Several years ago I noticed there appeared "switching cycles" (or numbers of starts) on CFL boxes.
I've picked a couple of values from two well-known manufactures:

Model Life (hrs.) Switching cycle Osram Duluxstar Stick 6,000 6,000 Osram Dulux Pro Mini Twist 8,000 10,000 Osram Dulux Superstar Stick 10,000 20,000 Osram Dulux Superstar Classic A 30,000 30,000 Osram Dulux Pro Stick 10,000 30,000 Osram Dulux Intelligent Facility 20,000 1,000,000 Model Life (hrs.) Switching cycle Philips EcoHome Mini 6,000 4,000 Philips Essential 6,000 5,000 Philips Economy Stick 6,000 6,000 Philips Genie 8,000 5,000 Philips Master PL Electronic 20,000 50,000 Philips Master Stairway 20,000 500,000 But it wasn't always so. I can remember the beginnings of CFLs with an electronic ballast (more than 15 years ago?). There was "frequent switching didn't matter" written in the ads. Now it seems it does. It's obviously a planned obsolescence. The lifetime hours printed on the box sell the product. But the switching often kills it prematurely. If you buy a cheap version, the switching cycle is strongly limited. Or am I wrong about the manufacturers' intention? Note: I have one Osram Duluxstar (E27/13W/2U) bought approx. 8 years ago. It's in the place where it already did at least 50,000 switching cycles. And it has worked flawlessly since today. It should have ended its life after 6,000 starts.

The wear related to switching cycles was always there, it was even worse than with newer lamps. But it was kept secret from customers (in other words the lamp makers were lying to customers all the time).
What forced makers to rate their products for switching cycles was not the sudden limitations, but the legislation: All products for home use were mandated to be life rated for two operating cycles: First one is aimed at "steady burning", the test pattern is repeating 2.75hours ON / 15 miutes OFF and log the number of hours burning (2.75hours per cycle), till 50% of the test population fail. This yield the rated life in hours. Second mandatory rating is aimed at power cycle life and the test should be done by repeating 15min ON/15min OFF pattern, again till 50% of the test population fails. This then yield the "switching cycles" rating.

Here are two things very important:
Both test cycles stress the tested lamps for both starting cycles, as well as burning hours. So an incandescent rating of "1000hours/4000cycles" does not mean the lamp suffer from power cycling in any way, but when you calculate the total amount of burned hours in the "cycle test", you get the same 1000hours as in the "life time" test, so that mean the lamp does not suffer from starting cycles at all (even when the 4000 "cycles" seems to be low at first glance).

And second, both tests mean the time/cycle stress till 50% of the population fails. So when reaching the rated life, already half of the lamps are supposed to be dead. With statistics about larger amount of lamps this rating tells, how many lamps you need to replace over given time, given you relamp just when the lamp fails.
But it is well possible, some lamps die after few hours, as well as some endure way longer than rated.
Now all lamps suffer from some wear, very frequently the wear is somehow traded with other performance parameters during lamp design, you can not have all at the same time, so an unusually long life actually mean the lamp was way inferior in other parameters, mainly light output, it was just not spotted.
The only signature of good quality control is the shape of what is called "lamp mortality curve" (there are actually two, one for each test), showing how lamps fail over time. With good quality control this curve is rather sharp, so you have ideally no failures before the rated life and virtually all the population failed just after that time. That mean, you may rely on these lamps for nearly all of their rated life.
But what is rather common mainly with the cycle life is quite flat curve, so that mean from very early in the life test you have failures and still many lamps survive long past the moment, when 50% are dead. The "many surviving" are then quite useless, when the earlly failures mean you can guarantee only very short time of flawless operation...

@Medved:
If the switching cycle is performed with 15min on and 15min off, how do you explain the endurance of the osram dulux facility (100000cycles=250000h?) and the phlips master stairway(500000cycles=125000h?)? They use a different measurement system?
Btw these two lamps use a true programmed start, probably  voltage rather than current controlled for better heating temperature accuracy with short preheating time.

The 15min ON/15min OFF cycle is in the standard the EU requires to be used for cycle life rating, nothing else.
So claims like "1000000" or "500000" switching cycle claims mean either the lamps are not rated according to that standard (what could be possible for older lamps - the requirement came in to effect short before the first incandescent bans).
The "half million cycles" were probably meant as an equivalent starting wear recalculated to the full lamp life, but has no backing in any standard.

But as I went through the Osram catalogue from 2008 and the half million claims were really a ridiculous nonsense there (with the footnote they claim "90min ON/15min OFF" cycle, what would mean burning 150khour with 10khour rated lamp), it was part of the "not guaranteed" marketing bull**it.
In the section extra devoted to lamp life they state 10000 cycles with footnote of 60min ON/15min OFF (it mean they did their own tests at time, when the standard was not yet there, 10k cycles that test can prove at least 10k cycles for the standardized 15minON/15min OFF), what mean the switching really does not wear the lamp at all..

The wear related to switching cycles was always there, it was even worse than with newer lamps. But it was kept secret from customers (in other words the lamp makers were lying to customers all the time).
What forced makers to rate their products for switching cycles was not the sudden limitations, but the legislation:

I still think that the quality of CFLs decreased during last 10 years. This is not just my observation or a regional issue, it could be so worldwide. And the switching cycle seems to be the main reason of their premature death.
If this is not on purpose (i.e. planned obsolescence), I can think of two other aspects:

1. In nineties, the manufacturers didn't offer that wide range of products. New affordable (economy) versions of lamps emerged later. The original versions became "Pro/Master/..." versions and the new economy versions introduced new parameters - shorter burning hours and much worse switching cycle. Manufacturers, driven by competition had to reduce their costs and started to use cheapest components available.
2. People weren't happy with very slow warm-ups of some CFLs. Some of them were so ridiculously dim for several tens of seconds that it limited their usability. In last years I don't see these very often. The warm-up time was reduced (or, at least start intensity significantly improved) but it's possible that it reduced the switching cycle (there could be a connection).

The electronic CFL circuit design of the early 90's was the same as used today on the cheapest versions. Even the life was similar: Around 6..8khours of mean life.
The CFL's were not in so wide use, so the chance to meet a bad piece was low. And with so high margin, the makers were preffering to replace the diffective pieces promptly, just to prevent the related publicity. Today, with so many pieces used around, it is natural you meet many of the defective ones as well.
There is a difference in the size of the ballast compartment: In order to be really usable as incandescent replacements, they had to be optically compatible. The ones from 90's were not at all, but most of the spirals are. That mean a drawback of tight space for the ballast, so higher operating temperatures of everything inside. That was only offset back by the technology development, no real improvement at all. But the benefit is, you can really use it.
What did changed was mainly the cost: In 90's, the retail price was around $50 of today's money, however the margin was rather high on these products. Today the selling price of technically similar products is around $5, but with way lower margins in the complete chain and nearly no development cost (compare to the early 90's).
What is annoying, when there are defects, the same defect is present in whole batch, so when you have the bad luck when installing many of them form that affected box, you have many failures at once. The reason behind is the mechanized production, so one mistake does not affect one lamp, but the whole batch. But on the other hand, the average failure rate isn't that different than it was in the 90's. It just get more concentrated around bad batches rather then distributed over the complete production.

And the few pieces lasting after that long time are no judges. From todays production will be similar portion alive after two decades. It will be just the lucky pieces not used as heavily or e.g. lucky to run at lower power (after that time nobody would question the inferior performance during the life of that piece anymore)

So my impression is, the CFL's are not that worse than they were, they are just not so much better as the marketing had promised. But that happened all the mankind, so I doubt it will change any time soon...


And there is another aspect: People remember just the good things from the past, never the bad ones. So the past always look better then the present.

I can accept everything but the last point: Lamp lives from 1997 to 2010 etc. - this is a fact, not any kind of nostalgia. But OK, I was a happy person having good batches of Duluxes and magnetically ballasted Prismatics.

Overheating the electronics due to reducing lamp size (so you can put them in small fittings) is a good suggestion. I also think that downward burning position (ballast above) makes the situation worse. Btw. several years ago, I read on the box of a stick CFL, that in that downward position is its luminous flux reduced to 90%). It was a stick CFL (I cannot remember more details now).

And I guess that problems with over/under voltage (there's occasional undervoltage in our area) and surges can kill the electronics, too.
The voltage fluctuations have reportedly worsen with so many solar power plants and wind turbines installed.

With the temperamental solar and wind power the stability maintenance became really a nightmare, but the over/undervoltages (you are describing on the other post in more detail) are usually linked to some fault between the transformer and your home, or even inside your home itself. It looks more like a loose Neutral problem.

When talking about magnetic vs electronic robustness, it depend on what over/under voltage we are talking about.
Undervoltages with magnetic ballasts cause the arc instability, so frequent starter actions, which then wear out rather quickly.
Mild, but long term overvoltages (~250..260V for hours or longer) quite "reliably" fry nearly any magnetic ballast (unless it features a resetable thermal cut out) just on the excessive losses, but it is handled by most of the electronic rather well (cumulatively it may affect it's life by the extra power with some designs, but way less than with the magnetic). Most of the simplest ballasts even do not mind at all (transistors are usually at least 600V rated, electrolytic capacitor 400V rated, but that mean it can tolerate a second or so of 600V surge as well, input rectifier diodes are usually of the 1kV rating, with the ferrite ring controlled selfoscillating designs the lamp arc current, so power, is quite well stabilized by the ring saturation, so no extra power at all).

But when speaking about short time (100's of us or so) high voltage surges, the electronic usually fail way easier on a voltage breakdown, while the magnetic components can tolerate kV pulses.


And in the case of the magnetic CFL's: I've seen them failing after just a few months as well: Once it was a defective batch (wrong ballast was installed, so it overheated), second time it were just the real Prismatic magnetic and incandescent epidemic failing, while electronic had no problem (it was a mains overvoltage due to loose Nedutral; there was a terrible mix of incandescents, magnetic and electronic CFL's, the electronic CFL's were the only thing what survived the event).
So bottom line really not much difference, just each of them have their different strong and weak points.

OK, thanks. Electronic ballast robustness is a good news for me.

@but the over/undervoltages (you are describing on the other post in more detail) are usually linked to some fault between the transformer and your home, or even inside your home itself

It seems that this is a problem in a wider area. At place where I currently work (this is less than 1 km from our home), there are problems, too. An expensive control computer of a CNC machine died several times before the undervoltage circuit-breaker was installed. And I found that few desktop computers with ATX power supplies were unable to turn on until the mains have been disconnected and reconnected again.

The power company doesn't find anything wrong with our power supply. If you are at least 95% of time within a 10% tolerance, you can't complain. What can happen in those 5%, it is not specified.

The overvoltage should not appear at all.
With the undervoltage, I'm not sure if that is allowed, but the only solution the power company has is to switch OFF the power.
In fact I would guess they could be aware about the problem you experience, desire to solve it (it formally isn't a problem, but it may indicate something is happening with their installation and they would like to prevent it from getting worse, when they would have to pay), but they are most likely unable to track the root cause.
If it happens frequently, you will need to log the voltage and how it evolves, capturing what happen during the "5%" and add it to the complaint and at the same time publish that, that will help a lot (the managers in the power company won't be able to pretend nothing happens and the technicians would have something to really track down the root cause)

And whether the magnetic or electronic are more robust, depend on the type of most frequent events and what protection is installed.

For the PC supplies: With an undervoltage they are not able to maintain the output, so the output undervoltage protection shuts them down. And this usually needs the input power to be cycled OFF and ON in order to reset it.
Holding the "power" button for >4seconds (= forced Power Off; that feature is in hardware, so independent on what the operating system does) may deactivate the "PowerOn" signal, so reset the protection as well, provided the 5VStBy supply haven't collapsed.