A place to cache linked articles (think custom and personal wayback machine)
選択できるのは25トピックまでです。 トピックは、先頭が英数字で、英数字とダッシュ('-')を使用した35文字以内のものにしてください。

index.html 75KB

  1. <!doctype html><!-- This is a valid HTML5 document. -->
  2. <!-- Screen readers, SEO, extensions and so on. -->
  3. <html lang=fr>
  4. <!-- Has to be within the first 1024 bytes, hence before the <title>
  5. See: https://www.w3.org/TR/2012/CR-html5-20121217/document-metadata.html#charset -->
  6. <meta charset=utf-8>
  7. <!-- Why no `X-UA-Compatible` meta: https://stackoverflow.com/a/6771584 -->
  8. <!-- The viewport meta is quite crowded and we are responsible for that.
  9. See: https://codepen.io/tigt/post/meta-viewport-for-2015 -->
  10. <meta name=viewport content="width=device-width,minimum-scale=1,initial-scale=1,shrink-to-fit=no">
  11. <!-- Required to make a valid HTML5 document. -->
  12. <title>24/192 Music Downloads (archive) — David Larlet</title>
  13. <!-- Generated from https://realfavicongenerator.net/ such a mess. -->
  14. <link rel="apple-touch-icon" sizes="180x180" href="/static/david/icons/apple-touch-icon.png">
  15. <link rel="icon" type="image/png" sizes="32x32" href="/static/david/icons/favicon-32x32.png">
  16. <link rel="icon" type="image/png" sizes="16x16" href="/static/david/icons/favicon-16x16.png">
  17. <link rel="manifest" href="/manifest.json">
  18. <link rel="mask-icon" href="/static/david/icons/safari-pinned-tab.svg" color="#5bbad5">
  19. <link rel="shortcut icon" href="/static/david/icons/favicon.ico">
  20. <meta name="apple-mobile-web-app-title" content="David Larlet">
  21. <meta name="application-name" content="David Larlet">
  22. <meta name="msapplication-TileColor" content="#da532c">
  23. <meta name="msapplication-config" content="/static/david/icons/browserconfig.xml">
  24. <meta name="theme-color" content="#f0f0ea">
  25. <!-- That good ol' feed, subscribe :p. -->
  26. <link rel=alternate type="application/atom+xml" title=Feed href="/david/log/">
  27. <meta name="robots" content="noindex, nofollow">
  28. <meta content="origin-when-cross-origin" name="referrer">
  29. <!-- Canonical URL for SEO purposes -->
  30. <link rel="canonical" href="https://www.xiph.org/~xiphmont/demo/neil-young.html">
  31. <style>
  32. /* http://meyerweb.com/eric/tools/css/reset/ */
  33. html, body, div, span,
  34. h1, h2, h3, h4, h5, h6, p, blockquote, pre,
  35. a, abbr, address, big, cite, code,
  36. del, dfn, em, img, ins,
  37. small, strike, strong, tt, var,
  38. dl, dt, dd, ol, ul, li,
  39. fieldset, form, label, legend,
  40. table, caption, tbody, tfoot, thead, tr, th, td,
  41. article, aside, canvas, details, embed,
  42. figure, figcaption, footer, header, hgroup,
  43. menu, nav, output, ruby, section, summary,
  44. time, mark, audio, video {
  45. margin: 0;
  46. padding: 0;
  47. border: 0;
  48. font-size: 100%;
  49. font: inherit;
  50. vertical-align: baseline;
  51. }
  52. /* HTML5 display-role reset for older browsers */
  53. article, aside, details, figcaption, figure,
  54. footer, header, hgroup, menu, nav, section { display: block; }
  55. body { line-height: 1; }
  56. blockquote, q { quotes: none; }
  57. blockquote:before, blockquote:after,
  58. q:before, q:after {
  59. content: '';
  60. content: none;
  61. }
  62. table {
  63. border-collapse: collapse;
  64. border-spacing: 0;
  65. }
  66. /* http://practicaltypography.com/equity.html */
  67. /* https://calendar.perfplanet.com/2016/no-font-face-bulletproof-syntax/ */
  68. /* https://www.filamentgroup.com/lab/js-web-fonts.html */
  69. @font-face {
  70. font-family: 'EquityTextB';
  71. src: url('/static/david/css/fonts/Equity-Text-B-Regular-webfont.woff2') format('woff2'),
  72. url('/static/david/css/fonts/Equity-Text-B-Regular-webfont.woff') format('woff');
  73. font-weight: 300;
  74. font-style: normal;
  75. font-display: swap;
  76. }
  77. @font-face {
  78. font-family: 'EquityTextB';
  79. src: url('/static/david/css/fonts/Equity-Text-B-Italic-webfont.woff2') format('woff2'),
  80. url('/static/david/css/fonts/Equity-Text-B-Italic-webfont.woff') format('woff');
  81. font-weight: 300;
  82. font-style: italic;
  83. font-display: swap;
  84. }
  85. @font-face {
  86. font-family: 'EquityTextB';
  87. src: url('/static/david/css/fonts/Equity-Text-B-Bold-webfont.woff2') format('woff2'),
  88. url('/static/david/css/fonts/Equity-Text-B-Bold-webfont.woff') format('woff');
  89. font-weight: 700;
  90. font-style: normal;
  91. font-display: swap;
  92. }
  93. @font-face {
  94. font-family: 'ConcourseT3';
  95. src: url('/static/david/css/fonts/concourse_t3_regular-webfont-20190806.woff2') format('woff2'),
  96. url('/static/david/css/fonts/concourse_t3_regular-webfont-20190806.woff') format('woff');
  97. font-weight: 300;
  98. font-style: normal;
  99. font-display: swap;
  100. }
  101. /* http://practice.typekit.com/lesson/caring-about-opentype-features/ */
  102. body {
  103. /* http://www.cssfontstack.com/ Palatino 99% Win 86% Mac */
  104. font-family: "EquityTextB", Palatino, serif;
  105. background-color: #f0f0ea;
  106. color: #07486c;
  107. font-kerning: normal;
  108. -moz-osx-font-smoothing: grayscale;
  109. -webkit-font-smoothing: subpixel-antialiased;
  110. text-rendering: optimizeLegibility;
  111. font-variant-ligatures: common-ligatures contextual;
  112. font-feature-settings: "kern", "liga", "clig", "calt";
  113. }
  114. pre, code, kbd, samp, var, tt {
  115. font-family: 'TriplicateT4c', monospace;
  116. }
  117. em {
  118. font-style: italic;
  119. color: #323a45;
  120. }
  121. strong {
  122. font-weight: bold;
  123. color: black;
  124. }
  125. nav {
  126. background-color: #323a45;
  127. color: #f0f0ea;
  128. display: flex;
  129. justify-content: space-around;
  130. padding: 1rem .5rem;
  131. }
  132. nav:last-child {
  133. border-bottom: 1vh solid #2d7474;
  134. }
  135. nav a {
  136. color: #f0f0ea;
  137. }
  138. nav abbr {
  139. border-bottom: 1px dotted white;
  140. }
  141. h1 {
  142. border-top: 1vh solid #2d7474;
  143. border-bottom: .2vh dotted #2d7474;
  144. background-color: #e3e1e1;
  145. color: #323a45;
  146. text-align: center;
  147. padding: 5rem 0 4rem 0;
  148. width: 100%;
  149. font-family: 'ConcourseT3';
  150. display: flex;
  151. flex-direction: column;
  152. }
  153. h1.single {
  154. padding-bottom: 10rem;
  155. }
  156. h1 span {
  157. position: absolute;
  158. top: 1vh;
  159. left: 20%;
  160. line-height: 0;
  161. }
  162. h1 span a {
  163. line-height: 1.7;
  164. padding: 1rem 1.2rem .6rem 1.2rem;
  165. border-radius: 0 0 6% 6%;
  166. background: #2d7474;
  167. font-size: 1.3rem;
  168. color: white;
  169. text-decoration: none;
  170. }
  171. h2 {
  172. margin: 4rem 0 1rem;
  173. border-top: .2vh solid #2d7474;
  174. padding-top: 1vh;
  175. }
  176. h3 {
  177. text-align: center;
  178. margin: 3rem 0 .75em;
  179. }
  180. hr {
  181. height: .4rem;
  182. width: .4rem;
  183. border-radius: .4rem;
  184. background: #07486c;
  185. margin: 2.5rem auto;
  186. }
  187. time {
  188. display: bloc;
  189. margin-left: 0 !important;
  190. }
  191. ul, ol {
  192. margin: 2rem;
  193. }
  194. ul {
  195. list-style-type: square;
  196. }
  197. a {
  198. text-decoration-skip-ink: auto;
  199. text-decoration-thickness: 0.05em;
  200. text-underline-offset: 0.09em;
  201. }
  202. article {
  203. max-width: 50rem;
  204. display: flex;
  205. flex-direction: column;
  206. margin: 2rem auto;
  207. }
  208. article.single {
  209. border-top: .2vh dotted #2d7474;
  210. margin: -6rem auto 1rem auto;
  211. background: #f0f0ea;
  212. padding: 2rem;
  213. }
  214. article p:last-child {
  215. margin-bottom: 1rem;
  216. }
  217. p {
  218. padding: 0 .5rem;
  219. margin-left: 3rem;
  220. }
  221. p + p,
  222. figure + p {
  223. margin-top: 2rem;
  224. }
  225. blockquote {
  226. background-color: #e3e1e1;
  227. border-left: .5vw solid #2d7474;
  228. display: flex;
  229. flex-direction: column;
  230. align-items: center;
  231. padding: 1rem;
  232. margin: 1.5rem;
  233. }
  234. blockquote cite {
  235. font-style: italic;
  236. }
  237. blockquote p {
  238. margin-left: 0;
  239. }
  240. figure {
  241. border-top: .2vh solid #2d7474;
  242. background-color: #e3e1e1;
  243. text-align: center;
  244. padding: 1.5rem 0;
  245. margin: 1rem 0 0;
  246. font-size: 1.5rem;
  247. width: 100%;
  248. }
  249. figure img {
  250. max-width: 250px;
  251. max-height: 250px;
  252. border: .5vw solid #323a45;
  253. padding: 1px;
  254. }
  255. figcaption {
  256. padding: 1rem;
  257. line-height: 1.4;
  258. }
  259. aside {
  260. display: flex;
  261. flex-direction: column;
  262. background-color: #e3e1e1;
  263. padding: 1rem 0;
  264. border-bottom: .2vh solid #07486c;
  265. }
  266. aside p {
  267. max-width: 50rem;
  268. margin: 0 auto;
  269. }
  270. /* https://fvsch.com/code/css-locks/ */
  271. p, li, pre, code, kbd, samp, var, tt, time, details, figcaption {
  272. font-size: 1rem;
  273. line-height: calc( 1.5em + 0.2 * 1rem );
  274. }
  275. h1 {
  276. font-size: 1.9rem;
  277. line-height: calc( 1.2em + 0.2 * 1rem );
  278. }
  279. h2 {
  280. font-size: 1.6rem;
  281. line-height: calc( 1.3em + 0.2 * 1rem );
  282. }
  283. h3 {
  284. font-size: 1.35rem;
  285. line-height: calc( 1.4em + 0.2 * 1rem );
  286. }
  287. @media (min-width: 20em) {
  288. /* The (100vw - 20rem) / (50 - 20) part
  289. resolves to 0-1rem, depending on the
  290. viewport width (between 20em and 50em). */
  291. p, li, pre, code, kbd, samp, var, tt, time, details, figcaption {
  292. font-size: calc( 1rem + .6 * (100vw - 20rem) / (50 - 20) );
  293. line-height: calc( 1.5em + 0.2 * (100vw - 50rem) / (20 - 50) );
  294. margin-left: 0;
  295. }
  296. h1 {
  297. font-size: calc( 1.9rem + 1.5 * (100vw - 20rem) / (50 - 20) );
  298. line-height: calc( 1.2em + 0.2 * (100vw - 50rem) / (20 - 50) );
  299. }
  300. h2 {
  301. font-size: calc( 1.5rem + 1.5 * (100vw - 20rem) / (50 - 20) );
  302. line-height: calc( 1.3em + 0.2 * (100vw - 50rem) / (20 - 50) );
  303. }
  304. h3 {
  305. font-size: calc( 1.35rem + 1.5 * (100vw - 20rem) / (50 - 20) );
  306. line-height: calc( 1.4em + 0.2 * (100vw - 50rem) / (20 - 50) );
  307. }
  308. }
  309. @media (min-width: 50em) {
  310. /* The right part of the addition *must* be a
  311. rem value. In this example we *could* change
  312. the whole declaration to font-size:2.5rem,
  313. but if our baseline value was not expressed
  314. in rem we would have to use calc. */
  315. p, li, pre, code, kbd, samp, var, tt, time, details, figcaption {
  316. font-size: calc( 1rem + .6 * 1rem );
  317. line-height: 1.5em;
  318. }
  319. p, li, pre, details {
  320. margin-left: 3rem;
  321. }
  322. h1 {
  323. font-size: calc( 1.9rem + 1.5 * 1rem );
  324. line-height: 1.2em;
  325. }
  326. h2 {
  327. font-size: calc( 1.5rem + 1.5 * 1rem );
  328. line-height: 1.3em;
  329. }
  330. h3 {
  331. font-size: calc( 1.35rem + 1.5 * 1rem );
  332. line-height: 1.4em;
  333. }
  334. figure img {
  335. max-width: 500px;
  336. max-height: 500px;
  337. }
  338. }
  339. figure.unsquared {
  340. margin-bottom: 1.5rem;
  341. }
  342. figure.unsquared img {
  343. height: inherit;
  344. }
  345. @media print {
  346. body { font-size: 100%; }
  347. a:after { content: " (" attr(href) ")"; }
  348. a, a:link, a:visited, a:after {
  349. text-decoration: underline;
  350. text-shadow: none !important;
  351. background-image: none !important;
  352. background: white;
  353. color: black;
  354. }
  355. abbr[title] { border-bottom: 0; }
  356. abbr[title]:after { content: " (" attr(title) ")"; }
  357. img { page-break-inside: avoid; }
  358. @page { margin: 2cm .5cm; }
  359. h1, h2, h3 { page-break-after: avoid; }
  360. p3 { orphans: 3; widows: 3; }
  361. img {
  362. max-width: 250px !important;
  363. max-height: 250px !important;
  364. }
  365. nav, aside { display: none; }
  366. }
  367. ul.with_columns {
  368. column-count: 1;
  369. }
  370. @media (min-width: 20em) {
  371. ul.with_columns {
  372. column-count: 2;
  373. }
  374. }
  375. @media (min-width: 50em) {
  376. ul.with_columns {
  377. column-count: 3;
  378. }
  379. }
  380. ul.with_two_columns {
  381. column-count: 1;
  382. }
  383. @media (min-width: 20em) {
  384. ul.with_two_columns {
  385. column-count: 1;
  386. }
  387. }
  388. @media (min-width: 50em) {
  389. ul.with_two_columns {
  390. column-count: 2;
  391. }
  392. }
  393. .gallery {
  394. display: flex;
  395. flex-wrap: wrap;
  396. justify-content: space-around;
  397. }
  398. .gallery figure img {
  399. margin-left: 1rem;
  400. margin-right: 1rem;
  401. }
  402. .gallery figure figcaption {
  403. font-family: 'ConcourseT3'
  404. }
  405. footer {
  406. font-family: 'ConcourseT3';
  407. display: flex;
  408. flex-direction: column;
  409. border-top: 3px solid white;
  410. padding: 4rem 0;
  411. background-color: #07486c;
  412. color: white;
  413. }
  414. footer > * {
  415. max-width: 50rem;
  416. margin: 0 auto;
  417. }
  418. footer a {
  419. color: #f1c40f;
  420. }
  421. footer .avatar {
  422. width: 200px;
  423. height: 200px;
  424. border-radius: 50%;
  425. float: left;
  426. -webkit-shape-outside: circle();
  427. shape-outside: circle();
  428. margin-right: 2rem;
  429. padding: 2px 5px 5px 2px;
  430. background: white;
  431. border-left: 1px solid #f1c40f;
  432. border-top: 1px solid #f1c40f;
  433. border-right: 5px solid #f1c40f;
  434. border-bottom: 5px solid #f1c40f;
  435. }
  436. </style>
  437. <h1>
  438. <span><a id="jumper" href="#jumpto" title="Un peu perdu ?">?</a></span>
  439. 24/192 Music Downloads (archive)
  440. <time>Pour la pérennité des contenus liés. Non-indexé, retrait sur simple email.</time>
  441. </h1>
  442. <section>
  443. <article>
  444. <h3><a href="https://www.xiph.org/~xiphmont/demo/neil-young.html">Source originale du contenu</a></h3>
  445. <div id="toc_intro">
  446. <img src="https://www.xiph.org/~xiphmont/demo/players2-small.jpg"/>
  447. <p class="aside">
  448. Also see Xiph.Org's new
  449. video, <a href="https://video.xiph.org/vid2.shtml">Digital Show
  450. &amp; Tell</a>, for detailed demonstrations of digital sampling
  451. in action on real equipment!
  452. </p>
  453. <p>Articles last month revealed that musician Neil
  454. Young and Apple's Steve Jobs discussed offering
  455. digital music downloads of 'uncompromised studio quality'.
  456. Much of the press and user commentary was particularly
  457. enthusiastic about the prospect of uncompressed 24 bit 192kHz
  458. downloads. 24/192 featured prominently in my own
  459. conversations with Mr. Young's group several months ago.</p>
  460. <p>Unfortunately, there is no point to distributing music in
  461. 24-bit/192kHz format. Its playback fidelity is slightly
  462. inferior to 16/44.1 or 16/48, and it takes up 6 times the
  463. space.</p>
  464. <p>There are a few real problems with the audio quality and
  465. 'experience' of digitally distributed music today. 24/192
  466. solves none of them. While everyone fixates on 24/192 as a
  467. magic bullet, we're not going to see any actual
  468. improvement.</p>
  469. </div>
  470. <div id="toc_ftbn">
  471. <h2>First, the bad news</h2>
  472. <p>In the past few weeks, I've had conversations with
  473. intelligent, scientifically minded individuals who believe
  474. in 24/192 downloads and want to know how anyone could
  475. possibly disagree. They asked good questions that deserve
  476. detailed answers.</p>
  477. <p>I was also interested in what motivated high-rate digital
  478. audio advocacy. Responses indicate that few people
  479. understand basic signal theory or the
  480. <a href="http://en.wikipedia.org/wiki/Sampling_theorem">sampling
  481. theorem</a>, which is hardly surprising. Misunderstandings
  482. of the mathematics, technology, and physiology arose in most
  483. of the conversations, often asserted by professionals who
  484. otherwise possessed significant audio expertise. Some even
  485. argued that the sampling theorem doesn't really explain how
  486. digital audio actually works [<a href="#foot1">1</a>].</p>
  487. <p>Misinformation and superstition only serve charlatans. So,
  488. let's cover some of the basics of why 24/192 distribution
  489. makes no sense before suggesting some improvements that
  490. actually do.</p>
  491. </div>
  492. <div id="toc_gmye">
  493. <h3>Gentlemen, meet your ears</h3>
  494. <p>The ear hears via hair cells that sit on the resonant
  495. basilar membrane in the cochlea. Each hair cell is
  496. effectively tuned to a narrow frequency band determined by
  497. its position on the membrane. Sensitivity peaks in the
  498. middle of the band and falls off to either side in a
  499. lopsided cone shape overlapping the bands of other nearby
  500. hair cells. A sound is inaudible if there are no hair cells
  501. tuned to hear it.</p>
  502. <img src="https://www.xiph.org/~xiphmont/demo/cochlea-and-responses.png"/>
  503. <div class="caption">
  504. <p>Above left: anatomical cutaway drawing of a human cochlea with the
  505. basilar membrane colored in beige. The membrane is
  506. tuned to resonate at different frequencies along its length,
  507. with higher frequencies near the base and lower frequencies
  508. at the apex. Approximate locations of several frequencies
  509. are marked.</p>
  510. <p>Above right: schematic diagram representing hair cell response
  511. along the basilar membrane as a bank of overlapping filters.</p>
  512. </div>
  513. <p>This is similar to an analog radio that picks up the
  514. frequency of a strong station near where the tuner is
  515. actually set. The farther off the station's frequency is,
  516. the weaker and more distorted it gets until it disappears
  517. completely, no matter how strong. There is an upper (and
  518. lower) audible frequency limit, past which the sensitivity
  519. of the last hair cells drops to zero, and hearing ends.</p>
  520. </div>
  521. <div id="toc_sratas">
  522. <h3>Sampling rate and the audible spectrum</h3>
  523. <p>I'm sure you've heard this many, many times: The human
  524. hearing range spans 20Hz to 20kHz. It's important to know
  525. how researchers arrive at those specific numbers.</p>
  526. <p>First, we measure the 'absolute threshold of hearing'
  527. across the entire audio range for a group of listeners.
  528. This gives us a curve representing the very quietest sound
  529. the human ear can perceive for any given frequency as
  530. measured in ideal circumstances on healthy ears. Anechoic
  531. surroundings, precision calibrated playback equipment, and
  532. rigorous statistical analysis are the easy part. Ears and
  533. auditory concentration both fatigue quickly, so testing must
  534. be done when a listener is fresh. That means lots of breaks
  535. and pauses. Testing takes anywhere from many hours to many
  536. days depending on the methodology.</p>
  537. <p>Then we collect data for the opposite extreme, the
  538. 'threshold of pain'. This is the point where the audio
  539. amplitude is so high that the ear's physical and neural
  540. hardware is not only completely overwhelmed by the input,
  541. but experiences physical pain. Collecting this data is
  542. trickier. You don't want to permanently damage anyone's
  543. hearing in the process.</p>
  544. <img src="https://www.xiph.org/~xiphmont/demo/ath-top.png"/>
  545. <div class="caption">
  546. <p>Above: Approximate equal loudness curves derived from
  547. Fletcher and Munson (1933) plus modern sources for
  548. frequencies &gt; 16kHz. The absolute threshold of hearing and
  549. threshold of pain curves are marked in red. Subsequent
  550. researchers refined these readings, culminating in the Phon
  551. scale and the ISO 226 standard equal loudness curves. Modern
  552. data indicates that the ear is significantly less sensitive
  553. to low frequencies than Fletcher and Munson's results. </p>
  554. </div>
  555. <p>The upper limit of the human audio range is defined to be
  556. where the absolute threshold of hearing curve crosses the
  557. threshold of pain. To even faintly perceive the audio at
  558. that point (or beyond), it must simultaneously be unbearably
  559. loud.</p>
  560. <p>At low frequencies, the cochlea works like a bass reflex
  561. cabinet. The <em>helicotrema</em> is an opening at the apex
  562. of the basilar membrane that acts as a port tuned to
  563. somewhere between 40Hz to 65Hz depending on the
  564. individual. Response rolls off steeply below this
  565. frequency.</p>
  566. <p>Thus, 20Hz - 20kHz is a generous range. It thoroughly
  567. covers the audible spectrum, an assertion backed by nearly a
  568. century of experimental data.</p>
  569. </div>
  570. <div id="toc_ggage">
  571. <h3>Genetic gifts and golden ears</h3>
  572. <p>Based on my correspondences, many people believe in
  573. individuals with extraordinary gifts of hearing. Do such
  574. 'golden ears' really exist?</p>
  575. <p>It depends on what you call a golden ear.</p>
  576. <p>Young, healthy ears hear better than old or damaged ears.
  577. Some people are exceptionally well trained to hear nuances
  578. in sound and music most people don't even know exist. There
  579. was a time in the 1990s when I could identify every major
  580. mp3 encoder by sound (back when they were all pretty bad),
  581. and could demonstrate this reliably in double-blind testing
  582. [<a href="#foot2">2</a>].</p>
  583. <p>When healthy ears combine with highly trained
  584. discrimination abilities, I would call that person a golden
  585. ear. Even so, below-average hearing can also be trained to
  586. notice details that escape untrained listeners. Golden ears
  587. are more about training than hearing beyond the physical
  588. ability of average mortals.</p>
  589. <p>Auditory researchers would love to find, test, and document
  590. individuals with truly exceptional hearing, such as a
  591. greatly extended hearing range. Normal people are nice and
  592. all, but everyone wants to find a genetic freak for a really
  593. juicy paper. We haven't found any such people in the
  594. past 100 years of testing, so they probably don't exist.
  595. Sorry. We'll keep looking.</p>
  596. </div>
  597. <div id="toc_s">
  598. <h3>Spectrophiles</h3>
  599. <p>Perhaps you're skeptical about everything I've just
  600. written; it certainly goes against most marketing material.
  601. Instead, let's consider a hypothetical Wide Spectrum Video
  602. craze that doesn't carry preexisting audiophile baggage.</p>
  603. <img src="https://www.xiph.org/~xiphmont/demo/visspec.png"/>
  604. <div class="caption">
  605. <p>Above: The approximate log scale response of the human
  606. eye's rods and cones, superimposed on the visible
  607. spectrum. These sensory organs respond to light in
  608. overlapping spectral bands, just as the ear's hair cells
  609. are tuned to respond to overlapping bands of sound
  610. frequencies.</p>
  611. </div>
  612. <p>The human eye sees a limited range of frequencies of
  613. light, aka, the visible spectrum. This is directly
  614. analogous to the audible spectrum of sound waves. Like the
  615. ear, the eye has sensory cells (rods and cones) that detect
  616. light in different but overlapping frequency bands.</p>
  617. <p>The visible spectrum extends from about 400THz (deep red)
  618. to 850THz (deep violet) [<a href="#foot3">3</a>].
  619. Perception falls off steeply at the edges. Beyond these
  620. approximate limits, the light power needed for the slightest
  621. perception can fry your retinas. Thus, this is a generous
  622. span even for young, healthy, genetically gifted
  623. individuals, analogous to the generous limits of the audible
  624. spectrum.</p>
  625. <p>In our hypothetical Wide Spectrum Video craze, consider a
  626. fervent group of Spectrophiles who believe these limits
  627. aren't generous enough. They propose that video represent
  628. not only the visible spectrum, but also infrared and
  629. ultraviolet. Continuing the comparison, there's an even
  630. more hardcore [and proud of it!] faction that insists this
  631. expanded range is yet insufficient, and that video feels so
  632. much more natural when it also includes microwaves and some
  633. of the X-ray spectrum. To a Golden Eye, they insist, the
  634. difference is night and day!</p>
  635. <p>Of course this is ludicrous.</p>
  636. <p>No one can see X-rays (or infrared, or ultraviolet, or
  637. microwaves). It doesn't matter how much a person believes
  638. he can. Retinas simply don't have the sensory hardware.</p>
  639. <p>Here's an experiment anyone can do: Go get your Apple IR
  640. remote. The LED emits at 980nm, or about 306THz, in the
  641. near-IR spectrum. This is not far outside of the visible
  642. range. Take the remote into the basement, or the darkest
  643. room in your house, in the middle of the night, with the
  644. lights off. Let your eyes adjust to the blackness.</p>
  645. <img src="https://www.xiph.org/~xiphmont/demo/apple-ir.jpg"/>
  646. <div class="caption">
  647. <p>Above: Apple IR remote photographed using a digital
  648. camera. Though the emitter is quite bright and the
  649. frequency emitted is not far past the red portion of
  650. the visible spectrum, it's completely invisible to the
  651. eye.</p>
  652. </div>
  653. <p>Can you see the Apple Remote's LED flash when you press a
  654. button [<a href="#foot4">4</a>]? No? Not even the tiniest
  655. amount? Try a few other IR remotes; many use an IR
  656. wavelength a bit closer to the visible band, around
  657. 310-350THz. You won't be able to see them either. The rest
  658. emit right at the edge of visibility from 350-380 THz and
  659. may be just barely visible in complete blackness with
  660. dark-adjusted eyes [<a href="#foot5">5</a>]. All would be
  661. blindingly, painfully bright if they were well inside the
  662. visible spectrum.</p>
  663. <p>These near-IR LEDs emit from the visible boundry to at most
  664. 20% beyond the visible frequency limit. 192kHz audio
  665. extends to 400% of the audible limit. Lest I be accused of
  666. comparing apples and oranges, auditory and visual perception
  667. drop off similarly toward the edges.</p>
  668. </div>
  669. <div id="toc_1ch">
  670. <h3>192kHz considered harmful</h3>
  671. <p>192kHz digital music files offer no benefits. They're not
  672. quite neutral either; practical fidelity is slightly worse.
  673. The ultrasonics are a liability during playback. </p>
  674. <p>Neither audio transducers nor power amplifiers are free of
  675. distortion, and distortion tends to increase rapidly at the
  676. lowest and highest frequencies. If the same transducer
  677. reproduces ultrasonics along with audible content, any
  678. nonlinearity will shift some of the ultrasonic content down
  679. into the audible range as an uncontrolled spray of
  680. intermodulation distortion products covering the entire
  681. audible spectrum. Nonlinearity in a power amplifier will
  682. produce the same effect. The effect is very slight, but
  683. listening tests have confirmed that both effects can be
  684. audible.</p>
  685. <img src="https://www.xiph.org/~xiphmont/demo/intermod.png"/>
  686. <div class="caption">
  687. <p>Above: Illustration of distortion products resulting
  688. from intermodulation of a 30kHz and a 33kHz tone in a
  689. theoretical amplifier with a nonvarying total harmonic
  690. distortion (THD) of about .09%. Distortion products
  691. appear throughout the spectrum, including at frequencies
  692. lower than either tone.</p>
  693. <p>Inaudible ultrasonics contribute to intermodulation
  694. distortion in the audible range (light blue area).
  695. Systems not designed to reproduce ultrasonics typically
  696. have much higher levels of distortion above 20kHz, further
  697. contributing to intermodulation. Widening a design's
  698. frequency range to account for ultrasonics requires
  699. compromises that decrease noise and distortion performance
  700. within the audible spectrum. Either way, unneccessary
  701. reproduction of ultrasonic content diminishes
  702. performance.</p>
  703. </div>
  704. <p>There are a few ways to avoid the extra distortion:</p>
  705. <ol>
  706. <li>
  707. <p>A dedicated ultrasonic-only speaker, amplifier, and
  708. crossover stage to separate and independently reproduce
  709. the ultrasonics you can't hear, just so they don't mess
  710. up the sounds you can.</p>
  711. </li>
  712. <li><p>Amplifiers and transducers designed for wider
  713. frequency reproduction, so ultrasonics don't cause audible
  714. intermodulation. Given equal expense and complexity, this
  715. additional frequency range must come at the cost of some
  716. performance reduction in the audible portion of the
  717. spectrum.</p></li>
  718. <li>
  719. <p>Speakers and amplifiers carefully designed not to reproduce
  720. ultrasonics anyway.</p>
  721. </li>
  722. <li>
  723. <p>Not encoding such a wide frequency range to begin with. You can't
  724. and won't have ultrasonic intermodulation distortion in the audible
  725. band if there's no ultrasonic content.</p>
  726. </li>
  727. </ol>
  728. <p>They all amount to the same thing, but only 4) makes any sense.</p>
  729. <p>If you're curious about the performance of your own system,
  730. the following samples contain a 30kHz and a 33kHz tone in a
  731. 24/96 WAV file, a longer version in a FLAC, some tri-tone
  732. warbles, and a normal song clip shifted up by 24kHz so that
  733. it's entirely in the ultrasonic range from 24kHz to 46kHz:</p>
  734. <p>Assuming your system is actually capable of full 96kHz
  735. playback [<a href="#foot6">6</a>], the above files should be
  736. completely silent with no audible noises, tones, whistles,
  737. clicks, or other sounds. If you hear anything, your system
  738. has a nonlinearity causing audible intermodulation of the
  739. ultrasonics. Be careful when increasing volume; running into
  740. digital or analog clipping, even soft clipping, will suddenly
  741. cause loud intermodulation tones.</p>
  742. <p>In summary, it's not certain that intermodulation from
  743. ultrasonics will be audible on a given system. The added
  744. distortion could be insignificant or it could be noticable.
  745. Either way, ultrasonic content is never a benefit, and on
  746. plenty of systems it will audibly hurt fidelity. On the
  747. systems it doesn't hurt, the cost and complexity of handling
  748. ultrasonics could have been saved, or spent on improved audible range
  749. performance instead.</p>
  750. </div>
  751. <div id="toc_sfam">
  752. <h3>Sampling fallacies and misconceptions</h3>
  753. <p>Sampling theory is often unintuitive without a signal processing
  754. background. It's not surprising most people, even brilliant PhDs in
  755. other fields, routinely misunderstand it. It's also not
  756. surprising many people don't even realize they have it wrong.</p>
  757. <img src="https://www.xiph.org/~xiphmont/demo/jaggy.png"/>
  758. <div class="caption">
  759. <p>Above: Sampled signals are often depicted as a rough
  760. stairstep (red) that seems a poor approximation of the
  761. original signal. However, the representation is
  762. mathematically exact and the signal recovers the exact
  763. smooth shape of the original (blue) when converted back to
  764. analog.</p>
  765. </div>
  766. <p>The most common misconception is that sampling is
  767. fundamentally rough and lossy. A sampled signal is often
  768. depicted as a jagged, hard-cornered stair-step facsimile of
  769. the original perfectly smooth waveform. If this is how you
  770. envision sampling working, you may believe that the faster
  771. the sampling rate (and more bits per sample), the finer the
  772. stair-step and the closer the approximation will be. The
  773. digital signal would sound closer and closer to the original
  774. analog signal as sampling rate approaches infinity.</p>
  775. <p>Similarly, many non-DSP people would look at the following:</p>
  776. <img src="https://www.xiph.org/~xiphmont/demo/jaggy2.png"/>
  777. <p>And say, "Ugh!&amp;quot It might appear that a sampled
  778. signal represents higher frequency analog waveforms
  779. badly. Or, that as audio frequency increases, the sampled
  780. quality falls and frequency response falls off, or becomes
  781. sensitive to input phase.</p>
  782. <p>Looks are deceiving. These beliefs are incorrect!</p>
  783. <p class="aside">
  784. <span class="caption">added 2013-04-04:</span><br/> As a
  785. followup to all the mail I got about digital waveforms and
  786. stairsteps, I demonstrate actual digital behavior on real
  787. equipment in our
  788. video <a href="https://video.xiph.org/vid2.shtml">Digital
  789. Show &amp; Tell</a> so you need not simply take me at my
  790. word here!
  791. </p>
  792. <p>All signals with content entirely below the Nyquist
  793. frequency (half the sampling rate) are captured perfectly
  794. and completely by sampling; an infinite sampling rate is not
  795. required. Sampling doesn't affect frequency response or
  796. phase. The analog signal can be reconstructed losslessly,
  797. smoothly, and with the exact timing of the original analog
  798. signal.</p>
  799. <p>So the math is ideal, but what of real world complications?
  800. The most notorious is the band-limiting requirement. Signals
  801. with content over the Nyquist frequency must be lowpassed
  802. before sampling to avoid aliasing distortion; this analog
  803. lowpass is the infamous antialiasing filter. Antialiasing
  804. can't be ideal in practice, but modern techniques bring it
  805. very close. ...and with that we come to oversampling.</p>
  806. </div>
  807. <div id="toc_o">
  808. <h3>Oversampling</h3>
  809. <p>Sampling rates over 48kHz are irrelevant to high fidelity
  810. audio data, but they are internally essential to several
  811. modern digital audio techniques. <em>Oversampling</em> is the
  812. most relevant example [<a href="#foot7">7</a>].</p>
  813. <p>Oversampling is simple and clever. You may recall from my
  814. <a href="http://www.xiph.org/video/vid1.shtml">A Digital
  815. Media Primer for Geeks</a> that high sampling rates
  816. provide a great deal more space between the highest
  817. frequency audio we care about (20kHz) and the Nyquist
  818. frequency (half the sampling rate).
  819. <a href="http://www.xiph.org/video/vid1.shtml?time=678.1">
  820. This allows for simpler, smoother, more reliable analog
  821. anti-aliasing filters, and thus higher fidelity</a>. This
  822. extra space between 20kHz and the Nyquist frequency is
  823. essentially just spectral padding for the analog
  824. filter.</p>
  825. <img src="https://www.xiph.org/~xiphmont/demo/filters.png"/>
  826. <div class="caption">
  827. <p>Above: Whiteboard diagram from <u>A Digital Media
  828. Primer for Geeks</u> illustrating the transition band
  829. width available for a 48kHz ADC/DAC (left) and a 96kHz
  830. ADC/DAC (right).</p>
  831. </div>
  832. <p>That's only half the story. Because digital filters have
  833. few of the practical limitations of an analog filter, we can
  834. complete the anti-aliasing process with greater efficiency
  835. and precision digitally. The very high rate raw digital
  836. signal passes through a digital anti-aliasing filter, which
  837. has no trouble fitting a transition band into a tight space.
  838. After this further digital anti-aliasing, the extra padding
  839. samples are simply thrown away. Oversampled playback
  840. approximately works in reverse.</p>
  841. <p>This means we can use low rate 44.1kHz or 48kHz audio with
  842. all the fidelity benefits of 192kHz or higher sampling
  843. (smooth frequency response, low aliasing) and none of the
  844. drawbacks (ultrasonics that cause intermodulation
  845. distortion, wasted space). Nearly all of today's
  846. analog-to-digital converters (ADCs) and digital-to-analog
  847. converters (DACs) oversample at very high rates. Few people
  848. realize this is happening because it's completely automatic
  849. and hidden.</p>
  850. <p>ADCs and DACs didn't always transparently
  851. oversample. Thirty years ago, some recording consoles
  852. recorded at high sampling rates using only analog filters,
  853. and production and mastering simply used that high rate
  854. signal. The digital anti-aliasing and decimation steps
  855. (resampling to a lower rate for CDs or DAT) happened in the
  856. final stages of mastering. This may well be one of the
  857. early reasons 96kHz and 192kHz became associated with
  858. professional music production [<a href="#foot8">8</a>].</p>
  859. </div>
  860. <div id="toc_1bv2b">
  861. <h3>16 bit vs 24 bit</h3>
  862. <p>OK, so 192kHz music files make no sense. Covered, done. What about
  863. 16 bit vs. 24 bit audio?</p>
  864. <p>It's true that 16 bit linear PCM audio does not quite cover
  865. the entire theoretical dynamic range of the human ear in
  866. ideal conditions. Also, there are (and always will be)
  867. reasons to use more than 16 bits in recording and
  868. production.</p>
  869. <p>None of that is relevant to playback; here 24 bit audio is
  870. as useless as 192kHz sampling. The good news is that at
  871. least 24 bit depth doesn't harm fidelity. It just doesn't
  872. help, and also wastes space.</p>
  873. </div>
  874. <div id="toc_rye">
  875. <h3>Revisiting your ears</h3>
  876. <p>We've discussed the frequency range of the ear, but what
  877. about the dynamic range from the softest possible sound to
  878. the loudest possible sound?</p>
  879. <p>One way to define absolute dynamic range would be to look
  880. again at the absolute threshold of hearing and threshold of
  881. pain curves. The distance between the highest point on the
  882. threshold of pain curve and the lowest point on the absolute
  883. threshold of hearing curve is about 140 decibels for a
  884. young, healthy listener. That wouldn't last long though;
  885. +130dB is loud enough to damage hearing permanently in
  886. seconds to minutes. For reference purposes, a jackhammer at
  887. one meter is only about 100-110dB.</p>
  888. <p>The absolute threshold of hearing increases with age and
  889. hearing loss. Interestingly, the threshold of pain decreases
  890. with age rather than increasing. The hair cells of the cochlea
  891. themselves posses only a fraction of the ear's 140dB range;
  892. musculature in the ear continuously adjust the amount of sound
  893. reaching the cochlea by shifting the ossicles, much as the iris
  894. regulates the amount of light entering the eye
  895. [<a href="#foot9">9</a>]. This mechanism stiffens with age,
  896. limiting the ear's dynamic range and reducing the effectiveness
  897. of its protection mechanisms [<a href="#foot10">10</a>].</p>
  898. </div>
  899. <div id="toc_en">
  900. <h3>Environmental noise</h3>
  901. <p>Few people realize how quiet the absolute threshold of
  902. hearing really is.</p>
  903. <p>The very quietest perceptible sound is about -8dbSPL
  904. [<a href="#foot11">11</a>]. Using an A-weighted scale, the
  905. hum from a 100 watt incandescent light bulb one meter away
  906. is about 10dBSPL, so about 18dB louder. The bulb will be
  907. much louder on a dimmer.</p>
  908. <p>20dBSPL (or 28dB louder than the quietest audible sound) is
  909. often quoted for an empty broadcasting/recording studio or
  910. sound isolation room. This is the baseline for an
  911. exceptionally quiet environment, and one reason you've
  912. probably never noticed hearing a light bulb.</p>
  913. </div>
  914. <div id="toc_tdro1b">
  915. <h3>The dynamic range of 16 bits</h3>
  916. <p>16 bit linear PCM has a dynamic range of 96dB according to
  917. the most common definition, which calculates dynamic range
  918. as (6*bits)dB. Many believe that 16 bit audio cannot
  919. represent arbitrary sounds quieter than -96dB. This is
  920. incorrect.</p>
  921. <p>I have linked to two 16 bit audio files here; one contains
  922. a 1kHz tone at 0 dB (where 0dB is the loudest possible tone)
  923. and the other a 1kHz tone at -105dB.</p>
  924. <img src="https://www.xiph.org/~xiphmont/demo/-105dB.png"/>
  925. <div class="caption">
  926. <p>Above: Spectral analysis of a -105dB tone encoded as 16
  927. bit / 48kHz PCM. 16 bit PCM is clearly deeper than 96dB,
  928. else a -105dB tone could not be represented, nor would
  929. it be audible.</p>
  930. </div>
  931. <p>How is it possible to encode this signal, encode it with no
  932. distortion, and encode it well above the noise floor, when
  933. its peak amplitude is one third of a bit?</p>
  934. <p>Part of this puzzle is solved by proper dither, which
  935. renders quantization noise independent of the input signal.
  936. By implication, this means that dithered quantization
  937. introduces no distortion, just uncorrelated noise. That in
  938. turn implies that we can encode signals of arbitrary depth,
  939. even those with peak amplitudes much smaller than one bit
  940. [<a href="#foot12">12</a>]. However, dither doesn't change
  941. the fact that once a signal sinks below the noise floor, it
  942. should effectively disappear. How is the -105dB tone
  943. still clearly audible above a -96dB noise floor?</p>
  944. <p>The answer: Our -96dB noise floor figure is effectively
  945. wrong; we're using an inappropriate definition of dynamic
  946. range. (6*bits)dB gives us the RMS noise of the entire
  947. broadband signal, but each hair cell in the ear is sensitive
  948. to only a narrow fraction of the total bandwidth. As each
  949. hair cell hears only a fraction of the total noise floor
  950. energy, the noise floor at that hair cell will be much lower
  951. than the broadband figure of -96dB.</p>
  952. <p>Thus, 16 bit audio can go considerably deeper than 96dB.
  953. With use of shaped dither, which moves quantization noise
  954. energy into frequencies where it's harder to hear, the
  955. effective dynamic range of 16 bit audio reaches 120dB in
  956. practice [<a href="#foot13">13</a>], more than fifteen times
  957. deeper than the 96dB claim.</p>
  958. <p>120dB is greater than the difference between a mosquito
  959. somewhere in the same room and a jackhammer a foot
  960. away.... or the difference between a deserted 'soundproof'
  961. room and a sound loud enough to cause hearing damage in
  962. seconds.</p>
  963. <p>16 bits is enough to store all we can hear, and will
  964. be enough forever.</p>
  965. </div>
  966. <div id="toc_stnr">
  967. <h3>Signal-to-noise ratio</h3>
  968. <p>It's worth mentioning briefly that the ear's S/N ratio is
  969. smaller than its absolute dynamic range. Within a given
  970. critical band, typical S/N is estimated to only be about 30dB.
  971. Relative S/N does not reach the full dynamic range even when
  972. considering widely spaced bands. This assures that linear
  973. 16 bit PCM offers higher resolution than is actually
  974. required.</p>
  975. <p>It is also worth mentioning that increasing the bit depth
  976. of the audio representation from 16 to 24 bits does not
  977. increase the perceptible resolution or 'fineness' of the
  978. audio. It only increases the dynamic range, the range
  979. between the softest possible and the loudest possible sound,
  980. by lowering the noise floor. However, a 16-bit noise floor is
  981. already below what we can hear.</p>
  982. </div>
  983. <div id="toc_wd2bm">
  984. <h3>When does 24 bit matter?</h3>
  985. <p>Professionals use 24 bit samples in recording and
  986. production [<a href="#foot14">14</a>] for headroom, noise
  987. floor, and convenience reasons.</p>
  988. <p>16 bits is enough to span the real hearing range with room
  989. to spare. It does not span the entire possible signal range
  990. of audio equipment. The primary reason to use 24 bits when
  991. recording is to prevent mistakes; rather than being careful
  992. to center 16 bit recording-- risking clipping if you guess
  993. too high and adding noise if you guess too low-- 24 bits
  994. allows an operator to set an approximate level and not worry
  995. too much about it. Missing the optimal gain setting by a
  996. few bits has no consequences, and effects that dynamically
  997. compress the recorded range have a deep floor to work
  998. with.</p>
  999. <p>An engineer also requires more than 16 bits during mixing
  1000. and mastering. Modern work flows may involve literally
  1001. thousands of effects and operations. The quantization noise
  1002. and noise floor of a 16 bit sample may be undetectable
  1003. during playback, but multiplying that noise by a few
  1004. thousand times eventually becomes noticeable. 24 bits keeps
  1005. the accumulated noise at a very low level. Once the music
  1006. is ready to distribute, there's no reason to keep more than
  1007. 16 bits.</p>
  1008. </div>
  1009. <div id="toc_lt">
  1010. <h3>Listening tests</h3>
  1011. <p>Understanding is where theory and reality meet. A matter is
  1012. settled only when the two agree.</p>
  1013. <p>Empirical evidence from listening tests backs up the
  1014. assertion that 44.1kHz/16 bit provides highest-possible
  1015. fidelity playback. There are numerous controlled tests
  1016. confirming this, but I'll plug a recent paper,
  1017. <a href="http://www.aes.org/e-lib/browse.cfm?elib=14195"><u>Audibility
  1018. of a CD-Standard A/D/A Loop Inserted into High-Resolution
  1019. Audio Playback</u></a>, done by local folks here at the
  1020. <a href="http://www.bostonaudiosociety.org/">Boston Audio
  1021. Society</a>.</p>
  1022. <p>Unfortunately, downloading the full paper requires an AES
  1023. membership. However it's been discussed widely in articles
  1024. and on forums, with the authors joining in. Here's a few
  1025. links:</p>
  1026. <p>This paper presented listeners with a choice between
  1027. high-rate DVD-A/SACD content, chosen by high-definition
  1028. audio advocates to show off high-def's superiority, and that
  1029. same content resampled on the spot down to 16-bit / 44.1kHz
  1030. Compact Disc rate. The listeners were challenged to
  1031. identify any difference whatsoever between the two using an
  1032. ABX methodology. BAS conducted the test using high-end
  1033. professional equipment in noise-isolated studio listening
  1034. environments with both amateur and trained professional
  1035. listeners.</p>
  1036. <p>In 554 trials, listeners chose correctly 49.8% of the
  1037. time. In other words, they were guessing. Not one listener
  1038. throughout the entire test was able to identify which was
  1039. 16/44.1 and which was high rate [<a href="#foot15">15</a>],
  1040. and the 16-bit signal wasn't even dithered!</p>
  1041. <p>Another recent study [<a href="#foot16">16</a>] investigated
  1042. the possibility that ultrasonics were audible, as earlier studies had
  1043. suggested. The test was constructed to maximize the possibility of
  1044. detection by placing the intermodulation products where they'd be most
  1045. audible. It found that the ultrasonic tones were not audible... but
  1046. the intermodulation distortion products introduced by the loudspeakers
  1047. could be.</p>
  1048. <p>This paper inspired a great deal of further research, much
  1049. of it with mixed results. Some of the ambiguity is
  1050. explained by finding that ultrasonics can induce more
  1051. intermodulation distortion than expected in power amplifiers
  1052. as well. For
  1053. example, <a href="http://www.davidgriesinger.com/intermod.ppt">David
  1054. Griesinger reproduced this experiment</a>
  1055. [<a href="#foot17">17</a>] and found that his loudspeaker
  1056. setup did not introduce audible intermodulation distortion
  1057. from ultrasonics, but his stereo amplifier did.</p>
  1058. </div>
  1059. <div id="toc_cl">
  1060. <h3>Caveat Lector</h3>
  1061. <p>It's important not to cherry-pick individual papers or
  1062. 'expert commentary' out of context or from self-interested
  1063. sources. Not all papers agree completely with these results
  1064. (and a few disagree in large part), so it's easy to find
  1065. minority opinions that appear to vindicate every imaginable
  1066. conclusion.
  1067. <em>Regardless, the papers and links above are
  1068. representative of the vast weight and breadth of the
  1069. experimental record.</em> No peer-reviewed paper that has
  1070. stood the test of time disagrees substantially with these
  1071. results. Controversy exists only within the consumer and
  1072. enthusiast audiophile communities.</p>
  1073. <p>If anything, the number of ambiguous, inconclusive, and
  1074. outright invalid experimental results available through
  1075. Google highlights how tricky it is to construct an accurate,
  1076. objective test. The differences researchers look for are
  1077. minute; they require rigorous statistical analysis to spot
  1078. subconscious choices that escape test subjects' awareness.
  1079. That we're likely trying to 'prove' something that doesn't
  1080. exist makes it even more difficult. Proving a null
  1081. hypothesis is akin to proving the halting problem; you
  1082. can't. You can only collect evidence that lends overwhelming
  1083. weight.</p>
  1084. <p>Despite this, papers that confirm the null hypothesis are
  1085. especially strong evidence; confirming inaudibility is far
  1086. more experimentally difficult than disputing
  1087. it. Undiscovered mistakes in test methodologies and
  1088. equipment nearly always produce false positive results (by
  1089. accidentally introducing audible differences) rather than
  1090. false negatives.</p>
  1091. <p>If professional researchers have such a hard time properly
  1092. testing for minute, isolated audible differences, you can
  1093. imagine how hard it is for amateurs.</p>
  1094. </div>
  1095. <div id="toc_htisualc">
  1096. <h3>How to [inadvertently] screw up a listening comparison</h3>
  1097. <p>The number one comment I heard from believers in super high
  1098. rate audio was [paraphrasing]: <i>"I've listened to high
  1099. rate audio myself and the improvement is obvious. Are you
  1100. seriously telling me not to trust my own ears?"</i></p>
  1101. <p>Of course you can trust your ears. It's brains that are
  1102. gullible. I don't mean that flippantly; as human beings,
  1103. we're all wired that way.</p>
  1104. </div>
  1105. <div id="toc_cbtpeadb">
  1106. <h3>Confirmation bias, the placebo effect, and double-blind</h3>
  1107. <p>In any test where a listener can tell two choices apart via
  1108. any means apart from listening, the results will usually be
  1109. what the listener expected in advance; this is
  1110. called <a href="http://en.wikipedia.org/wiki/Confirmation_bias">
  1111. confirmation bias</a> and it's similar to
  1112. the <a href="http://en.wikipedia.org/wiki/Placebo_effect">placebo
  1113. effect</a>. It means people 'hear' differences because of
  1114. subconscious cues and preferences that have nothing to do
  1115. with the audio, like preferring a more expensive (or more
  1116. attractive) amplifier over a cheaper option.</p>
  1117. <p> The human brain is designed to notice patterns and
  1118. differences, even where none exist. This tendency can't just
  1119. be turned off when a person is asked to make objective
  1120. decisions; it's completely subconscious. Nor can a bias be
  1121. defeated by mere skepticism. <em>Controlled experimentation
  1122. shows that awareness of confirmation bias can increase
  1123. rather than decreases the effect!</em> A test that doesn't
  1124. carefully eliminate confirmation bias is worthless
  1125. [<a href="#foot18">18</a>].</p>
  1126. <p>In <em>single-blind</em> testing, a listener knows nothing
  1127. in advance about the test choices, and receives no feedback
  1128. during the course of the test. Single-blind testing is
  1129. better than casual comparison, but it does not eliminate
  1130. the <a href="http://en.wikipedia.org/wiki/Experimenter%27s_bias">
  1131. experimenter's bias</a>. The test administrator can easily
  1132. inadvertently influence the test or transfer his own
  1133. subconscious bias to the listener through inadvertent cues
  1134. (eg, "Are you sure that's what you're hearing?", body
  1135. language indicating a 'wrong' choice, hesitating
  1136. inadvertently, etc). An experimenter's bias has also been
  1137. experimentally proven to influence a test subject's
  1138. results.</p>
  1139. <p><em>Double-blind</em> listening tests are the gold
  1140. standard; in these tests neither the test administrator nor
  1141. the testee have any knowledge of the test contents or
  1142. ongoing results. Computer-run ABX tests are the most famous
  1143. example, and there are freely available tools for performing
  1144. ABX tests on your own computer[<a href="#foot19">19</a>].
  1145. ABX is considered a minimum bar for a listening test to be
  1146. meaningful; reputable audio forums such
  1147. as <a href="http://www.hydrogenaudio.org/">Hydrogen
  1148. Audio</a>
  1149. often <a href="http://www.hydrogenaudio.org/forums/index.php?showtopic=3974#entry149481">do
  1150. not even allow discussion of listening results unless they
  1151. meet this minimum objectivity requirement</a>
  1152. [<a href="#foot20">20</a>].</p>
  1153. <img src="https://www.xiph.org/~xiphmont/demo/squishyball.png"/>
  1154. <div class="caption">
  1155. <p>Above: Squishyball, a simple command-line ABX tool, running in an xterm.</p>
  1156. </div>
  1157. <p>I personally don't do any quality comparison tests during
  1158. development, no matter how casual, without an ABX
  1159. tool. Science is science, no slacking.</p>
  1160. </div>
  1161. <div id="toc_lt">
  1162. <h3>Loudness tricks</h3>
  1163. <p>The human ear can consciously discriminate amplitude
  1164. differences of about 1dB, and experiments show subconscious
  1165. awareness of amplitude differences under .2dB. Humans
  1166. almost universally consider louder audio to sound better,
  1167. and .2dB is enough to establish this preference. Any
  1168. comparison that fails to carefully amplitude-match the
  1169. choices will see the louder choice preferred, even if the
  1170. amplitude difference is too small to consciously notice.
  1171. Stereo salesmen have known this trick for a long time.</p>
  1172. <p>The professional testing standard is to match sources to
  1173. within .1dB or better. This often requires use of an
  1174. oscilloscope or signal analyzer. Guessing by turning the
  1175. knobs until two sources sound about the same is not good
  1176. enough.</p>
  1177. </div>
  1178. <div id="toc_c">
  1179. <h3>Clipping</h3>
  1180. <p>Clipping is another easy mistake, sometimes obvious only in
  1181. retrospect. Even a few clipped samples or their aftereffects
  1182. are easy to hear compared to an unclipped signal.</p>
  1183. <p>The danger of clipping is especially pernicious in tests
  1184. that create, resample, or otherwise manipulate digital signals
  1185. on the fly. Suppose we want to compare the fidelity of 48kHz
  1186. sampling to a 192kHz source sample. A typical way is to
  1187. downsample from 192kHz to 48kHz, upsample it back to 192kHz,
  1188. and then compare it to the original 192kHz sample in an ABX
  1189. test [<a href="#foot21">21</a>]. This arrangement allows us
  1190. to eliminate any possibility of equipment variation or sample
  1191. switching influencing the results; we can use the same DAC to
  1192. play both samples and switch between without any hardware mode
  1193. changes.</p>
  1194. <p>Unfortunately, most samples are mastered to use the full
  1195. digital range. Naive resampling can and often will clip
  1196. occasionally. It is necessary to either monitor for clipping
  1197. (and discard clipped audio) or avoid clipping via some other
  1198. means such as attenuation.</p>
  1199. </div>
  1200. <div id="toc_dmdm">
  1201. <h3>Different media, different master</h3>
  1202. <p>I've run across a few articles and blog posts that declare
  1203. the virtues of 24 bit or 96/192kHz by comparing a CD to an
  1204. audio DVD (or SACD) of the 'same' recording. This comparison
  1205. is invalid; the masters are usually different.</p>
  1206. </div>
  1207. <div id="toc_ic">
  1208. <h3>Inadvertent cues</h3>
  1209. <p>Inadvertant audible cues are almost inescapable in older
  1210. analog and hybrid digital/analog testing setups. Purely
  1211. digital testing setups can completely eliminate the problem in
  1212. some forms of testing, but also multiply the potential of
  1213. complex software bugs. Such limitations and bugs have a long
  1214. history of causing false-positive results in testing
  1215. [<a href="#foot22">22</a>].</p>
  1216. <p><a href="http://www.bostonaudiosociety.org/bas_speaker/abx_testing2.htm"><u>The
  1217. Digital Challenge - More on ABX Testing</u></a>, tells a
  1218. fascinating story of a specific listening test conducted in
  1219. 1984 to rebut audiophile authorities of the time who asserted
  1220. that CDs were inherently inferior to vinyl. The article is
  1221. not concerned so much with the results of the test (which I
  1222. suspect you'll be able to guess), but the processes and
  1223. real-world messiness involved in conducting such a test. For
  1224. example, an error on the part of the testers inadvertantly
  1225. revealed that an invited audiophile expert had not been making
  1226. choices based on audio fidelity, but rather by listening to
  1227. the slightly different clicks produced by the ABX switch's
  1228. analog relays!</p>
  1229. <p>Anecdotes do not replace data, but this story is
  1230. instructive of the ease with which undiscovered flaws can bias
  1231. listening tests. Some of the audiophile beliefs discussed
  1232. within are also highly entertaining; one hopes that some
  1233. modern examples are considered just as silly 20 years from
  1234. now.</p>
  1235. </div>
  1236. <div id="toc_ftgn">
  1237. <h2>Finally, the good news</h2>
  1238. <p>What actually works to improve the quality of the digital
  1239. audio to which we're listening?</p>
  1240. </div>
  1241. <div id="toc_bh">
  1242. <h3>Better headphones</h3>
  1243. <p>The easiest fix isn't digital. The most dramatic possible
  1244. fidelity improvement for the cost comes from a good pair of
  1245. headphones. Over-ear, in ear, open or closed, it doesn't much
  1246. matter. They don't even need to be expensive, though expensive
  1247. headphones can be worth the money.</p>
  1248. <p>Keep in mind that some headphones are expensive because
  1249. they're well made, durable and sound great. Others are
  1250. expensive because they're $20 headphones under a several
  1251. hundred dollar layer of styling, brand name, and marketing. I
  1252. won't make specfic recommendations here, but I will say you're
  1253. not likely to find good headphones in a big box store, even if
  1254. it specializes in electronics or music. As in all other
  1255. aspects of consumer hi-fi, do your research (and caveat
  1256. emptor).</p>
  1257. </div>
  1258. <div id="toc_lf">
  1259. <h3>Lossless formats</h3>
  1260. <p>It's true enough that a properly encoded Ogg file (or MP3,
  1261. or AAC file) will be indistinguishable from the original at a
  1262. moderate bitrate.</p>
  1263. <p>But what of badly encoded files?</p>
  1264. <p>Twenty years ago, all mp3 encoders were really bad by
  1265. today's standards. Plenty of these old, bad encoders are
  1266. still in use, presumably because the licenses are cheaper and
  1267. most people can't tell or don't care about the difference
  1268. anyway. Why would any company spend money to fix what it's
  1269. completely unaware is broken?</p>
  1270. <p>Moving to a newer format
  1271. like <a href="http://www.vorbis.com">Vorbis</a> or AAC doesn't
  1272. necessarily help. For example, many companies and individuals
  1273. used (and still
  1274. use) <a href="http://xiphmont.livejournal.com/51160.html">FFmpeg's
  1275. very-low-quality built-in Vorbis encoder</a> because it was
  1276. the default in FFmpeg and they were unaware how bad it
  1277. was. AAC has an even longer history of widely-deployed,
  1278. low-quality encoders; all mainstream lossy formats do.</p>
  1279. <p>Lossless formats
  1280. like <a href="http://flac.sourceforge.net/">FLAC</a> avoid any
  1281. possibility of damaging audio fidelity
  1282. [<a href="#foot23">23</a>] with a poor quality lossy encoder,
  1283. or even by a good lossy encoder used incorrectly.</p>
  1284. <p>A second reason to distribute lossless formats is to avoid
  1285. generational loss. Each reencode or transcode loses more
  1286. data; even if the first encoding is transparent, it's very
  1287. possible the second will have audible artifacts. This matters
  1288. to anyone who might want to remix or sample from downloads. It
  1289. especially matters to us codec researchers; we need clean
  1290. audio to work with. </p>
  1291. </div>
  1292. <div id="toc_bm">
  1293. <h3>Better masters</h3>
  1294. <p>The <a href="http://www.aes.org/e-lib/browse.cfm?elib=14195">
  1295. BAS test I linked earlier</a> mentions as an aside that the
  1296. SACD version of a recording <em>can</em> sound substantially
  1297. better than the CD release. It's not because of increased
  1298. sample rate or depth but because the SACD used a
  1299. higher-quality master. When bounced to a CD-R, the SACD
  1300. version still sounds as good as the original SACD and
  1301. better than the CD release because the original audio used to
  1302. make the SACD was better. Good production and mastering
  1303. obviously contribute to the final quality of the music
  1304. [<a href="#foot24">24</a>].</p>
  1305. <p>The recent coverage of 'Mastered for iTunes' and similar
  1306. initiatives from other industry labels is somewhat
  1307. encouraging. What remains to be seen is whether or not Apple
  1308. and the others actually 'get it' or if this is merely a hook
  1309. for selling consumers yet another, more expensive copy of
  1310. music they already own.</p>
  1311. </div>
  1312. <div id="toc_su">
  1313. <h3>Surround</h3>
  1314. <p>Another possible 'sales hook', one I'd enthusiastically buy
  1315. into myself, is surround recordings. Unfortunately, there's
  1316. some technical peril here.</p>
  1317. <p>Old-style discrete surround with many channels (5.1, 7.1,
  1318. etc) is a technical relic dating back to the theaters of the
  1319. 1960s. It is inefficient, using more channels than competing
  1320. systems. The surround image is limited, and tends to collapse
  1321. toward the nearer speakers when a listener sits or shifts out of
  1322. position.</p>
  1323. <p>We can represent and encode excellent and robust
  1324. localization with systems like Ambisonics. The problems are
  1325. the cost of equipment for reproduction and the fact that
  1326. something encoded for a natural soundfield both sounds bad
  1327. when mixed down to stereo, and can't be created artificially
  1328. in a convincing way. It's hard to fake ambisonics or
  1329. holographic audio, sort of like how 3D video always seems to
  1330. degenerate into a gaudy gimmick that reliably makes 5% of
  1331. the population motion sick.</p>
  1332. <p>Binaural audio is similarly difficult. You can't simulate
  1333. it because it works slightly differently in every person.
  1334. It's a learned skill tuned to the self-assembling system of
  1335. the pinnae, ear canals, and neural processing, and it never
  1336. assembles exactly the same way in any two individuals.
  1337. People also subconsciously shift their heads to enhance
  1338. localization, and can't localize well unless they do.
  1339. That's something that can't be captured in a binaural
  1340. recording, though it can to an extent in fixed surround.</p>
  1341. <p>These are hardly impossible technical hurdles. Discrete
  1342. surround has a proven following in the marketplace, and I'm
  1343. personally especially excited by the possibilities offered
  1344. by Ambisonics.</p>
  1345. </div>
  1346. <div id="toc_outro">
  1347. <h2>Outro</h2>
  1348. <blockquote>
  1349. "I never did care for music much.<br/>
  1350. It's the high fidelity!"<br/>
  1351.      —Flanders &amp; Swann, <u>A Song of Reproduction</u>
  1352. </blockquote>
  1353. <p>The point is enjoying the music, right? Modern playback
  1354. fidelity is incomprehensibly better than the already excellent
  1355. analog systems available a generation ago. Is the logical
  1356. extreme any more than just
  1357. another <a href="http://www.youtube.com/watch?v=M3w1_E1V46M">first
  1358. world problem</a>? Perhaps, but bad mixes and
  1359. encodings <em>do</em> bother me; they distract me from the
  1360. music, and I'm probably not alone.</p>
  1361. <p>Why push back against 24/192? Because it's a solution to a
  1362. problem that doesn't exist, a business model based on
  1363. willful ignorance and scamming people. The more that
  1364. pseudoscience goes unchecked in the world at large, the
  1365. harder it is for truth to overcome truthiness... even if
  1366. this is a small and relatively insignificant example.</p>
  1367. <blockquote>
  1368. "For me, it is far better to grasp the Universe as it really
  1369. is than to persist in delusion, however satisfying and
  1370. reassuring."
  1371. <br/>     —Carl Sagan
  1372. </blockquote>
  1373. </div>
  1374. <div id="toc_more">
  1375. <h2>Further reading</h2>
  1376. <p>Readers have alerted me to a pair of excellent papers of
  1377. which I wasn't aware before beginning my own article. They
  1378. tackle many of the same points I do in greater detail.</p>
  1379. <ul>
  1380. <li>
  1381. <p><a href="http://www.meridian.co.uk/ara/coding2.pdf"><u>Coding
  1382. High Quality Digital Audio</u></a> by Bob Stuart
  1383. of Meridian Audio is beautifully concise despite
  1384. its greater length. Our conclusions differ
  1385. somewhat (he takes as given the need for a
  1386. slightly wider frequency range and bit depth
  1387. without much justification), but the presentation
  1388. is clear and easy to follow. <i>[Edit: I may not
  1389. agree with many of Mr. Stuart's other articles,
  1390. but I like this one a lot.]</i></p></li>
  1391. <li><p><a href="http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf">
  1392. <u>Sampling Theory For Digital Audio</u></a> [Updated link 2012-10-04] by Dan
  1393. Lavry of Lavry Engineering is another article that several
  1394. readers pointed out. It expands my two pages or so about
  1395. sampling, oversampling, and filtering into a more detailed
  1396. 27 page treatment. Worry not, there are plenty of graphs,
  1397. examples and references.</p></li>
  1398. </ul>
  1399. <p>Stephane Pigeon
  1400. of <a href="http://www.audiocheck.net/">audiocheck.net</a>
  1401. wrote to plug the browser-based listening tests featured on
  1402. his web site. The set of tests is relatively small as yet,
  1403. but several were directly relevant in the context of this
  1404. article. They worked well and I found the quality to be
  1405. quite good.</p>
  1406. </div>
  1407. <div id="toc_fn">
  1408. <h2>Footnotes</h2>
  1409. <ol>
  1410. <li id="foot2">
  1411. <p>If it wasn't the most boring
  1412. party trick ever, it was pretty close.
  1413. </p></li>
  1414. <li id="foot3">
  1415. <p>It's more typical to speak of
  1416. visible light as wavelengths measured in nanometers or
  1417. angstroms. I'm using frequency to be consistent with
  1418. sound. They're equivalent, as frequency is just the
  1419. inverse of wavelength.</p>
  1420. </li>
  1421. <li id="foot4">
  1422. <p>The LED experiment doesn't work
  1423. with 'ultraviolet' LEDs, mainly because they're not really
  1424. ultraviolet. They're deep enough violet to cause a little
  1425. bit of fluorescence, but they're still well within the
  1426. visible range. Real ultraviolet LEDs cost anywhere from
  1427. $100-$1000 apiece and would cause eye damage if used for
  1428. this test. Consumer grade not-really-UV LEDs also emit
  1429. some faint white light in order to appear brighter, so
  1430. you'd be able to see them even if the emission peak really
  1431. was in the ultraviolet.</p>
  1432. </li>
  1433. <li id="foot5">
  1434. <p>The original version of this article stated that IR
  1435. LEDs operate from 300-325THz (about 920-980nm),
  1436. wavelengths that are invisible. Quite a few readers wrote
  1437. to say that they could in fact just barely see the LEDs in
  1438. some (or all) of their remotes. Several were kind enough
  1439. to let me know which remotes these were, and I was able to
  1440. test several on a spectrometer. Lo and behold, these
  1441. remotes were using higher-frequency LEDs operating from
  1442. 350-380THz (800-850nm), just overlapping the extreme
  1443. edge of the visible range.
  1444. </p>
  1445. </li>
  1446. <li id="foot6">
  1447. <p>Many systems that cannot play back 96kHz samples will
  1448. silently downsample to 48kHz, rather than refuse to play
  1449. the file. In this case, the tones will not be played at
  1450. all and playback would be silent no matter how nonlinear
  1451. the system is.</p>
  1452. </li>
  1453. <li id="foot7">
  1454. <p>Oversampling is not the only
  1455. application for high sampling rates in signal
  1456. processing. There are a few theoretical advantages to
  1457. producing band-limited audio at a high sampling rate
  1458. eschewing decimation, even if it is to be downsampled
  1459. for distribution. It's not clear what if any are used
  1460. in practice, as the workings of most professional
  1461. consoles are trade secrets.</p>
  1462. </li>
  1463. <li id="foot8">
  1464. <p>Historical reasoning or not,
  1465. there's no question that many professionals today use high
  1466. rates because they mistakenly assume that retaining
  1467. content beyond 20kHz sounds better, just as consumers
  1468. do.</p>
  1469. </li>
  1470. <li id="foot9">
  1471. <p>The sensation of eardrums
  1472. 'uncringing' after turning off loud music is quite
  1473. real!</p>
  1474. </li>
  1475. <li id="foot10">
  1476. <p>Some nice diagrams can be found
  1477. at the HyperPhysics site:<br/>
  1478. <a href="http://hyperphysics.phy-astr.gsu.edu/hbase/sound/protect.html#c1">http://hyperphysics.phy-astr.gsu.edu/hbase/sound/protect.html#c1</a></p>
  1479. </li>
  1480. <li id="foot11">
  1481. <p>20µPa is commonly defined to be
  1482. 0dB for auditory measurement purposes; it is approximately
  1483. equal to the threshold of hearing at 1kHz. The ear is as
  1484. much as 8dB more sensitive between 2 and 4kHz however.</p>
  1485. </li>
  1486. <li id="foot12">
  1487. <p>The following paper has the best explanation of dither
  1488. that I've run across. Although it's about image dither,
  1489. the first half covers the theory and practice of dither in
  1490. audio before extending its use into images:</p>
  1491. <p>Cameron Nicklaus Christou,
  1492. <a href="http://uwspace.uwaterloo.ca/bitstream/10012/3867/1/thesis.pdf">
  1493. <u>Optimal Dither and Noise Shaping
  1494. in Image Processing</u></a>
  1495. </p>
  1496. </li>
  1497. <li id="foot13">
  1498. <p>DSP engineers may point out, as
  1499. one of my own smart-alec compatriots did, that 16 bit
  1500. audio has a theoretically infinite dynamic range for a
  1501. pure tone if you're allowed to use an infinite Fourier
  1502. transform to extract it; this concept is very important to
  1503. radio astronomy.</p>
  1504. <p>Although the ear works not entirely unlike a Fourier transform, its
  1505. resolution is relatively limited. This places a limit on the maximum
  1506. practical dynamic depth of 16 bit audio signals.</p>
  1507. </li>
  1508. <li id="foot14">
  1509. <p>Production increasingly uses 32
  1510. bit float, both because it's very convenient on modern
  1511. processors, and because it completely eliminates the
  1512. possibility of accidental clipping at any point going
  1513. undiscovered and ruining a mix.</p>
  1514. </li>
  1515. <li id="foot15">
  1516. <p>Several readers have wanted to know how, if ultrasonics
  1517. can cause audible intermodulation distortion, the Meyer
  1518. and Moran 2007 test could have produced a null result.</p>
  1519. <p>It should be obvious that 'can' and 'sometimes' are not
  1520. the same as 'will' and 'always'. Intermodulation
  1521. distortion from ultrasonics is a possibility, not a
  1522. certainty, in any given system for a given set of
  1523. material. The Meyer and Moran null result indicates that
  1524. intermodulation distortion was inaudible on the systems
  1525. used during the course of their testing.</p>
  1526. <p>Readers are invited to <a href="#toc_intermod">try the
  1527. simple ultrasonic intermodulation distortion test
  1528. above</a> for a quick check of the intermodulation
  1529. potential of their own equipment.</p>
  1530. </li>
  1531. <li id="foot16">
  1532. <p>Karou and Shogo, <u>Detection of
  1533. Threshold for tones above 22kHz</u> (2001). Convention paper
  1534. 5401 presented at the 110th Convention, May 12-15 2001,
  1535. Amsterdam.</p>
  1536. </li>
  1537. <li id="foot17">
  1538. <p>Griesinger, <a href="http://www.davidgriesinger.com/intermod.ppt"><u>Perception
  1539. of mid-frequency and high-frequency intermodulation
  1540. distortion in loudspeakers, and its relationship to
  1541. high definition audio</u></a></p>
  1542. </li>
  1543. <li id="foot18">
  1544. <p>Since publication, several commentators wrote to me with
  1545. similar versions of the same anecdote [paraphrased]: "I
  1546. once listened to some headphones / amps / recordings
  1547. expecting result [A] but was totally surprised to find
  1548. [B] instead! Confirmation bias is hooey!"</p>
  1549. <p>I offer two thoughts.</p>
  1550. <p>First, confirmation bias does not replace all correct
  1551. results with incorrect results. It skews the results in
  1552. some uncontrolled direction by an unknown amount. How
  1553. can you tell right or wrong <em>for sure</em> if the
  1554. test is rigged by your own subconscious? Let's say you
  1555. expected to hear a large difference but were shocked to
  1556. hear a small difference. What if there was actually no
  1557. difference at all? Or, maybe there <em>was</em> a
  1558. difference and, being aware of a potential bias, your
  1559. well meaning skepticism overcompensated? Or maybe you
  1560. were completely right? Objective testing, such as ABX,
  1561. eliminates all this uncertainty.</p>
  1562. <p>Second, "So you think you're not biased? Great!
  1563. Prove it!" The value of an objective test lies not only
  1564. in its ability to inform one's own understanding, but
  1565. also to convince others. Claims require proof.
  1566. Extraordinary claims require extraordinary proof.</p>
  1567. </li>
  1568. <li id="foot19">
  1569. <p>The easiest tools to use for ABX testing are
  1570. probably:</p>
  1571. </li>
  1572. <li id="foot20">
  1573. <p>At Hydrogen Audio, the objective testing requirement is
  1574. abbreviated <em>TOS8</em> as it's the eighth item in the
  1575. Terms Of Service.</p>
  1576. </li>
  1577. <li id="foot21">
  1578. <p>It is commonly assumed that resampling irreparably
  1579. damages a signal; this isn't the case. Unless one makes
  1580. an obvious mistake, such as causing clipping, the
  1581. downsampled and then upsampled signal will be audibly
  1582. indistinguishable from the original. This is the usual
  1583. test used to establish that higher sampling rates are
  1584. unneccessary.</p>
  1585. </li>
  1586. <li id="foot22">
  1587. <p>It may not be strictly audio related,
  1588. but... faster-than-light neutrinos, anyone?</p>
  1589. </li>
  1590. <li id="foot23">
  1591. <p><a href="http://www.wired.com/gadgetlab/2012/02/why-neil-young-hates-mp3-and-what-you-can-do-about-it/">Wired
  1592. magazine implies that lossless formats like FLAC
  1593. are not always completely lossless</a>:</p>
  1594. <blockquote>
  1595. "Some purists will tell you to skip FLACs altogether
  1596. and just buy WAVs. [...] By buying WAVs, you can avoid
  1597. the potential data loss incurred when the file is
  1598. compressed into a FLAC. This data loss is rare, but it
  1599. happens."
  1600. </blockquote>
  1601. <p>This is false. A lossless compression process never
  1602. alters the original data in any way, and FLAC is no
  1603. exception.</p>
  1604. <p>In the event that Wired was referring to hardware
  1605. corruption of data files (disk failure, memory failure,
  1606. sunspots), FLAC and WAV would both be affected. A FLAC
  1607. file, however, is checksummed and would detect the
  1608. corruption. The FLAC file is also smaller than the WAV,
  1609. and so a random corruption would be less likely because
  1610. there's less data that could be affected.</p>
  1611. </li>
  1612. <li id="foot24">
  1613. <p>
  1614. The <a href="http://en.wikipedia.org/wiki/Loudness_war">'Loudness
  1615. War'</a> is a commonly cited example of bad mastering
  1616. practices in the industry today, though it's not the
  1617. only one. Loudness is also an older phenomenon than the
  1618. Wikipedia article leads the reader to believe; as early
  1619. as the 1950s, artists and producers pushed for the
  1620. loudest possible recordings. Equipment vendors
  1621. increasingly researched and marketed new technology to
  1622. allow hotter and hotter masters. Advanced vinyl
  1623. mastering equipment in the 1970s and 1980s, for example,
  1624. tracked and nested groove envelopes when possible in
  1625. order to allow higher amplitudes than the groove spacing
  1626. would normally permit.</p>
  1627. <p>Today's digital technology has allowed loudness to be
  1628. pumped up to an absurd level. It's also provided a
  1629. plethora of automatic, highly complex, proprietary DAW
  1630. plugins that are deployed en-masse without a wide
  1631. understanding of how they work or what they're really
  1632. doing.</p>
  1633. </li>
  1634. </ol>
  1635. </div>
  1636. <hr/>
  1637. <div class="author">
  1638. <address>—Monty
  1639. (<a href="mailto:monty@xiph.org">monty@xiph.org)</a> March
  1640. 1, 2012
  1641. <br/><i>last revised March 25, 2012 to add improvements
  1642. suggested by readers.
  1643. <br/>Edits and corrections made after this date are marked inline, except for spelling errors
  1644. <br/>spotted on Dec 30, 2012 and March 15, 2014, and an extra 'is' removed on April 1, 2013]
  1645. </i></address>
  1646. </div>
  1647. <div class="et">
  1648. <div class="etleft">
  1649. <div class="etcontent">
  1650. <a href="http://et.redhat.com/"><img src="https://www.xiph.org/~xiphmont/demo/et.png"/></a>
  1651. </div>
  1652. </div>
  1653. <div class="etcenter">
  1654. <div class="etcontent">
  1655. <p>Monty's articles and demo work are sponsored by Red Hat Emerging Technologies.
  1656. <br/>(C) Copyright 2012 Red Hat Inc. and Xiph.Org
  1657. <br/>Special thanks to Gregory Maxwell for technical
  1658. contributions to this article</p>
  1659. </div>
  1660. </div>
  1661. </div>
  1662. <div>
  1663. <img src="https://www.xiph.org/~xiphmont/demo/brick-redhat.jpg"/>
  1664. </div>
  1665. </article>
  1666. </section>
  1667. <nav id="jumpto">
  1668. <p>
  1669. <a href="/david/blog/">Accueil du blog</a> |
  1670. <a href="https://www.xiph.org/~xiphmont/demo/neil-young.html">Source originale</a> |
  1671. <a href="/david/stream/2019/">Accueil du flux</a>
  1672. </p>
  1673. </nav>
  1674. <footer>
  1675. <div>
  1676. <img src="/static/david/david-larlet-avatar.jpg" loading="lazy" class="avatar" width="200" height="200">
  1677. <p>
  1678. Bonjour/Hi!
  1679. Je suis <a href="/david/" title="Profil public">David&nbsp;Larlet</a>, je vis actuellement à Montréal et j’alimente cet espace depuis 15 ans. <br>
  1680. Si tu as apprécié cette lecture, n’hésite pas à poursuivre ton exploration. Par exemple via les <a href="/david/blog/" title="Expériences bienveillantes">réflexions bimestrielles</a>, la <a href="/david/stream/2019/" title="Pensées (dés)articulées">veille hebdomadaire</a> ou en t’abonnant au <a href="/david/log/" title="S’abonner aux publications via RSS">flux RSS</a> (<a href="/david/blog/2019/flux-rss/" title="Tiens c’est quoi un flux RSS ?">so 2005</a>).
  1681. </p>
  1682. <p>
  1683. Je m’intéresse à la place que je peux avoir dans ce monde. En tant qu’humain, en tant que membre d’une famille et en tant qu’associé d’une coopérative. De temps en temps, je fais aussi des <a href="https://github.com/davidbgk" title="Principalement sur Github mais aussi ailleurs">trucs techniques</a>. Et encore plus rarement, <a href="/david/talks/" title="En ce moment je laisse plutôt la place aux autres">j’en parle</a>.
  1684. </p>
  1685. <p>
  1686. Voici quelques articles choisis :
  1687. <a href="/david/blog/2019/faire-equipe/" title="Accéder à l’article complet">Faire équipe</a>,
  1688. <a href="/david/blog/2018/bivouac-automnal/" title="Accéder à l’article complet">Bivouac automnal</a>,
  1689. <a href="/david/blog/2018/commodite-effondrement/" title="Accéder à l’article complet">Commodité et effondrement</a>,
  1690. <a href="/david/blog/2017/donnees-communs/" title="Accéder à l’article complet">Des données aux communs</a>,
  1691. <a href="/david/blog/2016/accompagner-enfant/" title="Accéder à l’article complet">Accompagner un enfant</a>,
  1692. <a href="/david/blog/2016/senior-developer/" title="Accéder à l’article complet">Senior developer</a>,
  1693. <a href="/david/blog/2016/illusion-sociale/" title="Accéder à l’article complet">L’illusion sociale</a>,
  1694. <a href="/david/blog/2016/instantane-scopyleft/" title="Accéder à l’article complet">Instantané Scopyleft</a>,
  1695. <a href="/david/blog/2016/enseigner-web/" title="Accéder à l’article complet">Enseigner le Web</a>,
  1696. <a href="/david/blog/2016/simplicite-defaut/" title="Accéder à l’article complet">Simplicité par défaut</a>,
  1697. <a href="/david/blog/2016/minimalisme-esthetique/" title="Accéder à l’article complet">Minimalisme et esthétique</a>,
  1698. <a href="/david/blog/2014/un-web-omni-present/" title="Accéder à l’article complet">Un web omni-présent</a>,
  1699. <a href="/david/blog/2014/manifeste-developpeur/" title="Accéder à l’article complet">Manifeste de développeur</a>,
  1700. <a href="/david/blog/2013/confort-convivialite/" title="Accéder à l’article complet">Confort et convivialité</a>,
  1701. <a href="/david/blog/2013/testament-numerique/" title="Accéder à l’article complet">Testament numérique</a>,
  1702. et <a href="/david/blog/" title="Accéder aux archives">bien d’autres…</a>
  1703. </p>
  1704. <p>
  1705. On peut <a href="mailto:david%40larlet.fr" title="Envoyer un courriel">échanger par courriel</a>. Si éventuellement tu souhaites que l’on travaille ensemble, tu devrais commencer par consulter le <a href="http://larlet.com">profil dédié à mon activité professionnelle</a> et/ou contacter directement <a href="http://scopyleft.fr/">scopyleft</a>, la <abbr title="Société coopérative et participative">SCOP</abbr> dont je fais partie depuis six ans. Je recommande au préalable de lire <a href="/david/blog/2018/cout-site/" title="Attention ce qui va suivre peut vous choquer">combien coûte un site</a> et pourquoi je suis plutôt favorable à une <a href="/david/pro/devis/" title="Discutons-en !">non-demande de devis</a>.
  1706. </p>
  1707. <p>
  1708. Je ne traque pas ta navigation mais mon
  1709. <abbr title="Alwaysdata, 62 rue Tiquetonne 75002 Paris, +33.184162340">hébergeur</abbr>
  1710. conserve des logs d’accès.
  1711. </p>
  1712. </div>
  1713. </footer>
  1714. <script type="text/javascript">
  1715. ;(_ => {
  1716. const jumper = document.getElementById('jumper')
  1717. jumper.addEventListener('click', e => {
  1718. e.preventDefault()
  1719. const anchor = e.target.getAttribute('href')
  1720. const targetEl = document.getElementById(anchor.substring(1))
  1721. targetEl.scrollIntoView({behavior: 'smooth'})
  1722. })
  1723. })()
  1724. </script>