next section in this report >>   |   previous section >>   |   report contents >>

PART ONE

Paul Kingsbury

INTRODUCTION

The Preservation Challenge: Changing Technologies for Recorded Sound

The recording and playback of sound began with Thomas Edison's invention of the phonograph in 1877. In the wake of that landmark innovation, history has seen the emergence of one innovative recording technology after another. Each new technology has rapidly supplanted its predecessor. Thus, 10 years after the arrival of Edison's phonograph, Emile Berliner patented his disc gramophone. And within a few decades, Edison's cylinder recordings were largely replaced by Berliner's more-convenient flat audio discs, recorded at approximately 78 revolutions per minute (rpm) and usually composed of hard but brittle shellac. Following World War II, the shellac 78 gave way to the almost-simultaneous introduction of the flexible vinyl 45-rpm single and 33-1/3–rpm long-playing (LP) record in 1948–1949 and to magnetic recording tape. Magnetic tape, developed in Germany and brought to the United States after World War II, came into widespread use in commercial recording sessions by the late 1940s.

By the mid-1960s, when record companies began to offer for sale prerecorded, continuous-loop, eight-track tapes, consumers began participating in the audio tape revolution in earnest. Smaller, more-convenient cassette tapes—both blank and prerecorded—reached the market by the end of the 1960s. The next major breakthrough in consumer playback came with the arrival of the first widely available digital carrier, the compact disc (CD), which was introduced in 1982. Since then, the parade of new playback and recording formats has continued. As digital audio has gained precedence, new digital carriers—digital versatile discs (DVDs) and MP3 players, among others—are already jostling the CD for preeminence.

As one considers the swift evolution and succession of recording and playback technology, it is clear that innovation and obsolescence are constants in audio recording. Cylinders, 78-rpm shellac records, and eight-track tapes are no longer commercially viable media. The LP disc and the cassette tape have seen declining sales for some time.

For many years after digital recording and playback came into wide use in the 1980s, there was an ongoing debate in the recording community about the merits of preserving audio programs in a digital, rather than analog, format. In 1997, in an in-depth, two-part series on problems of audio preservation and storage within the major record companies in the United States published in Billboard magazine, reporter Bill Holland stated that the consensus among leading audio engineers and such organizations as the Audio Engineering Society (AES), the National Academy of Recording Arts and Sciences (NARAS), and the Association for Recorded Sound Collections (ARSC) was that "because analog tape has been proved to last, generally, and because the shelf life of digital tape is unknown, recordings should be stored or backed up, at least in the analog tape format." ¹

Since that time, however, it has become increasingly clear that analog magnetic tape no longer provides the safe haven for preservation that it once may have. As of 2005, only one major manufacturer, Quantegy (formerly Ampex), still manufactures analog magnetic recording tape stock for the U.S. market. Only a handful of companies still manufacture the machines that play open-reel tapes. Some tapes manufactured for preservation reformatting, such as polyester tape, have been found to deteriorate over time. For example, they can be damaged by hydrolysis, the process by which the chemical that bonds the recording oxide to the polyester base absorbs moisture from the air. Upon playback, these tapes can break down and become unplayable.

Increasingly, leading audio engineers and audio preservationists believe that the future of audio preservation is in the digital arena. Unlike subsequent generations of analog dubs, each of which is farther removed from the original sound (much as a copy of a film photograph is removed in quality from that of the original image), each digital capture is capable of producing an identical copy of the original recording. Digital recordings are also easily transported and transmitted in a variety of ways, including the World Wide Web, making public access easy and cost-effective. Acknowledging that digital tape is as subject to deterioration as analog tape, some preservationists are developing systems to manage sound recordings as digital files, to be archived in repositories and periodically refreshed and migrated.

Addressing the Challenge of Preserving Our Audio Heritage

The importance of preserving and ensuring access to the nation's audio heritage is now widely recognized, and many in the public and private sectors have called for a coordinated national effort to address the preservation challenge.

In response, in 2000 the U.S. Congress enacted Public Law 106-474, creating the National Recording Preservation Board (NRPB) under the aegis of the Library of Congress (LC) and charging these bodies to identify and address the major challenges to audio preservation. The legislation specifically charges the LC and NRPB to conduct a study of "the current state of sound recording archiving, preservation, and restoration activities." Areas to be explored include "the methodology and standards needed to make the transition from analog 'open reel' preservation of sound recordings to digital preservation of sound recordings," "standards for access to preserved sound recordings by researchers, educators, and other interested parties," and "the establishment of clear standards for copying old sound recordings (including equipment specifications and equalization guidelines)." The study is intended to inform key players in the preservation of recorded sound and to be a prelude to a national plan.

As a first step in this process, in January 2004 LC and CLIR convened a group of leading audio engineers and audio preservation specialists, including several audio engineer members of the NRPB, to participate in a roundtable discussion of preferred methods of transferring analog audio for reformatting as digital files. (A list of participants is provided in Appendix 2.) The goals of the meeting were to determine what areas of agreement and disagreement exist among reformatting experts, to identify gaps in knowledge about crucial techniques in audio transfer, and to make recommendations to LC and the NRPB about actions to be taken to address those problems. This roundtable group was charged with focusing on migrating an audio signal from endangered analog carriers—disc and tape primarily—and the challenges of capturing that signal digitally. A second group of engineers and librarians met in March 2006 to discuss issues related to conversion to digital media and metadata.

Participants were asked to confine their discussions to the two most common forms of analog media—audio discs and audio tapes. (Less-prevalent forms, such as cylinders and magnetized steel wire recordings, were left for future, more-detailed study.) Participants worked from two workflow documents prepared by preservation reformatting experts at LC. By working through each step in the process of reformatting analog tapes and discs, participants were able to identify areas of accord on best practice; areas of disagreement; and areas where further study, or further development of effective solutions, is needed.

Although group members represented a fairly broad range of audio engineering specialties, they found themselves in agreement on nearly all the discussed aspects of transferring analog audio to digital. The following text summarizes these major points. For more details, readers are invited to turn to part two, which contains the original workflow documents along with extensive annotations made by meeting participants.

SUMMARY OF MEETING DISCUSSIONS

This section summarizes the main technical points covered in the discussions. It shows where participants agreed and disagreed on preservation procedures and what topics they felt needed further study. Roundtable members' recommendations on procedure are displayed in italics.

One point emerged clearly from the discussions: Although many aspects of transferring analog audio to digital media require hardware and software tools and some are amenable to automation and batch processing, there are many areas in which a trained ear and years of experience are by far the most important tools. "The ideal," one participant noted, "is to use ears in conjunction with measurement." Another engineer stated, "Technology will never replace the listener." Subjective as listening can be, there is still no substitute for the trained ear when reformatting sound recordings.

Mitigating Deterioration of the Original Analog Carrier

Audio Discs

Commercial audio discs date from 1894; two-sided discs first appeared in 1907–1908. In physical composition, audio discs can range from fragile forms such as rubber (the earliest disc recordings), acetate or lacquer (sometimes with glass, aluminum, or cardboard backings), to more-durable shellac and vinyl discs and the metal masters ("metal parts") used to stamp out commercial discs. The distinct physical characteristics of each disc type require different, often highly specialized, techniques to coax the sound from the carrier.

Although participants did discuss the possibility of using advanced materials-science technology (such as laser refraction and spectroscopy) to help identify disc composition, it appears that no archives in the preservation field currently uses such high-tech assessment tools. Roundtable members agreed that in nearly all cases, experienced audio engineers could readily identify the composition of a recorded disc. Thus, although use of such technologies may be desirable, these experts did not see it as urgent or essential.

The roundtable panel identified several best practices that hold true for virtually all disc formats. These practices are as follows:

Audio Tape

Audio tape is a German invention, perfected during World War II when electrical giant AEG joined forces with the chemical firm I. G. Farben to create recording tape covered with magnetized iron oxides. Tape recording machines manufactured by the Brush and Ampex companies made their way into some recording studios as early as 1947. The earliest audio tapes were paper based, followed not long after by tapes with a cellulose acetate base, which were in wide use from the 1940s into the early 1960s. Since that time, magnetic recording tapes have been produced primarily on polyester tape (sometimes known by the trade name Mylar) in the United States; polyvinyl chloride is also sometimes used in Europe.

Obtaining an Accurate Transfer

Audio Discs

Often, fragile shellac or glass discs suffer breakage, or some lacquer may peel off of instantaneous discs. Damage may make playback on a conventional turntable impossible. Lacquer shards, even if all of them have been saved, can sometimes no longer be pieced together because of shrinkage. In such instances, the roundtable group unanimously recommended to save all pieces of the broken record. Signal-reconstruction methods, such as optical imaging technologies that are now being developed, could recapture audio programs from such discs within the foreseeable future. With respect to both audio tape and audio discs, roundtable members recommended that damaged, unplayable recordings be stored until appropriate treatment becomes available.

Damaged discs. If an audio disc is damaged but not broken into pieces, various methods can be used to play it. Cracks and scratches should simply be played through, with the attendant noise noted in the metadata to accompany the digital preservation copy. Warps can sometimes be played either by using a slower speed or adjusting tracking weight. As always, thorough documentation of transfer techniques is highly recommended. One participant noted that Sony BMG Music Studios has had success using a special turntable fitted with a vacuum pump to correct for warps. However, this vacuum arrangement can be used only with flexible vinyl recordings and not with shellac or lacquer-coated instantaneous discs. Furthermore, it is expensive.

Groove abnormalities. A number of groove abnormalities can be compensated for. In the absence of a lead-in groove, the roundtable members recommended trial-and-error stylus drops until as much of the audio program as possible, including the needle drop, has been captured. The problem should be documented in accompanying metadata. Shallow or worn grooves can be dealt with by trial-and-error stylus selection, making careful comparisons of various dubs recorded from each stylus. Nonconcentric grooves (a manufacturing error) create an audio problem called wow—a low-pitch deviation of frequency resulting from irregular motion in the disc. Roundtable consensus was that adjustments for wow can be made only at the time of transfer by adjusting the disc on the turntable. No post-transfer tools are currently available to compensate for wow. Thus, the group agreed that wow should always be corrected for prior to transfer and that it is pointless to make transfers that simply preserve a disc's wow. Similarly, a disc whose spindle hole is punched off-center can be compensated for mechanically by using a smaller-than-standard spindle and adjusting the positioning of the disc to get an accurate playback.

Audio Tape

Choosing the playback speed. Once an original analog tape is in condition to play, the correct playback speed must be determined. The audio preservation engineer should have playback machines capable of playing at all known speeds with all known tape head configurations, plus a variable pitch control. When in doubt about a tape's speed, the engineer should start at 7-1/2 inches per second and listen.

Expert listening is the first step in determining playback speed. Roundtable members shared a few basic guidelines for determining correct speed. For example, piano recordings are usually tuned to a standard middle A (440 Hz). Music or language specialists may be able to determine correct pitch on the basis of their knowledge of the program content. In addition, the discussants recommended listening for low-level ambient, electrical-power-line hum (50 Hz for European recordings; 60 Hz for North American). Checking background electrical hum with test equipment may help determine the correct playback speed. Audio preservation engineers should be aware that recording speed can vary throughout the recording; this is particularly true of field recordings.

Preparing the tape for transfer. Roundtable members recommended that engineers undertake the following steps when transferring sound from analog tape:

Roundtable participants could not agree as to whether it is ever possible to transfer audio from an original source without introducing some level of signal alteration, however negligible. Some argued that systems do exist that will pass a signal unprocessed in its purest original form (although such systems were not identified). Members did agree on the necessity of monitoring whether noise has been unintentionally introduced during the transfer.

Best Practices for Digital Conversion/Considering a Sampling Standard

Once an analog disc or tape has been readied for transfer and the signal path has been properly tested and calibrated, the signal is ready for digital capture.

Roundtable members did not deal in depth with the details of digital conversion. However, one of the engineers attending submitted a suggested road map that outlined their recommended best practices for digital capture (Appendix 1). Following a brief, inconclusive discussion of the document, the group agreed that it was a useful work in progress and that it should be shared with the audio engineering community for future comment and possible group ratification.

It should be noted, however, that the first two recommendations on the road map document, namely, recommended digital bit depth and sample rate for archival preservation, were discussed at some length and that there was disagreement over them.

Choosing the sampling rate. Since the commercial introduction of the compact disc in 1982, the standard sampling rate for CDs (known within the audio engineering trade as the "red book standard") has been 44.1 kHz, with a bit depth of 16 bits. The sampling rate of 44.1 kHz was a deliberate compromise by the developers of the CD in which they balanced audio fidelity versus time capacity of the discs. In the 20 years since those compromises were made, the storage capacity for digital recording and digital carriers has increased considerably. The DVD is emerging as a carrier preferred by some, and it has been configured to handle 96 kHz, 24 bits. The International Association for Sound and Video Archives (IASA), based in Europe, has embraced 96/24 as its standard. Roundtable members noted that it would be beneficial for the United States to set a standard that is interoperable with that of Europe.

Near the conclusion of the meeting, discussions focused on recommended sampling rate and bit depth for audio preservation. There was considerable disagreement within the group about setting 96/24 as the new digital audio preservation standard. Those who disagreed noted that down sampling from 96 kHz to 44.1 kHz for audio CD-Rs has in the past not faithfully reproduced the original sound. They recommended 88.2 kHz instead of 96 kHz. Those in favor of the 96/24 standard noted that new digital converters are much better at down sampling, with negligible loss of audio information. In addition, they noted that no less than the 96/24 configuration would be recommended for preservation copies; if user access copies down sampled to 44.1/16 suffered a minuscule loss in audio quality, that seemed (to those in favor) an acceptable compromise to keep the preservation standard high. This issue will be investigated in greater detail at the next roundtable.

The Human Touch versus Automated Transfer

Although the process of digital capture itself was left mostly for future discussion, participants at the first roundtable did discuss, but not agree on, a proposal from one group member that a listener monitor the entire program every time an archival transfer is made to a preservation copy. This proposal, which seemed eminently sensible to several participants, inspired much debate. Ultimately, participants agreed that however desirable, such a standard is not practical at a time of real-world budget constraints and staffing exigencies. Batch processing, or high-throughput transferring, is often a necessary compromise for archives that possess large collections of fragile, often deteriorating, recordings and small staffs and budgets. The financial concerns are real for archives, and automation is one of the keys to affordable solutions.

However, roundtable members were wary of any reformatting operation that does not entail monitoring by trained, critical listeners. Several participants recommended strongly that when automation, or other techniques that reduce the amount of real-time monitoring, are introduced into a preservation program, there must be minimum standards for quality control employed to assure that a flat or straight-across transfer has been made. Tests for such standards can be scientifically established. Some of the participants suggested that statisticians or operations research scientists or both be enlisted to help establish guidelines when non-optimal monitoring conditions are necessary.

Deciding whether or not to use automated or high-throughput transfer practices can be a complex risk-assessment task. As a best practice, roundtable members recommended that before a large-scale digital transfer, an audio archive's staff carefully review recordings earmarked for transfer and create risk-assessment reports to determine which ones can be safely transferred through automated processes and which will need staff attention during the transfer.

Preservation archives doing automated transfers need guidelines and sources of expertise. Roundtable members recommended that guidelines be developed on how to select which types of recordings are appropriate for high-throughput preservation. Such guidelines would factor in the format, content, and condition of media to be transferred as well as the intended use of the copies. Members noted that tools also need to be developed to mitigate the damage of poorly done automated transfers.

Creating Metadata

Metadata is data about data. A digital recording can be accompanied by several kinds of metadata, including descriptive (e.g., track listings), administrative, and technical (e.g., a description of audio hardware used in digital transfer, hardware settings, and data compression used). Roundtable discussions of metadata were confined almost exclusively to administrative and technical metadata, with the understanding that metadata for preservation warrants a separate and more detailed discussion.

The roundtable group strongly recommended that, whenever possible, transfer engineers should note all documentation (e.g., box notations) that accompanies the analog source tape. During the preservation transfer process, transfer engineers should note anomalies in tape (splice problems/repairs, speed variations, blocking/shedding, etc.) as metadata to accompany the digital preservation copy. Such metadata could be embedded, eye legible, or both.

Slate announcements, i.e., brief, spoken, prefatory announcements commonly used in identifying analog preservation copies, need not be used in digital preservation copies if the engineer is already embedding identifying metadata as part of the transfer process. Including such announcements would be a needless redundancy.

Correct and sufficient documentation is of paramount importance in archiving digital recordings to produce an authentic digital copy that can be retrieved and migrated as necessary to new platforms and media. The need for proper identification and labeling cannot be overstated, and the methods for retaining metadata will depend on the destination media. Overall, the roundtable group agreed that transfer engineers should generate as much metadata as is reasonably possible about the nature of the tape, the original recording, and the transfer. Roundtable members noted that it is much easier to include metadata while one is working with a recording than to try to find it later.

Recommendations

In the final segment of the two-day roundtable meetings, participants made a series of broad recommendations for improving the practice of analog audio transfer for preservation. The recommendations were grouped into categories. The salient and most widely supported recommendations appear below. Each paragraph includes the recommendation's priority, as voted by the roundtable participants.

The group was also asked to assign priority for action on these recommendations. The sidebar on page 14 lists the 15 recommendations that members believed were most important, presented in order of priority

Resources/Tools Needed

Reference Materials

Research and Development

Infrastructure Needs

Standards

Developing Core Competencies in Audio Preservation Engineering

(Priority #1; see related additional #1 item on page 13, Web site, under "Reference Materials")

Roundtable members expressed concern that in some archives, fragile audio recordings are being handled, played, and transferred for digital preservation by staff who have limited experience working with audio recordings or little knowledge about the sonic characteristics and weaknesses of various audio formats. The group strongly recommended that audio preservation transfers be done by trained and experienced audio engineers.

Participants identified a number of core competencies that an audio preservation engineer should have. Among other things, these would include the ability to

Until recently, there have been few university programs to train audio preservation engineers. Roundtable participants agreed that audio preservation engineers should be trained in postgraduate audio engineering courses that include a practicum or laboratory component. They recommended that coursework be coupled with an apprenticeship with a skilled audio engineer experienced in audio preservation. "We should err toward professionalizing music archiving and music conservation," said one participant. "If we can just get these programs into universities, this would be genuine progress." Ideally, such coursework would include instruction in the following:

Conclusion

As new audio technologies evolve and supplant older ones, we risk losing decades of spoken-word and musical recordings that are valuable not only as commercial products but also as cultural touchstones that document who we are, what we feel, and how we experience our world. At present, there are both audio engineers and equipment capable of transferring even the oldest analog recordings safely to digital. But this will not be true for long. As one roundtable participant noted, "The pool of expertise is shrinking every day." If key technical knowledge is not passed along soon, thousands of recordings may not be accessible to America's listeners 20 or 30 years from now.

Roundtable members agreed that sharing their expertise with colleagues in audio archiving and audio engineering, both now and in the future, is of vital importance. Participants noted that some of the leading associations in the audio field—such as AES, ARSC, IASA, and NARAS—have unilateral efforts afoot that may lead to some progress in developing standards for digital preservation. They agreed that more communication across these groups—through the Web and group meetings—should be encouraged to facilitate the sharing of information and recommendations.

FOOTNOTES

¹ Holland, Bill. 1997. Upgrading Labels' Vaults No Easy Archival Task. Billboard (July 19): 99.

next section in this report >>   |   previous section >>   |   report contents >>

pub 137 abstract >>