Who Do You Believe?
by Greg Smith (firstname.lastname@example.org)
Back in March, I published a little article talking about inexpensive interconnects. It generated some useful feedback about what other people like, and ultimately that whole line of discussion built up to my going even more gonzo on expensive cable with the JPS Labs Superconductor. Way before that, though, I received e-mail from reader Yung Cheong asking me a very reasonable question. Here's an excerpt from that message, talking about me:
"I'm unclear as to his criteria for assessing the superiority of one interconnect over another. The only comments he seems to make are with respect to tonal balance: bass, treble, midrange. What about soundstage, detail, transparency, and rhythm? If he can't hear the difference, then he should tell his readers there's no difference in [those], only tonal balance"
Indeed, what about all those things? Enquiring minds would like to know. Our esteemed Editor felt that what the gentleman had asked was perfectly fair. In fact, as we've discussed many times, one of the problem with subjective reviewing is knowing, or at least understanding, a reviewer's measurement criteria. He suggested I respond with some brief comments about what exactly it is I'm looking for in cables. That seemed reasonable, so I started thinking about what the right priorities are for things like cable evaluation from my perspective.
Six months later, I'm still not even close to done thinking about that, but I do have some thoughts to share...
Where do words go...
Let's pick apart the terms in question for a second:
Now, the next word up for sale on "Is the Review Right?" is Rhythm. That's one of the big bass words, right next to Pace and Timing. Just what do any of these words mean in the context of audio, anyway? A trip to 1990's subjective review bible, The Audio Glossary from J. Gordon Holt, yields no help. As far as I can tell, the current fad in using those terms goes back to two columns from reviewer Martin Colloms. The December 1991 Stereophile goes into "Bass," while November 1992 yields "Pace, Rhythm, and Timing." I remember reading them when first published and being unenlightened. A recent return to the material didn't help much.
Regardless, I have to take exception to these terms. First off, audio components don't have pace, rhythm, or timing -- music does. One of the fundamental flaws of most audio reviewing that I read is the poor grasp between cause and effect for what people hear and report on. If I tell you a cable has bad rhythm, this gives you almost no useful information; I'm just throwing around a poorly defined term. Instead, I should be describing what it is that isn't coming through that makes the rhythm seem bad. If I tell you a cable rolls off the bottom end, you can bet that it's got bad "pace," too, but now I've given you an idea why. What's really a shame is that in these same reviews that confuse the reader by describing pace, rhythm, and timing, the real problems are mentioned -- things like speaker cabinet vibration coloration, often a cause of poor "timing." Why does a reviewer have to pick a trio of poorly defined words to describe that? Just say the cabinet rattles enough to smear the bass around already, don't pick a mysterious word that only true audiophiles are supposed to understand.
The whole style of subjective reviewing where sound quality is described with little concern for why doesn't do anything to help people who have issues with their system understand what it is they should be improving. It's no wonder people often read reviews and end up more confused than when they started!
The Meter Police
Some would say that the solution is easy: simply measure the equipment, then you know what the cause of any problems are. While this seems reasonable, there's one little problem. Nobody seems to be sure what to measure. Let's look at power amplifiers. For a long time, a significant section of the audio community has claimed that perceived differences between the way power amplifiers sound is strictly a fantasy on the part of the high-end devotee. If you lug the amplifier to the test bench and measure it into a resistor, you get the same graph of frequency response out of it as every other amplifier on the market. This is totally wrong. Fact is, power amplifiers do measure differently, but you need a more complicated test to detect it. In August of 1995, Stereophile made a landmark effort to quantify amplifier differences in the context of high-end reviews. John Atkinson adapted a speaker simulator, originally designed by speaker guru Ken Kantor, for use in Stereophile's amplifier testing. This circuit models the sort of irregularities that a real speaker gives to an amplifier, where the impedance varies between 6 and 20 ohms and the phase angle varies around +/-45 degrees (note that this is a typical but far from worst case speaker load). Ever since this test was introduced, you can see graphs showing blatant differences in the way some amplifiers react, differences totally masked in easier testing. I find this test very useful and commend Stereophile for taking this approach.
Unfortunately, the fact that you can measure how amplifier response breaks down into real-world loads still doesn't satisfy everyone that these differences are worth investigating. The Audio Critic has an excellent test they perform called the Power Cube. Given that amplifier response varies based on the load's resistance and reactance in addition to the input frequencies, their test maps the results across several areas into three-dimensional space. Unfortunately, despite that fact that they've got a method for testing what the real issues are, they still provide a rant about how high-end amplifiers are a rip-off as regular as their publishing schedule. This I just don't get; they have a test that starts to show what the more expensive amplifiers are doing better, and yet Aczel and friends still climb on their precarious soapbox and spew venom at those who feel amplifiers and (even worse) cables really matter.
"You can measure all you want, but a mass spectrometer isn't going to find a lot of difference between lunch at a high school cafeteria and the best dinner at a four-star restaurant." --Lynn Olson
Everyone has seen it happen with amplifiers. What's the next component that conventional measurement can't distinguish between now that will become more quantifiable in the future? The frequently quoted Clarke's third law (appearing in Lexicon ads as I write) tells you that something you don't have a technological grip on yet appears to be magic. I fully expect that some of the things that are scoffed at by the radical objectivist crowd now are going to be measurably better in the future. Coming back to interconnect cables again, I know I hear a variety of things that don't seem to correlate with any measurement currently being done. If I'm told the frequency response between two cables are the same, and that the transient response is perfect for slow signals like audio, yet still hear a difference, I see that as a failure on the measurement side. It's certainly not an excuse to deride those who hear the difference.
Part of the problem lies in sensitivity. The differences I note between cable are fairly obvious to me now, yet even a year ago I was just starting to really become convinced of their existence. There are two reasons for this. The first is system matching. You're not likely to note radical improvement by upgrading the cabling in an inexpensive system simply because the things that improve with the better cable aren't there in the first place. If your CD player doesn't accurately produce high frequencies when driving a preamp, two cables that sound different in the treble are likely to be indistinguishable. One of the worst results of this is a phenomenon I dubbed the cable unbeliever Catch-22. You take someone who believes that high-end audio is bunk. They switch to an expensive CD player, notice no difference. They switch to expensive cables, notice no difference, and decide the whole thing is a bunch of crap. But wait! Maybe both inexpensive CD player and inexpensive cable are flawed enough that either will remove enough musical information to make the better equipment they are matched with ineffective. I know it's far easier to pick out the subtle improvement better cables give me now that I've got components with very few weak points. I'll be the first to admit that spending big cable dollars relative to what the rest of your system costs is foolish, but at the same time I find the results from high-end cables well worth the investment even when spending hundreds of dollars on them in the context of a seriously expensive system.
The second thing dragging many equipment comparisons down is the whole issue of acclimation. It's often true that small changes like switching interconnects or adjusting component isolation are barely audible. In fact, I usually find that many small improvements aren't noticeable at all when I first make them. It's only after listening for a few months that I become conscious of what the superior configuration is doing better. It's a matter of training your ears to hear deeper into the music. After that, switching to the old way is obviously inferior. The question this begs is why you should even bother if differences are so minute. That's easy; while the individual changes are small and noticeable only to those who are really in tune to what they're hearing, the final cumulative results of many such small modifications can be obviously superior to anyone who listens without prejudice. I've had long stretches of time where I altered cables, changed speaker positioning, treated my room, all things that individually were small adjustments. Put them all together, and the results are worth it.
Gimme an A! Gimme a B! Gimme an X!
Some will claim that I'm deluding myself. Unfortunately, proving what's audible and what isn't remains another difficult problem. The de-facto method presented by many for determining what you can hear and what you're imagining is the double-blind ABX test. In this test, you take two components that are measurably the same. You let yourself get familiar with how both A and B sound in the context of the system they are in. Then, you get presented with a randomly determined sample X. You can compare X with both A and B, and are asked which it is. The results are collated and analyzed statistically to see if, in fact, you can distinguish A from B. While this is basically a good idea, the flaws in the usual implementation are numerous. First, the switching is often done by the infamous ABX box. Frankly, if you've got something that measures just like something else, and are splitting hairs over whether the pieces are audibly different, adding any extra circuitry to that system is likely to screw the whole thing up. The very notion that you can do something like compare whether premium "audiophile" capacitors are better than regular ones with an ABX box is utterly ludicrous. Whatever differences might appear under normal listening is likely to be swamped by the insertion loss from the extra box in the system.
Even if you could get a perfect switching mechanism, you're not home free yet. Level matching is the other hurdle to deal with. Some claim that differences as little as 0.1dB between components will cause people to favor the louder component as sounding "better" when, in fact, it's just louder. Accordingly, you need as exact of a match as possible to calibrate this style of comparison if you want it to be fair. Remember before when I said this was only good for things that were measurably the same? This is why. Try to level-match two different speakers. If the frequency response isn't the same, what do you match them up with? Using pink noise as a test signal gives a general average match, but you can end up with level mismatches based on source material. Using a single frequency to match with is obviously prone to error; if one speaker is -3dB at that frequency compared with the other, but is generally matched, you'll introduce a gross level error trying to match there. Generally, we need to measure the two devices pretty completely to even begin to level match them, and this brings us right back around to the measurement problems pointed out above. I wouldn't be surprised to find some attempts to match levels for blind comparisons are, in fact, introducing level errors because of poor up-front preparation for the test.
A Question of Balance
No, I still don't have a good answer for Yung Cheong. While some people believe that audio is a nicely solved problem where you take some measurements of the component pieces and know how they are going to work in a system, I just can't agree with that. To use an example I've heard Greg Weaver throw around, until someone can supply me with a way to measure a speaker in a room without listening to it and predict with high accuracy whether it's going to image well or not, I have to keep skeptical of relying solely on numbers for estimating how much I'm going to like something. At the same time, though, I refuse to use this as an excuse to totally abandon cause and effect and resort to the sort of totally subjective evaluations that some suggest. Even if I can't point out exactly what factor is at play when I hear something inexplicable, I can at least try to suggest patterns for the things that defy simple measurement. For me, the whole objective vs. subjective argument is like choosing a political party: I really don't want to be associated with either of those close-minded groups. It's only through being open to both views of the world that forward progress is going to happen, because nobody has all the answers yet. Dismissing anything as "impossible" because it doesn't fit into how you think things work is just an invitation to be proven wrong when our knowledge of the world expands. Out of the all the things I've brought up, that is one thing I am sure of.
Copyright © 1998 SoundStage!
All Rights Reserved