Header

Objective Reviews & Commentary - An Engineer's Perspective

July 13, 2011

CDs Are So 1980s

INTRO: Pick up an audiophile magazine in 2011 and you’re likely to find reviews of equipment for playing vinyl LP’s dating back to the 1940’s and the Compact Disc from the early 1980’s. In the last 25+ years, much has changed. So why do many hardcore audiophiles still cling to ancient formats? I’ll leave the Great Vinyl Debate for another article, but what’s the best digital format for audio playback? CD? SACD? FLAC? Apple? Dolby TrueHD?

COMPACT DISCS: I don’t have anything against CDs--I own about 6000 of them. All but a few, however, are safely packed away (mostly to keep the RIAA happy should they ever knock on my door). And the few that are readily accessible are either awaiting their turn to be ripped or used for for test purposes. Digital music files became my primary source 8 years ago when FLAC lossless compression was released.

THE LAST OF THE DINOSAURS: While players that only play CDs have all but disappeared from mainstream stores, there are still lots of audiophile offerings and even some new models. Granted there are lots of CDs in circulation, but increasingly, few are playing them anymore. They were ripped on a computer long ago and boxed up in a closet or sold. And the ones sold often get ripped and sold again. It’s no wonder CD sales have tanked. But, oddly, audiophiles seem way more likely to still play their actual CDs.

WHY DO AUDIOPHILES KEEP PLAYING CDs? It’s work to store, organize, play, put away, and generally pamper those little silver discs. If so many people are still doing it, they must have some good reasons, right? Let’s look a half dozen of the reasons:

  1. SOUND QUALITY: Many audiophiles still fussing with little silver discs will tell you it’s because they sound better than other forms of digital playback. But hardware has been available for over a decade that’s capable of delivering the exact same bitstream as a CD transport to the same expensive audiophile-approved DACs. Even taking jitter into account, there’s no reason playing a file off a hard drive cannot sound every bit as good as the same data played off a CD transport. A well designed DAC doesn’t care where the bits come from. You can get the exact same analog output from a DAC if they’re played from a CD transport, your PC’s hard drive, or beamed from outer space via satellite.
  2. LACK OF HIGH-END HARDWARE: While the choices were slim many years ago they’re fairly plentiful today. There a probably hundreds of DACs, from $30 to $30,000, that have USB jacks on them and can connect to any computer—PC or Mac. There are also a lot of network music and media players including several from high-end companies. And there are countless iPod docks including many high-end ones.
  3. COMPRESSION SOUNDS BAD: Lossy compression such as MP3 and AAC has often been done poorly, at substandard bitrates, etc. And this has generally given all compression a bad name among many audiophiles. But you don’t have to use any compression. And, as I’ll explain later, lossless compression delivers identical sound to playing a CD. So, in terms of sound quality, there’s no reason not to use it.
  4. DATA STORAGE IS A PROBLEM: Back in 1983, when the first CDs were sold, the amount of data on a single CD would have required 65 10 MB hard drives at a cost of around $120,000.00 to store the raw data! Today, you can put 3000 bit-accurate lossless CDs on a 1 Terabyte hard drive ($50) at a cost of $0.02 each. You can also get an inexpensive NAS—Network Attached Storage device that sits on your network using little power ready to serve up your music anytime.
  5. DATA (MUSIC) LOSS: When hard drives are around $50 and two cents per CD, it’s fairly trivial to make a couple redundant backups to store off site and then even if your house burns down your music is safe. That old CD collection, however, will be toast. There are also fireproof, shockproof and waterproof external drives like those from ioSafe—with or without redundancy. And there are variety of NAS units with redundancy. And most of this hardware costs significantly less than an audiophile approved CD player.
  6. TOO MUCH WORK: Granted ripping a large CD collection to a hard drive is time consuming. But with lossless compression, you can easily pay someone else to do it without having to worry about the quality of the result. There are many commercial services that specialize in CD ripping. And, regardless, it’s a one time effort that pays off in less work every time you play music compared to the hassle of CDs. You can instantly find and play anything in your collection, build playlists of anything you want, enjoy playback volume leveling, and much more. You never have to find, and put away, a CD again.

OTHER REASONS? Am I missing any other reasons, besides perhaps being nostalgic or a luddite, to keep using CDs?

DIGITAL FORMATS: I couldn’t find more recent numbers, but when the open freedb CD database was cleaned of duplicates in 2006 it had approximately 2 million unique CD titles. The CDDB/Gracenote database is supposedly much larger but apparently Sony (the current owner) doesn’t publish the data. It’s probably not a stretch to say there are 3+ million CD titles in circulation. And by 2007, over 200 billion CDs had been sold. So how does that measure up to other formats?

  • 99.7% CD 16 bit 44.1 khz 3 million titles
  • 0.2% SACD DSD 1 bit High Resolution 6500 titles (mostly classical/jazz)
  • 0.06% DVD-Audio Up To 24 bit 192 Khz 1800 titles (mostly classical/jazz)
  • 0.03% Hi-Res Downloads 24 bit 88/96 Khz 1000? (mostly classical/jazz)

So, 99.7% of the digital music commonly available is 16 bit 44 khz CD quality audio (or compressed versions of the same). If your goal is to listen to music, rather than equipment, there’s essentially only one native format right now as nothing else has really caught on.

HIGH RESOLUTION DISC FORMATS: Many smart people have demonstrated CD quality audio is sufficiently transparent for high quality music playback. SACD, as you can see above, has been the most successful of the hi-res formats—if you can call 0.2% “successful”. Meyer and Moran, in 2007 conducted a year long study with over 500 trials using audiophiles, recording engineers and students, and none of them, under normal listening conditions, were able to reliably tell when they were listening to an SACD or a CD quality 16/44 version. Even to the golden eared listeners, CD quality audio and SACD hi-res audio sounded the same. Sales volumes have been so low SACD was declared a “complete failure” in 2008. Considering SACD was originally trying to fix a problem that didn’t really exist, it’s easy to understand why it failed. Meyer and Moran, naturally, were largely attacked by the audiophile community. I discuss the study more in my Subjective vs Objective article.

THE MASTERING MYTH: I know many who compare a 2 channel SACD to the CD of the same title and plainly hear a difference. And there often is a real difference that survives a blind listening test. But, as Meyer and Moran demonstrated, it’s very likely not from the format, but from the SACD being mastered differently. There’s very little SACD content that’s not been altered when re-mastered for SACD. So it’s not a fair comparison.

HIGH RESOLUTION PART TWO: If you’re still convinced hi-res formats sound better, first I would suggest reading my Subjective vs Objective article providing lots of evidence we humans hear things that don’t exist. And there are many threads at Hydrogenaudio on the subject. There have been several trials where the free Foobar 2000 software player with ABX was used for blind comparisons of the same track in high res and CD formats using high quality 24/96 playback hardware. And, almost without fail, nobody can tell the difference. One AES study on sampling rate found some extremely slight differences that only a few of the listeners could hear and only with one piece of music. This study is discussed in a Hydrogenaudio thread.

HIGH RESOLUTION RECORDING FORMATS: When it comes to recording, hi-res formats, like 24/96 PCM and DSD, are clearly the way to go. Unlike with digital audio playback, there’s little dispute hi res allows for greater dynamic range and less degradation during editing. But, for playback on the consumer side of things, any benefits of hi-res are much less obvious and, at best, extremely subtle.

LOSSY COMPRESSION: This could be a topic by itself. Even audiophiles, in blind tests with high end gear, usually can’t tell lossy compressed audio (i.e. MP3, AAC, etc.) from the uncompressed version when it’s been properly encoded. But much of the compressed audio we listen to wasn’t properly encoded—including much of what’s available from iTunes, Amazon, etc. A lot of it has been encoded and/or transcoded in ways that optimize processing speed over audio quality. And, worse, some of it has been through more than one lossy operation. So audiophiles have some justification in their dislike of lossy compression. Typical high quality lossy compression is around one fifth the size of the original making storage on portable devices and streaming much easier. It’s too bad it hasn’t been used more carefully.

LOSSLESS COMPRESSION: FLAC and Apple Lossless are the two most popular lossless compression formats and quite a bit newer than MP3. They perform the same but FLAC is much more widely implemented than Apple’s in-house format. So FLAC provides vastly more playback options. The advantage of lossless is it really is lossless. The bits you end up with are the exact same bits in the uncompressed version. Audiophiles are finally starting to come around, but for a long time many thought any compression, including lossless, sounded worse. Ultimately, the DAC can’t tell the difference.

CONSUMER DEMAND: The masses are the 800 pound gorilla. So far, there hasn’t been much demand for higher quality audio formats. In fact, as demonstrated by the huge success of iTunes alone, the vast majority are not only happy with CD quality, but lossy compressed versions that often sound worse. Those who grew up listening to iPods have even expressed a preference for the sound of lossy compression. It’s what they’re used to, so for them, that’s what music is supposed to sound like. This doesn’t bode well for the future unless preferences change or the barriers to higher quality formats, most notably network speeds, bandwidth and copyright concerns, are greatly reduced.

DIGITAL INTERFACES & BITSTREAMS: From a strictly theoretical point of view, any interface with enough bandwidth can move digital audio around. From a practical perspective there are some differences. Here are most of the possibilities:

  • Self Contained Playback (I2S) – When you use use a CD player, iPod, or PC as a self contained playback device the digital audio bitstream never leaves the device. You get an analog output and all the bits are handled internally. In reality there is an internal digital audio interface, and the input to the DAC itself likely uses the I2S interface. I2S has the advantage of keeping the clock and the data separate so there’s no issue of “clock recovery” but it’s still prone to being implemented poorly. Because the storage format doesn’t matter, it’s entirely possible to design an iPod-like device that performs as well as a high-end CD player or even better. 
  • Ethernet/WiFi/Internet/3G/Etc. – When you “stream” audio over a network you can still get “bit perfect” data at the receiving end. Networks have very robust protocols that prevent errors. Software applications can crash if even a single bit is corrupted. Digital music, by comparison, is a walk in the park as long as there’s sufficient bandwidth to do it in real time. Contrary to myth, networks don’t create jitter, reduce the soundstage width, or remove the “air” from around Yo-Yo Ma’s cello. From the DAC’s perspective, the reassembled audio looks exactly the same as if it had come off a local hard drive, CD, or from flash memory. There are only two potential problem areas with network streaming:
    • Real Time Dropouts – Ever watch a YouTube video and had it pause during playback? That’s because there’s a bottle neck somewhere between you and YouTube’s servers that restricted the bandwidth to less than required for real time video. Fortunately audio usually needs less bandwidth than video, but the same thing can still happen—especially with WiFi or mobile connections with high quality audio. This isn’t some subtle audiophile quality issue, the audio will just stop playing nearly always in obvious ways. Larger buffers help prevent dropouts.
    • Real Time Synchronization – Generally this is a non-issue. For network audio playback there’s typically some delay in the audio as it’s buffered locally by whatever is playing it. You don’t care if the bits are leaving the source a half second earlier than when the sound reaches your ears. But if you’re trying to play the same internet radio station in two rooms of your apartment using two different players, now you have a potential problem. With analog FM radio this worked great. With digital you’ll likely have an echo between the two rooms due to different buffering delays. This is not a sound quality problem and you would also have a hard time trying to play the same CD in perfect sync in two different rooms on two different players. There are network players that address this issue (i.e. Sonos).
  • Optical (TOSLINK) & Coaxial S/PDIF – This interface has been maligned by audiophiles—primarily because the digital bit clock is embedded with the data and must be recovered at the other end. The professional AES3 (also called AES/EBU) digital interface gets much more respect. What some may not realize is they’re essentially the same standards. AES3 just uses a balanced 3 pin XLR connector. TOSLINK, coaxial S/PDIF and AES3 all embed the clock and move data in the same way. The vast majority of the music we listen to, even some on vinyl, has already been through the AES3 interface. So if embedding the clock corrupts the sound quality, consider most of your music already corrupted. S/PDIF was developed by Sony and Philips (the S/P) as a consumer version of AES3 with the introduction of the MiniDisc to allow recording digital sources. Likely because some CD players started including S/PDIF outputs the audiophile industry started using it between high-end CD transports and separate DACs. These interfaces—especially with long cable lengths--can contribute to jitter.
  • AES3 (AES/EBU) – This is the professional version of S/PDIF and in most applications the two interfaces perform about the same as they essentially are the same. The advantage of AES3 is it can be sent over longer cable distances with less degradation than S/PDIF due to the balanced cable. And that can reduce jitter in some circumstances.
  • Firewire – In the mid 90’s it became practical to record, edit and master digital audio on a PC instead of an expensive dedicated Digital Audio Workstation. Around that time the IEEE 1394 “Firewire” interface was popularized by Apple. Even though Firewire predates, just barely, USB it never really caught on in a big way outside of a few limited niches. One such niche is professional audio. USB has finally caught up, but for many years if you wanted to move more than two channels of high quality audio in and out of a PC, Firewire was the preferred way of doing it. But it appears to be a dying technology just like the ill fated Apple Desktop Bus before it. Apple has now gone out on a limb with Thunderbolt and many have suggested Thunderbolt and USB 3.0 will be the demise of Firewire. In any event, it’s use in audio has mainly been limited to professional, not audiophile, gear.
  • USB – Along with network streaming, USB is one of the most popular consumer digital audio interfaces. Prior to 1998 you generally had to install proprietary drivers to interface a computer to a USB audio device. But in 1998 the USB standards group published the USB Class Definition for Audio Devices (PDF). The new standard allowed native support of USB devices within operating systems including Windows, OS X, and Linux. USB audio then gave new meaning to “plug-and-play”. For more than a decade it’s been possible to plug a high quality external DAC into just about any PC and it magically Just Works. So, it’s all the more puzzling there are so many people still messing with little silver discs. But there are a few more things worth knowing about USB Audio:
    • USB Jitter? – There’s been a lot of talk about “asynchronous” USB DACs and various jitter issues with the USB interface. Part of the issue is that most USB DACs, unlike network players, don’t have a real digital audio buffer. Instead, the buffering is done in the operating system at the driver level in the PC. So the data is being streamed, more or less, in real time over the USB cable. Any jitter, however, is a function of the USB hardware within the DAC or other audio device as USB packets of data are completely different than a digital audio audio datastream. It’s the conversion process that’s tricky but the important thing to know is it’s an entirely different data format that either works or it doesn’t. USB problems are rarely subtle. For more details, see my jitter article.
    • USB Hi Res is Rare - There are only a few USB chips that can handle anything above 16/48 and some of them are difficult or costly (in licensing fees) to implement. As long as that’s true, the majority of USB audio devices, despite often having DAC chips that can do 24/96 or 24/192, can only operate up to 16/48 via USB. So be careful when shopping if you’re expecting hi res audio via USB. Some other devices, like the popular E-Mu 0202 and 0404 require proprietary drivers that can be problematic.
    • USB Cable Myth – There are now lots of “audiophile” USB cables but USB (and HDMI) data transmission either works or it doesn’t. When it doesn’t the results are generally obvious in the form of ticks, pops, dropouts, errors from your computer, the device locking up, etc. The cable doesn’t create jitter so a better cable can’t have less jitter, enhance the sound stage, or otherwise work any subtle magic on the sound quality.
    • USB Audio 1.x – This is the original 1998 standard and supports up to 24/96 two channel stereo audio in one direction (either recording OR playback but not both at once). It uses the USB 1.x Full Speed (12 Mbit) interface and uses less than 5 Mbits of bandwidth. It’s capable of flawless high quality audio as demonstrated by golden-ear approved products such as the Benchmark DAC1 and Grace Designs m903. This is sometimes known as “Class 1 USB Audio”.
    • USB Audio 2.0 – This was introduced in early 2009 and has been adopted by a few manufactures—mostly in pro audio so far. This “Class 2” interface uses the much faster 480 Mbit USB 2.0 standard and allows for 24/192 multi channel audio in both directions simultaneously. The few audiophile companies who have adopted 24/192 Class 2 have found it to be a can of worms. Ayre, for example, specifies never using USB hubs, keeping the USB cable to 3 feet or less, not sharing the USB interface with any other devices, and warns that not all PC’s and USB hardware will work correctly. Unless you’re doing multi-track recording, all this is rather unimportant as Class 1 24/96 is all anyone needs (Meyer and Moran have shown 16/44 playback is sufficiently transparent 99.9% of the time and 24/96 is overkill).
  • Internal PC Slots – Some still advocate putting a high quality sound card, like the Asus Xonar ST/STX, inside your PC in a PCI slot. But this is flawed in many ways including a much more electrically noisy environment, slots inside PCs are going away, and it generally requires proprietary drivers that are often buggy. Like CDs, it’s a very 1980s solution. I suspect the Xonar ST and STX are among the last of the audiophile-grade 2 channel sound card dinosaurs.
  • HDMI – Increasingly HDMI is being used to move digital audio as well as video. It supports many formats including plain old PCM 16/44 CD quality. And, like USB, the quality of the implementation is typically the limiting factor, not the cable or standard itself. The main advantage of HDMI is ease of use for A/V gear and it supports multi channel digital audio formats including SACD’s DSD, Dolby TrueHD and DTS-HD Master Audio—all of which are lossless. A modern Blu-Ray player can send many high quality audio formats directly to an A/V receiver or A/V processor/preamp.
  • Other Options – Some recording studios use externally clocked AES3 with a single central high-end clock. In this case a completely different cable carries the clock. This is generally used where multiple digital devices are best operated in sync with each other as with multitrack live recording for example. This generally isn’t applicable to home playback. As mentioned above, USB 3.0 and Thunderbolt will likely be used someday for audio but most likely for recording.

JITTER: I wrote an entire article on jitter but it’s sufficient to say it can be an audible problem. I’m not sure how often it actually is audible, but it seems reasonable to try and minimize it. Which brings me to my next point.

THE MULTI BOX MYTH: Audiophiles seem obsessed with dividing up the functionality of their systems. Some have a tuner, phono pre-amp, general pre-amp and one or more power amps instead of a receiver. Like many audiophile myths, this has the strong potential to make things worse rather than better. The “two box” CD player is a perfect example. There’s nothing about a CD transport that’s incompatible with having a DAC in the same device. Yet audiophiles seem to think splitting them apart makes for better sound. It usually requires sending audio through the very S/PDIF interface they don’t like. S/PDIF embeds and reconstructs the clock creating potential jitter. This also applies to network media players, external DACs for iPods, and USB-to-S/PDIF boxes. In general, it’s better to have one engineer (or team) responsible for as much of the playback chain as possible. That way you can be assured everything is designed to play nice together. And, more important, everything can run from one clock which can greatly improve jitter. In the headphone world, a self contained headphone DAC (one box) has the potential to perform much better than a USB-to-S/PDIF box connected to a DAC connected to a headphone amp (three boxes). But the more audiophiles spend in their pursuit of audio nirvana, the more likely they use multi-box configurations. Perhaps they would also prefer a car with the engine in a trailer?

HIGH RESOLUTION PART 3: Now that storage is cheap, and the Internet hasn’t yet been crushed under YouTube, Netflix and Spam, I suspect we’ll gradually see more hi-res audio. Apple has strongly hinted they’re working towards offering 24 bit downloads in their iTunes store. I’m all for Apple improving the quality of their music offerings, but that’s often more about what’s behind the scenes—lossy compression, transcoding and improper sample rate conversion—than bit depth. For the consumer, especially playing music on an iPod, 16 bits is plenty. What Apple 24 bit audio is really about is more profit for Apple. It would provide incentive for people to buy new Apple hardware with 24 bit support, and “upgrade” their music by re-buying it in the new format. It also could, no doubt, help swing many audiophiles into buying from iTunes rather than elsewhere. So it’s a great marketing move, and if it’s lossless, it might mean slightly better sound compared to lossy AAC. But lossless 24 bit files are comparatively huge. And, to do it right, they have to start from scratch with the original studio masters. So there are some serious hurdles.

MULTI CHANNEL: Multi channel audio, without video, has never really caught on. Most of the SACDs sold are multi-channel remixes of popular CDs. And, on a system optimized for multi channel music (versus movie) playback, some offer a very immersive and impressive sound. But not enough consumers cared for SACD (or the similar DVD-Audio) to ever get any real traction. I think part of the problem is a lot of music listening is either done casually around the home rather than sitting in the sweet spot. And many others listen with 2 channel headphones. Plus multi channel systems optimized for home theater are rarely ideal for multi channel audio. For movies you want a diffuse rear sound field and audio is usually mixed assuming a focused rear sound field. And now we have Dolby TrueHD and DTS-HD Master Audio--both of which support lossless multi channel audio and pound more nails into the SACD coffin. Many Blu-Ray players can output both via HDMI. With downloads replacing physical media it will be easier, at least in theory, to distribute multi channel audio in the future. The trick will be playing it how you want, not how say Apple wants you to.

DRM: Many consider DRM (Digital Rights Management) a bad thing. It’s basically copy protection and it does get in the way. But it also allows for some fairly amazing things like inexpensive subscription streaming music plans from Napster, Rhapsody, Spotify, etc. For around $10/month you can have access to a giant catalog of music with millions of tracks. Sure it might be in a lossy format, but much of it sounds relatively good. For those who haven’t tried it, it’s rather addictive to be able to listen to most anything, legally, anytime, with just a couple clicks or taps. Without DRM, it wouldn’t exist. There’s usually a way around most DRM but sometimes it’s not very practical, or convenient, and it’s usually illegal. So I see DRM as a worthwhile evil.

THE FUTURE: Two channel audio is pretty well covered in terms of formats. There are a few sources for lossless downloads, like HD Tracks, where the labels have agreed to it. But, much like SACD, the catalog of titles is so far very limited. The right DRM would help free up the log jamb as more music labels would then sign on but that will likely limit playback options. Apple’s 24 bit could be such a DRM format. Downloadable Dolby TrueHD and/or DTS-HD Master Audio multi channel lossless audio are also possibilities but there’s the same DRM issue. I can imagine a future where it’s easy to buy lots of high quality lossless music online in downloadable formats. But I bet only a minority will be DRM free. The rest will likely be restricted to authorized playback devices. Such devices will likely have a closed, self contained, architecture to prevent easy removal of the DRM. Apple, with their relatively closed proprietary world, is in a perfect position to blaze the trail. And, if that works out, we might get lucky and others will follow. But it will only be happen if there’s enough money to be made. And so far, the track record on that isn’t good. But if anyone will pony up for supposed higher quality, it will be Apple’s demographic. In the meantime, if you’re still fooling around with silver bits of plastic, consider an upgrade to 21st century technology.

MORE INFO: The Well Tempered Computer site has a lot of good information if you’re new to computer audio.

39 comments:

  1. Very interesting entry once again.I guess you are building up a huge fan base amongst subjectivists ;).

    I find the AAC 320Kbs codec a really nice compromise between quality and size. You really need a studio quality room and gear to notice the difference between it and flac. It's more efficient than mp3 specially on the mid region where our ears are more sensitive.

    ReplyDelete
  2. Another great article I see :)

    Regarding compression - even with decent bitrates it can be done badly... I have some Amazon MP3s at 256kbit/s where I can hear compression (clearly audible artefacts) while some others are compressed with a variable lower bitrate and sound fine as far as I can tell.

    Buying a CD removes that danger - you get the best they will sell you (which can still be bad - e.g. loudness war) and then decide for yourself what you do with it - plus it is a lovely backup.

    For practical purposes I use 320KBit/s MP3 for all my music from CDs (my MP3 player doesn't play FLAC, I never got MP4 tagging to work properly...) - streamed from a Windows Home Server (reripped pretty much nearly everything summer 2010 when I got it). I suppose I could use a lower bitrate an still get a good sound even from a lossy codec – but as you pointed out correctly, storage is cheap.
    -> This needs about 45KBit/s of bandwidth when streaming - home wi-fi with plenty of networks still get's 1MB/s + depending on interference (got up to 2,8MB/s once on g).
    Another neat trick for a home network is to use foobar and buffer the song - this means that foobar will prefetch the data and even if your network drops out for a few seconds the audio player will continue to play the music.

    And my CDs - backups just in case I lose the file (shouldn't though).
    Additionally in a lot of cases the price difference between MP3 and CDs on Amazon is small - but in some cases the CDs are impossible to get or (as imports) so overpriced that MP3s (or similar) are the only choice.
    Lastly: It is also nice to have a shelf full of CDs - in the same way that it is nice to have a shelf full of books.

    ReplyDelete
  3. Regarding differences between CD and SACD:
    Some months ago, on a big headphones forum, a L*nn Records staff member announced that they offered free to download/listen to CD and SACD version of the "same" track. So everyone interested could listen to the versions and clearly hear which one's better.

    Right, so I looked at the website where they offered the files. The first thing I noticed was that the overall tracks' waveforms looked different for both versions! ... Do I need to say more?
    Not much later the same guy admitted that they mastered the track differently for the (way more expensive) SACD.

    Format comparison FAIL.

    ReplyDelete
  4. Thanks for the comments so far. And sorry for the numerous errors and typos in the article which I'm still finding and fixing. I was a bit rushed trying to make my self imposed Wednesday deadline.

    As for L*nn Records, that's been exactly my experience trying to compare formats and files supplied by the people selling high resolution music, gear, etc. And that's one reason Meyer & Moran, in the big AES study, used the same SACD and simply "down-converted" them to 16/44 rather trying to find the exact same material in 16/44.

    Something similar might happen with Apple's 24 bit format. They could tweak the files in ways that intentionally make them sound different.

    As for lossy compression, it's a huge topic by itself. Hydrogenaudio is a great resource for comparing CODECs, bitrates, etc. In my opinion, the debate has lost a lot of steam now that lossless is so practical. Even my cheap Sansa Clip+ plays FLAC files. And, if for some reason I want to put a lot of music on a portable device, I just tell Media Monkey to transcode the huge playlist from FLAC to lossy on its way to the target.

    As for building up my subjectivist fan base, too many voices in this industry are already sugar coating the truth trying to appeal to the subjectivists who spend far more money than the objective crowd. But I don't want to be like the editor of Sound & Vision dancing around the Wired Wisdom article demonstrating expensive cables are a waste of money (see Subjective vs Objective). Unlike Sound & Vision et al., I don't have to worry about subscriber sales, ad sales, or product sales.

    ReplyDelete
  5. Flac on portable players:

    The only problem here is that most don't support it.
    You comment on your SansaClip - I guess that's just because the manufacturer actually chose to support it.

    Sony doesn't - pretty sure apple doesn't but then I don't care about apple anyway...
    Add to that that plenty of people buy many very cheap products, they also won't be graced with Flac support.

    Lastly:
    Transcoding - indeed, that is the solution, but considering what I see in terms of computer use... the older people seem to know less and less about computers and the younger ones know their consoles much better... computer use extends to Word, maybe some Excel and playing games...

    ReplyDelete
  6. My, what a novel.

    Re: lossy compression.

    Apparently the big labels like to store masters (or have them stored) in 384 kbit/s or thereabouts MP2. Now the MUSICAM/MP2 codec has been shown to be not too fond of 0dBFS+ levels (Nielsen/Lund). Oops.

    Re: internal soundcards.

    Good hardware with shitty driver support has unfortunately been a recurring theme in semi-pro soundcards for just about as long as they exist. Remember the Terratec DMX Win2k driver fiasco? And what was that 20-bit card with a similar fate again (Hercules? Hoontech?)?

    However, there is no way around internal cards if you want to keep overhead low. It wasn't uncommon to see 1% load during 16/44 playback 10 years ago.

    Re: formats.

    You forgot that people don't necessarily want convenience only. They're buying records because they're after an experience. There's a nice big cover to behold (a definite plus of LPs), and the whole listening process demands active participation. This is pretty much the antithesis to background music, which also is quite big these days.

    Ironically, the vinyl versions commonly do sound better than their CD counterparts these days because they are not compressed to death. (Random example.)

    That makes sense from a market point of view (active listening is more likely to take place with the big black discs, plus there's not much money to be made with CDs these days), but from a technical one can only be seen as proof that human stupidity is, indeed, infinite. First-rate vinyl playback is not easy or cheap to achieve at all, and even then the occasional crackles and lack of usage comfort would still be bothering me.

    Re: CD players.

    Assuming you do intend to stick with a standalone player in the first place, a dedicated CD player has a few definite advantages:
    It starts up almost instantly and accepts a CD in a matter of seconds, and it'll also have a number of dedicated playback controls at the device itself.
    (This kind of discussion, btw, regularly crops up in hifi forums.)

    Re: DRM.

    This has got to be the most unpopular technology in history. The public said pretty clearly, "We don't want no stinkin' DRM", and so even Apple eventually caved in and started selling unprotected tracks, following Amazon.

    Subscription services seem to be the last resort where it still finds use, with prices being low enough that people can accept content protection as a necessary evil.

    I don't think the Digital Radio Mondiale folks were too happy about this choice of acronym though. (That other DRM is a digital broadcasting standard for the AM ranges and was there first.)

    Re: lossless/lossy formats.

    Like many people these days, I keep my main music library in FLAC on the computer, plus most of it transcoded to MP3 -V 6 to -V 4 -q 0 for the Clip+.

    ReplyDelete
  7. The Sansas all support FLAC from the factory, as does Cowon and several others. And anything that can run RockBox supports FLAC and there are dozens of RockBox compatible players. To be honest, I have lots of portable players, but the Clip+ is the one I use 95% of the time. It's just a great player.

    Software like Media Monkey, in my opinion, is less confusing than iTunes. iTunes is designed to get you spend money at Apple rather than work with what you already have.

    Media Monkey is designed to let you listen to, and manage, your music collection however you want--not how Apple wants. You set up your player preferences once for your portable player(s). And you can specify if the bitrate is greater than some value (say > 320 Kbit) to automatically transcode the content when copying it to the portable. That way FLAC files get transcoded and MP3s don't.

    So after you set up your player once, it's just drag and drop. The transcoding goes on automatically in the background. I can drag an enormous playlist to the Clip+ or Fuze in just a couple clicks.

    ReplyDelete
  8. To Stephan, thanks for all the added comments. It's common to have digital album art available so I'm not sure the experience of playing a real CD vs the ripped version is all that different for most. But for those who like to browse the "liner notes" I can agree that's usually missing.

    The startup time is applicable for some devices but not others. The Sonos players, for example, are designed to be left on 24/7 and you have instant access from the Sonos remotes, iPhone, Android phone, iPod Touch, iPad, Android tablets, and any Mac or Windows PC, etc. And an A/V receiver that can play network files is (obviously) already powered up when you want to listen to music.

    ReplyDelete
  9. "I find the AAC 320Kbs codec a really nice compromise between quality and size. You really need a studio quality room and gear to notice the difference between it and flac. It's more efficient than mp3 specially on the mid region where our ears are more sensitive."

    I just wanted to post that AAC hasn't proven to be more efficient or transparent than LAME mp3 in recent years. . . kind of sick of audiophiles claiming AAC is significantly better than mp3. Neither CD's nor mp3's need to be replaced with anything "better."

    -Satellite

    ReplyDelete
  10. Re: "Those who grew up listening to iPods have even expressed a preference for the sound of lossy compression."

    http://seanolive.blogspot.com/2010/06/some-new-evidence-that-generation-y.html

    "Listening Test Results: Students Prefer Music in Lossless CD Versus MP3 Formats"

    ReplyDelete
  11. Thanks for the comments Satellite. Last I checked, AAC had an advantage at 128K and lower bit rates. But once you hit 192K I think it's harder to pick a winner. For me, it's worth it to use FLAC for archival purposes and piece of mind and transcode as needed for portable use. Lossless provides maximum options for the future.

    ReplyDelete
  12. NwAvGuy, another great write up. An audiophile here but one based on science and confirms everything I already knew. It's not the digital format or sampling rate that matter (for playback), but the dynamic range of the track within that sampling rate. For example, enjoy the great sound on 128kbps AAC!

    http://www.youtube.com/watch?v=Zi3i-HqDNFI

    Of course it's different for recording as you need the bandwidth to dither put noise and other sound enhancements, thus the need for 96kHz, which also cuts out any of the potential problems that lower sampling rates give. Dan Lavry, in his paper on digital sampling rates, thinks the optimal sampling rate is between 60-70kHz, as it accounts for every minute sound that any musical instrument can make and gets rid of any potential issues with 44.1 and 48 Khz sampling rates.

    Here's my comparison via a free spectrogram program called Spectro comparing Quicktime TrueVBR AAC @ ~192kbps and LAME 320kbps CBR mp3. The AAC sample has been converted to WAV only for it to work with the program otherwise it is 100% representative of the AAC sample.

    http://www.iforce.co.nz/View.aspx?i=5cyb1x4x.cfn.png

    And LAME VBR -v0 mp3 vs. AoTuV 5.7 Ogg Vorbis @ -q6

    (I know you dislike Head-fi but it's in the Sound Science section and I cannot be bothered rewriting it again)

    http://www.head-fi.org/forum/thread/484162/formats-bitrates-and-battery-consumption/15#post_6598535

    However Ogg Vorbis is a much more complex algorithm than AAC or mp3 thus even though it offers the best compression, it uses more CPU cycles to decode, which for portable devices, means shorter battery life. AAC is best as although it uses slightly more CPU cycles to decode than mp3, less bitrate is used for the same (or better) quality vs mp3 and mroe than negates tyhe extra cycles used to decode AAC. i.e. CPU cycles saved via less bitrate via AAC over mp3 is more than the extra CPU cycles used to decode AAC over mp3. AAC means in practical terms, particularly for portable devices such as iPods, DAPs etc.. more capacity and longer battery life.

    ReplyDelete
  13. Satellite, look at my spectrogram links. AAC is a far superior codec than mp3. Has more features, more efficient, more accurate and compresses better.... Basically ~192kbps AAC is scientifically better than VBR -V0 / 320kbps CBR mp3. VBR -v0 and 320kbps CBR mp3 are EXACTLY THE SAME bit-by-bit. There's little to no point of CBR mp3 these days. mp3 is an outdated codec. There's a good reason why ISO replaced mp3 with AAC as the audio codec in the MPEG-2 standard! Yes, that long ago! The most modern MPEG standard is MPEG-4. Remember of course mp3 stands for MPEG-1 Layer 3.

    ReplyDelete
  14. @Karl, spectrograms tell you nothing about perceived quality or how transparent an encoding is. A 18 kHz low pass filter may be less problematic than other compression artifacts which you do not see in a spectrogram.
    If you do not agree feel free to ask on hydrogenaudio.

    ReplyDelete
  15. "spectrograms tell you nothing about perceived quality or how transparent an encoding is" - They can if you compare them to the original lossless source's spectrogram like I have. Both Vorbis and AAC has less artifacting and better represents the lossless file. Perceived quality is another but really, regardless of perceived quality, AAC compresses better, resulting in better quality for less bitrate AND is more accurate. As stated, this results in practical terms, more songs in the same capacity and longer battery life as the CPU is processing less bitrate. That means really perceived quality is not really an issue here. Tbh, I should of ticked the low-pass filter option off as it's not relevant. You can do it in 5 minutes really like I did. There's a freeware spectrogram called Spectro. Just convert the AAC file to WAV as currently it doesn't support AAC. The result is more than repeatable across any song. Loudness war'ed songs tend to be better as they contain more digital infomation. Here's another test I did which has the original FLAC spectrogram to compare. You can easily tell the LAME 3.98.4. file has more artifacting or at the very least, much less accurate and representative of the FLAC file vs the Vorbis file.

    - http://www.head-fi.org/forum/thread/493173/320-aac-vs-flac/45#post_6667267

    AAC is much the same as Vorbis but if you really want, I can create another GIF for you. There will be other factors that are not measured on a spectrogram but really, you tell me what's the more accurate codec looking at those?

    ReplyDelete
  16. This is getting a bit off topic and the AAC vs MP3 debate is bit like Ford vs Chevy--it can go on forever. Perhaps you might want to take this to the head-fi thread referenced?

    For everyone else, just use FLAC. Storage is cheap. ;)

    ReplyDelete
  17. However FLAC has big negatives on things with restricted storage space and battery life like portable devices. Big bitrates = big loss on battery life mainly. Also larger files = less capacity. Backlighting is the biggest consumer of battery energy, followed by file bitrate in portable devices. FLAC and other lossless sources has no problems with PCs because as stated storage is cheap but also you have an 'unrestricted' power source in the power supply unit as it's supplied by mains power. Nothing wrong with a bit of computer science talk on an electronics blog is it? :P Audio has lots of different aspects to it.

    ReplyDelete
  18. There's nothing at all wrong with some computer science where applicable. Those are good points. Do you have a link to some hard numbers that compare battery life (or even better power consumption) of MP3 vs FLAC?

    As I said, at least with some music management software, it's relatively trivial to transcode an entire playlist with a few clicks from FLAC to the lossy format of your choice. Media Monkey can do it automatically in the background.

    I think lossless at home and lossy on the go is ideal in most ways.

    ReplyDelete
  19. Different Rockbox Builds but it gives a good general indication of codecs and bitrate and their influence on battery life, at least on Rockbox-ed Sansa models: http://www.rockbox.org/wiki/SansaRuntime

    As shown there, although codecs might not have a large influence, firmware and codec decoder does e.g. look at the e280's battery life with VBR mp3 and higher bitrate Ogg Vorbis with the same Rockbox build. Although codecs don't make a significant difference battery life wise, AAC and Ogg Vorbis still compress better than mp3, resulting in more songs in the same capacity, not to mention both Vorbis and AAC are more bit-accurate and artifacts less than mp3.

    Here's another academic report stating the difference in power consumption between different lossy audio codecs: http://islab.csie.thu.edu.tw/files/teacher/282.pdf

    ReplyDelete
  20. "Satellite, look at my spectrogram links. AAC is a far superior codec than mp3. Has more features, more efficient, more accurate and compresses better.... Basically ~192kbps AAC is scientifically better than VBR -V0 / 320kbps CBR mp3. VBR -v0 and 320kbps CBR mp3 are EXACTLY THE SAME bit-by-bit. There's little to no point of CBR mp3 these days. mp3 is an outdated codec. There's a good reason why ISO replaced mp3 with AAC as the audio codec in the MPEG-2 standard! Yes, that long ago! The most modern MPEG standard is MPEG-4. Remember of course mp3 stands for MPEG-1 Layer 3."

    I'm sorry but there is a lot of misinformation here. My statement is accurate, AAC hasn't proven to be any more efficient or transparent than mp3 recently, even at 128, surprisingly. A spectrogram is rather pointless, listening tests have to be done, and they have - see hydrogenaudio. v0 and 320 CBR are in fact NOT exactly the same bit-by-bit. By definition they cannot be. I agree that there is little point in CBR, however, no argument there.

    One more thing, I was told by saratoga @ ABI that Vorbis does not use up more battery power because it is a complex codec, and that is a myth, but I wouldn't know for sure.

    I barely use lossy at all anymore but anyway. . . just trying to set things straight.

    -Satellite

    ReplyDelete
  21. Energy consumption and codecs - if you follow Sony's spec there is a small difference.
    Personally I find that their playback spec for MP3s is pretty much spot on (for a new product)

    http://www.sony.co.uk/product/nws-x-series/nwz-x1060/tab/manual
    -> Open the English manual and check page 176

    Or just to copy the contents:
    Music
    Playback at MP3 128 kbps Approximately 33 hours
    Playback at WMA 128 kbps Approximately 31 hours
    Playback at AAC-LC 128 kbps Approximately 29
    Playback at Linear PCM 1,411 kbps Approximately 31 hours

    ReplyDelete
  22. NwAvGuy, thanks for the great blog, it's really nice to have a place to listen to reason. Trying to get objective info from many other places is a real challenge most of the time.

    I still prefer CDs to other digital formats. What was very interesting is what you didn't say; that the other formats produce better sound. Like others, I still enjoy the process of picking out and putting on a CD, I like having my source material handy. And while I do enjoy my iPod for portability and am starting to put my library on a music server, it is nice to know the CD is still unsurpassed for quality.

    ReplyDelete
  23. You're welcome Mark. You're correct in that CDs, and any lossless digital version of CD audio, is unsurpassed for playback of music from a subjective point of view (as demonstrated by Meyer/Moran and others). Lossless music on your iPod can sound every bit as good as your CD player. In terms of the format itself, both are equal. So it comes down to the rest of the signal chain. And I can say an iPod Touch 3G is actually a lot better than many of the early/cheap CD players and certainly better than many portable CD players.

    ReplyDelete
  24. I don't think DRM will exist in the future. It's pretty dead. As far as I know, it's restricted to subscription services nowadays. I hope Apple won't be evil enough to re-implement it with 24 bit audio. I just want lossless CD quality downloads, nevermind those huge 24/96 files. . . can't see it happening either way.

    Late thought.

    -Satellite

    ReplyDelete
  25. NwAvGuy, I just discovered your blog and I'm totally overwhelmed by this wealth of information, thanks a lot for all your efforts.

    Regarding your question why so many people still prefer CDA in its physical form - Well, like Mark I do too, because I think an album is more than the sum of its parts, e.g. the music and the artwork; it can be an integral work of art - many artist take great care about the production of the album as a whole, I know some independent artist who even send out limited editions of their albums in hand made editions, made out of special materials, with hand-written booklets etc. Such a lovingly produced artifact can never be replaced by a digital download, it gives the music its own, dedicated physical environment - it is so much more than just some profane "backup".

    Of course that's irrational, but so is art ;-)

    ReplyDelete
  26. The best lossy audio codec is Musepack. Period. Size of MP3s, quality of FLAC.

    ReplyDelete
  27. Hey anon. Lots of lossy codecs have been proven to be sufficiently transparent that are far more compatible. Having your music in a weird format is likely to be a big problem for many.

    ReplyDelete
  28. Sadly you're right... For a mysterious reason this format is not widely supported, even if it's open-source... It performs far better than OGG, but has no recognition.

    ReplyDelete
  29. Thank you all for your comments. I received one additional comment pointing out P2P file sharing has come a long way. But given the legal issues surrounding P2P and music, I'm not publishing the comment as I can be held responsible for the comments here. I did, however, revise the article.

    ReplyDelete
  30. In defense of hi-def formats such as SACD, the now defunct DVD-A and the 24bit files, may be the human ear can not perceive the higher resolution of this formats but much better care is taken in the production of music released on this formats. That's why they are popular with Jazz, Classic and less mainstream music. The dynamics for example are better preserved and even the re-releases of popular albums are far superior to the CD or mp3 release, a good example is Roxy Music's Avalon, it sounds fantastic on SACD. Of course it is not the norm, I could imagine a Metallica 24 bit release being as compressed as the 16 bit format.
    The problem now a days is not the format we choose but the quality of the music released, everything is compressed as hell.

    ReplyDelete
  31. @Christian, a very valid point and I (and others) agree. It's a sad (and expensive) means to better produced music. As an aside, even the CD of Avalon is way above average but then again it largely predates the worst of the loudness wars.

    ReplyDelete
  32. I'm becoming a bigger fan with each read. Keep up the good work.

    I engaged in a 4 year CD Player Build with a good Romanian EE friend of mine. You can follow it here:

    http://www.diyhifi.org/forums/viewtopic.php?f=10&t=1441

    You may need to setup an account to see the pictures..annoying but worth it.

    ReplyDelete
  33. Hello, Northwest AV guy. Awesome article.

    As a fairly recent reader of yours, I'd like to thank you for this amazingly interesting blog. The content here is so much above the internet/audiophile crowd, it's refreshing and enlightening. I love your honest and direct tone, and in-between-the-lines harmonics aren't bad either ; )

    It's great that you provide reality-proofed concepts when assessing various audiophile concerns, notwithstanding the impeccable scientific rigor you apply to your testing process. I, for one, always trusted my ears more than my readings, so it's nice—and refreshing—to see the two actually meet. I can only speak for myself, but I do see a lot of "truth" in your words: your blog does a really fine job teaching me exactly what matters in audio fidelity.

    I can count on one hand the publications (press, net, whatever) that provide(d) the same level of BS-free knowledge. Your work is simply brilliant, Sir.

    ReplyDelete
  34. Hello, Northwest AV,

    I'am really impressed with the quality of your articles. Thanks a lot. Regarding DIGITAL FORMATS, I have doubts about audio hi-res format in blueray discs. What is your opinion? Have you listen to those universal Disc Blue-ray units? I'am thinking of buying an Oppo model.

    ReplyDelete
    Replies
    1. Hi-Res audio has been shown not to have any significant audible advantages as a playback format. It is useful for use in recording studios. For more see my What We Hear article.

      Oppo makes some great players that do a great job playing music in any format. But don't buy a Blu Ray player just for audio.

      Delete
  35. Amazing article and one I needed to read as I was considering a CD player. I have a few noob questions if you don't mind.

    If I got a Macbook Air and stored my music on an external hard drive connected via USB would I lose sound quality? Will the quality be the same as if the files were stored on the Macbook?

    The part about Hi Res USB confused me. You say that USB can only handle 16/48 so does that mean that a USB DAC will only handle 16/48 when connected via USB despite the DAC being 24/192?

    I'm still learning about audio and equipment so I really appreciate this blog.

    ReplyDelete
    Replies
    1. The USB DAC situation is changing rapidly. A lot of older USB DACs, regardless of the DAC chip inside being 24/192, only supported 16/44 and 16/48 via USB. That's due to the USB interface itself. USB Audio Class 1 (i.e. UAC1) supports up to 24/96 over USB and some better, and newer DACs, support that. UAC2, or special proprietary drivers, sometimes allow up to 24/192. But, in reality, the higher sampling rates have never been proven to be audibly superior.

      Put another way, even 16/44 audio can be entirely audibly transparent (as demonstrated by the Meyer & Moran study and others). There is some benefit to 24 bit support (i.e. 24/44) if you want to control the volume in software. See my ODAC Article for more info.

      Delete
    2. Thanks for the reply... Very helpful. What do you think of computer audio software like Amarra? It allows you to catalogue using itunes but has flac playback features etc.

      Delete

PLEASE NOTE! If you see "LOAD MORE..." OR "LOADING..." just above this paragraph, you need to click that link to load the newest comments. It may take several seconds. When all the comments are displayed the Load More/Loading link disappears.

COMMENT APPROVAL: Comments may take anywhere from a few minutes to a few days to be approved. You might want to save a copy (highlight Ctrl-C) of your comments before submitting in case something goes wrong. The approval process is necessary to prevent obscenity, etc. Please keep comments on topic. Comments promoting commercial interests will not be approved and may be reported as SPAM.