Article: Why We Need Audiophiles, The subjective perspective Options

[2tec]

I'm interested in what the HA community thinks about this new Gizmodo article, or blog, about Michael Fremer, an audio reviewer from Stereophile, which clearly goes completely against the grain around here. For instance:

"We play my solid 256kbps VBR MP3 of "Heroes" off my iPod; it sounds like shit. Free of pops and crackles, yes, but completely lifeless, flat in every way. This is the detail that matters: Audiophiles are basically synesthesiacs. They "see" music in three-dimensional visual space. You close your eyes in Fremer's chair, and you can perceive a detailed 3D matrix of sound, with each element occupying its own special space in the air. It's crazy and I've never experienced anything like it." ~ Why we need audiophiles

[2Bdecided]

But good systems with excellently matched speakers (with excellent time and frequency domain responses) do "image" spectacularly better than lower quality stuff. The front/back depth of the sound stage is increased, the location of (say) the singer is focussed more tightly etc etc. You can also put the speakers further apart before the sound stage falls apart. It's not what the record producer intended (usually), but it's very impressive. Stereo is supposed to work with 60 degree speaker angle. I've heard it work stunningly well with 110 speaker angle - but only with very good speakers.

The photographs of that particular listening room are not impressive unless they misrepresent the reality - from what it looks like, I'd want the speakers much further away from the walls, and from everything else. The kind of early reflections I'd expect in that room would seriously damage the magical 3-d sound stage that's claimed to exist.

Cheers,
David.

[Arnold B. Kruege...]

But good systems with excellently matched speakers (with excellent time and frequency domain responses) do "image" spectacularly better than lower quality stuff.

The frequency domain part I'll agee with, but the time domain part - well numerous counter-examples exist. For example, one of the major innovations of the last 20 years in crossover design has been the so-called "Linkwistz-Riley" configuration. However Linkswitz-Riley agressively time domain response for improved frequency-domain response.

The front/back depth of the sound stage is increased, the location of (say) the singer is focussed more tightly etc etc.

That kind of poetry slides off the lips so gracefully...

You can also put the speakers further apart before the sound stage falls apart.

Whatever that means. The means by whioh it was verified by means of a DBT seem to be unknown...

It's not what the record producer intended (usually), but it's very impressive. Stereo is supposed to work with 60 degree speaker angle. I've heard it work stunningly well with 110 speaker angle - but only with very good speakers.

Most home speakers are non-directional enough that +/- 25degrees of toe shouldn't mean that much.

The photographs of that particular listening room are not impressive unless they misrepresent the reality - from what it looks like, I'd want the speakers much further away from the walls, and from everything else. The kind of early reflections I'd expect in that room would seriously damage the magical 3-d sound stage that's claimed to exist.

Looks like one of those too-small Manhattan apartments...

[2Bdecided]

The frequency domain part I'll agee with, but the time domain part - well numerous counter-examples exist. For example, one of the major innovations of the last 20 years in crossover design has been the so-called "Linkwistz-Riley" configuration. However Linkswitz-Riley agressively time domain response for improved frequency-domain response.

True.

However, in a digital world it's possible to make both excellent. I see no reason to compromise the time domain even if it's 100x less important. You don't need to make any trade off with digital cross overs, i.e. better time domain response doesn't implicitly make the frequency domain response worse. Of course, in both domains, it's what the errors are that matters, not just whether there are any - and in both cases, it is a real skill to correlate measurements with what you can or cannot hear. I don't have that skill - I'd do measurements and double-blind listening tests to try to understand the correlation. I think I've already reported passes and fails of such tests in both domains.

The front/back depth of the sound stage is increased, the location of (say) the singer is focussed more tightly etc etc.

That kind of poetry slides off the lips so gracefully...

Well, you have to describe what you hear somehow. Spatial width, depth, and source position uncertainty are real repeatable subjective quantities associated with audio. I made a psychoacoustic model that could measure some of them once. I make no claims of accuracy for it, but it was a start.

You can also put the speakers further apart before the sound stage falls apart.

Whatever that means. The means by which it was verified by means of a DBT seem to be unknown...

I don't have Harman's nice speaker test room to enable genuine double-blind testing of speakers.

However, I've tested many virtual surround sound algorithms, both via headphones and speakers. These tests were by necessity double blind - the listener can't see or know what algorithm they are listening to, and the PC doesn't know the contents of the files it is presenting. The qualities which you described as "poetic" are the exact kinds of details I was interested in, and reliably garnered from listeners. "It's further away" "It's closer" "It's over there" "It's not really anywhere" etc are the kinds of things that people say!


Let me be clear what I meant (though I'm convinced you know all this better than me): the stereo "trick" works when the perception of most listeners is that the singers, instruments etc which are present on both channels, sound like they are in the space between the speakers. The stereo "trick" can be said to be failing when that doesn't happen, and the perception of most listeners is that those same singers, instruments etc are difficult to locate - they're diffuse, or not particularly anywhere, or are clearly coming from the location of the speakers (rather than between them).

If you take a pair of speakers, and move them further and further apart, so increasing the angle at the listener, there comes a point where the stereo trick stops working. I'm not talking about toe-in (to keep a single variable in this test, you've got to keep the speakers pointing directly at the listener).

My experience is that the stereo trick keeps working at a greater angle for some speakers than for others. My guess is that this has something to do with the accuracy of the speakers, and something to do with how closely they approximate a point source, but I never had the chance to investigate further.


No, I have no ABX tests to back this up. Maybe Sean at Harman can do some.

However, there comes a point when is reasonable to assume that someone really does hear a difference, even without an ABX test. I suggest that moving the things that are actually making the sound by several feet is beyond the point where ABX tests are necessary - especially as there is already well-documented psychoacoustic data proving that humans can detect the location of something down to a few cm!


OTOH, I have evidence to prove that what people see affects their perception of sound location - putting an unconnected speaker at the location of a virtual source (whether the virtual source is presented via headphones or speakers!) changes the perception of that virtual source dramatically. However, having speakers visible which are well away from the virtual source has no detectable effect on localisation perception.

You could blindfold people to check this (I did) - but then if there was an interaction, you'd want people to be able to see the speakers, as that's how they usually listen.

Which is hilarious, and you'll hate it, but it's one example of where an audible difference isn't "placebo" in the traditional sense - it's due to a specific way in which our ears and eyes work together. When you see a speaker somewhere, you expect the sound to come from that location. You don't want to remove this effect from the test, so much as quantify it.

Cheers,
David.

This post has been edited by 2Bdecided: Apr 17 2009, 13:28