Sunday, November 25, 2012

Ultrasonic Speakers, Hi-Fi+, High Sample Rate Recording, The Last Flea Circus

I recently had a conversation with a college student about sound quality.  It started with him asking me some questions about vinyl records, a musical technology he mostly finds silly and impractical.  This lead to me briefly explaining the trade-offs between digital and analog sound, which lead to further discussion of digital formats and MP3 lossy-ness.

I wanted him to understand two things.  Mainly I wanted him to get that there are trade-offs.  Although the very best analog sound may be more accurate than the best digital sound[1] we usually end up also considering convenience and cost, and that's OK.  Consider whatever is important to you.  Secondly, I wanted him to understand that "better" does not always mean "better so that it matters to you."  For example, a stellar recording won't sound great through bottom-quality ear buds with shorts in the cord.  Thus, when it comes to my phone, the floor for acceptable recording quality is pretty low.

This conversation was still fresh in my mind when I saw Hi-Fi+ on a magazine rack last week and bought a copy.  Hi-Fi+ is a high-end consumer audio magazine from the UK.  It had been a long time since I read a consumer audio magazine, high-end or not.  Back in the 1980's I read Sound & Vision's predecessor, Stereo Review, from time to time.  I lost interest as the market shifted from music to home theater.  Around this time I also realized that even if I had money one day my future was probably full of factory installed car stereos and cheap headphones.  As a music lover and musician I cared about sound but doubted I would ever be an audiophile[2].  I was right.  All these years later I have a little money and I still don't own good headphones.  I have spent a hell-of-a-lot more money on equipment for making and recording music than for listening to it.

Through my misadventures as a musician I have come to know a lot more about sound than I did back when I was more interested in "good" stereo equipment.  I took two recording techniques classes in college and when I returned to electronic music production several years ago I started reading Mix, Electronic Musician, Sound On Sound and other production oriented magazines.  I also do extensive research before I buy gear.  I don't claim to be an expert but I am better informed than the average musician and this knowledge does provide context for all sorts of musical and technical decisions, not to mention nutty things other people write about audio equipment.

Speaking of which, my jaw dropped at some of what I read in Hi-Fi+.  The issue I purchased is mostly product reviews, some of them written by people who know precious little about where recorded music comes from.  Some of the language used to describe the expensive gear in Hi-Fi+ is similar to pro audio reviews I have read.  There are lots of references to "transparency" and descriptions of impossible-to-measure qualities of the sound stage and stereo image.  That was all fine.  Then I got to the claims about ultrasonics.

This passage is from a speaker review and follows the statement that the speaker's upper range goes to 40KHz (20KHz above human hearing but lower than some other high-end speakers). (sic) Indicates typos in the original. Also note […].
"These figures might not sound all that impressive but they are crucial if you want to(sic) the speaker to be able to produce natural sounding high frequencies.  Even though these break-up points are beyond the human hearing range what happens to the sound above this point has an effect on the audio band, much like they(sic) way that higher sampling rates give greater treble extension which results in cleaner midband. ... With high-res music files, we have material that theoretically goes up to 96Khz and anyone who has heard a good example of this on a revealing system will know that it sounds more open and relaxed than CD ever did." - Jason Kennedy, Equipment Review: Bowers & Wilkins PM1 Loudspeakers, Hi-Fi+, Issue 92, p 64.

It was timely that I picked up Hi-Fi+ around the same time I read one of Dan Lavry's white papers about sample rates.  Lavry argues that including sound data well above the range of human hearing actually causes distortion in the audible band.  Not everyone agrees with Lavry but he convincingly makes the point that if you record faster that 96KHz most of the additional data you are collecting isn't music.  For those of you not familiar with the relationship of bandwidth to sample rate, a recording at a given sample rate can capture sounds at half the frequency of sampling.  For example a 96KHz recording can capture a 48KHz "sound."  To reach 96KHz you need a 192KHz recording.  At 44.1KHz standard CD audio just clears the range of human hearing by topping out at 22.05KHz.

There is evidence that although we do not hear ultrasonic "sounds" (AKA "ultrasonics" or "ultrasound") we do perceive them and our brain processes them.  Google "ultrasonic hearing" etc. to check the research.  It can also be empirically shown that musical instruments create overtones in the ultrasonic range.  Thus, it is probably true that one of the experiential differences between being in the room with a pianist and listening to a 44.1KHz recording of the same performance will be the missing ultrasonic overtones.  That said, none of my reading or personal experience leads me to believe it is something that can be mitigated with ultrasonic treble extension at the end of the playback chain.

Remember what I said in the beginning about listening to a great recording on crummy ear buds?  The same thing is true about all points in the audio chain.  Although some microphones can capture sounds in the ultrasonic range that is not what they are designed for nor what they excel at.  How about studio mixers?  The same question can be asked at every point in the recording and playback chains from microphone to speaker.  If at any point a digital device drops to CD sample rate or an analog device filters out sound above 20KHz the ultrasonic overtones of the piano in our example are gone forever.

To his credit the reviewer above concedes (however indirectly) that if the recording in question is a CD the discussion of ultrasonics is over before it starts.  Not only is the 16bit 44.1KHz CD audio format incapable of faithfully reproducing sounds above 22.05KHz, the analog to digital conversion process for 16/44.1 includes clamping a hard low-pass filter on the sound to prevent sampling artifacts.  All standard A/D converters do this but for CD quality audio the filter typically starts near where human hearing stops.  This guarantees the almost[3] complete absence of ultrasound from the content of a CD.  Google "anti-aliasing filter" and "Nyquist frequency" for exactly why.

This leads to another question.  Assuming nobody is actually lying, what are the people who claim to "hear" the benefits of ultrasonics hearing?  Is it all a musical flea circus[4]?  Are they imagining it?  Are they hearing something other than music?  My guess--yes.  They are hearing noise and their brains register it as part of the music.

As I have more than hinted at, I think that the odds of ultrasonic musical content reaching our ears through commercially available recordings are minuscule.  If it does happen I suspect it is very quiet and the quality is terrible.  If it gets recorded (BIG "if") and survives the mixing process would it get past the mastering engineer? I don't want to belabor this point but I think it is worth thinking about how many barriers there are to preserving the ultrasonic overtones of a recorded musical performance.

Let's say our friends with really expensive speakers really are "hearing" ultrasound, perhaps, as some of the research indicates, transduced by the skull.  There are just as many chances for ultrasonic noise and distortion to be added to a recording as to strip out musical ultrasound.  In fact, it can happen at any stage where harmonic distortion occurs...and that's pretty much everywhere.  Every amplification, conversion, filtering or other processing point in recording, production, mastering, duplication, distribution and playback has the potential to introduce harmonic distortion[5].  

As when overdriving a guitar amp, harmonic distortion means adding overtones to the sound.  Although they may be subsequently removed by any of the means that would remove musical overtones there is no reason harmonic distortion would be limited to the audible range.  If it is added near the end of the signal path there are fewer opportunities for it to be removed.  In this case, one of the best things an audiophile could do is pair their ultrasonic-extension speakers with a noisy power-amp, provided the noise is above 20KHz.

*savoring the irony*

To return to the example of A/D conversion and the preparations for creating a CD, think about the introduction of dither, which turns quantization distortion into benign noise[6].  Perhaps reproducing inaudible noise (either present in high bandwidth recordings or generated by the playback system) partially recreates the experience of live performance.  It is implausible that it does so with any fidelity to the original performance, but like dither, sometimes a little noise helps by filling in an sonic gap with SOMETHING, even if it isn't what was there originally.  Specifically, in this case, it helps people justify spending many thousands of dollars on exclusive, high-end speakers.

It is theoretically possible that some day we will have commercially available recordings that capture ultrasonic energy that can be faithfully reproduced on our home audio equipment.  I don't think it would be worth the trouble, but most (if not all) of the technology already exists.  The problem, as I see it, is that people who don't understand the production process are telling consumers to buy equipment for it now, when few, if any, such recordings exist.  Most music is currently purchased in CD quality or similar, so at best people who buy such hardware are future-proofing their systems while getting little benefit now.

Unless, of course, they really enjoy very quiet ultrasonic noise.

[1] Notice my use of "may be."  I'm not arguing that this is currently true in practice.
[2] In that I would never bother to invest in audiophile components for listening to music at home. Using the literal definition, I do love sound.
[3] These filters vary, but they often start at or near 20,050Hz and slam down to near total rejection by 24,000.  Technically, it is allowing some ultrasound through, but precious little.
[4] There may or may not be real fleas under the glass but even if there are they aren't doing what the carnival barker says they are.  If you believe it is because you want to.
[5] And most of them really do, which is why short, simple signal chains are such a good thing. The good news is that many HD contributions are trivially small.
[6] The anti-aliasing filter reduces aliasing in the analog to digital conversion process.  Dither converts it to a less problematic type of sound later in the process.

No comments: