We Want Really Accurate Recordings, Right? Or Do We?

We've been ranting about high-resolution recording for lo, these many months. Nobody has questioned the desirability of high resolution. We all agree that we need such resolution in order to make accurate recordings. All I've been quibbling about is just how much resolution we really need.

But the assumption about accuracy needs to be examined too (is there nothing sacred?). Do we really need to make accurate recordings? Is that our goal? Is the most accurate recording the best-sounding one? If you read the ads, it sure seems like it. If you listen to all the audio pros and audiophiles holding forth, accuracy seems to be the Holy Grail. Why, if we could just get the recording to be accurate enough, we wouldn't be able to tell it from the real thing! Right? Well, ain't I right? Accuracy rules!


The dictionary defines accuracy as: "Conforming exactly to fact. Errorless." In recording, this is a no-brainer. The output from the loudspeaker should conform exactly to the sound detected at the microphone. Or, if we wanna be bold and visionary, what we perceive at our ears should be exactly what we would have perceived had we been where the microphone was when the recording was made. Now that's accurate! At least it seems that way.

Unfortunately, when we are dealing with high-resolution subjective detection systems like human hearing, this isn't easy to do. Even with dummy heads and in-ear headphones, it just doesn't sound exactly the same. Dang! Life ain't easy.

But even when we stay in the objective realm, we've got troubles. If we can't even get two loudspeakers to sound exactly the same, how are we gonna get either of them to sound anywhere close to a Martin guitar, much less both? How about a Steinway grand? If you're really looking for a goal, how about getting a loudspeaker to sound exactly like a symphony orchestra? Like Mom used to say, we aren’t there yet.


To make a truly accurate recording, we first have to define the recording parameters. Then we have to accurately record them. The primary parameters are amplitude, frequency and time. So, we buy microphones and loudspeakers that can accurately track amplitude changes over time, which also gives us frequency. Sounds reasonable, right?

But what do we mean by "accurately track"? Aside from the obvious issue that microphones aren't ears, the output of the microphone is a different energy medium than air. What electrical level most "accurately" represents a given air pressure?

It doesn't seem to matter very much, and there are lots of other design concerns that seem to drive that decision. So we've grown casual. We simply don't record the absolute sound pressure level. We only record relative levels, the relative changing of amplitude over time. Uh-oh! Ditto for time. We don't record the absolute time something happened, only a relative time sequence.

So, right out of the box, we've got major troubles with accuracy. We have no record of the actual sound pressure level we recorded and we have no record of when we recorded it! How bad is this? You can argue, of course, that the absolute time of the recording doesn't matter very much so long as we've got the relative time(s) of the recording correct.

Not so for amplitude. Thanks to the non-linearity of our hearing, when we change the absolute level of a sound, we change the EQ of our hearing (as described by the equal loudness contours). No recording can be even remotely "accurate" if it is played back at some other sound pressure level than the sound pressure level at which it was recorded.


The implications of this are a little scary, when you think about it, and they may cast the notion of accuracy in a whole new light for you — they did for me when I first thought about this seriously!

To make an accurate recording, we would need, at the very least, to record the absolute sound pressure level of the recording and also the absolute time at which we made it! That's right, to be accurate, we'd need to know that we recorded 103.6 dB SPL at 7:42:41.007 p.m., Wednesday, 3/15/00, the Ides of March.

Now, if you'd like a really accurate playback, you'd need to play back exactly 103.6 dB SPL, preferably at 7:42:41 PM, on a Wednesday, or perhaps on the Ides of March.

Admittedly, my playback absolute time requirement may be a little … well, over the top, but the SPL requirement sure isn't. And while absolute time may be hard to reproduce meaningfully, relative time — that fixed sequence of events that you recorded starting at, say, 7:38:06.497 p.m., Wednesday, 3/15/00, the Ides of March — should certainly be maintained as a sequence, if yer gonna be accurate about it.

What I'm getting at, as you've probably guessed, is that if we want to make really accurate recordings, we've got to give up level control and editing, at the very least. All level changes and reordering of time constitute major errors in the record/playback process. Forget about mixing. Forget about EQ, compression, reverb, panning. Forget about post production! There isn't any when you’re trying to be accurate!


It gets worse. There are other things that we should probably record in both absolute and relative terms, like spectra, absolute and relative direction, position of instruments and microphones in the room, temperature and humidity, to name a few. All these, in both absolute and relative terms, may have audible impacts on the perception of the recording. So, we should probably also concern ourselves with them during playback. This recording business is tough! I shudder to think about recording/playback accuracy in the visual realm. Gonna really be hard!

Naturally, this is all pretty silly. Fact is, we're not interested in accuracy at all. We're interested in illusion, the willing suspension of disbelief, under as wide a range of playback conditions as possible. "Accuracy" is our code word for "sounds good" or "sounds real." So when the sales weasel sidles up to you and murmurs, "Ya know, this speaker is really accurate," know that he wants you to think it sounds really cool.

When the audiophile starts ranting about the "clarity" of single-ended tube amps, keep in mind that he's actually promoting a big-time accumulation of errors. When the techno-dweeb Ambisonista starts talking about wavefront reconstruction, keep in mind that he's barely begun to address accuracy.

My suggestion? Forget accuracy! Concentrate on power of illusion!! That's where it's at, and where it will be for a long, long time. 'Nough said!

Thanks for, ah, listening.

Dave Moulton