2mTD: a capacitive dropper inside a Philips LED spotlight?!

[jitter]

Recently I dumpster dived a broken Philips GU10 4 W LED spotlight and obviously took it apart. What I found inside surprised me a bit from an A-brand like Philips.

It's this one that was introduced ca. late 2012.


Since I was likely going to break it permanently during the teardown, I turned it on first instead. The reasons why this was discarded was erratic behaviour of the light, and when lit it seemed too dull for the wattage (supposedly a 35 W halogen replacement). Probable cause: defective LED.

This is what's behind the lens:


200 V it says? Yep, when it lights, there's about 180 V across the two leads. There must be some shorted LEDs under that yellow goo and some others are intermittently dropping out completely.
The aluminium insert is not exposed on the outside, it's an all-plastic encasing.

Unscrew the top and you get to see one side of the driver-pcb:



What does "quasi-dc" mean?

But the most interesting find is a capacitive dropper:



Who'd have thought that in a Philips?

Some Big Clive style reverse engineering yields this schematic of the thing:



Some calculus with the values of C2 and C3 leads to a theoretical current of 28.6 mA, and sure enough, the specs printed on the side say "29 mA".

But some things seem underrated: C2 and C3 are rated 250 V, but I don't know if that's DC (too low) or AC (acceptable). Also, when the LED-chip drops out, the voltage across the electrolytic (C1) rises to more than 300 VDC, or more than 50 V over its rating.

The "driver" seems in good shape, the intermittent nature of the light is caused by the LED-chip. After some hours in the freezer, the problem became quite a bit more pronounced.
Let's conclude with the remark that I had expected better from a high priced Philips (and no, I have no reason to believe that this would be a counterfeit).

[microbug]

Given that these things claim to have a 10 year + life, it's pretty crap that they use sub-par components.

[jitter]

Given that these things claim to have a 10 year + life, it's pretty crap that they use sub-par components.

If it were some no name brand off eBay or something I would expect this (or worse), but from Philips?
I guess the dimmable GU10 LED lamps will have a proper driver, and I wonder how the current models are constructed.

[Zero999]

Could it be fake?

[SeanB]

No, real lamp, the dropper is used because it is cheap. the 10 year life is almost a joke, as if you look they are running the lamp for only a few hours per day to get that projected lifetime, or are using projected data for the led die inside the unit only, running in ideal conditions with a perfectly smooth and pure mains supply.

Real life they likely will do around 3 years if used for 4 hours a night 5 days a week, which is actually almost the same lifetime as the regular incandescent lamp they are replacing.

I actually had regular 20W halogen reflector lamps that ran for over a decade in service, running 7 days a week for 10 hours a day. They were the outliers, most would do 2 years though in that situation, though only the good brand names like Osram, Phillips and GE ( not china ones) has the long life.

[Pjotr]

Being not a fancy switching circuit doesn't mean it's crappy  Anyway no broken electronics, but broken leds!

[Jasper]

Looks like a perfectly fine driver to me. And isn't this dimmable, in fact?

It's the better SMPS/buck types that you can't really dim, afaict. Because the whole point of them is to output a fixed output voltage/current regardless of what their input is doing. The cheaper types aren't very good at that and will still get dimmer with diminishing voltage, but anything that's any good will reject all that nasty noise coming from the dimmer and manfully produce full output even from a chopped up 90Vrms input...

And given that it's the led that's broken, well.

Given the 200V forward voltage you measured we're talking about something like 60 series led chips in that one single cob package. Those must be tiiiiny chips, packed very densely. And then they're probably being driven at relatively high output.

PS: quasi-DC because the smoothing cap isn't nearly big enough to produce something that looks like "DC plus ripple" as a load waveform, but it's all in one direction so it's *technically* DC.

[rollatorwieltje]

How well does a capacitive dropper actually regulate the current? As already said these LEDs are probably driven very close to the edge. I wonder how well these power supplies respond to garbage on the mains. My experience is that LEDs die very quickly when driven even slightly over their rating.

Another thing to keep in mind is the cooling requirements of these LED replacements. Generally you can't use them in enclosed fittings, they need airflow.

Would be nice to see the difference between this one and their more expensive Master type. The Master range is supposedly more suitable for continuous operation.

[Gyro]

The problem with capacitive droppers is that their current is frequency dependent. They will pass considerably more current during spikes or when subject to sharp edges. The 4u7 electrolytic and series inductor will help a bit but the LEDs are probably still likely to see some overcurrent spikes.

[jitter]

That might be a good reason why this one failed early. Or maybe the LED was already weak from the factory, no way to know for sure.

I have no reason to believe it's fake, the quality of the whole lamp is on par with what I would expect from Philips, it's just that I would not have expected a capacitive dropper.
Looking on the internet for teardowns of Philips GU10 LED lamps I found one by Big Clive. And yes, that one too has a capacitive dropper.

[DmitryL]

By the way, a slightly unrelated question... Capacitors on the picture a marked with "250", which assumes 250V, I presume.
Is it 250V RMS (250V DC) or is it 250V AC (which is 350V amplitude for sinewave)?

[SeanB]

Should be a 400VDC capacitor, but they use a 250VDC capacitor, because the capacitor voltage rating is a minimum value and almost all units are tested at manufacture at a higher voltage. Thus your 250VDC unit is actually tested at 400VDC, and the 400VDC one is tested at 600VDC. This is to ensure they will still withstand 250VDc at high temperature, so the capacitor will operate at a peak of 330VDC on the mains peaks for a while before it degrades and starts self healing down in capacitance to failure as nearly open circuit.

400VDC capacitor is both larger physically and more expensive than a 250VDC unit, so they use the lower cost smaller one as it is good enough to last out the warranty period. As well the capacitors are not rated for direct connection across the mains, they should be a class X capacitor, which is made with 2 layers of plastic insulation between the electrodes ( so a flashover or void is not going to cause a current spike) and which is vacuum impregnated with resin so there are no voids, along with being tested to a much higher proof voltage during manufacture to ensure they will survive spikes on the mains. But those again are a lot larger and more expensive, so they use either a fusible resistor, fuse or fusible trace to provide protection against shorted capacitors in the units.

[Pjotr]

Should be a 400VDC capacitor, but they use a 250VDC capacitor, because the capacitor voltage rating is a minimum value and almost all units are tested at manufacture at a higher voltage. Thus your 250VDC unit is actually tested at 400VDC, and the 400VDC one is tested at 600VDC. This is to ensure they will still withstand 250VDc at high temperature, so the capacitor will operate at a peak of 330VDC on the mains peaks for a while before it degrades and starts self healing down in capacitance to failure as nearly open circuit.

Curious about the math you did use to find those numbers. Actually the peak voltage across the film capacitors will be: mains peak minus the the voltage across the 4.7uF el.cap. That is somewhere between 120V - 150V peak for 230V mains. Only at start-up when the el.cap is still discharged it will be somewhat higher for a few cycles.

There is simply not much wrong with this ballast and many led lamps < 5W use this kind of ballast: Reliable, small, low cost and virtually no EMI troubles.

[jitter]

Given the 200V forward voltage you measured we're talking about something like 60 series led chips in that one single cob package. Those must be tiiiiny chips, packed very densely. And then they're probably being driven at relatively high output.

I was able to get out most of the yellow goo without too much damage to the LED dies.

There are 4 dies, presumably of 1 W each as the lamp is rated at 4 W. Each die looks to have 15 junctions, so I'd say your educated guess of 60 LEDs is correct.

[Zero999]

Should be a 400VDC capacitor, but they use a 250VDC capacitor, because the capacitor voltage rating is a minimum value and almost all units are tested at manufacture at a higher voltage. Thus your 250VDC unit is actually tested at 400VDC, and the 400VDC one is tested at 600VDC. This is to ensure they will still withstand 250VDc at high temperature, so the capacitor will operate at a peak of 330VDC on the mains peaks for a while before it degrades and starts self healing down in capacitance to failure as nearly open circuit.

400VDC capacitor is both larger physically and more expensive than a 250VDC unit, so they use the lower cost smaller one as it is good enough to last out the warranty period. As well the capacitors are not rated for direct connection across the mains, they should be a class X capacitor, which is made with 2 layers of plastic insulation between the electrodes ( so a flashover or void is not going to cause a current spike) and which is vacuum impregnated with resin so there are no voids, along with being tested to a much higher proof voltage during manufacture to ensure they will survive spikes on the mains. But those again are a lot larger and more expensive, so they use either a fusible resistor, fuse or fusible trace to provide protection against shorted capacitors in the units.

I'd say it's more likely they used 250V capacitors because the nominal voltage across the capacitors is under 250V.

When the LEDs are working properly, the voltage across them is 180V and the voltage R2 & R3 will be 7.4V, which is fine for C1. The voltage across C2 & C3 will briefly exceed 250V at switch on but that shouldn't cause any problems as they're class X and will be designed to withstand the surge. Once it's settled to the steady state, the voltages across all the capacitors should be comfortably under 250V. If the LEDs fail then one or more capacitor may blow but the class X capacitors are flame retardant, the electrolytic is vented and is small and there's a fusible resistor so it will fail safe.

[Ice-Tea]

Have a bunch of Philips GU10's, albeit slightly different type. Been running 5+ years now. Same driver topology, so not a fake. Had one rail on me over the years.

AFAIK: most "brand" GU10's are rated fort 15 - 25000 hrs or so. Claims of 10y only hold if you don't run them all that much. Chinese vendors will tell you what ever you wanna hear.

[jitter]

Should be a 400VDC capacitor, but they use a 250VDC capacitor, because the capacitor voltage rating is a minimum value and almost all units are tested at manufacture at a higher voltage. Thus your 250VDC unit is actually tested at 400VDC, and the 400VDC one is tested at 600VDC. This is to ensure they will still withstand 250VDc at high temperature, so the capacitor will operate at a peak of 330VDC on the mains peaks for a while before it degrades and starts self healing down in capacitance to failure as nearly open circuit.

400VDC capacitor is both larger physically and more expensive than a 250VDC unit, so they use the lower cost smaller one as it is good enough to last out the warranty period. As well the capacitors are not rated for direct connection across the mains, they should be a class X capacitor, which is made with 2 layers of plastic insulation between the electrodes ( so a flashover or void is not going to cause a current spike) and which is vacuum impregnated with resin so there are no voids, along with being tested to a much higher proof voltage during manufacture to ensure they will survive spikes on the mains. But those again are a lot larger and more expensive, so they use either a fusible resistor, fuse or fusible trace to provide protection against shorted capacitors in the units.

I'd say it's more likely they used 250V capacitors because the nominal voltage across the capacitors is under 250V.

When the LEDs are working properly, the voltage across them is 180V and the voltage R2 & R3 will be 7.4V, which is fine for C1. The voltage across C2 & C3 will briefly exceed 250V at switch on but that shouldn't cause any problems as they're class X and will be designed to withstand the surge. Once it's settled to the steady state, the voltages across all the capacitors should be comfortably under 250V. If the LEDs fail then one or more capacitor may blow but the class X capacitors are flame retardant, the electrolytic is vented and is small and there's a fusible resistor so it will fail safe.

But that's the thing, I don't think that those red metallized film capacitors are rated X class at all.

[Kjelt]

They are metalised film caps with high self healing properties according to their datasheet, the bulbs are Tuv and UL and KemaKeur tested and approved, I don't see any problem?

[Seekonk]

I took many cap drop designs through UL and none of them were X types.

[jitter]

We're talking about a design here that has no exposed connections (other than the pins going in a socket, that is). If a cap fails short circuit it would blow the fusible resistor.

I'm pretty sure that (e.g.) a mains filter with caps directly between live, neutral and protective earth will need to be X or Y class to pass those tests.

[Ice-Tea]

Probably different because it is lightgear: when you have filaments connected between L and N, there is no additional "safety layer". I assume the same reasoning applies to the driver of GU10 bulbs.

 

[Gyro]

Actually, decent quality British standard filament lamps, Osram, Philips etc. used to include a fuse in the base (a little glass capsule). This was to prevent nuisance blowing of the lighting circuit fuse when the filament ruptured and the gas inside the lamp ionised.

Of course with cheap imports that went straight out of the window. 

[jitter]

I have never seen those fuses over here, not even in Philips bulbs. Perhaps the fact that we only use fuses/circuit breakers in the fusebox and not on sockets nor in plugs helps. A 16 A fuse isn't going to melt when a bulb blows.

The only nuisance blowing/tripping I experienced is inrush current on a 150 A MIG welder (running off 230 V single phase) on a 16 A circuit breaker. On a 16 A fuse it's fine though, those aren't quick enough.

[Gyro]

The fuse doesn't necessarily have to look like a fuse    The capsule type I mentioned is known as a 'Ballotini fuse'. Fusing can be implemented as a section of the lead-out wires (somehow, can't see the detail).  See:

http://lamptech.co.uk/Documents/IN%20Fusing.htm

Edit: I have personal experience of the 'secondary arcing' that is mentioned burning a hole in the side of the base of a cheap lamp, not surprisingly it also blew the lighting circuit fuse.

[Kjelt]

corrected, misread.