Relamping the shop

They are somewhat self-ballasting, as shown by the I-V curves.

Oh well then, your drill contraptions must surely be as great as you imagine!

No. Nothing wrong with marrying things. But calling it "design" is dopey.

LOL

Oh darn. Whatever will I do now?

Time for you to "design" a sawzall powered pogo stick, asshat.

I prefer not to do it that way, but for a manufacturer, batch-matching of the LEDs is the first step. Designing the lamp so that all stay at similar temperatures is the second. Maybe, too, the LEDs are deliberatelyl degenerated somewhat (aren't ideal diodes). In similar situations, I've seen fuses used, which both add series resistance and which prevent a single shorted (failed) unit from causing a total failure.

Unless I'm misintrepeting the curves, or looking in the wrong place, the graph shows a potential for runaway -- the slope of the curves increases with increasing voltage.

Here's a similar curve for a tungsten filament, which is self-ballasting.

This article talks about the hazards of paralleling strings of LEDs.

Or you could avoid the work-arounds and potential pitfalls: pitch the ballasts, and purchase lamps with built-in LED drivers that operate on line voltage.

What kind of 'built-in LED drivers' are you thinking of? Gizmos that switch OPEN when the current goes too high? Or, complete current-regulated DC power supplies? Something as reliable as a magnetic ballast is not going to look like a silicon IC, and take AC line input, and spikes, gracefully.

Try this:

The slope isn't straight like a resistor voltage divider, the reason I qualified them as "somewhat". It also shows their limited tolerance for reverse bias.

-jsw

No.

Yes. I haven't pulled one apart, but I'm pretty confident that the replacement tubes I purchased have constant current switching power supplies.

I've been designing, building, and installing automation and machine tools in industrial settings for 35 years. For most of that time the bulk of the low power loads in control systems have been powered by switchers with universal inputs; typically 100-265VAC. I've seen very few failures, so it's possible to build such supplies that survive in nasty electrical environments. Another example is electronic fluorescent ballasts, which are only slightly less reliable than magnetic ballasts.

The quality of the supplies in LED fluorescent replacements undoubtedly varies and is a legitimate question, but my experience so far has been more than satisfactory.

Yeah, it's sinking in. Slowly.

In the 1980's I designed and built a delicately balanced constant current string of LEDs, thermistors and Zeners that sensed and indicated four liquid ink levels in a rotating print head. It was a finely tuned version of the Low Fuel light in my car. At the time I understood the components' I-V curves well, but I haven't dealt as seriously with LEDs since.

The constant current limiter was a depletion mode JFET with its gate connected to the source like example 7. They were available as

-jsw

For Centronics? I recall you mentioning Centronics because in the early 90's I designed a couple pieces of automation for Presstek in Hudson NH. Presstek was started by the Howards and many of the employees had also worked at Centronics.

Same people, incorporated as "Howtek".

That car, a 78 Accord, had several unusual electrical and mechanical features I was curious about, such as current-operated reed relays in series with the brake lights that lit an indicator if a bulb burned out. Its extremely lean burning CVCC engine gave it ~38MPG on average and up to 44MPG on a long trip.