On Tue, 01 Nov 2016 00:14:43 -0400, Dan Espen
wrote:

Actually, 3 years wouldn't be useful for anything but the decidedly
"not realworld" test case of continuous use.

Well, you're entitled to invent a new testing standard, along with yet
another collection of artificial test conditions, that will satisfy
your vision of a "real world" test. I've only had three LED light
failures. All were in the bathroom, all were failures of the driver
electronics, two were mounted inverted (base up), and all were
retrofitted into incandescent fixtures with miserable ventilation.
Therefore, I propose a bathroom LED test, which includes heat,
condensing and non-condensing humidity, on-off cycle time, over
voltage, erratic power glitches by PG&E, limited ventilation, and dust
accumulation. Such a test will clearly define what might be expected
from "typical" bathroom LED service. The EU micro managers have specs
and tests for almost everything and will surely appreciate your
efforts on their behalf.

It would take considerably more than 3 years to test the common
on at night, off during the day test case.

I think you need a major dose of testing reality. Instead of LED's,
let's try drug testing. In order to release a new drug, one of the
tests that a pharmaceutical company must survive is a cancer test.
This is usually done with mice or rats. However, they're not ordinary
mice or rats. If such a cancer test were performed on the common and
ordinary breeds of mice and rats, the number of tumors found would be
very small and therefore statistically useless. In order to get
statistically significant numbers, mice and rats that are genetically
predisposed to developing cancerous tumors are used.

For LED testing, much the same trick is used. If you don't have a
sufficient number of failures during the test period, and you can't
extend the test period, you do whatever it takes to produce those
failures. The easiest is to elevate the temperature. For
incandescent lamps, raising the filament voltage also works. By
plotting a trend line of different temperatures or voltages, one can
extrapolate the graph to obtain a fairly good approximation of the
expected lifetime at more sane temperatures and voltages. That's how
one avoids multi-year tests.
http://www.powerelectronictips.com/seeing-light/

Instead, they run a HALT
(Highly Accelerated Life Test), which is faster, and presumably
produces the necessary inflated figures:

You could have left the word "inflated" out of that sentence.
It's an insult to the rather clever testing that you described.

There are quite a few products that suffer from inflated
specifications. Battery capacity (in particular 18650 cells),
flashlight output in lumens, wi-fi range/speed, laptop battery life,
laser printer toner cartridge pages, inkjet cartridge pages, etc. All
of these are characterized by inflated claims contrived to make the
numbers bigger. I can explain any of these in detail if you want to
know how it works. The reasons are competitive pressure and product
differentiation. Every manufacturer and vendor are trying to sell on
the basis of everything except price. So, they push service,
warranties, packaging, bonus junk, etc. Eventually, they run out of
these fringes, and start inflating the specifications on the
assumption that the typical customer doesn't understand the specs. I
think this thread demonstrates that this is true. Instead of
inflated, perhaps "grossly exaggerated" might be more accurate.

... snipped test description.

Sniff...

--
Jeff Liebermann
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558