Of Feedback & Wobbles

Many years ago I came up with an analogy for poorly damped feedback loops. The feedback loop that prompted this frankly tart analogy was indeed a horror. Anyone who remembers shame-faced their first attempts at clutch control resulting in the good old kangaroo launch across the lights only to stall in the middle of the crossroads will have some idea of the kind of data transfer lag involved in said system.

The analogy I made was that of a large blancmange. Untouched, a model of stasis and eminently stable and predictable; but prod the bugger and the resultant chaotic behaviour needs a very complex physical model and some very decent computing horsepower to predict its outputs, if at all.

The system in question was a large and complex audio-visual display at what eventually became 'The Electric Mountain' in Llanberis. We were ultimately contracted to maintain the thing having been witness to and peripherally involved in its installation. Certainly not involved in its design, or execution.

A 35mm cine film projected via a moving mirror traversed its image under software control across a cycloramic screen, 20 slide projectors, their images either complex dissolves projected onto smaller screens flown above the main screen or as static images projected onto the main screen itself. A central monitor carried the narrator's image, fed from a Laserdisc. The monitor itself was raised as a reveal on a hydraulic ram from the pit front of stage at the start of the show. I won't list the myriad scattershot special effects - there were lots; but you get a flavour of the complexity of the control system needed to manage the beast.

Unfortunately, manage the beast it did not do well. At it's barely beating heart was - cough! - a dead-timing system: this resulted only because the original plot to use the optical soundtrack of the 35mm film loop to feed timecode to drive the system failed due to an unfortunate characteristic of optical soundtracks: they can't reproduce square-waves. Given that timecode relies on decent waveform edges to work, you can imagine the (non) results. Add to this a frankly bonkers attempt to compensate for the differences in frame rate between the film (24 fps), Laserdisc video (30 fps) and CD audio (75 fps) by programming jumps in the Videodisc and using a completely surreal CD control system (I think four CD-players outputting doubled stereo, centre channel and sub-bass) programmed onto a BBC 'B' computer at the last minute. In the end, the rig was controlled by so many disparate and loosely connected pieces of kit, that it was completely incapable of prolonged stability. As soon as something went slightly off-kilter, the whole system would veer off drunkenly getting further and further out of sync until it fell over. By the way, as Joe & I repeatedly tried in vain to point out, the differences in frame rates were totally irrelevant due to the simple fact that a second of measured time is a second of measured time, no matter what each piece of kit uses as its time-base. A triumph of hubris and small knowledge in a high stakes environment - I think we've all heard that one - we experienced it first hand. Read part two to see how we made it right.

Oh my, doesn't this seem to be somewhat analogous to the Governments handling of the COVID-19 beast? But then Boris has always reminded me of a pink blancmange. Couldn't resist that.