9th January 2001 - AES-31 Audio File Interchange Standard

The topic of the January lecture of the British Section of the AES was the Audio File Interchange standard, AES-31. This standard has been developed over the last three years by the SC-06-01 AES standards committee. The lecture included an overview of the standard, and an example of it in practice with the transfer of material between SADiE and DAR machines. Mark Yonge from Solid State Logic introduced the lecture. He has been a key player in the development of the standard as he chaired the committee that developed the standard. He started by reviewing why there was a need for an audio file interchange standard and what capabilities this must have. He stated that the standard needed to be an agreed format common to all manufacturers, that it needed to contain adequate information to be practical to use, and that the performance needed to be predictable from machine to machine.

He argued that the standards approach was the ideal framework for this, being available to all and setting a single explicit specification. This would make the implementation and verification of the common format simpler. Mark explained that the AES standards are international standards, which are reached by consensus. Any individual can contribute to the working groups that create the standards, which makes it possible for opinions to be gathered from a wide range of interested parties.

The criteria that the standards had to meet were to enable digital audio file interchange for professional applications (such as production or post-production) in a manner that was predictable, reliable, implementable and universal. There was an important balance to be made between making the standard simple enough that it was practical to implement by manufacturers, yet comprehensive enough to be useful to audio professionals. The project was separated into four major parts that had a hierarchical structure. The first was the physical interchange medium. The second was the file format to be used on that medium. The third section was the simple project structure, whilst the final part was a more complex object-oriented project structure. The third section was published in December 1999 as AES31-3 and the final section is still under development.

Joe Bull of SADiE took the floor to explain more about the standard. He gave a historical overview of the problems of audio interchange. With magnetic tape there were a number of standards of tape formats, equalisation and noise reduction, though it was often possible to get a useable result with some effort. The advent of the digital age and computer-based workstations has made this problem more difficult.

He continued by questioning why we might want an interchange format. For the customers it would mean more freedom to choose the optimum facility or device for carrying out a certain part of the project. It would also make archiving easier, protecting against particular machines becoming obsolete and therefore losing the ability to replay certain audio files. For the manufacturers, Joe explained that the standardising of a single interchange format would mean the necessary development of only one format, rather than one for each group of manufacturers. This use of one standard would also make it easier to conform fully to that standard, would make compatibility checks simpler, and would therefore create a better product for the customers.

Joe explained that the reason why a standard interchange format had not been achieved already was due to the manufacturers. Firstly, it is an advantage for the manufacturer to use a proprietary format that ties the customer to that platform. Secondly, from a practical point of view, the proprietary formats are more efficient to implement with bespoke hardware, and easier to maintain, modify and control.

As explained before, there were three main sections to standardise – disks, files and a simple project structure. The disk section addressed the format the disk would take. For this, FAT32 was recommended because of its widespread use, simplicity to implement for all computer platforms, and its ability to address large disks (over 2 GB).

The file format was standardised on the BWF (broadcast wave format) which is an EBU standard. This has already been implemented in a number of workstations and is only a small modification to the widespread IBM/Microsoft WAV format.

The simple project structure or audio decision list (ADL) was slightly more complicated as there was no existing format on which it could be based. Mark Yonge went on to present information from Brooks Harris, who has been instrumental in writing the specification for the audio decision list protocol. The ADL had to be able to play back media files as a completed show with sample accuracy and be able to correctly locate the media files on any computer platform. The ADL had to define which part of which audio file is to be played, in what order and at what precise start time, which track to assign to, basic fade characteristics and coarse level information. It was also decided that the ADL should be human readable in order to be able to manually correct any errors in a file. In order to achieve this, and to enable complete platform and transport compatibility, the ADL had to be ASCII-based. It also had to be relatively simple as this would result in greater reliability and inexpensive implementation.

The most convenient way to achieve this was to use a system of HTML-like tags. Mark explained that there were a number of sections in the ADL He gave an in-depth summary of the format and capabilities of the ADL, including specifying all the tags used in the format. He explained that a large number of attributes can be specified, including complex fades. However, if the receiving machine is unable to recreate these, then the information in the list supports various fallback options, including the ability to substitute a rendered clip rather than recreate the process at all.

There are of course limitations to the format in order to keep it implementable. The settings of processing options such as compression and EQ are not reliably transferrable, as they cannot be described unambiguously so may sound different between machines even with the same numeric settings. However, such functions could be rendered as a new file and therefore transferred within the standard.

Joe Bull summarised the state of the AES31 audio file interchange standard. Currently a number of manufacturers are implementing the standard in their products, and the list of manufacturers was growing as they realised that they needed to implement the standard to put customers’ needs first. Joe urged the end users of these products to pressure the manufacturers of the products they use, to encourage them to implement AES31. This will help to spread the use of the standard and encourage the standard to become universal.

The lecture concluded with a demonstration of AES31 in practice. Digital editors from SADiE and DAR were used to show that material can be edited on either machine and transferred on a 2GB disk across to the other and played back identically. There were a number of questions from the floor before the meeting was closed with thanks to the presenters and demonstrators.

Russell Mason