source file: mills2.txt Date: Thu, 19 Dec 1996 10:09:01 -0800 Subject: Post from McLaren From: John Chalmers From: mclaren Subject: sound -- This post deals with the sound of xenharmonic recordings. As always, most of you won't like these opinions-- and, as always, many of you will claim that these statements are false, ignorant, untrue, delusional, etc. As usual, this is incorrect and yet another demonstration that most of you have ruined your hearing by listening to far too much rock music or far too many concerts of so-called serious electronic music at far too high amplification levels. (This proved a critical probem at the 1992 ICMC concerts. TA's often pumped up the volume so loud that the entire audience cringed, held their ears, and the speakers distorted.) A number of forum subscribers have claimed that most of the music we all hear is live, rather than recorded. This claim can only be explained as the result of mental illness or hard drug usage, since it is so indisputably contrary to everyday reality. "The pervasive assumption that support for new music and its composers equals support for performances...needs questioning. (..) Most music is heard at home, via records, radio and printed sheet music. Composers have been getting little support for creation or presentation of works via these media to the many people who depend on them." [Spiegel, Laurie, "A Non-Performance Viewpoint," New Music America '81 Festival program booklet, page 49, 1981] Since most music is heard at home via CD, cassette or radio, this means that the way in which microtonal music is recorded is *at least* as important as the microtonal music itself. First and most important is the issue of how the recording is mixed. The hard cold fact is that you can either mix for headphones, or for speakers--but you can't mix for both. A mix that sounds best when heard over headphones is likely to sound indistinct, with an exaggerated sound stage when heard over speakers. Contrariwise, a mix that sounds best when heard over speakers will seem too dry and too sharp-edged when heard over headphones. Alas, there is no way around this dilemma. You cannot mix so that the recorded sound is optimal on *both* headphones and speakers. Unfortunately, even the most expensive audiophile speakers sound utterly completely 100% totally different from headphones. There is no sonic comparison whatever. This becomes especially clear when you make binaural recordings using mikes placed about as far apart as the ears on a human head. A 2-mike binaural recording will sound truly startling in its realism when heard over headphones, but it will sound disastrously weird when you hear it through a pair of speakers. Headphones isolate one ear from the other; speakers blend the output of both channels in the air and throughout all reflecting surfaces in the room. The result? Speakers usually tolerate far less reverb in the mix. A subtle touch of BBE or Aphex aural enhancement will also greatly help the clarity of a mix heard through speakers. Even drastic processing like the Carver sonic holography process will sound acceptable. But the same "enhancements" will produce an unpleasant in-your-face edginess and dryness when heard over headphones. Thus, you must decide which source you're mixing for *before* you starting mastering your CD or cassette. Moreover, you must stick with that decision. Nothing sounds worse than one track of a CD optimized for listening over speakers and another optimized for listening over headphones. -- The next issue is dynamic range. Headphones take maximum advantage of a recording with a large dynamic range, while speakers cruelly punish a recording with vast dynamic range. Over speakers, much of a xenharmonic composition like William Schottstaedt's "Water Music" is either inaudible or ear-shattering...yet the same composition becomes clearly audible when heard over headphones. A recording with large dynamic range pretty much demands headphone listening (The alternative is a whisper-quiet listening environment with ultra-high-quality loudspeakers in a place where the neighbors won't protest jet-plane-takeoff sound levels. Very few listeners have such a loudspeaker environment available.). Thus, wide dynamic-range xenharmonic recordings intended to be heard over loudspeakers must, as a practical matter, employ some compression. Unfortunately, compression always changes the sound of a composition. This is especially noticeable in commercial pop music, where the maximum dynamic range is about 5 dB. In such recordings, changes in timbre take the place of changes in loudness, and this may require corresponding changes in you orchestration of a piece of music with a very wide dynamic range. One salient point about compression: it must be applied without regard for the settings of the equipment, but with concern only for whether the final result approximates the original sound of the live recording. My experience is that drastic alterations are often required to get a recording to sond like a live performance. Often, radical compression, sonic enhancement, dynamic noise reduction and gain riding are needed to produce a recording which when heard over loudspeakers approximates what you heard in live performance. Alas, loudspeakers are wildly non-linear transducers whose measured response curves bear no relation whatsoever to the sound you hear. (The reason is simple: graphs of loudspeaker response vs. frequency give data *only* on the magnitude of the frequency response, *never* on phase response at that frequency, and as Fourier's theorem tells us, both magnitude *and* phase data are required for a full reconstruction of the original signal. Equally important, loudspeaker response is measured in an anechoic chamber using a microphone which stays in a fixed position relative to the loudspeaker. In a real listening situation, each listener will hear the loudspeaker from a different position and in a different sound-absorbing and sound-reflecting environment. Thus, the graphs typically published in hi-fi magazines are meaningless as an index of the actual sound of a pair of loudspeakers. These graphs are as ludicrous as a graph of the reflectivity of 450 nm light from Rembrandt's "The Night Watch" from a photomoter fixed 5 cm from the painting's surface as an indication of the overall "look" of the entire painting.) Each loudspeaker type (acoustic reflex, ribbon, horn, electrostatic, etc.) and each placement and each listening position and set of amplifiers, recorders and preamps produces a different overall "sound." Using a vacuum tube amplifier as opposed to a solid state amp will significantly change the sound of a reproduced recording--changing the loudspeakers from, say, Klipsch horns to Magnaplanar IIIDs will transform the sound of the recording to an almost unrecognizable extent. This brings up the issue of boom boxes. Although it seems wildly insane, it remains a fact that most of your prosepctive audience will listen to the output of 50,000 dollar computer music workstations or tens of thousands of dollars of handcrafted xenharmonic instruments or a roomful of microtonal synthesizers over 40 dollar boom boxes. This is a development so weird as to border on the psychotic, yet it remains unquestionably true. Because of the rush-rush-rush nature of our society, most people listen to xenharmonic music either on a boom box or in their car. The best way to mix for this kind of peculiar playback situation is on a set of satellite loudspeakers with small cheap tweeters and the subwoofer EQ'd nearly off. This reproduces the extremely poor frequency response and dynamic range of the average boom box/car stereo. A piece of microtonal music mixed for boom box or car stereo playback must be compressed to about 20 dB dynamic range. Notice that this is not "20 dB of reduction in dynamic range," I mean that the entire range of the composition must be compressed to fit within the 20 dB of useful dynamic range available on the average cassette deck. This means that if you start with a composition which uses a dynamic range of 96 dB in 16-bit linear DAT format, you will have to compress the output of the DAT recording by at least 76 dB since (as we all know) most cassette decks produce garbage output if the recording level falls below -20 dB. Typical compressors useful for most musical material include the dbx 163x series. This kind of bizarrely extreme 76 dB compression will produce a large number of unfortunate artifacts which cannot, alas, be avoided. The alternative is to refuse to mix for cassette deck playback, which means that your recording will (to most people, whose hearing has been destroyed by excessive volume levels and who seem perfectly happy with the exercrable sound reproduction afforded by a boom box/car stereo) sound too reverberant, indistinct or "muddy," and inaudible throughout much of its length. -- The next issue to bear in mind when recording xenharmonic music is that computer music requires an entirely different mixing technique than live music. Computer music typically uses sine waves or other kinds of synthetic tones that cut through a listener's head like a knife; typical recording levels for a xenharmonic piece of computer music should be at least 20 dB lower than for live music. This is a classic case of the Fletcher-Munson curve at work, and when mixing computer music you should always ignore the meter levels. Instead, play back a live piece periodically to get a general loudness comparison and set the levels with complete disregard for the meter readings. -- Lastly, there remains the issue of Dolby encoding. As Dolby Labs has admitted, the Dolby decoding circuits of most cassette decks are badly miscalibrated at the factory, or simply left uncalibrated entirely. If you doubt this, ask yourself: when was the last time you saw an audio techician or a sound man with a Dolby B calibration tape? Try: "never." As a result, the poorly calibrated Dolby B decoding circuits in most cassette decks turn Dolby encoded recordings into mush. When poorly adjusted, Dolby B acts as a drastic low-pass filter. And since virtually all cassette decks use uncalibrated or misadjusted Dolby B encoding/decoding circuitry, this means that the best solution is always to *AVOID* Dolby encoding. Dolby C works well if the same cassette deck is used to play back the tape as was used for recording. Otherwise, Dolby C produces wretched artifacts. In effect Dolby C is so cassette-deck- specific as to prove useless in the real world. Dolby Spectral encoding is an excellent method of noise reduction which, alas, is featured on almost no casette decks because the fools at Dolby Labs overpriced the technology outrageously on its introduction. When DAT arrived, it proved far cheaper than adding Dolby S to existing 30 ips reel machines, and so Dolby S never caught on. Yet another example of "how not to introduce new technology." -- Because Dolby encoding (even when adjusted with a factory-authorized calibration tape) inevitably produces many unacceptable artifacts in the source material during playback, some hiss will be evident in all your recordings (since any sane person will avoid Dolby encoding like the plague). This means that it's vital to push the levels of the recording as high as they'll go during cassette dubbing so as to minimize noise during playback. In my experience, the most important element in a cassette dubbing chain is the source quality--preferably DAT--and after that, the cassette deck. This means that you will get far more bang out of your bucks by spending them on a good cassette deck than on a ton of signal enhancement gear. It remains a fact, alas, that Nakamichi still makes the best cassette decks around. Avoid the high-end Nakamichi decks and buy the cheapest new model available. The high-end Nakamichi cassette decks do such a freakishly good job of reproducing the execrable audio signal on the crappy cassette format that a Nakamichi Dragon (for example) will delude you into EQ'ing the high end of your recording down much lower than it should be. The best way to test your recording is to make the copy on a Nakamichi deck and play it back on a crappy K-Mart boom box, since this is the way virtually everyone will listen to your recordings. Such grotesquely poor quality playback will quickly bring to light any EQ or reverb problems with the master recording--generally you'll have to EQ the high end up or add considerable amounts of BBE or Aphex enhancement to make up for the fact that the average boom box has less bandwidth that a typical telephone. The analog compact cassette might not be the world's lowest-fidelity recording medium; it's possible that Edison wax cylinder recordings sounded worse. However, this question remains open. It is possible that the analog copmact cassette does actually the worst possible audio fidelity of any recording medium ever invented. Regardless of the outcome of that issue, it remains an obscene and outrageous fact that we will reach the 21st century with most people listening to microtonal music on ultra-lo-fi analog audio cassettes with lots of hiss, plenty of wow and flutter, and enormous amount of distortion. Thus it behooves us all to produce the best possible cassette recordings of microtonal music... a task akin to producing the best possible buggy whips we can make. Insane, senseless, bizarre, and outlandish--yet incontrovertibly necessary given the utter failure of DAT, MiniDisc or DCC to survive in the consumer marketplace as affordable viable mass recording/playback formats. --mclaren Received: from ns.ezh.nl [137.174.112.59] by vbv40.ezh.nl with SMTP-OpenVMS via TCP/IP; Thu, 19 Dec 1996 19:20 +0100 Received: by ns.ezh.nl; (5.65v3.2/1.3/10May95) id AA08559; Thu, 19 Dec 1996 19:22:23 +0100 Received: from eartha.mills.edu by ns (smtpxd); id XA08879 Received: from by eartha.mills.edu via SMTP (940816.SGI.8.6.9/930416.SGI) for id KAA24978; Thu, 19 Dec 1996 10:22:20 -0800 Date: Thu, 19 Dec 1996 10:22:20 -0800 Message-Id: Errors-To: madole@ella.mills.edu Reply-To: tuning@eartha.mills.edu Originator: tuning@eartha.mills.edu Sender: tuning@eartha.mills.edu