I always wondered this, why some clear mercury vapor lamps, mostly yardblasters, that I see are glowing a strong blue green while others are more white.  I know about white phosphor coated ones but Ive observed this with clear ones. I prefer the dimmer blue greeny ones that are old and dull.  There's one left by the hospital I work at and one at a restaurant.  The hospital is slowly changing over to LED :-/

Greener ones are older, and deposits on the inside of the arc tube block the light output. Newer ones which have a clearer arc tube produce light that appears whiter, especially when in proximity to older lamps. The same thing can be observed in coated lamps. Yardblasters, having cheap ballasts, may overdrive or underdrive lamps, which may also explain the difference.

One effect is the age/end of life:
Mercury lamps usually do not completely die at the end of their useful life, but throughout the whole life they gradually degrade. It mean gradual loss of lumens, till it become more economical to replace the lamp, than count for the further loss by the lighting system design (higher wattages and/or more fixtures to compensate for the output loss). As the lamp decay's gradually, the people moving there daily get used to the actual lamp state, so do not recognize directly the degradation reaching the end of life point. The lumen loss they do not see as a change, but usually blame the "old technology equipment" to be "so inefficient" (because they see their street being darker than they would like), instead of maintaining it properly (the habit "Why to replace the bulb, when is still lighting" mean the overaged bulbs, emitting very reduced light output, but still consuming full power are kept in service). And it is these overaged lamps, what count for the vast majority of the greenish glowing lanterns.

What happen is the electrode slowly evaporate/sputter (mainly during each start, but even during normal burning) and condense on the colder arctube. Here it form a kind of hazy layer (it is not homogenous, but consist of very small particles). And it are these particles, what block the shorter wavelengths at first and only as the coating progresses, they start to block longer as well.
In 70's a modification to the electrode coat composition was introduced, what left it's residues white (so less light absorbing), instead of the black bare metal deposit (it deposit in white oxides instead). This allow way thicker/denser layer to build up, before the light output degrade, so the useful lamp life was extended.
But the deposit whitening agent get consumed after some time (usually the designed lamp lifetime), so then only black layer get deposited again, causing steeper lumen droop.


So the lamp does not loose it's output uniformly across it's whole spectrum, but the loss start from shorter wavelengths. As the light from clear MV's consist from some blue peaks, some green and yellow (so they form about a white light), the blue get lost first.
That mean, than the green and yellow become more pronounced, so the light get greenish hue.

The same effect on coated lamps first reduce the arctube's output in the UV, reducing the light emited by the phosphor. As the phosphor usually emit in orange, blue and some green (so across  the complete spectrum), the lamp (beside the lumen loss) first loose it's CRI and then it shift in color to the green.

Wow, thanks a ton!  Interesting.  Yeah my favorite fixtures are those with greenish dimmer bulbs and loud ballasts.  Back in 1992 at a best childhood freind's house there was one such on a pole with a timed twist switch underneath.  I loved to start it as a kid. it was a yard blaster style but with the bulb at a 45 degree angle. I recently moved close to that areaup when I bought a house, it is still here! That house is abandoned now but some of the lighting around it seems to be on, maybe one day I will park my car and walk over and video turning the switch, it has been 20 years I wonder if it would light!! Or if I could get the fixture! It acted old back then!  

The EOL mechanism of MV lamps, seems to be the same as LEDs (Lamp gradually reduces intensity, until their light intensity is usefullness, but continues to consume the same amount of power. This is the exact EOL mechanism of LEDs).

Mercury lamps sometimes do fail "at once", usually after some time in the dim/useless mode but sometimes they fail when they still made quite a lot of light - In this case they are just unable to start at all, sometimes the starting electrode is glowing a bit but unable to start the main arc

This failure mode was common with mercury lamps of all manufacture dates, but the newer ones (with reduced quality) are more prone to this than good old lamps



LEDs have many failure modes depending on what happened :

First since they use a driver, there are all possible failures with the driver

Phosphorless LEDs running at low current and low temperature have pretty much infinite life. There are LED indicators in equipment that have been lighing continuously since the 80's which is over 200000 hours. Every blue LED power indicator on an electronic device from the early 2000s (when blue LEDs caught on) have done 80000 hours by now. There is no reason why lighting LEDs at low current would not last as long too

Phosphor LEDs running at decent current have slow wear of the phosphor - so color shiftng and brightness drop (in the orange color) with time. This is usually slow but can be very fast in some LEDs - Some pink LEDs have been reported to have severe color shifts within hours and days - most LEDs doing this are with organic phosphors. The phosphor decays and the LED becomes just blue LED, somewhat obscured by the non functioning phosphor

Higher running current and temperature, reverse voltage peaks etc cause LEDs electrical damage. In this case the LED itself will either go out at once (often shorting out - which means that other LEDs in series with it will be overpowered and it can start a cascading failure of a lamp or string with several LEDs), or slowly degrade in the form of "bad spots" - The corner of the chip stops lighting, then another, the dark spots expend and at the end there is little or no light from the chip anymore

Some pink LEDs have been reported to have severe color shifts within hours and days - most LEDs doing this are with organic phosphors. The phosphor decays and the LED becomes just blue LED

I had some magenta LED's do just that...I think it took a week or so of running 24/7 and they were fully blue.


---
LEDs in general (both phosphor-type and normal) I've noticed 4 main failure modes:
* Slowly becomes dim
* Flickers somewhat like an EOL fluorescent (can do this for months)
* Quits lighting, but still conducts (basically shorted)
* Quits lighting, and no longer conducts (open)

Quote from: Ash link

Some pink LEDs have been reported to have severe color shifts within hours and days - most LEDs doing this are with organic phosphors. The phosphor decays and the LED becomes just blue LED

I had some magenta LED's do just that...I think it took a week or so of running 24/7 and they were fully blue.


---
LEDs in general (both phosphor-type and normal) I've noticed 4 main failure modes:
* Slowly becomes dim
* Flickers somewhat like an EOL fluorescent (can do this for months)
* Quits lighting, but still conducts (basically shorted)
* Quits lighting, and no longer conducts (open)

I've seen even quit lighting, but electrical characteristics remain unchanged. Usually as result of an ESD event.
LED's, even with quite large chips, are extremely sensitive to electrostatic discharge, way more than most of the dense integrated circuits (and here I mean the dense CMOS stuff, so high integration microcontrollers,...)

"had some magenta LED's do just that...I think it took a week or so of running 24/7 and they were fully blue"

Thats phosphor degradation. I had another case with a LED, where the chip in a white LED failed within few days of intermittent use (about 24 hours total lifetime). This was a USB light for laptop with 1 5mm white LED. When i took it apart later i found a 33 ohm resistor inside. (5V-3.4V)/33ohm = 48mA. With a LED which is rated to 20. No wonder it failed. I replaced the LED and put a resistor which makes the new one run at about 13mA



Slowly become dim = For phosphorless LEDs its chip degradation from overcurrent/overheating. The rated 20mA for most LEDs are too much. i prefer to run them at no more than 10-13mA (10mA best, 13mA compromise when not want to use too much LEDs)

Flickers = broken bond wire (sometimes tapping the LED or applying force to the leads will put the LED in full working or full not working condition, untill it is disturbed again)

The bondwires tend to detach by electro-migration, when operated too hot. What degrade here is not the bondwire itself, but the metal layer on the semiconductor device forming there the interconnects. Most frequently it fail on bondwires, where the current flow from the chip to the package lead (so LED cathode, IC ground connection,...).
The major factor is the temperature, second the operating current.
With "20mA" and other small LED's this frequently happen, because these devices usually have poor thermal properties, so tend to run quite hot.

Just wanted to revitalize this thread for minute; all I wanted to say is thanks. This thread right here hooked me on Lighting-Gallery... for that I thank you!!

Same! This was one of the first ones I read, lol.

so in theory  if designed "properly" LED could last nearly indefinitely outside of "mechanical damage" OR Phosphor degradation on coated LEDS

Interesting that this thread went from MVL's to LED's...

there is a parking lot I am watching that is mostly abandoned but the lights are still powered on and only 2 light at all one is stuck on the starting electrode and the other is as dim as with almost zero light making it to the ground
a month ago BOTH were very dim but LIT properly