2. PREVIOUS SYSTEMS

I have made several attempts to build a truly dynamic surround sound system for the dome, and the current configuration represents the third generation in my thinking. Over the years, I have changed tack several times with regard to the spatial distribution techniques, and I am confidence that the current system employs the most efficient and versatile approach available. The core of the system has been called the spatial distributor, a device with a small number of inputs and many outputs connected to the speakers in the dome. Under appropriate software control, it was intended to move sounds around by varying the volumes sent to the speakers, much like ordinary stereo panning.

Our initial attempts involved building hardware for generating digital staircase signals for controlling 16 bit multipliers, that modulated the amplitude of 16 bit digital audio. This approach was found to be cumbersome, due to the need for high quality A-D-A stages with filtering in large numbers. For instance a two input sixteen output device would have required 18 filters alone, not to mention the A-D-A stages and related control hardware. In 1987, the cost of such circuitry was prohibitive and to achieve the requisite quality in practice was found to have necessitated expertise and test equipment not available to us.

The next approach did away with A-D-A conversion and multipole filters and relied on keeping the audio in the analog domain, but controlling its amplitude with multiplying DACs (MDACs). This involves a digital word controlling the gain of a DAC with the audio fed in to the reference voltage pin. This was a much more appropriate technique. Dedicated ramping and control hardware was replaced by a 680x0 based computer called the Applix 1616, and plug-in cards containing the MDACs, audio I/O and bus interface components were made. A MIDI interface was added to the 1616, which allowed a Macintosh running user interface software to control the sound movement, either algorithmically or in real time with a mouse. All programming was in the Forth language.

Although this system worked, it was certainly not optimal. For a start, each MDAC channel card had one audio input and only six outputs, due to lack of board space. Due to the five-fold geometry of the dome, five outputs normally fed the lower ring of five speakers and the sixth was connected to the speaker at the zenith position. Panning for height effects gave quite indefinite localisation with this patch. Images tended to "stick" to the space around the speakers and the 1.5dB steps in gain control afforded by the MDACs were often audible when moving a sound in real time. A more conceptual worry also arose - the software was a descendant of stereo/quad panpot techniques and in retrospect, it was found to be unsuitable for a surround environment. Speakers were grouped together into five triplets, and panning only occurred within a particular triplet. The paradigm was one of trying to define a phantom point image as in intensity stereo. Waves or moving masses of sound were not achievable, as was the situation where every speaker was to be equally excited. It became clear that a new approach and a third stage of development was required.