General comments on digital reproductions of
sound recordings for American Memory
Introduction
The sound recordings in the American Memory collections
available include early phonograph recordings of political speeches and vaudeville routines and
two collections of recordings gathered in the field by folklife researchers. These recordings
include songs, poems, and
narratives.
Digital formats and resolutions
The Library of Congress does not yet make computer-based digital recordings for archival
purposes. The recordings digitized for American Memory were first copied to specialized audio
formats for archival purposes. The current practice is to make two DATs (Digital Audio Tape)
and one 3/4 inch U-matic video tape for safety. Digital "service" versions are derived from a DAT
tape. The Library expects to create new service versions over time, as network bandwidth to
homes and schools increases and the technology on typical desktops improves.
Two qualities of service version are currently in use. WAVE RIFF (.wav)
files are derived by sampling from the DAT at 16 bits, 22,500 times per second. To reach the
widest possible
audience, lower quality RealAudio files (aimed at 14.4 Kbits/sec modems) are generated in
batches from the WAVE files (including transfer of corresponding information held in headers).
Each RealAudio recording comprises two files; the .ra file contains the recording proper and the
.ram (RealAudio metadata) file contains the metadata that facilitates the streaming capability.
A collection of political speeches, digitized earlier, was captured at lower sampling rates and
is also available in .au format.
Descriptive records
The descriptive records available for these sound recordings vary widely in content and
format. For two collections, item-level MARC
records are available. In one field of the record, the
combination of subfields $d and $f provides a unique identifier for the associated digital
reproduction. For another collection, a non-MARC record includes an equivalent identifier.
Files representing the recordings for each collection are available in a directory structure (known
at the
Library of Congress as an "aggregate") for which $d identifies the root. Filenames for
the different digital versions of a recording are created by combining the $f value (which identifies
the
item) with distinguishing file extensions. |