It’s going to be a teardown double post, although not a happy one for me for a number of reasons.
IKEA Sparsam 15W CFL
This particular CFL is a young globe – only just installed prior to my holidays but purchased back in 2015 with other globes from IKEA. Unfortunately, when I got back from holidays, I was told that it was dead and flickered intermittently before failing. Sure enough, it was dead, but it wasn’t entirely at fault (see section at the end).
As this is a dimmable CFL, the ballast complexity is much greater than that of your “garden variety” el-cheapo CFLs.
Unfortunately, nasty stuff has definitely happened, but in a subtle way. On the one hand, everything including the woven insulation is covered in an oily substance, and it smells very much like a vented capacitor. The ballast design is made of two PCBs intersecting at 90 degrees to make the most of the space inside the base, and features two transformers/large inductors, and a number of polyester capacitors.
It seems that this unit may well have a double-safety of sorts – a fuse in line with one side, and probably a fusible resistor in the other (or at least, an RF filtering choke).
However, it does seem to have leaked out of the bottom – staining the two insulating sleeves. That being said, it still does achieve its rated capacitance value – but it may not reach all operational specifications (e.g. be so safe at withstanding operating voltage).
The underside has a few diodes, capacitors and resistors along with a bridge rectifier. While none of the fuses have blown, the suspected ultimate failure was the opening of the surface mounted glass diode package on the bottom left corner – this may have been responsible for the preheating of the filaments, and without it, an arc cannot be struck.
The globe itself is surprisingly smart – the vertical PCB contains three main ICs, one of which is a Microchip PIC16F616 8-bit microcontroller, a ST TD221ID Gate Driver and Two Point Regulator and a ST L6574D Ballast driver.
After desoldering the PCB clean, we find four ST D2NK60Z 600V MOSFETs.
The filament, unsurprisingly, is intact with the tube showing no real signs of distress. So what happened to the globe?
The cause of failure was most likely the fault of the luminaire fitting. The Edison Screw holders have a habit of fatiguing over time, so the center pin contact is only intermittent. The result is that the intermittency “tricks” the globe into pre-heating the filaments almost constantly because it thinks it’s being switched on and off. In doing so, it seems to have overstressed the diode responsible for preheating – and if it didn’t, it probably would have eventually burnt a filament out. The constant powering on and off probably resulted in high in-rush currents into the capacitor which appear as excessively high ripple current, which probably overheated it and caused it to release its contents inside the globe as well. An external cause of failure – something that’s often overlooked.
The good news is that I’ve cut out the old holder and fitted the “guts” from a regular bayonet cap holder into the remains of the downlight fitting, so the kitchen can maintain its original look while also eliminating the problematic Edison screw holder in the first place. Definitely a win here, and it’s still going today. Just a shame that globe was ruined by the fitting.
OSRAM Lightify CLA60 RGBW
This was also a big disappointment to see it failed – I received this for review back in 2015, and while it was “noisy”, it operated just fine. Less than two years later, it’s completely failed. Initially, it was placed into a desk decorative light fitting, rarely used but powered in case the need arose. After a lighting rationalization, I swapped it into my overhead ceiling fitting where it remained for a month before silently going out while I was doing work.
The globe refused to be recognized over Zigbee, but was not completely dead. Instead of consuming 500mW at idle, it consumed just 60mW, suggesting something catastrophic happened to the power circuitry or microcontroller inside.
As we noted last time, the front panel has LEDs but no screws. So how can we get in?
Scraping the plastic up the sides didn’t give us any answers.
A bit of reluctance had me eventually prying the LED panel off the base, where we find thermal-adhesive tape.
A bigger disappointment was to see the internals of the base are potted through – there’s no way to get into it from here it seems.
Out of desperation, I tried to get in from the other end, only to find that the whole thing down to the cap was potted. There’s absolutely no way to service these globes or repair them, and it seems a bit of a waste. The fault may have been trivial to fix, but now, the whole thing is waste (well, I could reuse the LED panel).
If I had paid the AU$60-75 retail cost per globe at the time it was launched, this would have me very much disappointed. Surprisingly, it was this “Made in Italy” globe that failed first – the accompanying “Made in China” Tunable White globe is still performing okay. Prior to its failure, it did occasionally do strange things – power up when not commanded to do so, drop-out of the Zigbee network and then come back. There’s a good chance that the power supply may have been based around a capacitive dropper which has a capacitor that’s failed.
In short, it seems the kitchen is bad luck … and smart globes can be quite an expensive investment if they fail so soon. Just like a doctor that couldn’t save their patient, I feel a deep disappointment that I couldn’t do anything more to save this globe. The odds were against us from the start.