Noise-Related Audio Performance
Four of our standard audio measurements are so closely related I’ll discuss
them together. Our noise-level, excess-noise, noise-modulation, and linearity-error
tests combine to make things difficult for a player. It’s hard to perform
well on all four tests simultaneously, and if a player does poorly in one of
these tests, at least one of the others will usually show nonideal behavior
as well.
The most conventional and easiest to understand is noise level, which measures
how far the player’s inherent background noise (consisting of hiss as well
as hum and buzzes from the power supply) falls below our standard reference
level, which is not full output but 20 dB below that (–20 dBFS). All of
our noise-level measurements use A-weighting, which (approximately) corrects
for how the ear’s sensitivity to low-level sounds varies with frequency.
We also use a special “dither” signal (random extremely low-level
noise) for this test, not an all-zeroes data stream as in other methods of measuring
noise, so the figure representing theoretically perfect performance is not minus
infinity (–•) but is instead set by the resolution of the medium.
The resulting noise-level figure should be as close as possible to the theoretical
minimum without going lower. For 16-bit CD audio, perfect performance is –75.9
dB. For DVD-Audio using 24-bit data, perfect performance would be 48.16 dB lower,
or –124.1 dB. I don’t expect any DVD-Audio player ever to measure
that low, however. Most players will be closer to 20-bit performance, or -100.0
dB, at which level a perfect DVD-Audio recording should have absolutely no audible
background hiss at any reasonable listening volume.
Excess-noise figures explicitly measure a player’s departure from an ideal
level of background noise. For a CD player, this test — which so far is
exclusive to Sound & Vision — quantifies how much higher (or lower)
than perfect the player’s output level is with four types of signal: 16-bit
(EN16) and quasi-20-bit (EN20) dither, both with and without a low-level sine-wave
signal superimposed, in a frequency band that’s squarely within the ear’s
region of greatest sensitivity. Ideally, all excess-noise figures should be
0 dB, meaning that the player adds no noise or distortion to the dither signal
in the frequency band we look at. In practice, we’ve seen 0-dB EN16 and
single-digit EN20 performance with CDs played on the latest DVD-Audio gear.
If the results aren’t exactly zero, then the “with sine tone”
figures should be within 1 dB of the “without sine tone” figures.
When it comes to DVD-Audio playback, the excess-noise figure represents, again,
the amount of noise added to a theoretically perfect 24-bit signal. I don’t
expect ever to see 0-dB EN24 results. Probably the best that can be expected
is approximately 20-bit performance, or excess noise around +24.1 dB. As long
as a player comes close to that figure, its inherent noise is unlikely to ever
be audible.
Noise modulation measures how much the background-noise and distortion levels
change as the signal level changes over a very wide range. Ideally, it should
be 0 dB. Typical, and excellent, values for CD playback are less than 0.5 dB,
with noise modulation figures as great as 3 dB usually being of no audible consequence.
Linearity error measures the level difference, plus or minus, between a –90-dB
signal on our test CD-R and the player’s analog output as it plays that
track. Such very low-level recorded signals often prove difficult to reproduce
accurately, but errors are still usually much less than a decibel (zero error
is ideal). A relatively large error (greater than 1 dB) often shows up also
as noise modulation (which can be generated by the same mechanisms as a negative
linearity error) or higher-than-theoretical noise or excess-noise figures (a
positive error).
Ranking the Readings
Given the generally high level of performance we’ve been seeing from recent
DVD players, not all the measurements we’ve been making are now of equal
importance for your buying decision. You can usually skim over the audio frequency
response and distortion measurements. All of these are unlikely to be less than
very good. The notes at the end of “in the lab” will mention anything
unusual.
Some other measurements might be irrelevant for a different reason, not because
every player is equally good but because they don’t apply to your viewing
situation. If your TV doesn’t have component-video connections, the two
component-output measurements have no bearing. In-player letterboxing is very
important if you’re watching on a 4:3 screen, irrelevant with a widescreen
16:9 set.
If I had to pare down the lab-test figures to the minimum that could be decisive
in choosing one player over another, I’d rely on luminance frequency response
and in-player letterboxing for video. On the audio side, the excess noise and
noise modulation results should hold sway because, of all our audio tests, they’re
the hardest things to get right.
Of course, such a minimalist view assumes that your mind hasn’t already
been made up by some other, more subjective aspects of the players’ performance
— things we can’t measure in the lab but that our reviewers always
discuss in detail in their reports. How easy it is to use the remote control,
for instance, or to find what you need in the onscreen menus are things that
should weigh just as heavily in your purchase decision as the most important
lab results, and maybe even more so.
