Abstract

Over a period of several decades, the major studios have accumulated a vast amount of valuable analog audio assets. As has been proven time and time again, these audio assets can be reused as new and better consumer release formats become available. This is true for both Ԩ�lm and music. In many cases, the sonic scope of analog tape recorders and sprocket driven mag recorders exceed the capabilities of the consumer playback systems they targeted. In some important ways, they even exceed the capabilities of state-of-the-art digital PCM recorders. This is a fact that is undisputed among audiophiles.

Realizing the limitations of current PCM technology as it relates to archiving, DSD, (Direct Stream Digital) was developed several years ago by Sony and Philips as an audio recording technology capable of archiving every nuance of the analog master. It is common sense to choose the best technology available at the time of archival, and we will make the case that the best technology is DSD.

Why DSD? Why not PCM?

A good question. What can DSD get from analog master that PCM cannot. The Ԩ�rst thing that one notices is that DSD can capture realistic tape hiss. It is the very low level audio signals that PCM has the most trouble with. DSD does not. This translates to a much more realistic and true listening experience for the consumer. Room ambience sounds far more real from a DSD master than from a PCM master giving the music listener or movie goers a far more realistic experience. The die-hard audiophile has a much simpler take on the matter. They feel that there is something not right about PCM. While this is

Super Audio Center, LLC �Ģ 3063 Sterling Circle East, Unit 5 �Ģ Boulder, Colorado 80301 �Ģ Tel 303-443-9855 �Ģ Fax 303-581-7941

a very controversial topic among audio equipment design engineers, it is not that controversial among critical listeners. Most agree that DSD offers a kind of realism that still can��t be captured with the best PCM recorders. It could explain why even after 20 years of perfecting PCM conversion technology, DSD playbacks consistently win listening tests, even after only a few years working on conversion technology. Certainly on paper, if we believe that our hearing is only sensitive between 20Hz and 20kHz, then 8fs and 4fs PCM should sufԨ�ce It should sound as good as DSD. According to critical listeners, this is not the case. So there is still much to learn about the way we hear sound. It is obvious now that sample rate has an impact on realism. Clearly a 2.8MHZ DSD recorder captures more time domain information than does an 8fs 384kHz recorder. It��s better to be safe than sorry. Why decimate the archive when it is not necessary?

Flexible Distribution Options

A more important reason to consider DSD for archival use is the fact that from a DSD archive, all of the common PCM formats can be generated. DSD is a recording technology that sits ahead of the Ԩ�lter that decimates the sigma-delta stream to produce multiple PCM formats; therefore, from the DSD master, various PCM formats can be generated by changing the decimation Ԩ�lter without quality loss. (Quality is less than the DSD archive but not worse than a parallel PCM master made by a PCM recorder, and in many cases, the down-conversion from DSD to PCM is actually better than would have been produced with a PCM recorder because there are compute-intensive decimating options that are not available in standalone A to D converters.)

�� _{��}

��

The front end of oversampled PCM A to D converters is a one bit sigma delta converter. As you can see from the graphic, the DSD archive is captured prior to the decimation Ԩ�lter. Rather than locking into a particular sample rate and word length, storing the archive as a DSD stream offers future PCM options such as word length, sample rate, type of decimation Ԩ�lter and noise shape. There is also the option to decimate with a much higher quality Ԩ�lter because once the DSD data is on disk, decimation can be done on a powerful number cruncher that can do a better job. (SBM Direct used today to create PCM Red Book masters from SA-CD masters is an example of the latter) In fact, it is also quite possible that future motion picture release formats may be able to play the un-decimated audio just as the SA-CD format does now. DSD is an option in the Blue Ray format.

DSD Standard File Format

The DSDIFF Ԩ�le format has become the standard Ԩ�le format for professional DSD audio recording. It is currently being used in multitrack recording applications, music mastering, and SA-CD Edit Master delivery to authoring stations. Supported by all manufacturers producing professional DSD tools, DSDIFF is a capable and Ԩ�exible backbone for archival applications.

Interleaving

DSDIFF offers both mono and interleaved Ԩ�le options for any number of channels. Mono Ԩ�les are best suited for initial production work. Interleaved Ԩ�les are best suited for archival applications where sync between tracks must remain unchanged. Interleaving also helps playback of multichannel audio from optical media with slow access time. (Mono Ԩ�les scattered on disk media require a faster access time). Should the choice be made to archive to DVD-R, or Blue Ray disk, interleaved Ԩ�les would allow direct playback from the archived media.

Meta Data

DSDIFF offers a very capable meta-data system for keeping track of asset history, adding notes, unique Ԩ�le identiԨ�ers, search criteria, and other information to the Ԩ�le. (See attached DSDIFF speciԨ�cation)

Easy Adoption

In deԨ�ning a new Ԩ�le format for DSD production tools, it was useful to reuse as much as possible of the standard Ԩ�le formats that the audio industry uses today. A commonly used Ԩ�le format for PCM is the Audio Interchange File Format, [AIFF], which conforms to the Electronics Arts Interchange File Format, [EA-IFF 85]. It enables storage of uncompressed or compressed sampled sound. Therefore the DSD interchange Ԩ�le format has been deԨ�ned which conforms, as much as possible, to the principles of AIFF and EA-IFF 85. This makes it possible to re-use large quantities of source code and libraries available today.

Large Ԩ�le support

DSDIFF, unlike AIFF, is not restricted to a 2GB Ԩ�le size. This is important when dealing with long interleaved multichannel Ԩ�les which can get very large.

Storage Media Considerations

Embarking upon a large archive project requires that careful consideration be given to the storage strategy. Multiplatform audio Ԩ�le standards such as DSDIFF and AES-32 remove restrictions that would lock in a particular physical storage system. This is good. It allows a choice among multiple platforms supporting multiple storage devices. It also makes possible the option of having a ��living archive�� which is a large redundant hard disk array based server offering instant access to the archived assets.

Hard Disk

It��s quite common to see hard disks containing audio masters sitting on shelves, either waiting to be archived, or serving as the archive themselves. The low storage cost of consumer hard drives makes archiving to hard disk attractive. But this is VERY DANGEROUS! Hard drives WILL fail, and consumer hard drives will FAIL FASTER. One is ill-advised to store their master on a hard drive, even for the short period of time during an active project. A hard drive can fail at any time. A hard drive sitting on a shelf can fail over time from loss of lubricant. A hard drive powered and running for a long period can fail from physical ware or component failure.

Tape

Tape wins the cost per bit war. As an example, 400GB of uncompressed data can Ԩ�t on an LTO-3 tape which costs about $100. That��s 1/4^th the cost of consumer hard disk. At speeds of 60MB per second uncompressed, duplicating valuable assets is quick and cheap. The LTO format is of interest because it is fast becoming the standard in the IT industry. This bodes well for long term LTO support. New LTO incarnations are backward compatible with old tapes, and because it has growing support in the IT industry, chances are better that the format will enjoy a long life. The down side of tape is that the shelf life is shorter than that of optical media. Physical ware occurs upon use, and there is always the risk of a drive malfunction that can damage an archive. (A good reason to create redundant tape archives. A system to duplicate tape archives can be constructed to be automatic and painless.)

Optical Disk

Optical storage offers a very long shelf life which makes it an attractive archival solution. Unlike tape, playing back optical media does not cause media ware. Since the introduction of CD, new optical devices have maintained backward compatibility with mainstream optical formats (CD, CD-R, CD-RW, DVD, DVD-R, DVD+R, DVD-RW, etc�Ķ This trend is expected to continue. It is going on 20 years and current devices can still read CD. It stands to reason that Blue Ray, for example, will have a long supported life as well.

Another attractive aspect of archiving 6 track mag assets to optical media is that interleaved DSD Ԩ�les on optical media can be played directly. One reel on one disk �� It��s a more traditional approach to library management and it is very cost effective. It is also very easy to create multiple optical archives simultaneously. The main disadvantages to optical storage are capacity and access time. Neither present a problem if the material to be archived is 6 track or less. Multitrack DSD assets require a larger capacity.

Living Archive

Budget permitting, the best solution would be a living archive system �� a self healing, self backing redundant array of hard disk storage serving clients who need instant access to all the assets. From this living archive, optical disks and tape backups can be generated as needed. The living archive tape backup process can be automated and can be as redundant as necessary to assure adequate protection of assets. The living archive is platform independent and can serve any type of operating system. Labor cost is lower because transfer to and from the living archive is fast. Access to assets for production use can be real-time (no copy needed.) As operating systems are upgraded, so is the living archive.

Available Tools and Services

The Sonoma DSD multitrack made by Super Audio Center, LLC along with Meitner DSD conversion from EMM Labs will serve as the DSD archival recorder. It is capable of creating 6 track interleaved (or monoԨ�le) DSD archives. It is also capable of archiving 24 channel music assets to DSD. The Sonoma is a capable editor should picture changes require audio modiԨ�cation, and it can edit up to 24 channels of DSD audio. It can play and record over gigabyte Ethernet making it an ideal DSD acquisition and editing device in a living archive environment.

Super Audio Center��s engineering team is available for consultation regarding system issues (such as living archive design and implementation.) SAC can also provide modiԨ�cations to the Sonoma for special requirements related to the archival process such as custom metadata support as well as control and automation design.

Sonoma Editing Interface

System Example

This example shows the living archive as the center of operation. Connected to the archive are multiple transfer stations which can transfer either 6 track mag masters to DSD or 24 track music masters to DSD. Initial meta data is entered at the time of transfer. A dedicated library management station handles incoming and outgoing asset orders, and manages and updates catalog information. The library manager can open asset Ԩ�les and add to the metadata. A down conversion station equipped with computers running down conversion algorithms receives orders from the library manager and creates deliverables of various formats, (PCM, DSD, or analog) on various media (Optical, Tape, Hard Disk, and WAN transmission). Assets that were originally PCM are stored in the archive in their native format, and can converted at the down conversion station for delivery in other formats.

Transfer Station

An eight track Sonoma DSD recorder is shown here resolved to the same house video reference as the six track mag playback system. The Mag is driven via a Magnatec MLB. Meitner MKIV A to D is clocked by a Rosenthal box which produces 44.056 from Video. (The entire system is pulled down to the video rate for the transfer.) The Meitner A to D drives the Sonoma Computer via optical clock. A dedicated operator sits at the Sonoma managing the transfer and entry of meta data as recordings are made.

Recommendation for an Analog Audio Asset Archival Systemfor Motion Picture and Music

Analog Audio Asset Archival A Recommendation