55
113
Sound Directions
Best Practices For Audio Preservation
T
his chapter explores the audio preservation systems and workflows employed at both
Harvard University and Indiana University. This chapter also addresses selection for
preservation and quality control, two key parts of preservation workflow that are not discussed
elsewhere.
7.1 Preservation Overview
Audio Preservation Systems
An audio preservation program may be usefully conceptualized as a system using principles
from General Systems Theory. A system may be defined simply as a set of interacting units
or elements that form an integrated whole intended to perform some function or attain a
goal. Each element ideally contributes to an efficient workflow that maximizes throughput
while resulting in high-quality output of preservation-worthy digital objects. Thinking in
system terms can aid in evaluating an audio preservation system’s design, effectiveness,
completeness, behavior, and sustainability. The OAIS reference model provides a useful and
detailed examination of the functional entities of a preservation and access system operated
by an archive that is charged with preserving content.
107
Selection for Preservation
One part of an audio preservation system that requires significant attention is the task of
selecting recordings for preservation. It has been estimated that over 50 million hours of
audio recordings exist worldwide; most of them are analog, many are unique, and none
reside on permanent carriers.
108
Obsolescence—of playback machines, technical expertise,
tools, and audio formats—combined with degradation of carriers represent twin evils
that impede archivists’ race against time to preserve important holdings.
109
In addition,
preservation transfer work can be expensive and not easily afforded by many institutions.
Within this context it is easy to understand the need for careful prioritization of preservation
work. Selection often consists of an assessment of both research value and preservation
condition. The first involves careful evaluation of the depth and breadth of documentation
provided by an audio collection, assessing its potential value to researchers both now and
in the future. The second requires an analysis of the extent of the risk borne by a collection,
including the level of deterioration that is either manifest or expected based on its specific
format, storage history, or current condition. Collection managers may also need to take into
account political, economic, technical, donor-related, and other issues in making selection
decisions. It is important to create a selection process that responds to shifts in priorities
as perception of research value and knowledge of preservation condition changes over
time. Key tools to aid the selection process for audio recordings include the IASA selection
document available on the web, and FACET software and associated documents developed
by the Sound Directions project.
110
107 CCSDS, OAIS, sec. 4 and p. 148.
108 Schüller, “Preserving the Facts for the Future,” 618. See also the survey of the holdings of broadcast archives in
Europe completed by the Presto project: Richard Wright and Adrian Williams, PRESTO—Preservation Technologies
for European Broadcast Archives, PRESTO-W2-BBC-001218 Archive Preservation and Exploitation Requirements
(6 June 2001), http://presto.joanneum.ac.at/Public/D2.pdf.
109 IASA, Selection Criteria.
110 IASA, Selection Criteria; FACET—the Field Audio Collection Evaluation Tool—will be available as open source
software from the Sound Directions website later in 2007.
7 Audio Preservation Systems and Workflows
44
114
Sound Directions
Best Practices For Audio Preservation
Quality Control and Quality Assurance
Quality control is fundamental to the successful operation of an audio preservation system
and must permeate each part. The terms ‘quality control’ and ‘quality assurance’ are often
used interchangeably, although their meanings are substantially different. One meaning of
control is to make certain or to verify, whereas assurance may be defined as the act of giving
confidence. Quality control (QC), then, is the techniques and activities used to ensure that
a product or service adheres to a defined set of quality criteria or requirements. Quality
assurance (QA) refers to the activities implemented within a system to provide confidence or
evidence that a product or service under development or being produced will meet specified
requirements.
111
The first is reactive, and the second is proactive. For example, the procedures
and software tools used by an audio engineer to check a digital file fall into the realm of
quality control, while the use of an audio engineer within a preservation system for transfer
work is quality assurance.
Quality control and quality assurance are both necessary in preservation systems and too
often neglected in project planning and implementation, with consequent reductions in the
quality of the output. Many things can be done to provide quality assurance that preservation
output is both accurate and high quality, such as defining the characteristics of preservation
files, using professional equipment, employing professional personnel to play back the
recordings, and verifying data integrity with checksums. It is imperative, however, to check
and verify the actual output itself before certifying it as preservation-worthy and sending it
to long-term preservation storage. It may be useful, conceptually, to distinguish between
housekeeping QC (e.g., are all files present and named correctly?) and reproduction QC
(e.g., is the quality of the audio acceptable?). Quality control is both expensive and time
consuming. One vendor with a robust quality control program estimates that these activities
consume 18% of his yearly budget.
112
Mistakes and errors are inevitable and it is important
to devote sufficient resources to discovering and correcting them.
Quality control must be performed both by technical and/or project staff from either the
institution or the vendor. In addition, permanent staff from the organization supplying
the content for preservation must be involved in quality control, although possibly at a
less technical level. Quality assurance is inherent in the decisions made by institutional
management staff or vendor owners in developing preservation systems, programs, and
services.
111 See American Society for Quality, “Basic Concepts: Quality Assurance and Quality Control,”
http://www.asq.org/learn-about-quality/quality-assurance-quality-control/overview/overview.html and “Basic
Concepts: Glossary,” http://www.asq.org/glossary/q.html.
112 George Blood, “Commentary: Quality Control in Audio Reformatting,” ARSC Newsletter, No. 111 (Summer
2006), 7. Also available online: http://www.arsc-audio.org/nslr111.pdf.
44
115
Sound Directions
Best Practices For Audio Preservation
7.2 Recommended Technical Practices
7.2.1 Selection for Preservation
7.2.1.1 Best Practices
Best Practice 40: Develop a prioritized list of recordings and/or collections for preservation
treatment based on, at a minimum, an analysis of research value and preservation
condition.
7.2.1.2 Rationale
The rationale for this best practice is presented in the preservation overview section, above.
7.2.1.3 Selection at Indiana
For several years the ATM has engaged in a selection process that deeply examines both the
research value and preservation condition of its field collections. This process consists of
three steps:
1. Assessment of Research Value
As part of the Sound Directions project we are building the Cultures in Conflict Digital Archive
(CCDA) which will focus on cultural practices that have been threatened or abolished as a
result of conflict. Accordingly, our criteria for evaluating the research value and intellectual
merit of field collections were couched in terms of conflict. Specifically, we assessed whether
a collection
documented a cultural practice or language that has been lost, endangered, or
systematically oppressed as a result of conflict;
included expressive culture that was directly related to a specific conflict or was tied
to a particular event or period before, during or after conflict;
documented traditions, practices, or languages that have changed significantly as a
result of conflict;
represented cultural practices undergoing active revival;
provided especially deep or wide documentation of any tradition(s) related to the
above criteria.
Points were awarded in each category so that a collection would receive more than five
points only in exceptional circumstances. A collection’s total score then placed it in one of
the following categories:
5 points: The collection has exceptional research value as judged by the above
criteria
4 points: Solid research value
3 points: Moderate research value
2 points: Minor research value
1 point: Minimal or no research value
42
116
Sound Directions
Best Practices For Audio Preservation
2. Assessment of Preservation Condition
Over the past three years the ATM’s Associate Director for Recording Services, Mike Casey,
developed a preservation risk assessment instrument called FACET (Field Audio Collection
Evaluation Tool), a point-based software tool for ranking field collections for the degree
of deterioration they exhibit and the level of risk they carry. FACET currently assesses the
characteristics of, preservation problems with, and modes of deterioration of the following
formats: open reel tape (polyester, acetate, paper, and PVC bases), analog audio cassettes,
DAT (Digital Audio Tape), wire recordings, lacquer discs, and aluminum discs. This tool
helps collection managers construct a list of collections prioritized by the level of risk they
represent, enabling informed selection for preservation. FACET was further developed during
the Sound Directions project through collaboration with a number of audio preservation
engineers, archivists, and collection managers. FACET includes an approximately 80-
page/35-photo document that details degradation mechanisms and processes for the audio
formats covered by the software.
113
At the ATM, data for input into the FACET software was generated from a preservation survey
of open reel tape holdings conducted in 2002-03, documentation provided by collectors,
and new inspections of collections as needed.
3. Selection Process
Hundreds of ATM collections were assessed for both research value and preservation condition
using the procedures described above. The research value and FACET scales carried equal
weight and were then combined to provide an overall score that enabled us to generate one
ranking for ATM collections that reflected an analysis of both research value and preservation
condition. The Director and the ATM Archivist then reviewed and assessed the results in a
painstaking process of comparing collections and considering other factors specific to our
institution to make final choices.
7.2.1.4 Selection at Harvard
The responsibility for a decision to preserve a collection or an item within a collection
rests ultimately with the curator who relies upon the audio engineer to assess the technical
attributes of the objects being considered. The research-value-and-object-condition-based
principles and criteria we use for analysis resemble those used with, and in, the FACET
system. However, since Harvard’s Archive of World Music is a relatively small collection
compared with Indiana University’s Archives of Traditional Music, those decision principles
and criteria can be applied manually—rather than through the use of a formal survey tool
such as FACET. It is the collaborative effort between curator and engineer that we value most
in assuring successful, efficient preservation. Therefore as our systems evolve, we look for
ways to facilitate that collaboration.
113 The FACET software and associated documents will be publicly released as open source software during the
fall of 2007.
41
117
Sound Directions
Best Practices For Audio Preservation
7.2.2 Quality Control and Quality Assurance
7.2.2.1 Best Practices
Best Practice 41: Develop a formal, preferably written, plan for both quality control and
quality assurance.
Best Practice 42: All digital files and all metadata produced as part of the preservation
process must be subjected to quality control.
Best Practice 43: Permanent staff at the management level from the institution providing the
target content for preservation should conduct some quality control.
Best Practice 44: Initiate one full round of quality control as soon as possible after beginning
a new project, collection, or format, as well as after installing new equipment.
Best Practice 45: Match quality control tasks with appropriate personnel by the level of
expertise required, and perform these tasks in the appropriate environment with the
appropriate tools.
7.2.2.2 Rationale
A well-defined quality control (QC) program is essential for guaranteeing the quality of
the output of both preservation transfer and metadata collection, as discussed above. All
data bound for long-term preservation storage must be checked and verified. Some of this
verification is undertaken by permanent managerial staff with ultimate responsibility for the
outcome of preservation work. These staff members are able to evaluate with larger issues
in mind. One particularly important time for QC is at the start of something new, such as
when a project is set up and key parameters are established, or any time there is a change
in equipment or project specification. It is critical to discover and correct basic mistakes
at the beginning before they are propagated through an entire project or collection and
a significant amount of work must be re-done. Involving management staff at this stage
assures that appropriate high-level decisions are made and helps emphasize for all staff the
importance of the QC function. Quality control tasks are best assigned to personnel by the
level of expertise that the tasks require. For example, an audio engineer may not be the best
person to evaluate metadata while a content specialist may not be the best choice for deep
assessment of the audio quality from a preservation transfer. In addition, quality control tasks
that involve qualifying audio cannot be effectively performed by staff in a noisy, acoustically
challenging environment or with equipment that cannot accurately reproduce the audio
content—whereas quantifying QC tasks require a less specialized acoustic environment and
rely upon measuring or analysis devices.
7.2.2.3 Background
The distinction between quality control and quality assurance is explored in the preservation
overview to this chapter.
Indiana and Harvard share a number of quality assurance functions and procedures. To
assure the overall quality of the preservation process we do the following:
62
118
Sound Directions
Best Practices For Audio Preservation
Use professional audio engineers for preservation transfer work
Use professional equipment and audio engineering procedures in the preservation
studio
Use a test/calibration chain in the preservation studio that is able to verify a number of
factors related to audio quality, such as the studio noise floor, for example
Formally implement relevant standards and best practices in our work
Maintain and use a document that details the characteristics and uses of each type
of file created during the preservation process, including procedures used in creating
them
Maintain and use a document that details quality control procedures for all objects
created during the preservation process
Use checksums for all files
Use a local (interim) storage strategy that always maintains backup copies of all files
7.2.2.4 Quality Control at Indiana
7.2.2.4.1 Background
At the ATM, the beginning of a new collection triggers a special and particularly intensive
round of quality control by the Project Engineer, Project Assistant, and the ATM Associate
Director for Recording Services. This round of QC is initiated after the first few recordings—
usually three to five—of the collection have been transferred. One goal is to catch and correct
basic, repetitive mistakes early so that they do not affect the remainder of the collection as
it is transferred. Another goal is to make sure that project parameters are adequate and
appropriate. Here are a few issues that have surfaced during this stage of QC:
A mistake was found in one element of a filename in the Preservation Master Files. It
was important to catch this early as renaming a file also means recreating the ADL,
re-parsing the file, regenerating the checksum, and deleting and re-entering data into
our metadata software
Discovery of an odd looking and sounding waveform was traced to an incorrect setting
in the software from a research project that had concluded the previous day
Discovery of a tone produced by a pitch pipe on collection recordings generated
decisions on what speed to use for playback and how to document our decision
This last issue required the involvement of permanent ATM staff, in this case the Associate
Director. The role of this position in the QC program is to look not only for basic mistakes,
but to address their causes when appropriate. It is also to verify that appropriate procedures
are used and that correct interpretations are made, serving as a check on staff understanding
of pertinent issues. Here are a few of the issues that the Associate Director has addressed
after QC:
How to document unusual events, such as sudden level jumps, in the metadata
software
How much unrecorded space/time to leave at the start of a Preservation Master or a
Production Master File. How to maintain the reference timeline while addressing this
issue
What information about device settings that are likely to vary from recording to
recording within a given collection should be communicated by the engineer to other
project staff to facilitate checking of metadata
Quality control procedures are also necessary for assessing the performance of new pieces
50
119
Sound Directions
Best Practices For Audio Preservation
of equipment while continuing to verify older units. ATM QC procedures using WaveLab’s
global analysis tools uncovered occasional, random errors or glitches from one brand of A/D
converter that manifested as transient tics or clicks added to the audio file in one channel
only. There were typically two or three errors per hour-long file, although in some files there
were none. This low number, plus their transient nature, would have made discovery through
listening difficult, at least initially.
ATM QC procedures are presented below by the target object. QC tasks within our workflow
are presented later in this chapter.
7.2.2.4.2 Digital File QC
For each Preservation Master File, the audio engineer
visually inspects the waveform representation of the audio in the file, scanning its
entirety for anything that looks unusual such as very large peaks or no signal;
listens to the beginning and end of each file, verifying the start and finish of content;
checks random places in the middle, listening to around 6-10 minutes of audio in a
30-minute file including a 3-minute continuous stretch, verifying (among other things)
completeness of content;
analyzes the file using the error tab on WaveLab’s global analysis tool. Threshold and
sensitivity settings in the tool are manipulated to return about 100 points for analysis.
The engineer searches for unusual looking things in the waveform, such as spiky and
jagged samples, that do not appear as if they could be from a tape machine (for tape
transfers) and also looks for events that occur in one channel only.
We do not specifically analyze for clipping as the engineer is attending each transfer and
sees clips held on meters as they happen, and restarts the transfer as necessary.
The Project Assistant, using clones of Preservation Master Files
verifies that all files that should exist according to ATMC (Audio Technical Metadata
Collector software) and collection documentation do exist and have been uploaded to
the ATM NAS (Network Attached Storage device);
verifies the checksum;
listens to a substantial portion of each file, and sometimes nearly all of it, during the
process of marking areas of interest. The person in this position, after some training,
has the ability to recognize obvious problems, at least, and sometimes more subtle
ones;
notes problems such as extremely low levels, distortion, or other unusual events and
checks ATMC to see if these are documented;
checks all metadata entered into the BWF <bext> chunk;
checks the filename.
If errors are discovered a report is sent to the engineer.
58
120
Sound Directions
Best Practices For Audio Preservation
The Associate Director
performs QC for a percentage of all files created;
114
follows the audio engineer’s procedures for visually inspecting, listening, and analyzing
with software;
checks BWF metadata;
checks the filename.
7.2.2.4.3 Technical Metadata QC
The audio engineer (who has primary responsibility for entering technical metadata)
communicates to the Project Assistant the data elements that are likely to vary over a
collection. For example: settings on devices in the signal chain such as preamp gain
or tonearm tracking force.
This facilitates QC by a non-audio engineer (Project Assistant, below) who may not be able
to judge the values for these settings but can note their presence, absence, and repetition
over many items.
All records for all audio objects created using our technical metadata software are checked
by the Project Assistant who
matches the data entered to what is known about the source recording and what is
heard while listening;
checks all technical metadata that has been entered for the source audio object, the
digital files created from the preservation transfer, and the processing history instances
that document this transfer;
matches the number of Faces (see the discussion of audio object structure in Chapter
4) entered for the audio object to the number of files produced;
checks that durations are entered for each Face and checks that the data for multiple
Faces of one object matches when appropriate;
checks the signal chain in the processing history instance, focusing on the data
elements that are likely to vary as reported by the project engineer, looking for patterns
that indicate obvious mistakes—things that are not changing within a collection that
probably should and possibly have been forgotten;
examines notes fields and the evaluation section of ATMC for problems with the source
recording or the transfer indicated by the engineer. This provides information on what
to expect when listening to the file.
The Associate Director
examines 20% of all records entered into the metadata collection software;
searches mainly for errors of interpretation or documentation;
verifies that required technical metadata for the format and/or collection is in place.
114 The target is currently 20% of all files. Because Phase 1 of Sound Directions was an R&D project that required
an enormous portion of the Associate Director’s time, we are not yet sure if this is attainable under a more normal
workflow that does not include R&D work. In the next project we plan to explore the use of statistical analysis
tools to help us consider what percentage of files should be checked at this stage.
63
121
Sound Directions
Best Practices For Audio Preservation
7.2.2.4.4 Workflow Tasks
The Project Assistant manages an item-level workflow database that matches tasks with
personnel and tracks their completion.
7.2.2.4.5 Studio Signal Chain
The following tasks are undertaken to assess the devices in the studio signal chain:
At the beginning of a new collection or at least once a month, full alignment and
frequency response tests are done for the specific tape machines to be used. This data
is saved and linked to the collection about to undergo transfer work
At the start of a new collection, or at least once per month, a basic verification of the
entire signal chain is done by sending test tones through the system and looking at
channel balance, frequency response after the converter, and other items
7.2.2.5 Quality Control at Harvard
Our workflow affords us the opportunity to perform quality checks at many points and by
different sets of ears, eyes and devices. In approximate order of our workflow we do the
following:
Use MRL calibration tapes to verify the performance of our open reel playback machine
before a project is started and whenever the head stack is changed
Use SpectraFoo audio analyzing software during transfers to visually check for
aberrations in the signal
Use the hardware-based Audio Precision audio analyzing system to track down
signal chain problems when they are encountered, and to quantify signal-to-noise
performance and frequency response when there is a change, when there has been
maintenance performed, or when there has been an addition to the equipment in the
signal chain
Our software tools perform message digests on audio data and metadata to ensure
their integrity
Our software application AudioObjectManager will warn in the command line
interface of any exceptions during original object metadata creation and will validate
the metadata document against a local schema (based upon the as-yet-unreleased
draft AES-X098-B)
115
when the document is saved
Our process history software application APXE has a quality assurance tab that records
QA information such as checksums or message digests for events, and also will validate
the process history against a local schema (based upon the as-yet-unreleased draft
AES-X098-C) when the digiprov document is saved
Use an XML QC viewing tool to compare and examine large amounts of metadata for
errors before creation of the deposit package
Use the process history and tracking sheets to enable any operator to take-up a project
and know exactly what has been done up to that point
Divide the workflow into transfer and archival creation, de-noising, production and
deliverable creation, and overall QC and deposit so that more than one person has the
opportunity to check another’s work
In the process of creating the deposit package, the Dmart (DRS METS archive tool),
115 Harvard validates its metadata against local copies of schemas based upon early drafts of the emerging AES
standards until such time as these standards are published. At that time we will choose a date to implement the
published standard for all future work, and all metadata in the DRS created prior to that date will be automatically
migrated using a script.
34
122
Sound Directions
Best Practices For Audio Preservation
which is called by the “makedeposit” script, checks for any malformed items
The DRS loader will reject deposits that do not conform to the batch control document
presented for the deposit and will generate error messages that are sent via e-mail to
the depositor
View the objects in the DRS and confirm that the data is correct
If necessary, download the deposited files and compare with originals
7.2.3 Audio Preservation Systems and Workflows
7.2.3.1 Best Practices
Best Practice 46: Specify the elements in the system designed for audio preservation including
the function of each.
Best Practice 47: Confirm the quality of the output of the audio preservation system through
an assessment of the contribution made by each part.
7.2.3.2 Rationale
Successful audio preservation requires the completion of many tasks and functions as described
in this section, below. These range from planning and selection to long-term storage. Each
step contributes to the quality of the final product. An audio preservation system may span
units within institutions and may also include other institutions and/or commercial vendors;
system elements must be fully functional across these divisions.
7.2.3.3 Audio Preservation System at Indiana
The Archives of Traditional Music built its part of the IU audio preservation system largely
from scratch beginning with the Sound Directions project. We found it useful during various
stages of development to reflect on some of the basic principles by which systems operate,
particularly the influence of system parts on the system as a whole, in order to gain insight into
the strengths and weaknesses of our efforts. Our focus was usually directed to the “preservation
chain” or sequence of actions that are typically employed for audio preservation projects,
particularly as they constitute a preservation workflow. Elements in the IU audio preservation
chain, organized by the steps within the various stages of a typical preservation project, are
presented in Figure 26 (next page).
Documents you may be interested
Documents you may be interested
