Presentation – Victoria Sloyan – 07/07/2010

Archiving Digital Audio Files

Introduction

Hello and welcome to my presentation “Archiving digital audio files”.

I am the trainee for the futureArch Project, which aims to expand the Bodleian’s capability to archive and preserve born-digital material. As part of this we have been developing a procedure for extracting digital files from deposited media such as floppy disks and moving them into our secure repository. This procedure involves the use of forensic imaging equipment to create bit-by-bit duplicates of files, but it is only suitable for data files. Therefore my project brief was to research an effective way of extracting and preserving audio files.

Audio material requires its own method for three reasons:1. Unlike data disks, audio CDs do not hold the audio data within a file system,

and so forensic imaging kits have difficulty creating an image file. 2. Audio files have different metadata. Some of the metadata fields are the same,

such as ‘title’ and ‘creation date’, but other metadata is specific to audio files, such as ‘duration’ and ‘file format’.

3. The final difference involves delivery to users: the way you access a word document is different from the way you access a sound recording.

Therefore, the overall aim of this project can be divided into three objectives:1. First is selecting the most appropriate format to store the audio in. There are

many formats available such as MP3, FLAC and WAVE files and many others. So, the first decision to be made was which to use.

2. Secondly, since imaging audio disks does not work I had to find a way to extract the files from their original media and move them onto the secure server.

3. Finally, I had to devise an effective way of delivering digital audio to users.

Selecting a format

The most important consideration when choosing a format is finding one which is uncompressed. Basically, formats fall into three categories: uncompressed formats are, as the name suggests, uncompressed, which means sound and silence are encoded at the same bit/time rate. Lossless compression is where the file is compressed to shrink the file size, but done without reducing the sound quality. This is typically achieved by compressing any silence. Lossy compression is where the whole file is compressed. This can significantly reduce the file size, though often the reduction in quality is unnoticeable by ear. For archiving purposes you want an exact replica of the original data, therefore an uncompressed format must be used.

The second consideration is to find an open source format in order to aid accessibility, because open source means you do not have to worry about licensing and legal issues.

Also, a standardised format is preferred, as this tends to mean it has been thoroughly reviewed and is more likely to be longer lasting. After evaluating all the formats I concluded the best one to use is WAVE. Crucially it is an uncompressed format, also it is open source and standardised and it is recommended by the International Association of Sound and Audiovisual Archives, the Library of Congress and the British Library Sound Archive.

Capture audio and extract metadata

So, once the format was decided I needed a way of extracting files from a CD and the best way is one you may well be familiar with: ripping, although the technical term for it is ‘digital audio extraction’. However, normal ripping, like that done by Windows Media Player is what is known as ‘fast ripping’, whereas for archiving we wanted a ‘secure ripper’. The main difference between the two is that secure rippers perform various validation tests to ensure maximum accuracy.

There are quite a few secure ripping programmes available but the one I decided to use is Exact Audio Copy. It is well regarded by audio professionals, it works with Windows OS, and it is free to download and rips to WAVE.

Here is a screen shot of EAC.• Down the left-hand side you can see the available ripping options including

ripping as a MP3 and burning to a CD, but for my project I was only interested in the first icon – ripping as a WAVE file.

• Along the toolbar there are various options to further specify the format. For instance you can state whether the recording is ripped in mono or stereo.

• You can also specify the sample rate to use, although the maximum sample rate available for WAVE files is 44.1 kHz. This is because Red Book Audio states that audio on CDs should be recorded at a sample rate of 44.1 kHz with a 16 bit-depth, thus there is little benefit to ripping audio at a higher rate. Moreover, from an archiving perspective, the British Library stipulates that audio transferred from one medium to another should retain the same sample rate.

Once EAC has ripped the files it will produce a report recording the rip result. Under each track it will say either OK or Finished. If it says Finished you know the rip was achieved but the resulting file is not identical to the original. This could occur for several reasons, the most common being if the disk is dirty or scratched.

Once the disk has been ripped, metadata needs recording in a spreadsheet. This is based on the futureArch project’s metadata spreadsheet for digital files, but has been modified to suit audio material. Here you can see all the fields that need to be completed. The ones in bold are ones that have been added for audio files.

Delivering audio

Archives exist not only to preserve material, but to also make it available to researchers. Therefore, audio files need to be delivered in an efficient and effective way. There are two potential problems with audio files. Firstly, they can be extremely large, particularly if they are in an uncompressed format and secondly, the quality of the recording can be quite poor. So, in order to combat this two issues I created two versions of each file: I already had the master file so I processed this to create a processed WAVE file and an optimised MP3 file. The table illustrates the intended use for each derivative. MP3 files are lossy; therefore the file size is significantly reduced.

Processing was done using Audacity, which looks like this. The processing done to a file will depend on its content and quality, but the tools I most often used were:

• Silencing and cutting to trim the beginnings and ends of recordings and remove long pauses.

• Noise removal tool to either remove or reduce the volume of background noise, like high pitched hissing on poor quality recordings.

It is very important to record every change that is made to the master file, so I created a Process History Spreadsheet to record these changes. This includes the name of the original file (22cd), all actions done to it in detail (such as two second noise cut at 22:54) and the name of the resulting file (22cd_mp3).

After processing was finished each file was exported from Audacity, first as in the WAVE format and then the MP3 format.

Once the files are processed the MP3 versions and possibly the processed WAVE files can be made available to listen to in the reading room. The master WAVE files will be stored in the repository and will not be touched be users. All digital material, both data and audio, will be accessed via a specific laptop in the reading room. This laptop will have a specially designed interface similar in feel to an internet browser and audio will be streamed, so accessing audio will be a similar experience to using something like MySpace.

Conclusion

So, to sum up very briefly, if we go back to the three aims you can see I’ve pretty much answered them:

1. The best format to use is WAVE2. The way to capture audio is by securely ripping it3. Audio will be processed and compressed and will be accessed by streaming

the files through a self-contained interface within the reading room.