This article documents a second failure of a 24W LED oyster. The luminaire was purchased complete on eBay for about $45.
After about two years use, the light became sensitive to switching transients on the mains, visibly blinking when other appliances were turn off or on. After some time, this progressed to oscillating on and off for a few seconds on a cold startup, but on hot startup it was stable.
These are exactly the same symptoms as the first failure… no surprises, it is the same driver board that had a faulty capacitor replaced… and worked fine for a couple of years.
Above is the failed driver board with the replaced 105° 6.8µF input filter capacitor (at the right). The capacitor should have an ESR around 5Ω, but now cannot be measured by my ESR meters (both upper limit ~100Ω). The other capacitor also has excessive ESR.
It was replaced by a new module of a different type, a little larger physically.
Above is a thermograph of the LED plate and driver immediately after the power was switched off (it had been running with the diffuser installed).
Ceiling temperature above is 22.5°, driver is 57.8°, a temperature rise of 35°. We might expect the ceiling to reach over 40° in summer, so operating temperature greater than 75°.
It has been working for some days now, it is only a matter of time before it fails again… weeks or years.
Input filter analysis
A different, but similarly rated module was measured, specifically the voltage across the 15µF input filter capacitor.
We can calculate the approximate RMS ripple current by considering that this wave is approximated well as a triangular wave of amplitude 45V, rise time of 2ms and decay time of 8ms. Starting with \(i=C\frac{de}{dt}\), we can calculate ripple current to be 169mA, and if ESR=5Ω, dissipation would be 142mW.. That is sustainable, but as the capacitor ages, ESR increases which increases dissipation and operating temperature which accelerates aging.
For example, a Panasonic EE-A series high endurance (10,000h) 105° 15µF 400V aluminium electrolytic capacitor is rated for 200mA ripple current @ 100Hz and cost around $1000/1000. Of course these are not made with those capacitors, and it is doubtful that the cheap Chinese capacitors used have comparable quality.
Most of the modules that I have inspected have a smaller input filter capacitor, 6.8µF is common which results in approximately double the ripple voltage, ripple current is much the same (increased ripple voltage is offset by smaller C), but a slightly smaller package to dissipate heat.
This gives insight as to why these only last a couple of thousand hours at best.
Conclusions
I have studied failure of LED lighting over 10 years or so, and just like the Compact Fluorescent Lights before them, they just do not live up to claims of life, efficiency, cost effectiveness. Post mortem analysis informs that conclusion.