Maker Pro
Maker Pro

Effect of starting on lifespan of general-lighting HID lamps?

T

Travis Evans

Jan 1, 1970
0
I've done some searching a number of times in the past on exactly how
adverse an effect on life frequent starting has on various (otherwise
properly installed/operated) types of lamps, general-purpose-lighting
HID lamps in particular (or more specifically, HPM, HPS, and MH lamps
around the 50-100 watt range; the kind usually available in common "home
center" stores in the US), but without much luck. I've seen at least
one website or document say that HPM lamps are the most tolerant of
frequent starting, but no cited sources or any actual figures.

As a hobby/for fun, I have some fixtures/lumiaires of these types
intended to be used as dawn-to-dusk "security lights" that I instead
switch manually and use indoors in my home. As such, the "operating
hours to starts" ratio tends to be quite a bit lower than the 10 that
HID lamps are typically spec'd at. I have no way of knowing how low
exactly, but I very roughly estimate the average is probably somewhere
in the range of maybe 40 minutes to around 4 hours per start, give or
take. The thing is, I haven't been able to get an idea of how much I
really need to worry about how often the lamps are started, if at all.

The most heavily used light is a 100-watt MH (ANSI M90). The ballast,
ignitor, and original lamp, used almost daily for approaching 3 years
now, still appear to be holding up well to this pattern of usage so far.
A 50-watt HPS lamp used over a slightly longer period of time still
works well, and a Philips H38 phosphored HPM lamp (used somewhat less,
though I'm unsure by how much) also seems fine. Finally, my 70-watt HPS
and 70-watt MH units still work, though I haven't yet had these two for
very long. All of these are on conventional iron ballasts (with
ignitors for the HPS and MH ones), except for the 50-watt HPS Lights of
America fixture which instead appears to use some sort of electronic
ballast circuit on a board (mostly discrete components with maybe an IC
or two).

On the other hand, I've had a clear H38 mercury (the one shipped with my
particular corresponding luminaire) mysteriously no longer start without
warning after operating apparently perfectly normally the last 2-3
years, despite almost certainly not having had 6000 hours, if half that;
a couple of HPS ignitors fail prematurely (though this perhaps was just
a bad manufacturing run as both were installed in fixtures bought the
same day from the same store (one died on the very first startup; the
one in the exchanged unit lasted about half a year); and a dirt-cheap
Lights of America 50-watt HPS luminaire's ballast electronics fail after
around 9 months [though in my experience, this seems to be the rule
rather than the exception with LOA products, so no big surprise there
:) ], though the lamp itself was fine (the original lamp and
replacement luminaire, since reinstalled outside the house and returned
to dusk-to-dawn operation, currently still work after a little over 2
additional years). But as with various CFLs that I've seen fail early
over the years, I just have no way of knowing whether or not failures
like these could have directly or indirectly had anything to do with
number of starts.

Does anyone have any experience with using these kinds of lamps on
fairly short cycles like this over their entire lifespan [and perhaps,
whether the end-of-life behavior tends to be any different from the
"normal" 10 or more hours per start usage, just to satisfy my curiosity
:) ]?
 
T

Travis Evans

Jan 1, 1970
0
The lighting companies usually publish survival curves for x hrs/start for all
HIDs. For this, you'd need to get a hold of an engineering catalog for the
corresponding company.

Fascinating. Is this something that an average consumer would be able
to easily obtain?
Because of sputtering and other side factors during the start, the ideal
performance is usually one lighting-continuous, and all other modes, such as x/y
hrs/start fall below the curve for continuous burning, in terms of overall
population survival.

However, understand that the published curves are probability distributions
showing only survival of 50% of the sample, as such they cannot be taken as
individual indicators, as they apply to the average/mean samples and not to
every single individual from the population.
[...]

This makes sense. So if I understand correctly, if I were dealing with
a fairly large number of lamps, I could expect most of them to last
roughly somewhere around what would be shown in general, but in terms of
individual lamps (especially in very small sample sizes like what I am
dealing with), one could consider it more or less "anything goes", right?

Would the percentage of lamps that have lifespans at the extremes
generally tend to increase as the average hours per start decreases, or
is it always about the same?

I'm hoping that at least some of my lamps will have the more gradual
death that I've typically seen in streetlighting and other installations
(cycling in the case of HPS, dimming/color shift with HPM/MH, etc.)
since I've never had a real opportunity to observe and follow the entire
end-of-life process of these lamps over time, nor certain aspects of
behavior (warmup, hot restrike time, etc.) during this stage. Though
in a sense, the unpredictability of what failure mode and when kind of
makes it somewhat exciting. :)

For a while, I have also wondered if there were some measurable
indicators in behavior that could be used to roughly estimate about how
far in lifespan a lamp is, but in light of what you've said and the
possibility of "unpredictable" failures, it would seem that if such an
indicator exists, it might not necessarily be meaningful in a particular
case.

Thanks for the response!
 
T

TKM

Jan 1, 1970
0
Travis Evans said:
The lighting companies usually publish survival curves for x hrs/start
for all
HIDs. For this, you'd need to get a hold of an engineering catalog for
the
corresponding company.

Fascinating. Is this something that an average consumer would be able
to easily obtain?
Because of sputtering and other side factors during the start, the ideal
performance is usually one lighting-continuous, and all other modes, such
as x/y
hrs/start fall below the curve for continuous burning, in terms of
overall
population survival.

However, understand that the published curves are probability
distributions
showing only survival of 50% of the sample, as such they cannot be taken
as
individual indicators, as they apply to the average/mean samples and not
to
every single individual from the population.
[...]

This makes sense. So if I understand correctly, if I were dealing with
a fairly large number of lamps, I could expect most of them to last
roughly somewhere around what would be shown in general, but in terms of
individual lamps (especially in very small sample sizes like what I am
dealing with), one could consider it more or less "anything goes", right?

Would the percentage of lamps that have lifespans at the extremes
generally tend to increase as the average hours per start decreases, or
is it always about the same?

I'm hoping that at least some of my lamps will have the more gradual
death that I've typically seen in streetlighting and other installations
(cycling in the case of HPS, dimming/color shift with HPM/MH, etc.)
since I've never had a real opportunity to observe and follow the entire
end-of-life process of these lamps over time, nor certain aspects of
behavior (warmup, hot restrike time, etc.) during this stage. Though
in a sense, the unpredictability of what failure mode and when kind of
makes it somewhat exciting. :)

For a while, I have also wondered if there were some measurable
indicators in behavior that could be used to roughly estimate about how
far in lifespan a lamp is, but in light of what you've said and the
possibility of "unpredictable" failures, it would seem that if such an
indicator exists, it might not necessarily be meaningful in a particular
case.

Thanks for the response!

There are two indicators of lamp life that are fairly simple to measure:
(1) Lamp light output over time and (2) Lamp operating voltage.

Manufacturers publish lamp depreciation curves so you could track the drop
in light output of your test lamp over time. If your lamp is at or above
the published curve, rated (or longer) lamp life is likely. But, a lower
measured output than published or a sudden drop in output could indicate arc
tube problems such as a leaking seal and thus imminent lamp failure

As lamp cathodes deteriorate due to lamp starts or burning hours, the
voltage required to start the lamp rises. There comes a point in time when
the starting voltage that the ballast can supply is lower than what the lamp
requires. The lamp doesn't start and that's the end-of-life point. So,
measure the lamp voltage (when the lamp is fully warmed up) over time.
Manufacturers also publish lamp voltage values; but the data are harder to
get. Ballast manufacturers have to have such data too, of course.

HPS lamps are exceptions because they utilize high voltage starting
circuits. Those circuits get them started OK, but then, if the lamp is near
the end-of-life and as the lamp warms up, the lamp voltage goes up beyond
what the ballast can supply, the lamp goes out and the cycle repeats.

Terry McGowan
 
T

Travis Evans

Jan 1, 1970
0
The Venture Lighting International HID Systems Sourcebook published in
2000 has the following data on page 142:

Compared to Rated Life at 10 hours per start:

5 hrs/start approx. life 75% of rating.
2.5 hrs/start approx. life 55% of rating.
1.25 hrs/start approx. life 40% of rating.

I cannot find any similar data in the 2011 PDF version of the same
catalog.

Thanks. It would seem that more frequent starts do generally
significantly reduce life expectancy, though to be honest, perhaps not
quite as badly as I was supposing they might. :)

One other thing I'm curious about in this regard are 120-volt self-
(tungsten-) ballasted HPM lamps. I'm told the 120-volt versions use a
bimetallic element to preheat the arc tube electrodes for several
moments in order to strike the arc at such a low line voltage. The two
SBM lamps that I added to my collection fairly recently indeed appear to
behave this way.

My thought is that very short periods of operation (where the lamp is
switched on and then off again before the arc-striking attempt begins)
would not affect life significantly since the only active component in
this case would presumably be the incandescent filament. For longer
periods where the arc does begin operating, I wonder if the electrode
preheating would have the side-effect of reducing sputtering on strike
in the same vein as with fluorescent lamp electrodes. If this is the
case, it would seem that these lamps' lifespans would be somewhat less
impacted by starts compared to ordinary HPM or HID where the electrodes
are not preheated?
 
Top