The Wave Scanner
The film strips in the Oramics machine say which note to play and how loud. They say nothing about what it sounds like. The timbre came from four waveforms Daphne Oram painted by hand on glass slides, and the wave scanner is the part that turns paint into sound.
What nothing does
The usual description of this machine has something tracing the painted curve, reading off its height, and using that as the waveform. That cannot be what happened. Oram's surviving slides carry thick, irregular ribbons of paint, several millimetres across in places, with edges that double back. A ribbon like that is not a single-valued function of position: ask what height the curve has at a given point and there is no one answer. There is nothing there to trace.
What the machine actually does is look for light. A dot on a cathode ray tube sweeps across behind the slide, a photomultiplier on the far side watches for anything getting through, and a feedback loop drives the dot upward for as long as the paint is in the way. The moment light reaches the sensor the loop stops pushing. So the dot settles on the top edge of the paint, and it does so without anything ever needing to know the shape of what was painted. The voltage holding the dot in place, rising and falling as the sweep crosses the slide, is the audio signal.
That is why the thick irregular paint is not a problem. Only its upper boundary is ever measured. Flying spots and photomultipliers covers the two components this is assembled from.
The diagram
Six steps: paint, sweep, look, follow, tap, hear. Turning the sound on and then redrawing the slide is the quickest way to get the idea, and raising the pitch while it plays will show the upper harmonics dropping off the scope as the sweep gets too fast to resolve the fine detail. There is also a switch between filled and outlined drawings, which changes what the circuit looks for and leaves the sound alone.
Paint a slide, sweep a spot of light across it, and listen to the voltage that keeps the spot at the edge of the paint.
Open the diagramIt needs a wider screen than this one.
One sweep per cycle
The sweep is not running at some fixed scanning rate. It crosses the slide once per cycle of the note being played, so a 440 Hz note sweeps it four hundred and forty times a second and an octave higher sweeps it twice as often. The slide holds one cycle of a wave, and reading it repeatedly at the note's frequency is what produces a pitched tone at all.
It also means the timbre is not independent of the register. At a low note the loop has time to settle against every small feature of the painted edge, and all of that detail ends up in the sound as upper harmonics. At a high note the sweep outruns the loop, the fine detail is smoothed over, and the tone simplifies. An acoustic instrument behaves the same way for different reasons, and it is part of why the machine sounds like an instrument rather than an oscillator.
The modern reconstruction of this in Daphne runs the same feedback loop at audio rate across a two-dimensional opacity field rather than reading a wavetable, for exactly this reason: a wavetable lookup would give the same timbre at every pitch, which is the one behaviour the original could not have had.
Filled and outlined
The glass timbre slides were filled in solid, so the edge the loop finds is the boundary of a shape. The control films carrying loudness and vibrato were different: until 1972 those were hand-drawn lines, read by small cathode ray tubes running the same kind of loop.
Those line-following circuits needed a steady mains supply, and the mains at Tower Folly did not provide one. On 1 February 1972 Oram wrote in her logbook that the mains was down to −13, that the volume tracks were hopeless, and that she was giving up on the cathode ray tube spot-followers in despair. Two days later she removed them and fitted plain lamps and photocells. From then on every graph had to be filled in solid with ink or tape rather than drawn, which was slower and harder to correct, and her logbooks stop not long afterwards.
The line-following circuits were Tom Richards' discovery while working through the machine and the archive for his thesis. No earlier account of the instrument mentions that they existed.
The rest of the series
- Flying spots and photomultipliers The two components this is assembled from, and the loop that ties them together.
- The Oramics machine Who Oram was, how the machine came to exist, and what was actually new about it.
- How the machine read a pitch Four film strips holding one decimal digit each, read as binary-coded decimal.
- Drawn sound on film before Oram Avraamov, Sholpo and McLaren, and the point at which Oram departs from them.
Common questions
How did the Oramics wave scanner work?
A cathode ray tube behind a painted glass slide produced a dot of light that swept across the slide once per cycle of the note being played. A photomultiplier watched for light coming through. Wherever the paint blocked the light, a feedback loop drove the dot upward until light got through again, so the dot came to rest on the top edge of the paint. The voltage driving it up and down as it swept is the audio waveform.
Did the Oramics machine trace the drawn waveform?
No, and this is the detail most accounts get wrong. Nothing follows the drawn curve. Oram's surviving slides carry thick, irregular ribbons of paint which are not a single-valued function of position, so there is no curve to trace. The feedback loop sidesteps the problem by finding the edge of the paint rather than interpreting its shape.
Why does the timbre change as the pitch changes?
Because the sweep rate is the note frequency. A higher note sweeps the slide faster, so the loop has less time to settle against each feature of the paint, and the fine detail that produced the upper harmonics gets smoothed away. The character of the sound is tied to the register it is played in, as it is on an acoustic instrument.
Why were some Oramics drawings filled in and others drawn as lines?
The glass timbre slides were always filled in solid. The control films used drawn lines until February 1972, when the cathode ray tubes that followed those lines were defeated by the unstable mains at Oram's studio and were replaced with ordinary lamps and photocells. After that every graph had to be filled in, which took longer and was harder to correct.
Sources
Tom Richards, Oramics: Precedents, Technology and Influence, PhD thesis, Goldsmiths, University of London, 2018. Graham Wrench built the wave scanner between 1964 and the summer of 1966 and described the work in an interview with Sound on Sound. The logbook quotations are from the Daphne Oram Archive at Goldsmiths. Written for Drawing Sound.