🚀 go-pugleaf

On 10/03/2012 09:41 PM, Jeff Liebermann wrote:
> On Wed, 03 Oct 2012 10:32:57 -0400, Phil Hobbs
> <pcdhSpamMeSenseless@electrooptical.net>  wrote:
>
>> I don't know of any data that supports this common idea, but I'd be
>> interested in reading about it if anybody's actually done the experiment
>> carefully.
>
> It's an accelerated life test.  The deration curve of the incandescent
> light bulb is well known and assumed to be
>    (Vapplied/Vdesign)^-12 to ^-16 * Life at design voltage
> <http://www.welchallyn.com/documents/Lighting/OEM_Halogen_Lighting/MC3544HPX_Catalog_2_11_09.pdf>
> See Fig 5 on Pg 5 for the graph.  Nobody wants to wait 1000 hours for
> a bulb to blow.  So, they increase the applied voltage, which
> dramatically decreases the lifetime down to reasonable test times.
> Using a rack of bulbs, they obtain an average (or median) lifetime at
> the higher voltage.  Then, they work backwards on the curve to
> estimate what it would be at the design voltage.

You can't run an accelerated life test when the exponent isn't known 
more accurately than 12 to 16.

>
> When I was specifying lamps for a direction finder for the USCG, I had
> to deal with minimum lifetime specs.  I asked the vendor (Dialight)
> how they tested their T-1 3/4 bulbs and was told that they did an
> accelerated lifetime test on a few bulbs from each lot to insure
> adequate lifetime along with the usual sampled 1.5% AQL failure test.
>
>> Electromigration is a smaller effect in an AC bulb, since
>> the leading order effect cancels.
>
> Yep.  As I understand it (possible wrong), AC filaments break in the
> middle, mostly from vibration flexing.

I don't think so, because there's no mechanism for that, as I said.  The 
wire is fully annealed at all times, so there's no possibility of 
progressive fatigue failure.

>
>> I suspect that the notion that cycling is hard on bulbs comes from the
>> way that the bulb often fails at turn-on, when the thinnest hot spot
>> vapourizes before the rest of the filament has a chance to come up to
>> temperature and reduce the inrush current.
>
> Yep.  See my comments on the relatively high failure rate on the
> 40watt theater marquee lamps due to cycling.  The same lamps in the
> lobby and foyer were not cycled and seemed to last forever.

I was actually disagreeing with you.  There are lots of possible reasons 
for the marquee lights failing prematurely.  I'm not a tungsten expert 
myself, so I'd be very interested in seeing actual data that shows a 
dramatic shortening of life due to cycling.  I'm not saying it's 
impossible, just that I haven't seen any such data.

>
>> The tungsten in the lamp is run within a few hundred kelvins of its
>> melting point, so it's always in the fully annealed state, which ought
>> to mean that there are no metal fatigue mechanisms operating, just
>> material migration due to sublimation.
>
> Yep, but different failure mode.  When the extremely thin layer of
> tungsten plating evaporates, the light becomes dimmer.  Below some
> brightness level, it is considered to have failed.  However, most such
> tungsten coated filaments fail due to corrosion of the base steel
> alloy wire which is exposed to the internal gases inside the bulb
> after the tungsten evaporates.  The gases (mostly nitrogen and some
> argon) are inert, but there's a little water vapor outgassing from
> heating the glass envelope, which eventually corrodes the filament.
> Other failure modes are hot spots and notches caused by manufacturing
> variations and tungsten evaporation.
>

The filament isn't tungsten-plated, it's pure tungsten or a low alloy. 
The brightness drop comes from tungsten condensing on the envelope.

And the connecting wire isn't plain steel, it's generally Dumet, 
http://www.jlcelectromet.com/dumetwire.htm

which is a 42% Ni steel with OFHC copper or nickel plating.

You're making a lot of that up.  I'd still like to see 
carefully-collected data.

Cheers

Phil Hobbs

-- 
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net