PURE AUDIO BLU-RAY
Stefan Bock’s background has included DVD-Audio, Video, SACD, and Blu-ray. Recently he has been working on Pure Audio Blu-ray, which he called a new way to bring high-definition audio to a wide range of homes. There’s an option for multichannel audio and MP3 content, as well as additional content that the artist wants to include. These disks will play on any standard Blu-ray player, as opposed to t h e f o r m e r s i t u a t i o n w i t h D V D t h a t required a special DVD-Audio player to play DVD-A disks. It’s important, said Bock, that the user doesn’t have to turn on his television to control the disk replay—it should be possible simply to put in a disk and play it. However without some means of display and control it can be difficult to select the audio stream to be replayed, so a solution was needed. Every player has the standard four col- ored buttons on the remote control, so Bock had found a way of using these but- tons to select the audio format (e.g., 2-channel stereo, 5.1 surround). There is also a display available for use on a TV screen if one is connected. Java code stored on the disks is used to implement some of the Pure-Audio-specific features. At the moment the only company that can offer this option in mastering is msm, but they are working closely with others to try to make it available elsewhere.
Pure Audio Blu-ray can work at up to 192-kHz/24-bit resolution and there are a l s o t h e l o s s l e s s l y c o m p r e s s e d H D formats introduced by Dolby and DTS, as well as FLAC and MP3, but it’s possible to h a v e a l l t h i s c o n t e n t , i n c l u d i n g 7 . 1 surround, on one Blu-ray disk without
running out of space as one can store around 50 Gbytes of data. mShuttle is an o p t i o n t h a t w a s i n t r o d u c e d f o r P u r e Audio Blu-ray disks, which turns the player into a small web server, enabling audio files to be served from the player over the home network to other devices. That way the content can be used on portable media devices or played out of alternative file-based audio players. This requires a player working to “Profile 2.0,” w h i c h w a s i n t r o d u c e d i n 2 0 0 9 a n d i n c l u d e s t h e p r o v i s i o n o f n e t w o r k functionality.
ISSUES WITH AAC AND MP3
Robert Bleidt of Fraunhofer spoke about “how to make great-sounding tracks using AAC and MP3.” (AAC and MP3 are both lossy compressed audio formats.) AAC has an installed base of about 5 bil- lion devices, said Bleidt, across all plat- forms, players, and browsers. When mas- tering for compressed tracks it’s a good idea to start off with the highest quality master possible and back off on any clip- ping or hard limiting. Using the best encoding software possible is another important factor, as not all systems are created equal, and it’s also important to check the sound quality on a simple decoder to get an idea of how it will sound on older phones or music players.
D i s t r i b u t o r s d o n ’ t u s u a l l y a c c e p t encoded bit streams, so there are differ- ent degrees of control that you may have over the delivered content. Depending on the nature of the mix and the mastering style employed, it’s possible that you may not need to make separate CD and AAC masters. A best-case scenario when it comes to control over the content can be found in Apple’s “Mastered for iTunes” program (see below). Before this came a l o n g , y o u s e n t a C D t o A p p l e , t h e y encoded it, and you hoped it sounded good. With the new system you can pres- ent a 24/96 master recording to a copy of the encoding software yourself, listen to the result of the encoding/decoding, and tweak the master accordingly. The origi- nal file is then sent to Apple, which encodes it using the same process, so that what the customer gets is exactly the s a m e a s w h a t y o u h e a r d . T h e Sonnox/Fraunhofer plug-in also allows you to audition the results of encoding and decoding in real time, using conven- tional workstation software. As far as Bleidt was aware, Apple is the only opera- tion to offer such a mastering program, so it’s more difficult to tell with other
services what the quality of the distrib- uted audio will be.
If you can hear coding artifacts when using data-compressed audio it’s very hard to offer generic advice about what to d o t o i m p r o v e m a t t e r s , s a i d B l e i d t . Fraunhofer is always interested to hear examples but has yet to perfect a “remove artifacts” feature in its encoders, he quipped. Sometimes it can help to back off the peak levels a little in order to avoid internal clipping problems that can lead to artifacts. This rather suggests the need to avoid the tempting level-maxi- mization options often offered in work- station software. The clipping behavior of d e c o d e r s v a r i e s , w i t h s o m e o f t h e m including mechanisms for avoiding clip- ping when signal levels are very hot. Simple decoders on older players, such as the VLC player for PC, tend to exhibit more noticeable clipping under these circumstances, and the Sonnox decoder can be used to audition the worsts effects of decoder clipping. It’s also likely that the effects of any encoder clipping will be more audible the lower the bit rate, so it’s particularly important to keep the levels down when compressing material for very low bit-rate streaming, for example.
The most challenging tracks for encod- ing are usually single instruments, which don’t give rise to a lot of the masking that arises in a busier track. If you reduce the bit rate far enough with any material then there is no doubt you will hear arti- facts, but at the bit rates typically used for downloads these days artifacts are more rare. (Whereas previously a bit rate of 128 kbit/s was used for iTunes tracks, they are now encoded at 256 kbit/s.) Double-blind comparison tools are avail- able in the Sonnox plug-in that enable u s e r s t o c o m p a r e t h e o r i g i n a l a n d encoded versions without knowing which
Robert Bleidt of Fraunhofer USA discusses issues with AAC and MP3.
Stefan Bock, the leading light of the Pure Audio Blu-ray format
is which. This allows a much more reli- able way of determining whether artifacts are audible, as people are notoriously unreliable when they think they know what they’re listening to. It’s best to home in on short sections where artifacts are thought to be audible, then do an ABX comparison to discover what percentage of times the encoded and original can be distinguished from each other.
KATZ COMMENTS
Lossy coding gives rise to a number of challenges for mastering engineers, said Bob Katz. Lossy-coded material should never be used as original source material, he suggested, because re-encoding causes a build up of artifacts that will make the sound quality worse. It’s therefore really important to hang on to the original PCM masters of projects, rather than relying on MP3 or AAC files that might otherwise be used if nothing else is available. Robert Bleidt commented that in the broadcast- ing world they do use AAC as a contribu- t i o n c o d e c f o r s o u r c e m a t e r i a l , a n d tandem coding (multiple generations of coding) is done on a more regular basis. The quality can be satisfactory for the purpose up to several generations of encoding if a reasonably high bit rate is used. Jim Kaiser noted that the quality requirements differ between applications, s o o n e s h o u l d n ’ t n e c e s s a r i l y e x p e c t broadcasting to be the same as music distribution.
The importance of avoiding clipping when using lossy formats was reinforced by Bob Katz—it can give rise to very nasty artifacts, worse than with linear PCM. Even original signals just below
clipping can give rise to problems, either i n t e r n a l l y — w i t h i n t h e e n c o d e r, o n decoding, or during oversampled D/A conversion.
MASTERED FOR ITUNES
Bob Katz was also a speaker at Bob Lud- wig’s Platinum Mastering session on Mas- tered for iTunes. Eric Boulanger of The Mastering Lab, the second speaker, had m a s t e r e d t h e v e r y f i r s t M a s t e r e d f o r iTunes title by Colby Caillat.
Before Mastered for iTunes (MfiT), tracks for iTunes were either simply ripped from CDs or taken from the major record company servers and loaded into iTunes Producer. With MfiT, AAC encod- ing is done from 24-bit masters, often with lowered level to avoid clipping and get a much cleaner result. (The aim is to get the best results out of the 256 kbit/s constrained variable bit-rate (CVBR) of the iTunes Plus format.) All the encoding for an iTunes release is done by Apple, and there is identical free Apple software called “afconvert” that enables users to do the same thing themselves before submit- ting masters. The first process in this software is Sound Check, which looks at the relative loudness levels of songs to be encoded and attempts to determine how much their levels should be raised or lowered on replay to make their loudness comparable. It adds metadata that can be used by players to avoid loudness differ- ences when tracks are played alongside each other. If the track is at a higher sampling frequency than 44.1 kHz it is down-sampled to 44.1 kHz, otherwise it is left alone. There is also a process that will convert the AAC-encoded track back to PCM so that you can h e a r t h e d e c o d e d version. “afclip” looks at the likely on-sample and inter-sample clips, behaving like a true p e a k - r e a d i n g m e t e r, enabling the user to determine the poten- t i a l f o r e n c o d e r a n d postdecoder clipping. Even when no clips are i n d i c a t e d , L u d w i g suggested that it can be necessary to experi- m e n t w i t h d i f f e r e n t a m o u n t s o f l e v e l reduction and compare the encoded result to the 24-bit master, as the sonic differences
between settings can be quite dramatic. After the track is transferred to Apple, it is encoded in exactly the same way as the user would have done. “Test pressings” are then returned to the record company to confirm what is about to be released on iTunes. Usually these turn out to be bit- for-bit the same as the final encoding created by the mastering engineer, which confirms the integrity of the process.
According to Ludwig, MfiT has been a tremendous success, having been embraced by all the major record compa- nies, and with substantial support from recording and mastering engineers, as well as there being options for independ- ent artists. When properly done the results are excellent and the advantages are clear, with quality that is much closer to what the artist and engineer intended. While it might be asked why Apple would not go directly to using lossless down- loads, Ludwig pointed out that the installed base of some 400 million Apple music players, as well as Windows players that access iTunes made it necessary to retain compatibility. Apple had spent 18 months figuring out the best possible AAC encoding method to achieve excellent sound quality, considering the “player landscape” in question. John Coltrane’s “Blue Train” track was a good example of the file-size advantages of using AAC over lossless—the AAC track was 3.2 times smaller than the lossless version—so it can be seen that going to lossless would make downloads much longer and make the entire experience rather sluggish, as
MfiT SAMPLE RATE CONVERSION
Apple’s own documentation states that it prefers to receive high-resolution masters at sampling frequencies above 44.1 kHz, preferably 96 kHz. That way the encoding process uses its mastering- quality sample-rate conversion that generates 32-bit floating point CAF files as the input to AAC encoding. It’s claimed that this avoids the need for redithering and preserves all the dynamic range inherent in the original file, avoiding the potential for aliasing or clipping that can otherwise arise in sample-rate conversion.
If you supply 44.1-kHz files to Apple, the advantages of the above process are bypassed as the sample rate conversion is not initiated.
Bob Ludwig (center) with Bob Katz (left) and Eric Boulanger (right), before the "Mastered for iTunes" Platinum Mastering workshop.
FEATURE ARTICLE
APPLE’S MfiT TOOLS
The tools contained in the current free mastering suite that can be downloaded from Apple include the Master for iTunes Droplet, which is used to automate the creation of iTunes Plus masters. The suite of tools requires at least the Snow Leopard (10.6) version of OS X to run. The droplet needs either AIFF or WAVE files to be provided as source material and converts them temporarily to Apple’s Core Audio Format (CAF) with a Sound Check metadata profile attached that can normalize the relative loudness levels of songs on replay. AAC files are then encoded.
afconvert is a command-line utility that enables more direct control
over all of the above MfiT encoding operations.
AURoundTripAAC is an Audio Unit (AU) that allows the comparison
of encoded audio against the original source file, which also includes clip and peak detection (see screen shots below). There is a similar listening facility to that provided in the Sonnox/Fraunhofer Pro Codec plug-in that allows a simple double-blind ABX test to be set up, in order that users can check whether they can reliably tell the difference between source and encoded versions. The plug-in can be used with workstation software that conforms to the AU plug- in format, such as Logic, or alternatively the AU Lab application can be used to run the process.
AU Lab is a free standalone digital mixer utility that lets you use AU-
type plug-ins without needing an AU-compatible DAW.
afclip is a Unix command line tool that can be used to check a file
for on–sample and inter-sample clipping. Inter-sample clipping can
arise in oversampling D/A converters used after decoding, for example. (Four-times oversampling is used to estimate sample values in afclip.) When mastering a track for iTunes that peaks very close to digital maximum, it’s necessary to check it using this tool and reduce the level slightly until an acceptable number of clips is indicated (which may be zero, unless a small number turn out to be inaudible). If there’s any on-sample clipping the output of this process is an audio file (.wav) where the left channel data is the original audio and the right channel contains impulses where the audio is clipped, so that clips can be quickly located visually in a digital audio editor. There’s also a table that comes up in the Terminal window (see screen shots above) to show the timing locations of clips and the amount by which the samples exceed the clipping point. “Pinned samples” can also be reported—that is any in a series with a digital level of exactly ±1.0 (peak level), which suggests on-sample clipping may have occurred.
Finally, the Audio to WAVE Droplet converts files that are in other audio file formats (any supported by Mac OS X) to the WAVE format. This only works at 44.1k/24 bits at the moment.
Tables showing instances of clipping, total number of clipped samples and pinned samples
Display of WAVE file generated by afclip showing impulses marking clipped samples in the right (lower) channel
well as consuming more processing power, memory, and so on.
Most laptops, phones, and tablets only use 44.1 and 48 kHz for D/A conversion so are likely to down-sample anything else to get it to play from their analog outputs. Mac Snow Leopard can send digital audio data out at any sampling frequency up to 192 kHz and Windows 7 can go up to 96 kHz. However, in order to make this happen on the Mac it’s impor- tant to remember to change the output sampling frequency in the Audio/MIDI setup, close iTunes and reopen it, other- wise iTunes will continue to attempt to convert the material to the previous sampling frequency. (In other words, iTunes only appears to read the new output sampling frequency from the audio preferences when it opens, not if it is changed subsequently.) A number of o t h e r p l a y e r s a r e a v a i l a b l e , t h a t a r e compatible with iTunes, that interface directly with the Core Audio functionality of the Mac, that may provide better qual- i t y s o n i c r e s u l t s , a n d w h i c h s w i t c h sampling frequencies automatically.
KATZ COMMENTS
Bob Katz had commented on the Mas- tered for iTunes project in the previously m e n t i o n e d w o r k s h o p , n o t i n g t h a t although the iTunes Plus bit rate is nomi- nally 256 kbit/s it is normally a con- strained variable bit-rate format. This means the bit rate can go higher than that on occasions to allow for sections that are difficult to encode. It’s the first lossy format that he hasn’t felt embar- rassed to use, he said, and the quality is surprisingly good. Because 24-bit PCM audio files are supposed to be submitted to Apple for encoding, it’s important to remember to dither the signal correctly at that resolution when rendering mate- rial from a digital audio workstation whose internal resolution may be 32-bit floating point. Although there are people around who claim the effects of dither at this level are inaudible, Katz reinforced the importance of doing it correctly for optimum sound field depth and audio quality. It’s not always clear whether workstation software is doing this auto- matically, and some surprises can be encountered.
Apple stores everything for posterity using its own lossless format called ALAC (Apple Lossless Audio Coding), Katz said. So 24-bit masters submitted to them are first converted into this format for long- term archiving purposes, then they can
be rendered for delivery at whatever rate is required. This lossless storage is partly what allowed them to reissue earlier m a t e r i a l a t t h e h i g h e r b i t r a t e o f 256 kbit/s used in iTunes Plus. (iTunes tracks were encoded at 128 kbit/s until around 2009.)
MfiT IN PRACTICE
Eric Boulanger debunked the myth that MfiT is a new format. In fact AAC encod- ing has been around for a long time and MfiT is just a new process for preparing and encoding AAC files. It puts mastering
for digital downloads into the foreground, instead of it being an afterthought to the CD mastering process. Now that digital downloads have overtaken CD sales, par- ticularly for some artists, this seems important.
A p p l e ’s S o u n d C h e c k p r o c e s s d o e s something similar to Dolby’s dialnorm and the recent ITU BS 1770 standard, in that it attempts to measure the average
l o u d n e s s o f s o n g s a n d a d d l o u d n e s s normalization metadata based on aiming for about 16 dB of headroom between the average loudness level and the peak signal level. It has the potential to end the loud- ness wars and restore the concept of headroom, suggested Boulanger, if it becomes enabled as a default function in iTunes. In fact if someone masters a song too loud its level will automatically be reduced by this process when the finished iTunes files is replayed, as long as Sound Check is active in the player. So it tends to reward good practice and discourage the recent tendency to “slam” levels. Ludwig suggested all his clients should listen to both a Sound Check optimized version of a master and a more slammed version on an iTunes player with Sound Check turned on. In almost all cases it will persuade them to go with the opti- mized version as the slammed version j u s t s o u n d s m e s s y w h e n t h e l e v e l i s reduced.
Bob Katz explained that the independ- ent music website, CD Baby (www.cdbaby .com) is now able to work with MfiT to publish higher quality independent music t h r o u g h i Tu n e s . Tw o l i c e n s e s a r e r e q u i r e d , o n e f o r s u b m i t t i n g 1 6 - b i t masters for normal releases and one for submitting the 24-bit masters required by MfiT. There is no guarantee that your