The Adventure of the Camera-Shy Computer

Years ago, I worked on a megawatt scale inverter product whose H-bridge was controlled by fiber optics instead of wires in order to avoid interference problems. A unit under test was left connected to a big rig sized DC simulator when somebody removed the control board in order to perform some kind of upgrade. During our lunch break, a marketing guy saw all of this neat electronic stuff (and didn't know that part of it was missing), and decided to take a picture. The light from his camera's flash travelled up the dangling fiber optic cables, resulting in *all* of the H-bridge's IGBTs coming to life, and thus shorting the massive DC simulator. We laugh only because the debris trajectory cleared his head...

In 1978, following a trade show, I visited a Radio Station in LA that had just taken delivery of a competitors "automation system".
When I arrived the station was running in automation mode. A Z80 processor was controlling four reel to reel tape players (music), four carousel type broadcast cartridge machines (commercials and announcements), and two standard broadcast cartridge machines (time of day). A full days "events" had been programmed by hand. I asked if I could take a couple of photos and was given permission by the station engineer. The first photos were uneventful, then the engineer opened the access door to the electronics for one last photo.
The flash caused the entire system to go quite crazy -- all the tape machines started running, error reports filled the screen and all the programmed events were lost.
On inspection, we found that non of the eproms had their windows covered!

In the mid-80's I actually was an EPROM covered in paper tape. I could see. Somedays I would notice the quick illuminations of my diety, it was wonderful. Then on day I was blinded by foil and left in the dark. Saddened I passed but am now reincarnates as a traffic light camera. DON'T RUN IT I SEE YOU!

You guys are nerdz.

In the mid-80's, I was working on the design of a new low-cost business phone where I used an infrared photo-interrupter to replace the mechanical switchhook. We put the first prototypes in the team members' offices. One manager started complaining because his phone would go offhook everyday between 9am and 10am. It turned out that his phone got the full morning sun. That in combination with the white, ABS, plastic housing and a small opening in the seam between the two housing halves allowed the sunlight to reach the photodetector. The fix was using a rubber bootie over the plastic interrupter and modulating the LED signal. One bad phone out some twenty+ installed, I thought I was nuts until I wired up an IR flashlight to find the "hot spots". One more item for the experience list.

Years ago I designed a control system for a Demo-sized MagLev train. This system used striped reflective tape on the bottom of the cars, both for sensing position and velocity. When the customer first heard the description, they pointed out that visitors would be taking flash pictures and so the system suddenly also had to be immune to multiple photoflashes. This was achieved by going to a modulated beam from the IRLEDs and full synchronous detection. We demonstrated complete stability for our customer using a General Radio Strobotach, and it worked well.

How about a computer that was afraid of the dark. The current drawn by a LED changes based on the amount of light impinging on it. They draw more in the dark (like 3x -- at least mid-70's red ones did) Back in the day, a PDP-11 had a row of "teeth", and several rows of LED's. Add a power supply with an overly sensitive current limit crowbar, and you had a machine that would crash when you turned the room lights off.

I had a more explosive experience along the same lines. A 180kW power supply was designed with optically isolated gate drive circuits. The fiber optics worked just fine unitil marketing wanted a picture of the insides. The unit exploded just when the picture was taken and was caught on film. We discovered the sheath for the fiber was not light tight and the flash turned on all the IGBTs.

Industry’s first Integrated Substation Protection and Control system was designed as an EPRI project in early to mid 1980s using the then state of the art microprocessor 8086 16-bit technology. Due to processing limita-tions, (8086 operating at 6MHz and memory of 128 kbytes) multiple processors (up to 6) shared the compu-tational loads and were configured in shared memory clusters.
The Station Computer provided operator interface for substation management and was connected to Protec-tion Clusters via coaxial cable Data Highway communicating at 1Mbps. Protection Clusters’ function was protection and control; Data Acquisition Units interfaced to the power system and were connected via fiber optic Data Links operating at 1Mbps to Protection Clusters.
Fiber optic Clock and Arbitration Bus synchronized the system and allowed for intercluster communication for time critical control operations.
The system was designed with interoperability in mind; the elements shown in dotted lines were provided by another manufacturer and were successfully integrated into the system during final installation at a major
500 kV substation.
In the final phase of testing at the factory, the system suddenly started to “crash” several times per day.
Extensive use of fiber optics assured very high degree of noise immunity, thus exonerating system hardware.
The interaction of multiple processors in clusters was suspected and analyzed over the period of days; no ob-vious reasons were found.
It was observed later that the crashes occurred during late morning, and were later correlated to the passing of a Mail Robot vehicle near the system. The vehicle has a motor that generates EM interference; the system hardware became a prime suspect.
Very soon the problem was solved:
A robot (in addition to motor) had a safety strobe light that was penetrating one of the Protection Clusters via partially open enclosure rear door.
These light pulses were “read” as spurious interrupts by a fiber optic receiver left open for the integration of other manufacturer’s cluster into the system.
A strategically placed piece of electrical tape cured the problem.

COOL!

NickB is right about old tape drives. In the 70s and 80s 1600 and 6250 bpi tape drives were used extensively in the seismic data recording industry aboard maine vessels. The ships would navigate carefully scripted lines on the map and guns would fire every ten seconds and aqusition hardware would stream a thousand channels of data to tape drives. Usually they had four drives... one recording, one rewinding, one ready to go at BOT and one hot spare. Sometimes they would have another bank of four on another system for larger jobs. Plus one or two more recording the navigational data without which the seismic data was just numbers. When recording it was just blowing and going, a carefully orchestrated sequence of events between man, navigator, vessels, airgun sources and recording system.
Anyway, most all the drives (as many as 10 in operation) had the EOT photo sensors as Nick described. If there was a system failure it would take four hours to circle the boat and all the streamer sensor cables behind it back to the location of the faiulre to repeat the data. The first time I was on a boat in the instrument room I whipped out a camera and about 5 guys jumped me and told me not to use a flash - it was well known to them that all the drives would unload and rewind and data be lost resulting in a 4-hour circle at about $2000/hour operating costs and cost them their production bonuses.

I recognize this problem. Around 1985, I developed a very low power circuit with a CMOS (windowed) EPROM. The breadboard had a socket and I did not put a label on the EPROM during development. My bench was near the window and when there was direct sunlight shining on the EPROM the circuit fully blocked (the rest was logic, no uC). When the sun was gone, it worked again. I blamed the Photovoltaic effect of the chip, generating voltages that blocked signals in the low power circuit (all together used some 50uA).
Ton

Interesting. I would have thought the laser scanner might have absorbed a command to shut off if doused with a flash light freq. In general robotics in the warfighter bomb bots using weapons, this might be a neat shut off feature for safety purposes when they go full autonomous.

Reminds me of a similar occurance from the same era, when our engineering group put together our first microcomputer system for radio dispatch. It included a reel-to-reel data tape drive for backup. (this was before hard disks were commonly available and floppies were 8-inch and low-capacity). A camera flash would disrupt the operation of the tape drive, but the cause was easier to determine. Optical sensors were used to detect the end of tape. The flash would make the drive think it was at the end and ready for a rewind! Unfortunately, hard to fix without a redesign.