Now I'm aware lots of starting decreases lifetime on all lamps but incandescent, LED, neon, and Cold-cathode fluorescent, but by how much? I've read somewhere for example that a CFL start takes about 15 minutes off of lifetime. Is this about right? What are the figures for Xenon short arc, HPS, LPS, mercury, metal halide, ceramic metal halide, and xenon metal halide? Is a hot restrike worse than a regular start for the lamp? By how much? I've heard of "preheat," "rapid start," and "instant start" fluorescent, but which is better for for the life of the lamp, and by how much?

Just curious, thanks.

The problem is with electrodes of hot cathode discharge lamps running in cold-cathode mode just upon startup, before they heat up to normal operating temperature (usually a second after strike).
On HID's there is no other mean mean to start, then cold start and let the cathode fall dissipation heat the electrodes. Here it does not matter, if it is cold start or hot restart, as electrodes cool down in matter of few seconds, so when these lamps are able to restart, electrodes are already long time cold.

On fluorescent lamps electrodes could be preheated, so such damage avoided by proper ballast. But everything came with it's cost, so there are different types of ballast, mainly in terms of treating electrodes during the starting process.

"Preheat": Switch (starter) short out electrodes so, the ballast current flow only trough the filament, then open and let the lamp ignite. If not successful, it repeat. This method is theoretically capable to start the lamp without damage, but starter tolerances are too wide for this. Moreover there is market pressure to shorten the starting process, so starter attempt to strike the lamp usually too soon, so still some damage persist.
However this method suffice with simple series choke as ballast (all lamps for 230V, lower voltage types for 120V), while these ballasts are the most efficient magnetic types.

"Rapid start" The filament is heated from an auxiliary secondary, while the ballast open circuity voltage is set to be not enough to start lamp with cold electrodes, but enough when they warm up. At first glance this is the best method (controlled directly by electrode temperature), but there is problem with strong dependence of striking voltage on the tube cold-spot temperature: When too cold, it does not strike even with hot electrodes, when too hot, it "cold-start". Furthermore the starting is sensitive to electric field around the lamp, so to e.g. moisture or missing reflector. These ballasts need moderate OCV, what ask for the usage of step-up transformer even for 230V areas, so it is used only in 120V, where the transformer would be needed anyway.

"Instant start" Lamp is ignited solely by the high enough voltage. This method damage electrodes at most - their heating rely only on cathode fall losses.

"Programmed start" is a form of "preheat", where the heating is controlled more precisely. Usually this type does not use the starter, the whole lamp circuity might be wired in many different ways, depending on actual ballasting concept. The precision of the preheating might be even so high, then the switching has no influence on the lamp life at all.
This starting concept is usually used with high frequency electronic ballasts, as there is possible to control the start sequence by simple frequency control, without any extra power element. But some electronic starter replacements are on the market what do precise programmed start sequencing on present "starter-preheat" installations, but they are becoming obsolete.

The disadvantage (from general public point of view) is, then the light need some time after power on, before it strikes, so many people prefer instant start, as it come on instantly (as they are used to see on incandescents).

It is indeed true that starting frequency has an impact on a fluorescent lamp's life, regardless of whether it is being operated on a preheat, rapid-start, or instant-start circuit.  The lamp will last longer if starts are kept to a minimum.  The reason for this is the fact that every time the lamp is started, a certain amount of emitter material is spurned from the electrodes.  (The spurning off of emitter is what results in blackening around the ends of the lamp as the lamp advances in its age.)  This is not to say that a lamp that is in constant operation won't eventually fail, as emitter material is minutely spurned during the lamp's operation, however not nearly as much as it is during the starting action.

As far as on what type of circuit a fluorescent lamp will last the longest is truly debatable.  Theoretically, based upon operating principles, one could see that an instant-start circuit, which literally "jolts" the lamp with a high voltage spike to start it instantly, could be the most harmful for a lamp from a longevity standpoint.  However, it could also be theorized that a preheat circuit could be harsher on a lamp, due to the fact that the ends are preheated, sometimes for several seconds before the arc is struck.  Additionally, rapid start circuits aren't much different than preheat circuits, in that they heat the electrodes as well for a few seconds during the time the lamp is flickering to life before the resistance of the tube has dropped enough to allow full current to flow.

With regard to HID sources, the electrodes don't play a role at all in how long the lamp lasts; therefore, frequency of starts is irrelevant in determining how long an HID lamp will last.  How long the gases within the burner remain concentrated enough to sustain a continuous arc in the lamp is the deciding factor in how long the lamp will maintain a useful life, by either staying bright enough (as in MV), not cycling on and off (as in HPS), and not failing altogether (as in MH).  In summary, an HID lamp's life is solely determined by how many hours it has been run for.

Xytrell, I hope this answers your question!

@Brian:
The sputtering is result of high energy ions bombarding the electrode and it happen during whole lamp operation. But it's rate depend on ion's amount (obviously) and energy. The amount is given by the lamp current: higher current = more ions, but there you might do nothing about. Cation energy is given by the fill ionisation energy and is usually very small, but energy of anions come directly from cathode fall ( described on Figure 31 ): The potential difference between end of the anode column accelerate them towards the cathode and when their energy is high enough, they kick away an atom from the electrode.
The cathode fall depend mainly on it's emission capability: If the cathode would emit electrons freely (is hot), the cathode fall would be very small (few V), but if it does not (cold), the cathode fall might be 100's of V. At the same time the cathode fall causes power dissipation, what in fact heat the cathode during normal operation and if this is the only cathode heating mechanism (instant start, switch or starter preheat ballasts, HID's), it settle to the temperature, when the cathode fall causes just enough power dissipation to keep the electrode on such temperature (colder => larger fall => larger dissipation => heat up), what yield typical ~10V fall during operation. Note, then the cathode fall might increase, if the current overload it, e.g. by capacitor discharge current upon arc (re)ignition. This is the reason, why capacitive lead-phase ballasts yield shorter lamp life (more accurately their very high current crest factor).
Of course there is other phenomenon responsible for electrode loss: Evaporation. And this is faster, when the electrode is hotter, but this phenomenon is not as steep (towards the temperature) as the sputtering, so quite significant margin exist. So there is overall temperature optimum for longest electrode life.
And here come the influence of frequent starting: If the used starting method heat the cathode to the operating temperature before and there is no current spike upon ignition, there is no extra electrode wear. If the starting sequence overheat electrodes, they evaporate. If it does not heat them enough, they sputter away, before the cathode fall heat them to correct temperature.

  You are very wrong with statement, then sputtering does not influence HID life (LPS i don't count as HID here).
You are right, then electrodes have huge excess of material to not make the limiting factor, but when sputtered away, the material does not disappear. It settle on arctube wall and cause it's blackening, on MH's this is only slowed down by the selfcleaning halogen cycle.
On all lamps the blackening absorb the radiation output, so reduce the lumen output, while at the same time it increase the arctube temperature, mainly in the electrode region (it convert light to heat). This temperature rise has very minor influence on MV (as these use unsaturated vapors), but it does increase the pressure in both HPS and MH's (they have usually "cold spot" behind the electrode). This pressure increase cause the arc voltage to rise, what increase the power delivered to the lamp (the extent depend on the exact ballast characteristic), what increase the arctube temperature even more (and on HPS this power increase compensate the lumen output for the blackening, yielding good overall lumen maintenance over life). After some point this yield to thermal runaway, what result in lamp cycling (voltage rise so high, the ballast is not able to keep the arc), arctube leakage or (on QMH) even explosion ("normal" end of life of these lamps).

Thanks for the replies. So if metal halide lamps have a self-cleaning halogen cycle, why don't they last longer than they do? Why don't halogen lamps get blacker with age like metal halides?

Thanks for the replies. So if metal halide lamps have a self-cleaning halogen cycle, why don't they last longer than they do?

They are tailored to color properties and efficacy, what mean the hottest operation the material can handle, so e.g. the quartz crystallize much faster then on MV's.
And second the required color properties call for substances chemically aggressive towards arctube material, e.g. sodium in QMH (Hot quartz is acid, what promptly react with the metal forming dark brown opaque material) or halides in CMH (the ceramics is soluble within molted halides). The sodium problem is attacked by the electrical field, to avoid sodium ions being attracted from the discharge towards arctube wall - connection via wire loop, so it's field does not attract ions, using insulated quartz shroud as photo-effect charged positive electrode repelling sodium ions into the discharge center, operating bulbs at negative DC potential if possible (low frequency electronic ballasts),...

Why don't halogen lamps get blacker with age like metal halides?

Because in halogen incandescent is no discharge sputtering electrodes, neither free sodium reacting with quartz.

Now I'm aware lots of starting decreases lifetime on all lamps but incandescent, LED, neon, and Cold-cathode fluorescent, but by how much? I've read somewhere for example that a CFL start takes about 15 minutes off of lifetime. Is this about right? What are the figures for Xenon short arc, HPS, LPS, mercury, metal halide, ceramic metal halide, and xenon metal halide? Is a hot restrike worse than a regular start for the lamp? By how much? I've heard of "preheat," "rapid start," and "instant start" fluorescent, but which is better for for the life of the lamp, and by how much?

Just curious, thanks.

I had read a long time ago that...
Preheat fluorescents loose 5 hrs per start
Rapid start fluorescents loose 3 hrs per start
HIDs loose 10 hrs per start

With regard to HID sources, the electrodes don't play a role at all in how long the lamp lasts; therefore, frequency of starts is irrelevant in determining how long an HID lamp will last.  How long the gases within the burner remain concentrated enough to sustain a continuous arc in the lamp is the deciding factor in how long the lamp will maintain a useful life, by either staying bright enough (as in MV), not cycling on and off (as in HPS), and not failing altogether (as in MH).  In summary, an HID lamp's life is solely determined by how many hours it has been run for.

If you cycle a MH on and off several times, the tube will begin to blacken quickly. They will often get addicted to this behavior and continue to cycle a few times on there own.