected to be in the region of -0.1dB per 1m of additional propagation distance (rule of thumb at 10K, you probably know the precise formula by heart). For frequencies from 0.5K to 8K this is even less, and in rooms of the usual size of a listening room expected to be negligible.

Put the two together, and even an omni, dipole or dipole loudspeaker will show attenuation of the steady-state sound field as a function of frequency.

Cannot confirm this from own experience with constant directivity speakers for frequency bands below 8K. Even the simplified in-room FR estimations on spinorama page lead to a pretty balanced result with such a speaker within the frequency range responsibly for reverb. Example:

That is a fullrange constant directivity speaker with cardioid bass/midrange. Between 600 and 8,000Hz, the predicted in-room-response stays almost linear within a +-1dB corridor without showing a major tendency of tilt. Below 600Hz it gets a bit higher due to lower d.i. of the cardioid. I found this to be achievable in practice, and personally perceive the difference between such a linear off-axis soundfield and a ´dull´, decreasing reverb level towards higher frequencies in the region presented in your chart (around -7dB in the same range) to be clearly audible and decisive for tonality judgment.

I fully agree with Siegfried Linkwitz on the importance of this matter, although I would limit the frequency range to strive for a perfectly even directivity/indirect soundfield to something like this window (maybe 400 to 8,000Hz), which brings other concepts than a fullrange dipole into the equation. Would you agree with him as well, or do I understand you correctly that you find this (i.e. coloration in the region of -7dB over the range of 4 octaves compared to a linear soundfield) to be irrelevant?

 

[goat76]

First I would like to thank Dr Toole for his time on this forum and Amir for his usual calm approach those naysayers and those of us learning.

Who are the “naysayers” in this thread? Just because some of us like to broaden the discussion and take it further, doesn't mean we are “naysayers”. Science hardly never stands still, new knowledge is often won by letting the discussion be open, and that will unavoidably lead to different views on the matter.

 

[Heinrich]

Answer is in Dr. Toole's book and papers. Audibility doesn't just depend on level but also Q (how broad it is). The wider the deviation, the more chances that musical notes hit it and show the deficiency. At the extreme, eve 0.5 dB deviation with low Q can be audible. This is why I build EQs and experiment to detect level of audibility of frequency response errors.

This is circulated for decades, and a well respected word of caution. I recently tested with a third person the audibility of changes in bass level. Just 1dB made a difference when listening pretty loud. Too thick, a tad too lean were the comments, depending on the record. (Speaker and amplification far from their limits. No 'blind' statistical testing needed, too obvious.)

I don't want to bother you with any more anecdotes, I found my answer on spiorama.org - 0,6 on the score. Everything below this is not significant or shows equivalence. I was worried because without equivalence (aka measurement error) a scale has no meaning, that's the way it is.

Now we often see that when tuning speakers, the score increases by less than this 0.6. Everyone has to make up their own mind about that.

 

[thewas]

Between 600 and 8,000Hz, the predicted in-room-response stays almost linear within a +-1dB corridor without showing a major tendency of tilt. Below 600Hz it gets a bit higher due to lower d.i. of the cardioid. I found this to be achievable in practice, and personally perceive the difference between such a linear off-axis soundfield and a ´dull´, decreasing reverb level towards higher frequencies in the region presented in your chart (around -7dB in the same range) to be clearly audible and decisive for tonality judgment.

You don't need though a special and expensive design to reach such a corridor in that region, even a decently implemented low budget mass market design with continuously increasing DI can reach similar when equalised to flatish listening window response as the biggest SP change happens lower, exemplary done on the KEF Q7 Meta with 5 filters to make the LW flatish, this is corresponding EIR with those filters: