TAPIIR, A SOFTWARE MULTITAP DELAY
Maarten de BoerAudiovisual Institute
Pompeu Fabra University
Barcelona, Spain
mdeboer@iua.upf.es
This paper presents a dedicated software implementation of a flexible multi-delay, that aims to combine flexibility and high accuracy with high quality audio and usability. At the same time, several important issues in software effect processing will be addressed.
This article appears in Proceedings of the COST G-6 Conference on Digital Audio Effects (DAFX-01), Limerick, Ireland, December 6-8, 2001
Figure 1. Tape delay.
These techniques were not only used in studio recordings, but also for live performances. Stockhausen's Solo (1966) is a work for a melody instrument and a complex tape feedback system, where four assistents control six playback heads along a tape of several feet length. Outback is played back on speakers and fed back into the tape.
A direct descendent of the Morphophone, the Copycat, could be used to create more complex echo effects and even reverbation, with the use of five playbacks placed on a tapeloop to produce irregular patterns of delays.
One would expect that digital techniques made these tape techniques obsolete. This is often the case, as standard digital effects processors provide delays, echo's and reverbation. However, they also have limitations, which makes them less flexible than the tape delays, and which even inhibit some of the techniques described above.
This cross-feeding of audio signals throughout the system of delay-lines and
mixers, allows the user to create a very large variety of stereo delay effects.
Very simple echos or ping-pong effects can be achieved easily, but more complex
effects such as early reflection echo's, reverbs, complex rhythmic and
arrhythmic patterns and even Karplus-Strong like synthesis is also possible. It
is important to observe, that these more complex effects are only possible by
using sample accurate processing.
Figure 2. TAPIIR flow diagram.
Obviously, this limitation in hardware effect processors is deliberate, both out of technical concerns or marketing. Most users are not interested in higher accuracy, and the standard user interface of hardware effects processors - buttons or at the most an alpha-dial - would make it a painful job to adjust. Also, one can imagine that lower accuracy means less computational cost, and therefore lower overall cost of the effect processing hardware.
For advanced users however, this limitation can be annoying. Of course, many of the effects obtained with very short delay times, such as reverb or filtering, are usually also implemented in the same hardware, but it can be very interesting to combine all these with longer delay-time effects; it would be necessary to use several processors connected together to do this.
The implementation of TAPIIR, however, is sample accurate. This means that extremely short delay times can be used, 0.023 msec when using a sample-rate of 44100 Hz. In addition to this fine control over delay-lengths, the sample accuracy is also implented for feedback and even cross-feeding between the various delay-lines, This is achieved by the fact that the internal processing core of TAPIIR is written in such a way, that the input and output values of the delay-lines and mixers are passed on 1 at a time, instead of buffer-by-buffer.
The maximum delay-length that can be achieved is only limited by the physical RAM memory of the system TAPIIR runs on. To give an example, with 32 MB of free memory, a total delay-length of more than 6 minutes can be used. While this might seem rather useless for normal effect processing, it clearly has musical applications. Several compositions have been written that make use of long delay times. Originally performed with the use of tape-delays, they could take great profit of the use of digital techniques for sound quality. The use of hard disk space with sufficient fast access would take away time limitation even more.
Figure 3. Screenshot of TAPIIR delay control
The system requirements of TAPIIR have been kept very low, being a single-purpose application. This makes it perfectly possible to run TAPIIR in combination with other applications.
TAPIIR has been implemented on the Linux operating system. Low-latency is a hot issue among many Linux audio developers. Even though Linux is not a true Real-time operating system, it is very suitable for applications that have high scheduling requirements. The multitasking facilities allow time critical tasks such as audio input, output and processing, to be separated from less critical tasks such as file i/o and graphical user interfacing. Linux has not been designed with low-latency audio applications in mind, and the standard time slicing shows this. However, and here we see one of the advantages of open-source software, a patch has been written for the Linux kernel to greatly improve this - you can achieve consistent worse-case scheduling latencies of 0.5 milliseconds on a 500MHz machine [3][4].
Linux can be considered the operating system of choice for applications such as TAPIIR. The other way around, TAPIIR is a nice demonstration application of Linux' Real-time effect processing capabilities.
The author hopes that TAPIIR will be used by composers and performers, and that is can facilitate the performance of existing works which otherwise would be difficult to accomplish.