For some strange reason, it seems like it’s recently been CFL failure season. It could be the weather, but more likely, it’s just a random co-incidence. Today’s teardown candidate is another one of IKEA’s old series of CFL globes – I just so happen to be quite a fan of some of their products because they take the time to pre-heat the filaments and generally feature decent internals and endure the test of time.
This one was one of their fancy round frosted “GLS lookalikes”, model GSU111, rated at 11W for 530 lumens (48lm/W), made in China in week 35 of 2007. Judging from its production date, it had lived a full life. Already, from the product code, it appears to be OEMed by Megaman, a brand of Neonlite Electronic & Lighting Ltd. Hong Kong, as part of their CFL Classic series.
The symptoms of failure were obvious – no reaction at all from the light globe. Upon opening the luminaire, a nice oily brown liquid was found dripping from the globe – its obvious that an electrolytic capacitor had gone “bang” and vented its insides. The smell was somewhat pungent – a bit like toasty burnt nail-polish.
The volume of electrolyte vented was quite a surprise, as upon closer inspection, it had pooled near the threads as well.
The globe itself had a frosted outer appearance, and this frosting was performed by a “laminated” texturised translucent rubber sheet on the globe itself. It had started to peel at the end, and eventually I peeled it off completely to reveal a transparent outer glass envelope. I guess this rubber sheeting has a secondary function of “shatter protection”, to prevent the shards from going everywhere.
Of course, I didn’t actually know it was glass at this time. It looked a little hazy for glass, so I started chipping away at the plastic, trying to pry the dome off. Eventually it came off, with a sharp snap … at which point, I was glad it didn’t just completely explode into shards all over my workbench. The remnant glass can still be seen, glued to the remaining collar which I didn’t savage.
Internally, to make this globe compact, the tube itself has an intricate design. It is both folded in the U-shape sense, as well as “terminated” at square ends joined together (out of sight at the underside). There is a small “nipple” at the bottom of the U pictured above – I wonder if this is a cold-spot to keep the mercury vapour pressure regulated.
To get to the electronics required a lot of cutting at plastic and then breaking of wires. A 3A in-line fuse can be seen hanging above – it’s usually 2A for lighting products, so it seemed a little strange. The high voltage main capacitor has obviously been the one to vent, and it seems like it was under quite a lot of stress. It’s rated at 105 degrees C, but its heatshrink wrapper is deformed, indicating that it had overheated (possibly just before exploding).
I suppose it is a little surprising that the vented cap is a quality product – a Rubycon CFX series electrolytic. IKEA globes have, in the past, featured quality capacitors, so this might just be a one-off failure. The Rubycon CFX datasheet specifies the capacitor for a load life of 5,000 hours at 105 degrees C, which might sound short compared to the 10,000 hour expected lifetime of the globe, but is actually more than adequate (provided it is correctly sized for ripple) as the actual operating temperature is likely to be 50-85 degrees C. Using the Illinois Capacitor Lifetime Calculator, the predicted life is 20,000 to 226,274 hours.
Taking the PCB away from the globe, we can see the use of pinned connections to rivets in the PCB as a way to connect the filaments to the PCB without wire-wrapping or soldering. There are burn marks on the plastic frame, which suggest other component failures occurred.
A close look at the tube ends shows one rounded side, with the other side appearing to be a chipped cylinder of glass. This is the second time I’ve paid close attention to these ends, so maybe that’s normal … I’m not entirely sure. The adhesive sure looks like it’s been roasted long enough though!
A quick check of the filaments seem to show that they both have a high-ish resistance of 13.5 and 12.6 ohms respectively. The resistance is higher than expected, so maybe the filaments have themselves degraded over time with the unequal readings suggesting the tube may be suspect.
A look at the underside of the PCB shows numerous surface mount resistors have exceeded their rated dissipation and failed by “smoking”.
Aside from the vented cap and the burnt resistors, the rest of the PCB looks unremarkable. We can see four regular diodes being arranged in a full wave bridge, inductors used for RF suppression, and a thermistor, possibly for soft starting.
In the end, it seems the bulb’s lack of action was due to the protective opening up of the fusible resistor, which registered an open circuit.
As a result, I think the likely failure scenario is as follows:
- Capacitor may have been subject to a surge or a degradation in its working voltage due to impurities or contamination in the electrolyte.
- This would have led to increased temperature due to increased leakage current causing heating of the capacitor.
- When the temperature increases, the degradation rate also increases. This process could happen over a long or a short period of time, but the accumulated heating and leakage current effects cause the capacitor to overheat and vent from overpressure, spattering its guts everywhere.
- It is likely, and somewhat possible, that the loss of electrolyte compromised the insulation within the capacitor causing it to fail almost dead-short, causing the fusible resistor to open up as a reaction to overload.
- Simultaneously, operation without a capacitor may have subject the rest of the ballast to high ripple currents which may have caused the circuit to operate continuously as if it were pre-heating and starting the filaments – this would have led to continuous heating of the filaments and associated components (resistors which got toasted possibly).
I suppose the venting of the capacitor might be a one-off component failure, but it is important to note the lack of “headroom” in the working voltage. This isn’t a factor that’s specific to the IKEA globes, but generally, a 400v capacitor seems to be the norm due to economic choices. With the sine wave having 1.414x the RMS voltage as a peak, if supplied with Australian 240v, the capacitor would be experiencing peak voltage of 340v. If you’re living at the edge of acceptability, you could be seeing 264v RMS which translates to a peak voltage of 373v, which is quite close to the 400v rating. More headroom may translate to less possibility of seeing such a failure in the field, but given the lamp had lived a “full” life, the fact that we see different failure modes on the lamps suggests that there is some “optimization” of failure modes so as not to overengineer and overspend on one aspect where the complete system would not see a lifetime benefit.
In all, even though it was a catastrophic failure, it was a graceful one with protective mechanisms operating as designed to ensure the safety of the product. It’s always interesting to see just what the insides look like, and do an examination as to the cause of failure.