At a glance

Archive of Recorded Sound

Archive news

Steinway Red Welte (1922), detail

In support of a major initiative to bring attention to the study of roll playing musical instruments, the Stanford Archive of Recorded Sound and Department of Music have acquired the Denis Condon Collection of Reproducing Pianos and Rolls, a collection of over 7500 rolls and ten players. The Condon Collection has long been known as one of the most important collections of reproducing pianos and piano rolls in private hands. Leading figures in the field of rolls and players are working along with Stanford faculty and staff on the project. The initiative will include roll preservation through scanning and digitization, restoration of instruments for playback, item level cataloging to allow for content discovery, and research into under-represented or rare systems and rolls. Plans for the collection include making streaming audio files of the recordings available to the public at large.

A compact cassette from the Clayman institute on Gender Studies SC0705

While often at the Stanford Media Preservation Lab we work with sound recordings Stanford acquired long ago sometimes we have the opportunity to work on media freshly acquired on the premise of immediately serving faculty and students. Recently a case like this occurred.


The Stanford Media Preservation Lab has kicked around the idea of building a dedicated Digital Audio Tape (DAT) "ripping" workstation around a Digital Data Storage (DDS) drive for a few years, but we never pursued it in earnest. We assumed the benefits of using a computer drive to read audio DATs largely centered around extraction time and reporting. Transferring a DAT in a conventional deck is done in real-time, whereas a DDS drive, we were told, would rougly cut the time in half depending on the speed of the drive (Peter Oleksik's retrospective on the Fugazi archive mentioned speeds up to 4x real-time using the DDS method with a Sony SDT-9000 drive and DATXtract). We also liked the idea of accompanying logs identifying where dropouts occurred. Still, we were skeptical whether such a system would be better than one designed around multiple conventional decks. Was there a way to test both methods without investing a bunch of money in late 1990s computer components?