Must We Test? Yes, We Must! Page 3

Looking first at the dynamic range of real music, Mr. Fielder experimented with the lowest possible sound level people could hear, using gated white noise. It turns out to be around 4dB spl, well below typical background noise levels of 20-40dB spl. He then both researched the literature and carried out his own measurements on the peak levels featured by live music. Typically this appears to range from around 112dB peak for classical music to 129dB peak for rock, which implies that for domestic playback, a dynamic range of 98dB is necessary, while for professional playback a 122dB (!) range will be necessary.

The next part of the research involved examining the human ear's acuity. Although Clark Johnsen strongly disagrees in this month's "Letters" column, Fielder is convinced that the ear/brain does act as a multi-band spectrum analyzer, with the audio band split up into 25 distinct, different-sized regions, these the so-called "critical" bands. A sound with all of its energy in one critical band will increasingly fail to mask sounds in critical bands whose center frequencies are increasingly removed from the first. This is why hum is so irritating with chamber music recordings: the music has very little content in the critical band containing the hum, so utterly fails to mask it. Combining the concepts of critical bands and masking with the measured sensitivity of the ear across the whole band gives the primary test of whether distortion will be audible or not in the presence of the fundamental signal.

To consider an example given in Fielder's paper (his fig.23), take a consumer playback system with a maximum sinewave output level of 102dB spl (equivalent to 105dB peak), which is not untypical of the full-range high-end systems I have experienced. Play a 500Hz tone at a level of -60dB reference to 102dB—ie, at 52dB spl—and examine the audibility of its harmonic components with respect to the criteria outlined in the above paragraph. Relatively high levels of low-order harmonics will nevertheless be inaudible, due to the combination of the human threshold of hearing and masking by the primary signal. Even 1% of second harmonic will not be heard.

However, because the human ear's acuity increases to a maximum between 2.5kHz and 5kHz and because any signal content in this region lies in critical bands well away from the one containing the 500Hz signal, even very low amounts of spuriae will be audible. The seventh harmonic at 3500Hz, for example of a tone already 60dB down from the loudest level, must be at least another 50dB down in level; ie, lower than -110dB, if it is not to be audible. And if the playback system has a varying noise background rather than a steady one, as CD players are prone to have, then any content in this region must also be -110dB with respect to maximum level if the noise is not to be heard.

Consider a typical CD player. Yes, it has vanishingly low distortion at full level, and in any case that distortion will be very effectively masked by the high fundamental level. But for lower midrange tones at -60dB, this is certainly not the case. By Fielder's criteria, even if the player's distortion lies 40dB below the fundamental, which is typical, as it is comprised of predominantly high-order harmonics, it will be audible nevertheless and the player is inadequate.

If you have ever wondered why a tube amplifier with relatively high amounts of distortion at high levels sounds less distorted than a CD player with no distortion at high levels, then there is the reason why. The amplifier distortion is both masked through most of its dynamic range window and drops with level, while the CD player introduces more and more of the most audible, least masked type of distortion as the level drops to regions where, traditionally, it is held that such distortion is inconsequential.

In view of the above discussion, you may find it surprising that I intend to increase the amount of traditional measurements carried out in support of Stereophile's fundamentally subjective equipment reports. In recent months, we have spent a five-figure sum on test equipment, including a complete Audio Precision two-channel measurement system, and hired a Technical Editor, Robert Harley, to run our new laboratory. (Turn to this month's "Industry Update" for Bob's discussion on the vexing subject of CD Soundrings.) Bob will routinely measure every component submitted to the magazine for review, taking into account the most up-to-date research (like Mr. Fielder's) we can find.

Unlike the "slicks," however, where measurements are supposed to define or even replace the experience, Bob's work will be in order to: a) show up "dogs" before we send them out to the selected reviewer, b) build up a measurement database that will eventually throw up correlations with what is heard, and c) ensure that there isn't some simple reason for a component to sound the way it does. (A preamp with a rising phono response will always sound bright, for example, and spending a lot of money on a component with a built-in tone control is not my idea of a good deal.) I have to also admit—look at that, you can split an infinitive without even thinking about it—that I love the Great Detective side of the reviewing process, as typified by the measurements in my review of the MartinLogan Sequel II in this issue:

Watson: [For it is he.] Dash it, Holmes, there's no obvious reason for the Sequel to sound this way—but it does!

Holmes: (footnote 8) [Deep in thought: he and the woman, Mrs. Irene Adler, are committing the ultimate crime in order to finance purchase of a WAMM system on the Audio Anarchist's recommendation in this month's column.] Hah! [Quick half-smile to stage left—women in the audience faint.]

It just so happens, Watson, that I wrote a monograph on this very subject. Investigating the impulse response of the electrostatic panel and the woofer individually will make everything clear. [Frowns quixotically.] Precisely. [Men in the audience rise to their feet, cheering.]

I deduce that Professor Moriarty—neé Harry Pearson?—is going to be a very unhappy man. [Sucks hard on cigarette; reaches for violin, &c...]

Perhaps that particular wrinkle in the technique didn't make it absolutely clear why the Sequel II should sound the way it does—Holmes was lucky in having a handy deus ex machina present in the form of Conan Doyle who could usefully simplify the world—but measurements do have their role. They provide an underlying framework for a report based on the effect of a component on the music, though what is heard will always be more important than what is measured.

"We don't believe in magic," said Michael Riggs back in January. But with the exception of the work being carried out by Louis Fielder and his ilk, the current state of sound reproduction does remind me of that famous dictum by science-fiction author Arthur C. Clarke: "To a primitive people, any sufficiently advanced technology is indistinguishable from magic." Measurements, yes, "properly" performed but, more importantly, correctly conceived, will lift us above the level of the primitive. Reproduced music will not then be magic; it will just sound magic.



Footnote 8: As played by Jeremy Brett, not by Basil Rathbone. Did I ever tell you that Mr. Brett once waved at my wife? No? Maybe one day...

COMMENTS
MattJ's picture

I like the idea of using dB for power ratings. Benchmark has similar ideas to use dB for distortion tests.

https://benchmarkmedia.com/blogs/application_notes/interpreting-thd-measurements-think-db-not-percent

X