Why evaluating the sound of a single speaker is essential

[janbth]

And my post links to it.

So it does! Post #6 - sorry I missed that! Thanks! I’m also embarrassed to see that this is Figure 3.15 in the 3rd edition of Sound Reproduction, which I own…

 

[RigorDude]

Just use any of the higher-scoring pieces listed in post #6 of this thread, played in summed stereo.

That would be Pink Noise, Fast Car, a few studio-pop tracks and a Sousa march with lots of bass-drum thwacks -- correct? Here's the table:

Am I alone in finding this an unconvincing proof-of-concept for single-speaker listening tests? You probably disagree; in any case, we're left with five or six tracks as the basis for our single-speaker evaluation method. Out of millions. It's a great playlist -- for Gitmo.

A more damning point is that these tracks -- not just Toole's list but also Amir's choices for everyday reviewing -- are all well-produced studio pop for which the engineer made sure to "check the mono." That's why they sound great on your hi-fi but also pretty good on a JBL portable usb speaker playing from your phone.

Listening through one speaker for evaluation purposes is a nifty idea. It reduces room reflections that muddy the waters, and we're less liable to be distracted by spatial effects. In practice, however, single-speaker listening has ended up heavily reliant on studio pop, i.e. artificial sound. That's because, unlike studio pop, a stereo recording made in real space with real instruments and voices sounds crappy when summed to mono. (Try it! All you need is some y-adapters.) Moving in the opposite direction, from the crappy mono back to stereo, the change is a transformation for the better. The gross differences between mono and stereo overwhelm the differences between speakers, and the "correct judgments" percentages plummet.

When I first posted to this thread, I hoped that ASR folks would be able to suggest some natural recordings that work well when summed to mono. That way, those of us who don't listen to studio pop all the time could check out single-speaker listening. Lightnin' Hopkins, Prestige jazz, pre-stereo RCAs and Mercury Living Presence -- these are all excellent recordings, but they are 60 or 70 years old. Perhaps there is something more recent? Sadly, the response so far is "Pink Noise and Fast Car. You don't need anything else."

 

[kyuu]

The whining about the song choices used for loudspeaker evaluation is tiresome. You don't like the music selections. Ok? Catering to your (or anyone's) personal taste in music is not the point.

 

[RigorDude]

Okay, so what is the point of the music selections? To be representative of what people actually listen to? I think not.

 

[No. 5]

That would be Pink Noise, Fast Car, a few studio-pop tracks and a Sousa march with lots of bass-drum thwacks -- correct? Here's the table:

Am I alone in finding this an unconvincing proof-of-concept for single-speaker listening tests? You probably disagree; in any case, we're left with five or six tracks as the basis for our single-speaker evaluation method. Out of millions. It's a great playlist -- for Gitmo.

A more damning point is that these tracks -- not just Toole's list but also Amir's choices for everyday reviewing -- are all well-produced studio pop for which the engineer made sure to "check the mono." That's why they sound great on your hi-fi but also pretty good on a JBL portable usb speaker playing from your phone.

Listening through one speaker for evaluation purposes is a nifty idea. It reduces room reflections that muddy the waters, and we're less liable to be distracted by spatial effects. In practice, however, single-speaker listening has ended up heavily reliant on studio pop, i.e. artificial sound. That's because, unlike studio pop, a stereo recording made in real space with real instruments and voices sounds crappy when summed to mono. (Try it! All you need is some y-adapters.) Moving in the opposite direction, from the crappy mono back to stereo, the change is a transformation for the better. The gross differences between mono and stereo overwhelm the differences between speakers, and the "correct judgments" percentages plummet.

When I first posted to this thread, I hoped that ASR folks would be able to suggest some natural recordings that work well when summed to mono. That way, those of us who don't listen to studio pop all the time could check out single-speaker listening. Lightnin' Hopkins, Prestige jazz, pre-stereo RCAs and Mercury Living Presence -- these are all excellent recordings, but they are 60 or 70 years old. Perhaps there is something more recent? Sadly, the response so far is "Pink Noise and Fast Car. You don't need anything else."

It must be understood that the recommendations are made in the context of a specific purpose: reveal audibly objectionable resonances and the spectral balance of a loudspeaker. It is no different than using selected video clips to ascertain a TVs de-interlacing and color accuracy, the purpose of the signal is to reveal flaws in the device under test, not provide artistic enjoyment to the person performing the test.

If you are interested in preforming the kind of testing that Dr. Toole is talking about, any music you wish to use that has dense spectrum and wide bandwidth appears to make a suitable test signal, but as Sean Olive remarked in an interview that one should probably not select music you like for this kind of testing because you’re going to listen to it so frequently that you’ll hate it.

 

[Safal gupta]

Evaluating the sound of a single speaker is essential because it helps you understand its true tonal character, clarity, and balance without the influence of other speakers. By testing one speaker at a time, you can:

Identify any distortion, coloration, or imbalance in the driver or crossover.

Ensure proper phase and wiring, which directly affect stereo imaging.

Hear the natural frequency response and tonal accuracy before pairing or calibrating with others.

Make more accurate decisions when matching speakers in a multi-channel setup.

In short, listening to a single speaker isolates variables, letting you judge its individual performance before it becomes part of a larger system.

 

[Beave]

Evaluating the sound of a single speaker is essential because it helps you understand its true tonal character, clarity, and balance without the influence of other speakers. By testing one speaker at a time, you can:

Identify any distortion, coloration, or imbalance in the driver or crossover.

Ensure proper phase and wiring, which directly affect stereo imaging.

Hear the natural frequency response and tonal accuracy before pairing or calibrating with others.

Make more accurate decisions when matching speakers in a multi-channel setup.

In short, listening to a single speaker isolates variables, letting you judge its individual performance before it becomes part of a larger system.

Gee, thanks, Captain Obvious. Which AI program did you use to write this post for you?

 

[MAB]

Okay, so what is the point of the music selections? To be representative of what people actually listen to? I think not.

You should read Toole's post right above the table that was in the link you followed.

"Here is the result of evaluations by Sean Olive in 1994,(Olive, S.E. (1994). “A Method for Training Listeners and Selecting Program Material for Listening Tests”, Audio Eng. Soc. 97th Convention, preprint 3893.) and as reported in Figure 3.15 in the 3rd edition of my book. This shows the success rate of listeners reporting hearing resonant flaws in loudspeakers. You will see poor correlation with entertainment value."

It's hard to write it more clearly than Toole. It has nothing to do with what people find entertaining. It's about finding tracks that are optimal for hearing resonances and determining differences. Many of our personal favorite songs are not good for critical evaluation.

 

[LSPhil]

I've been away from home for work for a while, so I quickly put together a 2-way speaker using a CSS10DX and a Peerless tweeter (1"). It’s crossed over at 1.5 kHz with an acoustic Butterworth filter. I know that’s a bit of a daring setup, but with a 42 dB/octave slope it actually works quite well. The woofer can’t really go any higher, and the tweeter doesn’t want to go any lower.
To evaluate the box, I’ve been using a really good recording of Grace Jones’ Slave to the Rhythm. That track is actually capable of revealing the acoustic characteristics quite clearly, even when played back on an ordina

ry smartphone. Pretty cool, isn’t it?

 

[Salt]

Maybe, but what does this contribute to the topic?

 

[LSPhil]

Maybe, but what does this contribute to the topic?

If a recording like the one I mentioned produces a linear response, it can be considered a reference, since it generates the same amplitude across all occurring frequencies. I claim that AC/DC does not.

 

[RigorDude]

You should read Toole's post right above the table that was in the link you followed.

It's hard to write it more clearly than Toole. It has nothing to do with what people find entertaining. It's about finding tracks that are optimal for hearing resonances and determining differences. Many of our personal favorite songs are not good for critical evaluation.

I'm not sure where you got the idea that I want to use my "favorite songs," though I doubt it's anything I said. I do think it would be unwise to buy speakers without hearing them with any music you really like. After all, that is what you are most familiar with and thus might have some idea of how it "should" sound. "This speaker is the best by far with Tuvan throat singing, and though I've never heard Tuvan throat singing before in my life, I'll take it." Seems a bit off, doesn't it? Personally, I listen to classical music a lot, so I have always tried to hear a solo instrument or voice track or two, something with string quartet (highly revealing), and an orchestral piece of some kind. There are thousands of good recordings of each of these kinds of music, so there's no need to use "favorites."

But shouldn't they be recordings that sound good? The vast majority of classical recordings are in stereo or multichannel, and they sound just plain wrong when summed to mono. I know because I've tried it. The reasons ought to be obvious, and the effect is discussed at length in the ASR thread "Sonic impact of downmixing stereo recordings to mono."

That is what, initially, inspired me to ask for suggestions for music to use in single-speaker listening. @Floyd Toole writes that "the most revealing music in multiple-loudspeaker comparison tests tends to have complex instrumentation - a dense spectrum - wide bandwidth - especially bass extension - and reverberation." He offers no specific examples, but this seems like it could describe a lot of classical recordings.

So far, the only specific tracks tossed into the ring are those that appear in a bar chart in Sean Olive's 1994 AES paper. Toole includes this chart in his response to another ASR poster asking for specific music suggestions. Short of buying the paper, one can't even tell whether the listed samples were heard in mono or stereo. Toole's main takeaway is, Music that works for single-speaker listening may be unappealing. Okay, but shouldn't it at least be something I'm familiar with?

Both Toole and @amirm have preached the single-speaker method to ASR readers. In the first post in this thread, Toole states, "to determine how good your loudspeakers are, do comparison listening tests in mono." Heck, the title of this thread says that mono listening is "essential." In his video on the topic, Amir says that "mono is much more useful and powerful in determining the fidelity of a speaker than stereo or multi-channel."

It's hard to process that claims of this import about the how of listening are accompanied by such a glaring lack of interest in the what of it.

 

[Floyd Toole]

I'm not sure where you got the idea that I want to use my "favorite songs," though I doubt it's anything I said. I do think it would be unwise to buy speakers without hearing them with any music you really like. After all, that is what you are most familiar with and thus might have some idea of how it "should" sound. "This speaker is the best by far with Tuvan throat singing, and though I've never heard Tuvan throat singing before in my life, I'll take it." Seems a bit off, doesn't it? Personally, I listen to classical music a lot, so I have always tried to hear a solo instrument or voice track or two, something with string quartet (highly revealing), and an orchestral piece of some kind. There are thousands of good recordings of each of these kinds of music, so there's no need to use "favorites."

But shouldn't they be recordings that sound good? The vast majority of classical recordings are in stereo or multichannel, and they sound just plain wrong when summed to mono. I know because I've tried it. The reasons ought to be obvious, and the effect is discussed at length in the ASR thread "Sonic impact of downmixing stereo recordings to mono."

That is what, initially, inspired me to ask for suggestions for music to use in single-speaker listening. @Floyd Toole writes that "the most revealing music in multiple-loudspeaker comparison tests tends to have complex instrumentation - a dense spectrum - wide bandwidth - especially bass extension - and reverberation." He offers no specific examples, but this seems like it could describe a lot of classical recordings.

So far, the only specific tracks tossed into the ring are those that appear in a bar chart in Sean Olive's 1994 AES paper. Toole includes this chart in his response to another ASR poster asking for specific music suggestions. Short of buying the paper, one can't even tell whether the listed samples were heard in mono or stereo. Toole's main takeaway is, Music that works for single-speaker listening may be unappealing. Okay, but shouldn't it at least be something I'm familiar with?

Both Toole and @amirm have preached the single-speaker method to ASR readers. In the first post in this thread, Toole states, "to determine how good your loudspeakers are, do comparison listening tests in mono." Heck, the title of this thread says that mono listening is "essential." In his video on the topic, Amir says that "mono is much more useful and powerful in determining the fidelity of a speaker than stereo or multi-channel."

It's hard to process that claims of this import about the how of listening are accompanied by such a glaring lack of interest in the what of it.

What is missing from this meandering discussion is the knowledge/recognition that these musical tracks were ONLY used in monophonic, multiple-loudspeaker (3 or 4 at a time, loudness matched, double-blind) listening tests in which the objective is to eliminate the room as a variable (by position substitution) and the music (by editing it into 30 s loops so that all loudspeakers were heard with exactly the same sound). Pleasure in or excellence of the music were irrelevant.

It is the most powerful and revealing method for identifying timbral flaws (mostly resonances) in loudspeakers, and the most revealing signal is not music at all, but pink noise. The reason why music was used is to put the audibility of the resonances into realistic contexts. As was shown in the Sean Olive list of tracks the genre of music is unimportant, except that simply mixed pop recordings are less revealing than more complex ones, classical music, for all its compositional and performative excellence is no better. Solo sounds, voices and instruments are not useful because of limited spectral extent and density.

Because these very expensive and tedious tests were done for decades at the NRCC and at Harman we now have identified measurements that reveal the timbral flaws heard by listeners. The most neutral loudspeakers turn out to be the most technically accurate ones, and the technical information necessary to describe them is in the spinorama data set. Learning to interpret spinorama data is a more reliable method of identifying timbrally neutral loudspeakers than listening in stereo in the normal "take it home and listen to it" method that is open to numerous biasing influences. Now that the hard work has been done, and the measurement method standardized, we can move on to other problems - like listening room resonances.

Does finding a neutral loudspeaker ensure satisfying listening? No, because recordings are variable, but at least you will be hearing what was recorded and will know who or what to credit or blame. However, listening through loudspeakers that don't monotonously color everything in the same way is a genuine pleasure, whatever one's musical tastes.

So, this entire discussion of mono compatibility is irrelevant - nobody listened to music in mono to enjoy it, it was a test signal. Even a flawed stereo downmix can be useful, albeit annoying to listen to.

This is explained in my books, and rather painlessly in a YouTube video of a recent lecture I gave at the Toronto section of the AES. The message on the title slide is key.

 

[olieb]

It is the most powerful and revealing method for identifying timbral flaws (mostly resonances) in loudspeakers, and the most revealing signal is not music at all, but pink noise.

I guess much of the discussion comes from the difficulty to stomach this (use of pink noise in mono) and the recommendation of learning to understand spinorama instead of listening.
This just goes against almost everything people are used to. So there are a lot of questions and a lot of discussion.

About the mono-single speaker method. Did you ever contemplate to use mono-double speaker in the speaker comparison?
As to compare not the hard panned sound but the phantom center of different speaker models?

EDIT: As I am just reading the 4th edition, I just stumbled over the answer to my question on page 152.
Phantom center with 2 speakers creates a situation with a much more complex soundfield than one speaker and therefore results in a less critical hearing/jugdement of the sound impression from the speakers. Therefore one speaker is by all means more revealing.

 

[RexrothPigeon]

instead of listening.

The thing is, unless we're doing very controlled, blind listening, we probably aren't really going by what we hear. We are so ocularcentric and so prone to bias from non-auditory influences, that there's just no other way to really know that we're judging by our listening and not other factors.

In other words, when we go to a shop to listen to speakers, when we take them home to audition them, when we hear them in someone else's home or studio, we are subject to a lot of factors that have nothing to do with the sound and the equipment, and our impressions in those conditions just aren't reliable.

 

[Floyd Toole]

I guess much of the discussion comes from the difficulty to stomach this (use of pink noise in mono) and the recommendation of learning to understand spinorama instead of listening.
This just goes against almost everything people are used to. So there are a lot of questions and a lot of discussion.

About the mono-single speaker method. Did you ever contemplate to use mono-double speaker in the speaker comparison?
As to compare not the hard panned sound but the phantom center of different speaker models?

The thing is, unless we're doing very controlled, blind listening, we probably aren't really going by what we hear. We are so ocularcentric and so prone to bias from non-auditory influences, that there's just no other way to really know that we're judging by our listening and not other factors.

In other words, when we go to a shop to listen to speakers, when we take them home to audition them, when we hear them in someone else's home or studio, we are subject to a lot of factors that have nothing to do with the sound and the equipment, and our impressions in those conditions just aren't reliable.

Yes, as I emphasize in the lecture and books (especially in the 4th ed) the audio industry has been burdened by a fundamental problem since its inception. Demonstrably unreliable opinions have guided both the manufacturing of loudspeakers and their selection by users - both consumers and professionals. That is why there have been so many very different sounding products over the years. Few bothered to apply the scientific method to the subjective evaluations. I/we did, and were fortunate enough to have excellent technical facilities and sponsors with deep pockets at the time. It was a rare and fleeting opportunity and it has passed, but not before some very important facts were established.

We don't make a fuss about the timbre of electronics and wire because they sound the same when they measure the same. Loudspeakers have been the weak link, but now we have insights into the meaning of the appropriate measurements. Getting the audio populous to accept the fact that we can now reliable identify timbrally neutral loudspeakers in measurements is a challenge - we all have opinions, and of course they are correct - sigh!

As the title slide says: liking the loudspeakers is the answer to the wrong question. Get some neutral loudspeakers and find art you enjoy. Tone controls and accessible EQ are still legitimate manipulations, but one can always default to neutral to hear what was recorded. It might be good.

 

[LSPhil]

Because these very expensive and tedious tests were done for decades at the NRCC and at Harman we now have identified measurements that reveal the timbral flaws heard by listeners.

I must admit I'm rather sceptical.

The idea that someone could identify a specific driver resonance by listening to a speaker playing pink noise is astonishing to me. With so many frequencies present, picking out the resonance of a single membrane seems a tall order.

Naturally, anyone can hear a drastic difference in a two-way speaker when you reverse the phase of the tweeter; that's obvious. However, I seriously doubt that anyone could immediately identify the specific polarity of a 5th-order Butterworth filter. That seems highly questionable to me.

When designing a speaker, the 'zing' of an aluminium diaphragm will eventually disappear if you cross it over low enough. But the average consumer lacks this experience and is perfectly well-served by choosing a speaker that simply sounds good with their music. Furthermore, it's impossible to hear the extent to which the drivers influence each other.

That said, some things *are* very clear to the ear: whether a crossover is very steep, or if it's the acoustically poorest-sounding LR2 filter. The speed of the membranes, or a driver's ability to radiate acoustic energy without storing it, are also quite audible.

So, with all due respect, I struggle to see any practical advantage in assessing a speaker using pink noise.

 

[Floyd Toole]

I must admit I'm rather sceptical.

The idea that someone could identify a specific driver resonance by listening to a speaker playing pink noise is astonishing to me. With so many frequencies present, picking out the resonance of a single membrane seems a tall order.

Naturally, anyone can hear a drastic difference in a two-way speaker when you reverse the phase of the tweeter; that's obvious. However, I seriously doubt that anyone could immediately identify the specific polarity of a 5th-order Butterworth filter. That seems highly questionable to me.

When designing a speaker, the 'zing' of an aluminium diaphragm will eventually disappear if you cross it over low enough. But the average consumer lacks this experience and is perfectly well-served by choosing a speaker that simply sounds good with their music. Furthermore, it's impossible to hear the extent to which the drivers influence each other.

That said, some things *are* very clear to the ear: whether a crossover is very steep, or if it's the acoustically poorest-sounding LR2 filter. The speed of the membranes, or a driver's ability to radiate acoustic energy without storing it, are also quite audible.

So, with all due respect, I struggle to see any practical advantage in assessing a speaker using pink noise.

I didn't say that we "assessed" the performance of loudspeakers using pink noise. It was one of many test signals used to determine whether there were audible resonances in 3 or 4-way comparison tests. It used to be that no two loudspeakers would sound the same through pink noise, now the good ones are getting close - as the measurements improve, loudspeakers become more neutral.

You said: "With so many frequencies present, picking out the resonance of a single membrane seems a tall order." Not at all. All listeners can do it, some more reliably than others, and Sean Olive developed a listener training exercise that taught listeners to recognize and report estimated Q and centre frequencies of resonances. Human hearing is remarkable when given a chance to perform.

If you read the published literature and remain sceptical I would be surprised. None of this has to do with crossovers, by the way, the resonances at issue are in the individual transducers and enclosures. When you say: "a driver's ability to radiate acoustic energy without storing it, are also quite audible." you are talking about resonances that absorb and release energy in the time domain. Did you really watch the video?

 

[Duke]

The idea that someone could identify a specific driver resonance by listening to a speaker playing pink noise is astonishing to me. With so many frequencies present, picking out the resonance of a single membrane seems a tall order...

So, with all due respect, I struggle to see any practical advantage in assessing a speaker using pink noise.

Ime it's easier to quickly hear the differences between speakers with pink noise than with music. And it's also easier to quickly hear that there is a problem which needs to be diagnosed and fixed.

 

[MaxwellsEq]

So, with all due respect, I struggle to see any practical advantage in assessing a speaker using pink noise

Give it a go. You have nothing to lose!