AUC Academic Conference 'From Virtual to Reality' The University of
Queensland 1996AUC Academic Conference 'From Virtual to Reality' The University of
Queensland 1996
Back to Home Page ...... by Presenter ...... by Paper Title ...... by Keyword
Paper Title:
Digital Audio Studio
Presenter:
Trevor Hobbs, University of New South Wales, ADFA
Authors:
Trevor Hobbs and Zoran Milutinovic
School of Electrical Engineering, University College,
University of New South Wales, Australian Defence Force Academy
Keywords: Audio, Synthesis
Faculty area: Electrical Engineering
Introduction
Those in the music making and recording industry can choose from a seemingly endless array of digital and analogue audio equipment. It is quite common for a serious musician or composer to have at their finger tips a considerable number of dedicated hardware units as well as a computer that are used together to develop songs, develop sound timbres, write music and produce a master copy of their work. The equipment available includes sound synthesising musical keyboards, effects units for reverb, echo, filtering etc, a digital event sequencer to record and play back MIDI (Musical Instruments Digital Interface) events that were generated from synthesiser keyboards, electronic drum units, a CD player, a DAT (digital audio tape) recorder and a resampling unit that converts 44.1 k samples per second of digital audio to 48 k samples per second. Each piece of equipment may cost $1000 (1996) or more. The musical industry does not have an equivalent in hardware to the software of a desktop workstation where different software packages provide functionality and may be updated and reconfigured by the user. However, some recent products are starting to integrate the computer with external hardware to assist the composer in his or her work. It is possible to provide a single piece of general purpose hardware that serves many music generating, processing and recording needs. The Digital Audio Studio project has this aim.
The Conventional Audio Studio
A small audio studio is shown in figure 1. A major feature of most existing studios is the large mixing consol seen in the centre of Figure 1. Often 24 channels are provided where signals from 24 different sources around the studio, including live input from singers and acoustic instruments, are mixed together to produce the desired balance between the various sound sources. The sound is then recorded on a multi-track tape recorder such as the 8 track analogue recorder or the multi-track DAT recorder (DAT : Digital Audio Tape) shown in Figure 1. A major feature to observe is that most of the signals that pass into or out of the mixer unit are analogue in spite of the increasing presence of digital units such as the DAT recorder and the effects processor. These units have very high performance and sound quality achievable with digital techniques but then are connected to the mixer using old fashion analogue signals with inferior signal to noise and distortion characteristics.
An important part of most music composition systems is the computer and electronic musical keyboard. The keyboard generates a digital output called MIDI (Musical Instruments Digital Interface). The most common use for the computer is to run a software program, called a sequencer, that records the MIDI events from the studio. Some recently available MIDI sequencing programs have started to provide not only the traditional recording and play back of MIDI data but the management of both MIDI and digital audio sound tracks with the help of external hardware.
Towards the All-Digital Audio Studio
The aim of the current work is to develop a system that uses a fully digital mixing system. As was seen in Figure 1, virtually all of the signals that pass into or out of the mixer are in analogue form even when the external equipment is designed using digital technology. Digital mixers are starting to appear on the market but most of their external interfaces are still analogue, partly for compatibility with existing studios but also due to the fact that mixing together multiple streams of digital audio inputs (ie SPDIF or AES/EBU) causes a resampling problem solved in this work.
In the current work, the aim was to place the hardware of the mixer in a single unit with the effects processor, sound modules and sampler unit so that all these activities can be done in the digital domain. The host computer is then provided with an application that lets the musician control the synthesis of sound, the audio effects and so on, from the host computer's graphical user interface. A bi-product of this goal was that these hitherto separate units could be "wired" together, not by a mass of cables as is currently the case in any audio studio, but using software and a computer screen based activity. By incorporating sufficient processing power in a single external unit and appropriate host computer software, a studio may be configured and reconfigured in a versatile way. Configurations may be save on the computer and re-established almost instantly.
The system comprises three major parts.
(1) The external hardware based around digital signal processors (DSPs) that provides both the compute power and appropriate audio studio signal interfaces. This unit is called DASP (Digital Audio Signal Processor).
(2) The real time software system running on the DSPs which execute a variety of algorithms including sound synthesis, audio effects, digital mixing algorithms, oversampling, hard disk recording and host communication.
(3) The host computer software which provides a graphical user interface to the musician to permit control and configuration of the external hardware.
Work to date (June 1996) has resulted in the completion and testing of the DASP hardware (discussed in a later section) and part of the real time software system.
An Audio Studio "In a Box"
This section describes the requirements of an external hardware unit that will permit one to carry out many digital audio studio functions including mixing. The aim in the design of this unit was to provide all the usual hardware interfaces so that it could be connected into existing studios and could interface to any musical equipment that can be purchased. The table in Figure 2 shows the various classes of studio equipment across the top and both hardware interfaces and internal capabilities as resources down the left side. The body of the table of Figure 2 shows the result of a survey of existing equipment. The following paragraphs describe the nature of the audio equipment to which this table refers.
One of the most obviously musical pieces of equipment found in the studio is the musical key board with a span of notes covering typically 4 octaves (49 notes), 5 octaves (61 notes) and sometimes the full 7+ octaves (88 note). This unit may generate no sound, just a stream of digital events as the musician strikes the keys. In such a case the unit is called a keyboard controller (right hand side column in Figure 2). The digital data is usually placed on 5 pin DIN connectors at the rear of the keyboard. These signals belong to a standard called MIDI (Musical Instruments Digital Interface) which is recognised by virtually all manufacturers of musical equipment. Such MIDI input and output connections are listed in the resources column at the left of Figure 2. If the keyboard makes a sound by itself it would be called a synthesiser or and electronic piano and it would then contain a "sound module" (or music synthesiser) within the unit. Figure 2 separately lists the sound module (or synthesiser) as it can be purchased as an individual unit which inputs MIDI events and outputs synthesised sounds.
The "Sampler" shown at the top of Figure 2 is a unit that takes analogue sounds such as those coming from a microphone or from any unit with analogue output such as a radio, a CD player or a tape recorder and digitally captures sounds for later playback. The unit then behaves like a sound module, receiving a MIDI input stream and generating sounds, often many instruments all playing at once. The hard disk recorder ("H/D Recorder", in Figure 2) is a hardware unit with the action of a sampler but it is used to record a full musical performance onto a computer disk. Hard disk recorders are now taking the place of tape recorders for studio work. The "multi effects" unit is a digital processor that applies filter functions, reverberation, chorus and many other effects to audio signals. The "Sequencer" ("H/W Sequencer" in Figure 2) is a unit that records streams of MIDI events for later editing and play back. This function, however, is most often done in software on the host computer.
Down the left hand side if Figure 2 is found the internal and interface resources of various audio studio units. There is often a digital processor of one form or other which is characterised by various amounts of processing power, memory capacity (RAM and ROM) and possibly a hard disk. These units must connect to other equipment in the audio studio through interfaces which are listed as the remainder of items in the resources column to the left of Figure 2.
AES-EBU (or often called SPDIF : Sony-Phillips Digital Interface Format) is a standard digital bit stream that is used in the more expensive CD players and on all digital tape recorders. The "+++" symbol in Figure 2 indicates that these interfaces are not present on commercial units as yet and will be provided in the DASP hardware being developed here. In the AES-EBU standard, stereo audio information is carried as digital samples in a sequence of left, right, left, right (and so on) samples on a serial protocol running at about 3 million bits per second. These are often carried on the same familiar 1/4 inch phono plugs that you find on domestic hifi amplifiers and music systems.
The analogue input and output listed in Figure 2 is the familiar 0.3 V or 1 volt (RMS) analogue signals that are used in the audio industry, both domestic and professional. Finally, Figure 2 shows two interfaces in the resources column for interconnection to the host computer, a serial one (RS232) and a parallel one (SCSI, Small Computer Systems Interface). The aim is to control this studio-in-a-box from a host computer, typically a Macintosh.
The body of Figure 2 shows the typical compute power (MIPS = Millions of Instructions Per Second), data storage (MByte = Millions of Bytes) and available interfaces for units found in the market place.
DASP Architecture
The requirements for various audio functions was shown in Figure 2. The architecture for DASP (Digital Audio Signal Processor) is shown in Figure 3. To meet the processing requirements, two digital signal processors (TMS320C31) were designed into the processing core of this system. As the units must do several things at once, the dual DSP architecture provides a convenient way to implement two (or more) activities in parallel. More significant was the need to do real time digital mixing of the results of any sound synthesis, effects processing or hard disk play-back that DASP would provide.
A digital mixer using a multi-rate signal processing algorithm was developed so that it consumed no more than 100% of the processing power of one only of the digital signal processors (DSPs). This allowed the second DSP to be used for sound synthesis or effects and other audio management activities. Due to space restrictions, the details of the mixing algorithms are not described in this paper. In summary, the algorithm permits in excess of 30 (internal) channels of digital audio to be mixed together in the same way as would be done in the multi-channel mixing console found in an audio studio. The mixed output is a stereo channel of 44.1 kHz sampled sound or any such desired sample rate. The 30 input channels can be sourced from synthesised instruments or any other processed channels of sound, say from the play back of a hard disk recording from hard disk. A novel feature of this mixing algorithm is that it can handle a different sample rate on each of its 30 input channels and can even accept sample rates on each input channel that vary in time. These are virtual channels, not currently connected directly to external hardware. However, an expansion of this architecture could permit such a possibility and thus provide a fully digital equivalent to the analogue mixing console.
The two digital signal processors each have a bank of zero wait state memory (SRAM : Static Random Access Memory). This permits them to execute in parallel at full speed. These Texas Instruments DSPs provide a super scaler structure that permits two floating (or integer) operations to be executed each clock cycle. The clock rate varies with the version of silicon purchased and may be 16.6, 20, 25 or even 30 million cycles per second. This provides up to 60 MFLOPS (Millions of Floating Operations Per Second) of compute performance, or over 100 MFLOPS for the dual DSP-based DASP architecture.
A large share memory structure can be seen in the top-centre of Figure 3. It is a dual bank structure that permits near conflict free access for both processors for appropriately written software that runs on the DSPs. The peripheral subsystem shown at the bottom of Figure 3 provides all the interfaces listed in Figure 2. The AES-EBU (or SPDIF) digital audio signals are directly coupled to the serial inputs lines on each of the two DSPs and is not shown in the figure. All other peripherals, including ADC (Analogue to Digital Converter) samplers, DACs (Digital to Analogue Converters) for play-back of sound, MIDI, RS232 and SCSI are connected via a peripheral bus structure available to both processors. It is the DSP software which controls which processor will access which peripheral.
Signals may be sampled from analogue form, passed into the processing core for application of effects or for storage while one of the DSPs acts a sound synthesiser that is playing back sounds in response to the MIDI input events. The digital mixing described above prepares a stereo output data stream for both live play-back using the DAC outputs and for sending via SCSI to the host computer for hard disk recording of the mixed result. Such an array of activities would require several traditional digital audio units but can be supported here within a single DASP unit. The architecture may be expanded (or contracted) to provide differing numbers of peripheral channels. The current system resides on a 6-layer printed circuit card measuring 240 x 210 mm and supporting about 60 semiconductor devices including DSPs, SCSI controller, AES-EBU controllers, programmable clock sources, memory and programmable logic.
System Software
As described earlier, the system comprises one layer of hardware and two layers of software. To date, the hardware has been designed, constructed and tested and work has commenced on the software systems. A real time software system running on the DSPs provide processing, audio and computer interface activities while software on the host computer provides a graphical user interface to the musician. The hardware has been designed to permit connection to any host computer with a SCSI interface. A Macintosh tends to be the computer of choice in the music studio and is the host computer being used in the current system. A Macintosh 8500 was provided as part of an AUDF grant to help support this work.
The real time system on the DSPs is being implemented in C and using the Texas Instruments C compiler and libraries to target the C31 digital signal processors. For the host computer (a Macintosh), Prograph CPX is being used to build the graphical user interface while Metrowerks' CodeWarrior is being used to develop various lower level modules written in C. This work will be carried out throughout the remainder of 1996 and 1997.
Conclusion
Through the provision of DSP compute power dedicated towards the mixing of digital audio data, a system has been devised that permits several of the functions of a digital audio studio to be carried out within the one hardware unit when supported by a host computer used for control and to provide an interface to the musician. This system provides a cost effective solution to those wishing to set up a small music studio but with the power and capabilities found in much larger studios.
Back to Home Page ...... by Presenter ...... by Paper Title ...... by Keyword
Bibliography
ANSI S.4.40 - 1992 / AES3 - 1992, AES Recommended practice for digital audio engineering - Serial transmission format for two-channel linearly represented digital audio data.
TMS320C3x Users Guide, Texas Instruments, 1991.
TMS320C31 Addendum to the TMS320C3x User's Guide, Texas Instruments, 1991.
Dr Trevor Hobbs, Senior Lecturer
School of Electrical Engineering,
University College, University of New South Wales,
ADFA. Australian Defence Force Academy
Northcott Drive, Campbell,
ACT, 2600Ph ++61 6 268 8219
Fax ++61 6 268 8443email: trevor@ee.adfa.oz.au
Back to Home Page ...... by Presenter ...... by Paper Title ...... by Keyword
