Post
by Devonavar » Tue May 10, 2005 11:22 pm
Ok, just because I feel like being argumentative, I think I'll wade into this (excellent) discussion.
I think everyone here has agreed that a good audio system should reproduce the original as closely as possible. Most of the disagreement has been around the extent to which expensive audio equipment actually does this and/or how much a measureable improvement actually translates into an audible one.
My question is this: What is the original? If the goal is to reproduce a recording, the rationalists should win this discussion hands down. However, if you're trying to reproduce a piece of music or a sound as accurately as possible, it's not so simple. While there are very scientifically acceptible ways of measuring how well a particular setup reproduces a recording, reproducing music is a much more subtle task. Keep in mind that a recording is itself a reproduction of an original performance. If the goal of high end audio equipment is to reproduce the original performance, not the recording, there's a lot of flex room for equipment that "inaccurately" reproduces the original recording but may (on average) come closer to the original performance.
I suppose that the ideal reproduction of a performance would have these attributes:
-Recording would be done with two (and only two) mics positioned as close as possible to the relative position of the eventual listener's ears.
-The pickup pattern of microphones would perfectly emulate the frequency response of the listener's ear.
-No mixing or mastering would take place before or after the recording is imprinted onto whatever medium is used to store it.
-All transmission of the recording data, and the recording medium itself, would be completely lossless.
-The recording would be listened to on headphones (ideally the in-ear kind) that reproduce exactly the recording they have been given.
Only the last two points have been covered in this debate, which the general consensus (I think) that it is possible to meet these goals within the limits of human perception, and that this is possible without going overboard on the amount of money spent.
There has been discussion about the importance of room acoustics, and I think this holds the key to the point I am about to make. I specified that headphones should be used in order to eliminate the variance of room acoustics, but this variable will affect every evaluation of an audio system's quality, subjective or objective. No matter how much "objective" testing is done, there is always a subjective element that creeps in regarding how the measurement is done.
Our standards of objective measurement are nothing more than conventions that imitate our subjective impressions. When evaluating how well a recording is reproduced, scientific method requires that it be "captured" with an instrument of some kind. How this instrument is used is defined conventionally according to what best reproduces our subjective impressions.
The most obvious conventional variable is the position of the instrument. Different setups are likely to have different "sweet spots" where they reproduce the recording most accurately. These will vary based on the specific components in the system, and the room in which they are tested. As was mentioned, the quality (which I take to mean accuracy) of an audio system varies considerably from room to room. Since it's unfeasible to test every system in its most optimal conditions, scientific testing must set a conventional standard of measurement that takes into account both how the system is intended to be used and how it is actually going to be used. This conventional standard is going to be subjective; there's no way around it. Determining what standard will produce the most subjectively "accurate" objective results is itself an art, not a science.
A scientifically feasible standard of measurement will necessarily favour some systems over others according to the distance of the instrument of measurement, the acoustics in which it is tested and a host of other factors.
So much for our ability to objectively measure a system's accuracy of reproduction. However, the conventions inherent in trying to measure a sound are doubly crucial when trying to record it. I think I'm correct in saying that the purpose of a recording is to reproduce a performance as accurately as possible. This purpose is similar the purpose of measuring the accuracy of a playback system: In both cases the idea is to reproduce a source as "accurately" as possible in a different medium. In the case of making a scientific measurement, the source noise is the output of the speakers (and any background noise) at a particular point in space, and the new medium is the data output by the instrument of measurement (typically, another recording which can be compared — bit by bit — against the original). When recording a performance, the source is the performance, and the new medium is the final recording (technically, I think it's the mode of representation, but that's irrelevant for my point). In both cases, the original source is "correct" by definition, and the end result is judged according to how well we feel the recording matches the source. This judgement is subjective, even in scientific analysis. In the case of a recording, a "good" recording is one that reproduces our original subjective impression of the performance (assuming a perfect playback system); for an objective measurement, a "good" system of measurement is one that produces results that we feel (subjectively) represents the original source.
You still with me? Good. Now, here's my real point:
I will assume that a musical recording is recorded in a manner that is designed to bring the recording as close as possible. As someone mentioned, in pop music, the recording process is often just as much a part of the final product as the performance, in which case my critique needs revising. For the moment however, I will assume that the recording process is intended to be as transparent as possible.
I mentioned at the beginning of my post that an ideal recording system would position the mics at the position of the final listener's ears, and would use microphones that perfectly imitate what the listener hears. However, virtually no recordings of quality are made in this way. Microphones do not emulate the human ear perfectly, so compromises are made in the name of subjectively improving the final recording. Instead of the dual mic setup I suggested, each instrument/voice is routinely recorded separately from the others and mixed together afterwards in a way that better imitates what the mixer originally heard (or, more likely, what he/she wants to hear. In this respect, recording engineers are like the scorned audiophiles who distort the source because it sounds "better". However, this rejects the assumption that the goal of a recording is an accurate reproduction of an original).
Again assuming that accurate reproduction is the goal of a recording, the unrealistic practices of mixing and studio mic-ing are necessary because microphone technology is not the same as the human ear. There may be a bit of cross-contamination going on — microphones are designed for specific purposes, such as studio recording — but when it comes down to it, the most subjectively "accurate" recordings are achieved in setups that are completely untrue to how the performance is actually heard.
This brings up another source of subjectivity in recording: In order to determine what the recording should emulate, a particular listening position must be selected. Since there is rarely an objective basis for preferring one position over another, the recording engineer simply picks the one that sounds "best". In other words, a recording is mixed according to how the engineer imagines it should sound. In most cases, this is probably whatever gives him the most pleasure to listen to (since I presume this is the goal of music).
Now, its quite obvious that there are imperfections in the "accuracy" of most musical recordings when compared to a live performance. These imperfections creep into the music in both the recording and the playback. However, when most people listen to music, the only part of the experience they have control over is the playback system. So, it seems perfectly natural to me that a playback system that can "correct" some common flaws in the recording system might well be more true to the original performance than the recording it is playing back. And, if this is true, I can understand why people are willing to pay enormous amounts for "inaccurate" audio systems and even why they claim it plays back the audio more accurately when objective measurement says otherwise. Likewise, I can understand why people feel vinyl has higher fidelity than digial audio.
Obviously, all recordings are not made in the same way, so an audio system that sounds fantastic for some recordings may sound terrible for others, but this is true of even low-fi systems. The trick is to find some average that sounds reasonably good no matter what is played on it. Obviously, you can "tune" a system towards a particular genre of music (or "recording style"), but this introduces a further subjective element: The intended use of the system. There are few touchstones for attaining this average. A system that is perfectly faithful to the source recording (not the original performance) may be one of these. Sound clarity may be another. (Incidentally, is there a scientifically acceptible way of measuring clarity? It doesn't seem that either frequency or amplitude reproduction fully captures it. Measuring timing is probably part of it, but I'm not sure this captures it either.)
I think it's not implausible that there might be other attributes of audio gear that might compensate for recording errors while reducing their ability to accurately reproduce a recording. My headphone amp has a "sound enhancer" feature that introduces a small amount of crosstalk and delay between the stereo channels. There's no doubt that this distorts the source recording. There's also no doubt in my mind that it does "enhance" the sound; it adds a sense of "presence" that is often lacking. Ostensibly, the feature is to simulate the crosstalk that naturally occurs between our right and left ears when listening to loudspeakers (which is missing in headphones). In this case, the goal is to bring the listening experience more in line with the original "performance" (or at least, the way the music was intended to be heard, since most music is mixed for loudspeakers) at the expense of accurate reproduction of the recording. I believe there is some precedent using this technique during the recording process to add "presence" to an otherwise flat recording. In this case, the recording itself is modified to make it more "faithful" to the original performance.
Ultimately, I'm not sure I buy the rationalistic argument that the most "accurate" reproduction of a recording (or even a performance) is best. No matter how many objective results proving the transparency of a particular audio system are thrown in my face, I have to ask what does the system "objectively" reproduce? Music is subjective to begin with; it's purpose (most of the time) is to give us pleasure. Pleasure is a sublimely subjective experience. Unless it's possible to somehow "objectively" reproduce a subjective experience, I have difficulty understanding why I should care whether my stereo system can accurately reproduce a recording.
As I mentioned above, I think this ability is a good guideline for finding a system that will produce decent sound over a wide sample of recordings, and short of listening to every song I own on a given system, it may well be the criterion I use to buy an audio system, but I don't see how this refutes the "long term" subjective impressions that I have of a system. Ultimately, my judge of a sound system's quality will be how much pleasure it gives me from the music I listen to.