The difference between good and great

[Inner Space]

I also think that most recordings are pretty mediocre (at least the ones I listen to). And I don't think any speaker can address that. And better speakers might make it a bit worse.

I would quibble with "mediocre" ... I think these days most recordings are expertly and artfully aimed at "portability", i.e. they should sound good on a most-people-most-of-the-time basis, i.e. they're aimed at earbuds and soundbars and kitchen-counter pod speakers. Which means yes, our better systems make them sound worse, because we're not most people ... in fact we're a tiny, irrelevant minority.

I think a lot about this quote from Dr. Toole's book, and it's fundamental to my perspective on speaker reviews: "How do listeners approach the problem of judging sound quality? Most likely the dimensions and criteria of subjective evaluation are traceable to a lifetime accumulation of experiences with live sound, even simple conversation. If we hear things in reproduced sound that do not occur in nature, or that defy some kind of perceptual logic, we seem to be able to identify it ... "

I disagree with Toole here. I think after more than a century and four or five generations of reproduced sound, most people have accommodated a kind of mental category division between "music" and "loudspeaker music". The two are very dissimilar, and subconsciously are judged by different criteria. I would rewrite Toole's sentence thus: "Most likely the dimensions and criteria of subjective evaluation are traceable to a lifetime accumulation of experiences with reproduced sound, including TV soundtracks and voices on the radio. If we hear things in reproduced sound that are different from what we've heard before, we seem to be able to identify it - and might reject it, even if it's objectively closer to reality."

 

[Duke]

The question though is what happens if something that perhaps measures a little bit worse, ends up, for some reason, sounding a little better. Just wondering about other people's experiences.

My understanding is that spinorama data is something like 87% reliable in predicting loudspeaker preference, at least among monopole loudspeakers. I share your interest in explanation(s) for the other 13%.

Edit: See andreasmaaan's illuminating comment below.

 

[napilopez]

I disagree with Toole here. I think after more than a century and four or five generations of reproduced sound, most people have accommodated a kind of mental category division between "music" and "loudspeaker music". The two are very dissimilar, and subconsciously are judged by different criteria. I would rewrite Toole's sentence thus: "Most likely the dimensions and criteria of subjective evaluation are traceable to a lifetime accumulation of experiences with reproduced sound, including TV soundtracks and voices on the radio. If we hear things in reproduced sound that are different from what we've heard before, we seem to be able to identify it - and might reject it, even if it's objectively closer to reality."

That doesn't make sense though. The whole point of Toole's research is showing that, for the vast majority cases, people prefer neutrality, which on average is closer to reality.

 

[Inner Space]

That doesn't make sense though. The whole point of Toole's research is showing that, for the vast majority cases, people prefer neutrality, which is closer to reality.

I disagree again - another completely valid way of describing Toole's research is that it shows, for the vast majority of cases, people prefer something a little clearer but basically the same as what they've heard before, which isn't necessarily closer to reality.

 

[Pearljam5000]

I think a lot about this quote from Dr. Toole's book, and it's fundamental to my perspective on speaker reviews:

"How do listeners approach the problem of judging sound quality? Most likely the dimensions and criteria of subjective evaluation are traceable to a lifetime accumulation of experiences with live sound, even simple conversation. If we hear things in reproduced sound that do not occur in nature, or that defy some kind of perceptual logic, we seem to be able to identify it. By that standard, the best sounding audio product is the one that exhibits the fewest audible flaws. Perhaps this is how we are able to make such insightful comments about sound quality based on recordings that either had no existence as a “live” performance, or that we have no personal experience with."

Emphasis mine.

Most audio reviews and impressions tend to be written in an additive sense. Before reading Dr Toole's book, that's how I approached things too. I thought the best product would be the most euphoric, the one at which I could throw the most positive adjectives. "Transparent," "like a live concert," "the most detailed treble," "the smoothest midrange I've heard," "I heard things I'd never heard before," It plays well with the naive misconception that more expensive products have to sound better. But I think there's a rough limit to how "good" or "real" a speaker can sound, after which the best you can do is "different."

So now I think of things differently. If the best speakers are the ones with the fewest audible flaws me that means the ones least likely to break the illusion of real music. So when I think about products I'd consider to be among the best I've heard over extended listening, like the D&D 8C, I think of the products that do everything right.

For example, the Neumann KH80 is a fantastic speaker, but compared to the 8C, its obvious flaws are bass extension and SPL handling. Inevitably, those flaws will make themselves known when I try to play music with lots of bass loudly. And without very good room correction, the 8C automatiaclly beats everything else I've heard by virtue of having such controlled room interaction built in.

Basically, I'm not looking for something to blow my mind, just for something that doesn't break my suspension of disbelief.

I think it's made me a better audiophile, honestly. Instead of looking for speakers that deliver audio nirvana, I look for the speakers that most let me just listen to the music without the distraction of something sounding "off." The best a speaker can really do, after all, is just sound convincingly real.



I strongly disagree with this. Perhaps my perspective is different because I get to listen to and measure more speakers in my own home than the average person but I can very much get a good idea of how much I'll like a speaker based on comprehensive measurements alone.

I do measure after listening, but in cases where measurements are already available correlating impressions with the data is really feels like second nature at this point. And 99 percent of time impressions can be explained by Frequency respnose, Directivity, and power output.

The one thing I sometimes have trouble fully quantifying is "dynamics," but that's more likely simply because dynamics can be a combination of multiple measurable factors (distortion, compression, directivity, maybe other stuff).

I think it's just a matter of having the right measurements and knowing how to interpret them based on your preferences and listening environment. And I do not think the relationship between measurements and sound is so complicated that they can't be understood by the average audio enthusiast.

Yeha but for example both the KH120 and 8030 are rather flat and neutral, but I like the Genelecs a lot more.
I actually came to the conclusion that the most important thing in a a speaker sound is the tonality, especially the highs, if it sounds even 1% off to your ears then you won't like it , it will always sound off to you.
Another example the the Solo 6 don't measure perfectly and do have some sort of a smiley curve to their sound but I liked it.
There are many things beyond the science and measurements, we're not robots, that's also why people like vinyl and tube amps, accuracy is less important to them they like warm sound

 

[napilopez]

I disagree again - another completely valid way of describing Toole's research is that it shows, for the vast majority of cases, people prefer something a little clearer but basically the same as what they've heard before, which isn't necessarily closer to reality.

How do you arrive at that conclusion from "flat frequency response is preferred?" I also am not sure how you're saying it's "basically the same as what they've heard before" when it seems like before his research, a lot more speakers were designed with the philosophy of flat power response, uneven directivity, smiley curves, etc.

Put another way, how could you get a speaker "closer to reality" without aiming for a flattish frequency response?

Then again, I disagree with your central premise that reproduced music and live music are judged by different criteria. I'm also not sure how you arrive at that conclusion. I want my speaker music to sound like live music. Elaborate? What are these different criteria? Not trying to be snarky, genuinely curious as to your arguments.

 

[andreasmaaan]

My understanding is that spinorama data is something like 87% reliable in predicting loudspeaker preference, at least among monopole loudspeakers. I share your interest in explanation(s) for the other 13%.

Actually, it's 87% reliable when applied to the very sample of speakers from which it was derived. The figure is therefore a best-case scenario when applied to the population of speakers beyond that sample.

 

[phoenixdogfan]

Charles Sprinkle commented that spin-o-rama measurements and their relationship to listener preference was simply the lowest-hanging fruit in the quest to identify objective measurements that explain subjective preference... But the research needed to begin somewhere, and in-room frequency response, resonances, and timber which can all be identified with the spin-o-rama data set was the logical starting point, but by no means exhaustive.

I also believe it's a mistaken view by some that objective measurements are all that's needed. Clearly not even Harman believes this given the great effort/expense Harman puts into collecting high quality subjective data via controlled double-blind listening tests.

I'm watching to see what/where the next breakthroughs in objectively quantifying loudspeaker performance will be.

I agree, but we need a starting point, and, I think Harman has given it to us.

 

[napilopez]

Yeha but for example both the KH120 and 8030 are rather flat and neutral, but I like the Genelecs a lot more.
I actually came to the conclusion that the most important thing in a a speaker sound is the tonality, especially the highs, if it sounds even 1% off to your ears then you won't like it , it will always sound off to you.
Another example the the Solo 6 don't measure perfectly and do have some sort of a smiley curve to their sound but I liked it.
There are many things beyond the science and measurements, we're not robots, that's also why people like vinyl and tube amps, accuracy is less important to them they like warm sound

And why couldn't you tell that tonality from the measurements? The science also tells us that being off by 1 dB can make a difference in preference.

I want to be clear that I don't think every speaker needs to measure the flattest or have the smoothest every directivity to be preferred. There's a difference between saying "the best measuring/flattest/smoothest speakers will always be preferred" and "measurements can explain what you are hearing." I'm saying the latter. I listen to vinyl often. My favorite speakers aren't always the absolutely flattest.

But I do have a good sense of which things affect my preference and which don't. For example, I tend to prefer wider directivity even if there are narrower speakers that appear to have smoother directivity (as is usually the case, as with waveguides that will smooth and narrow directivity). Speaking of...

One occasion was the Neumann KH80 against the Genelec G2, the latter I've recently measured in detail. Having compared them to the factory measurements I dare to say I'm not far off reality. I listened to them (mono) in a large auditorium, timbrally they weren't miles apart. I can't say I preferred one over the other in this aspect. The G2 however simply had a larger, more spacious quality to it. The KH80 sounded 'smaller' with a more noticable shift in timbre as I moved off-axis. There is little to nothing however in the measurements that supports or explains that experience.

I'd still be willing to bet this has something to do with directivity. It's hard to compare measurements from different sources somethimes, and the G2 does not appear to have particularly wider directivity than the neumann, but it does seem to show more constant directivity behavior after 2kHz , which I find I tend to prefer in terms of tonality. Lots of waveguided/horn speakers will instead have a frequency response that becomes increasingly tilted off-axis.

 

[Bugal1998]

I agree, but we need a starting point, and, I think Harman has given it to us.

Indeed they have

 

[Inner Space]

Then again, I disagree with your central premise that reproduced music and live music are judged by different criteria. I'm also not sure how you arrive at that conclusion. Elaborate? What are these different criteria? Not trying to be snarky, genuinely curious as to your arguments.

All based on observation, of music, reproduction, and people. I'm sure you'd agree that even good reproduced sound is a million miles from the real thing, in terms of scale, bandwidth, dynamics, impact, volume, tone, timbre, and a dozen other things. Yet most folks seem very happy with what they're getting at home. How is that even possible without two separate but parallel sets of criteria?

E.g., my granddad went to the symphony when he could, and enjoyed chatting with his pals at the union hall, and at home loved his tabletop radio, which thumped away in the corner, all warm and chesty. He was a sentient human being, so he must have known that in real life orchestras and voices sounded different than his radio, yet he seemed to score both very highly, as if separately, in and of themselves - as if "music" was one thing, capable of being good or bad, and "loudspeaker music" was another thing entirely, to be judged on a different scale.

Reproduced sound is inevitably miniaturized, compressed, and limited ... to say it can be compared to the real thing on the same 1 - 10 scale seems to me absurd. There have to be two separate 1 - 10 scales, surely. A great orchestra in a great hall might score 9/10, and a great stereo in a great room might score 9/10 also - but on different scales completely. Do you disagree?

 

[restorer-john]

My understanding is that spinorama data is something like 87% reliable in predicting loudspeaker preference, at least among monopole loudspeakers.

Brian Fantana:

 

[andreasmaaan]

All based on observation, of music, reproduction, and people. I'm sure you'd agree that even good reproduced sound is a million miles from the real thing, in terms of scale, bandwidth, dynamics, impact, volume, tone, timbre, and a dozen other things.

Reproduced sound is inevitably miniaturized, compressed, and limited ... to say it can be compared to the real thing on the same 1 - 10 scale seems to me absurd. There have to be two separate 1 - 10 scales, surely. A great orchestra in a great hall might score 9/10, and a great stereo in a great room might score 9/10 also - but on different scales completely. Do you disagree?

I agree with your conclusion, but not with a number of the premises. A big system can do dynamics, volume, tone, timbre etc quite commensurately with a live orchestra.

Where it falls down is in terms of spatial reproduction IMHO.

Ofc, I'm not talking about your granddad's tabletop radio here

 

[March Audio]

My understanding is that spinorama data is something like 87% reliable in predicting loudspeaker preference, at least among monopole loudspeakers. I share your interest in explanation(s) for the other 13%.

Edit: See andreasmaaan's illuminating comment below.

From memory a lot has to do with the bass response. IIRC the accuracy of the prediction goes up to 95% with smaller (more bass limited) speakers.

 

[SKBubba]

The last $2.

 

[Inner Space]

I agree with your conclusion, but not with a number of the premises. A big system can do dynamics, volume, tone, timbre etc quite commensurately with a live orchestra.

Honestly, and with great respect, I think you're falling into the same confusing-the-two-scales trap here. I sincerely doubt if there's a system on earth that could reproduce even a single tympanum strike commensurate with a live percussionist.

Ofc, I'm not talking about your granddad's tabletop radio here

That radio became the heart of my first record player, with a junkshop Garrard changer, a homemade plinth, and a ceramic cartridge. I used it for nine years - until my other granddad died and left me a small legacy that bought me my first "proper" system.

 

[andreasmaaan]

Honestly, and with great respect, I think you're falling into the same confusing-the-two-scales trap here. I sincerely doubt if there's a system on earth that could reproduce even a single tympanum strike commensurate with a live percussionist.

At what distance? The starting point for me is a seat in the auditorium.

 

[andreasmaaan]

That radio became the heart of my first record player, with a junkshop Garrard changer, a homemade plinth, and a ceramic cartridge. I used it for nine years - until my other granddad died and left me a small legacy that bought me my first "proper" system.

Sounds like a lot of fun

 

[Inner Space]

Sounds like a lot of fun

It was by far the best system I ever had - in that the biggest jump is always from nothing to something.

 

[HooStat]

Actually, it's 87% reliable when applied to the very sample of speakers from which it was derived. The figure is therefore a best-case scenario when applied to the population of speakers beyond that sample.

It is actually not even that good. Remember that the data from the top speakers was used to define one of the inputs to the model (optimal predicted in room response?), so some of the correlation is induced by the methodology, not by the results (i.e., it is a self-fulfilling calculation). Plus, 87% isn't interpreted as a measure of explanatory power. That would be the square of the correlation coefficient (r-squared), or 76%. And even that, as mentioned above, is only relevant for the internal sample. Its relevance to current samples is unknown and is almost certainly optimistically estimated for reasons beyond those mentioned.