Neon Logic

[Al Kossow]

from 
http://www.computerhistory.org/events/lectures/johnniac_09151998/johnniac_xscript.shtml

Well, that bell rang frequently. Finally, we had to convince Harriet -- by the 
way, this lady worked graveyard shift, midnight to eight in the morning -- to 
stay over a little extra, and we would meet with her, and find out what was 
going on. We asked Harriet to please go through, in gory detail, exactly what 
had happened when the payroll failed. By the way, she couldn’t give me the data, 
or anybody else, because Harriet Pierson, besides top management, was the only 
[person] allowed to look at payroll data, so -- it was worse than top secret. It 
was compartmented like you don’t want to know. Anyway, Harriet stayed over, and 
came in, and went through this exercise, and for the exercise she had created 
dummy data, just in case we might look at it.

Well, nothing failed, and she couldn’t believe it. So we said, "Tell us exactly 
what you did." So we backed up, and she went through the whole thing, and I 
said, "Just go back out into the machine room like you normally would, after you 
loaded the machine." And she did. The first thing she did is, she walked by the 
door, is turn the lights off. But the drapes were open, and there was lots of 
light in the room, and so it didn’t have any effect as far as we could tell. 
But, when we sat down and thought about all this, we said, "Gee, maybe you’d 
better simulate the whole situation." So we closed the drapes, and ran the 
payroll program. And sure enough, after about 15 or 20 cards, we got an echo 
check error. The damn machine was afraid of the dark. Open the drapes 
[laughter], turn the lights on, and the machine ran fine! [Laughter] Makes no 
sense. Until Dick Stahl, one of the technicians on the machine, remembered that 
the neons were an active part of the circuit, and apparently by running a little 
test he determined that without any sunlight coming through in the windows, or 
fluorescent light from the overheads, which provided just enough ionization to 
keep them active, they deionized to the point where they would no longer 
conduct. [Laughter] Now, the question was, how do we fix this? There were 
something like two hundred and some odd -- how many neons were there? Well over 
200. And nobody wanted to get in there and unsolder and resolder 200 -- the 
machine would have been down for a week at that point.

Then somebody had a brilliant idea. Down here, where the air ducts for the 
return air from the air conditioner were, they put a bank of fluorescent lights 
on each side of the machine. If the filaments were on, they were on, the machine 
never had to run in the dark again. It never did. [Laughter] It never was afraid 
of the dark again.