1. Speakers are not smooth enough in the top end.A smooth, flat top end has been a nominal goal of audio for so long that it might seem boring to mention it. But the advent of digital EQ devices like the Z-System's rdp-1 has made it possible for the first time for everybody to check out what is really going on here. Before, we could only compare different speakers or different crossover adjustments in the same speaker. Now we can make specified small changes of frequency response in a fixed speaker and see what happens. The rdp-1 will let you punch in and our peaks and dips as small as 0.2 dB and of varying widths ("Q" factor). Trying this our with pink noise is startling. The +/- 0.2 dB changes are quite obvious, especially in the region of maximum hearing sensitivity centered around 3 to 4 kHz, from 1 to 10 k, say. And the kinds of errors that tweeters, even quite good tweeters, typically make, on the order of +/- 1 dB , are gross. On music's ever-changing signal, it takes longer to hear the effects. But you'll get there. And once you have heard what music sounds like with the peaks in your tweeter massaged out by DSP, once you have heard the marvelously relaxing and beautiful sound of a truly smooth top end, you won't want to go back. The best tweeters nowadays are good, but even the best can be made a little better. And others are really in need of help, or replacement (bad ones cannot be fixed even by the DSP). We have all lived too long with abuse where our ears care the most. And +/- 1 dB is not good enough, not when +/0.2 is so easy to hear
.2. Speakers are too noisy.When a speaker has no input, it is silent, so we tend not to think of it as a source of noise in the sense that a hissing preamp is. But as soon as a speaker gets an input signal, it starts doing things it shouldn't and starts making noise, not just the music it should be making. Cones and surrounds flexing, mechanical structures vibrating, cabinets flexing in unpredicted and unpredictable ways, air flowing turbulently, electrostatic diaphragms vibrating chaotically on the scale of small areas even if they are moving regularly on a large scale, such sources of noise are everywhere. You can see all this in the chaotic tail ends of "waterfall" plots, after the big signal and the resonance ridges have decayed. You can see (and hear) it in the decay of the sound if a large signal input to the speaker is suddenly switched off. And you can see it in the "spectral noise contamination test," devised by the late Deane Jensen and Dr. Gary Sokolich, in which the input is a number of sine waves at spaced frequencies that are notched our of the measuring mike pickup signal, leaving the noise exposed as a broad-band, lower-level signal. (This test is available commercially in the Sys/ld software.) How much noise are we talking about here? A lot, a whole lot by the standards of noise levels in electronics and recording systems. Speaker noise appears only 20 to 30 dB down from signal in some cases, and even the cleanest speakers I know do not get the noise down much more than 55 dB or so. (See my review of the Mordaunt Short MS30, Issue 103, for a discussion of what happens in a good situation.) In a world where we worry about noise products in electronics 80, 90, 100 dB down, maybe we should worry a little more about the noise of speakers that is much louder than that.
3. Speakers are not flat enough in rooms from the midrange down.This is a familiar problem I have mentioned often, especially in terms of using digital-signal processing to correct it. (See my review of the SigTech, Issue 113; Accuphase, Issue 120; Tact RCS, this issue.) Still, it is shocking to measure the actual performance of systems whose owners are assuming that because their speaker is anechoically flat, it will be reasonably flat in-room. All you have to do is to look at in-room response curves to see what an illusion this usually is. Try it yourself, with warble tones and an SPL meter (even a non-calibrated inexpensive one will be sufficient to reveal the gross problems that usually occur). Remember how sensitive the ear is to response errors, and be appalled. If you can get +/-2 dB from 1 kHz on down to about 40 Hz, count yourself wildly lucky. And then remember that that is nowhere near flat enough for perfection in audible terms. Without DSP correction, it is nearly hopeless to expect reproducible high fidelity in any reasonable sense. Lest we forget.
