Compact Fluorescent Bulb Autopsy

I fished the failed CFL bulb from the recycling box:

Failed CFL - case damage
Failed CFL – case damage

The straight-ish crack between the tube ends looks like it happened as the (yellowed) plastic ruptured and hardened.

Not wanting to get a face full of glass fragments spiced with metallic mercury, I wrapped a blast shield around the spiral tube:

Failed CFL - tube wrap - shattered base
Failed CFL – tube wrap – shattered base

The terminal ends fit loosely in the crumbling base at the start of this operation, leaving the tube wobbling above the base. The plastic cracked as I wrapped the tube, so, for lack of anything smarter, I applied a pin punch to break away the rest of the upper base.

The tube doesn’t fit into a socket, of course, and terminates in four wire connections:

Failed CFL - tube terminals
Failed CFL – tube terminals

Those wires pass through notches on the edge of the PCB, bend around the board, pass through vias, and get soldered to pads. The solder side faces the tube, with all the components nestled into the base toward the screw terminals:

Failed CFL - PCB solder side faces upward
Failed CFL – PCB solder side faces upward

The component side sports a surprising number of parts:

Failed CFL - PCB components - 2
Failed CFL – PCB components – 2

A view from the other direction, where you can see the tube wires curling around the edge:

Failed CFL - PCB components - 1
Failed CFL – PCB components – 1

I generally harvest inductors & suchlike, but it got really really hot in there and, methinks, cooked the life out of the parts:

Failed CFL - overheated capacitor
Failed CFL – overheated capacitor

The PCB date code stamp could be “730”, suggesting either 1997 or 2007. In any event, it’s been a while.

I hope LED bulbs outlast these things, but I have my doubts …


6 thoughts on “Compact Fluorescent Bulb Autopsy

  1. II’ve been looking at a Costco (Feit or Lights of America) 23W Similar design, though no heat sinks on the NPN transistors. The base haves snap together (knocked one apart working in the barn; snapped it back and it was good). It’s a challenge to figure the schematic, but mine is similar to your example; no heat sinks on the DD128 NPNs, and the 22uF and a ceramic cap are on 1″ wires for real estate.

    The filaments are in series at the start; an MZ6 thermistor connects them until it gets warm. Interesting approach, and I guess it explains why they let the base get so hot. The small transformer (the coil-looking one in yours) is a toroid; 2, 2, 6 turns on the winding.

    So far, no failures in the LED installations. I think we’re going on 5 years for the older ones.

    1. I’m sure the heatsinks got optimized out: if the solder doesn’t get too melty, it’s all good!

    2. Out of some 10 deployed LEDs I’ve had one fail completely and one developed a barely perceptible flicker. Both were from IKEA, they replaced the first one without any fuss and I’m too lazy to even take out the flickering one :)
      That’s over some 3 years.
      They do get little hotter then I’d like but overall I like them very much. Now if only those pesky filament ones in rarely used places would finally burn out so I’d have a good reason to replace them as well :)
      Banggood had a sale on LED COB’s the other day and I ordered a few. Some of them have either an integrated driver or enough emitters to run straight off the mains with almost no circuitry which peeked my interest.

    1. Perfect description: “transistors usually [don’t] survive longer temperature overloading”. The one I have looks like a knockoff Philips.

      The LandLite (near the bottom) has a voltage doubler from 120 to 240, although all those extra parts seem too expensive. It’s most likely a get-a-foot-in-the-market design and they eventually rebuilt the main circuit to work from half the voltage (or got crushed like a bug by some other company).

      Thanks for the link!

Comments are closed.