There are many standing jokes about changing light bulbs, usually involving the Irish.  Here is a recent experience where a simple fluorescent tube replacement ended up being a protracted electronic ballast repair.  Maybe you've had a similar experience?

Another day, and another worn out appliance.  The fluorescent tube in my illuminated magnifying desk lamp, which I use every few days, has died.  The lamp is of Chinese manufacture and I can’t understand the Chinese writing on the product label other than F6008, 240 V, 50 Hz, 0.05A and 12 W.  The lamp was purchased new locally but is no longer stocked or supported by the retailer.

Figure 1.  Desk Lamp

This isn’t the first time that the tube has gone West.  The last time it happened I purchased two replacement 12 W, 220V T4 circular 6400K ring tubes (external diameter 120 mm).  So I went to my spare bulb collection and found the replacement.

Darn, the replacement tube is not 120 mm in diameter.  It’s OD is 140 mm!  My fault entirely because I should have checked it when it arrived all those years ago and sent it back for the proper size.

I bit of hunting on the Internet and I found some replacement tubes at www.replacementlightbulbs.com.  I ordered three this time, one to replace the dead tube and two spares.  Even with freight these were going to be cheaper than a replacement LED magnifying desk lamp.

The new tubes duly arrived and they are all the correct electrical and physical size.

With the lamp unplugged, and the removal of just five screws, the bulb cover came off and I replaced the tube. But with the power re-applied I still had no light.  Darn!  I was certain this failure was tube related because the ends by the filament had gone appreciably dark suggesting that the tube has gone gassy and the filaments had probably burnt out.

Figure 2.  Black Tube Ends

Time to remove the power, put the new tube somewhere safe, disassemble the base and get at the electronic ballast.

The first thing I noticed was the ‘crappy’ earth connection to the cast iron counter weight.  Where I come from earth connections do not double as attachment points (they are dedicated to earthing) and they are positive acting.  The arrangement in this lamp simply does not comply.

Figure 3.  Earth - NOT

The second thing was a large washer floating around on the inside of the base.  I have no idea where this came from and it does not fit any component in the lamp.  There is no way it fell inside so it must have been incorporated in the design during manufacture.

Time for Component Level Servicing

The first thing to do is to discharge the two electrolytic capacitors and the large Mylar capacitor on the board with a clip lead and a resistor (about 1K Ohm is good).  There are no bleed resistors on the electrolytic capacitors, and although they are not huge, they are charged to around 200 V DC.  The Mylar capacitor charge status is a little less certain, but better safe than sorry.

Now to check ‘the obvious’.  I measured the resistance of phase (through the switch) and neutral from the plug through to the respective board terminations.  All good.  Next check the wiring continuity from the board to the lamp head.  All good too.

On to the board.  There is nothing obviously burnt, or smelling of being burnt, and no obvious dry solder joints.  There are the usual blank spaces where components have been omitted during assembly and one area where the board had been manually re-worked.  Missing components and the rework are likely due to this board being manufactured for either 110 V or 240 V operation.  I have the 240 V version.

Figure 4.  Board

There is a component fuse which is serviceable (measured as a dead short) followed by a 100 Ohm 2 W resistor.  The resistor doesn't look burnt and checks out okay at about 100 Ohms.  I would have expected the fuse to have failed or signs of the resistor having been hot if something had failed resulting in excessive current draw.

The basic circuit is a half wave ballast using the following general circuit configuration (reproduced without permission from On Semiconductor Application Note AN1543D, January 2009, Figure 11).  Please refer to the application note for details of the circuit’s operation.  It is quite clever - particularly how filament heating occurs.  Note that I have not actually drawn out the specific circuit from the board during this repair.  The general configuration is enough to understand what the actual components are doing, what might reasonably have failed, and typical test voltages and waveforms

Figure 5.  Typical Half Wave Circuit Diagram

I quickly checked all of the diodes and inductors in-circuit with a multimeter.  While these are not principle suspects for a fault they are easy to check in-circuit.  They are all okay.

Next came the electrolytic capacitors.  There are only two of these.  Electrolytic capacitors have a tendency to go leaky or loose capacitance over time.  Their values check out fine.  Next I applied and removed the power, and measured their residual charge at about 190 V. This circuit is not mains isolated so take some care about what you touch when it is energized, and remember to discharge stuff after the circuit is turned off.  These capacitors are okay.

While I suspect that the problem could be the driving transistors I set about measuring the non-electrolytic capacitors, again with a simple multimeter.  Their rated voltages (printed on the components) are all adequate.  The test required lifting of at least one lead from the board to remove the influence of any other stuff in parallel  – nothing is untoward as the component values within coo’ee of their package ratings.   Ah!  I have found at least one problem.  One of the capacitor terminations appears to have a dry joint.  A quick clean and resolder and this joint is now electrically and physically secure.

With the capacitors soldered back in place I reconnected the tube and applied power again in case the dry joint was the problem.  Still no joy (and no light).  If the capacitor joint was dry then this was not the only problem.  Time to discharge stuff again and check out the transistors.

The two transistors are identical NPN types rated at more than 400 Vce and 4 Amps Ic at up to 70W.  There is no way they will handle 70 W without heat sinks and they have none fitted.  I figure their actual dissipation is perhaps a Watt or so in normal operation.  The base of this lamp never got warm in operation.  Well designed transistor circuits are very reliable so if one of these is dead I will need to determine the actual cause - probably too much power dissipation - before turning the lamp back on.   I removed the transistors from the board because in-circuit measurements would be meaningless due to the low impedance transformer drives.

The base-collector and base-emitter junctions are intact and there is no leakage or reverse leakage from the collector to the emitter.  While the transistors are out of circuit I quickly measured their transistor action and gain with a couple of resistors and a 12 Volt DC supply.  The transistors are fine.  That's good because I don’t have any equivalent devices in stock.

I am running out of stuff to test.  With the transistors still out-of-circuit I checked the windings on the feedback transformer and the associated base resistors with an Ohm meter.  They are all good.

There is only the Diac, its charging resistor and two small ceramic capacitors left untested.  If either the Diac or its charging resistor have failed then the transistors will never start oscillating.  The Diac probably requires more than 30 V DC to check so I started with the 470K Ohm charging resistor.  This component is absolutely open circuit.  It doesn't look burnt or cracked but it is buggered.  I replaced it and put the transistors back on the board.

Figure 6.  Dead Resistor

As a final check I tested the printed circuit board track continuity and for adjacent track leakage.

With the lamp still unassembled I re-fitted the old tube which immediately struck and my lamp is back in business.

Before reassembly I tidied up a couple of the existing solder joints, removing what I considered to be excessive solder.  I cleaned the circuit board with some isopropyl alcohol, put a proper earth termination on the base plate, cleaned the lamp housing and reassembled everything.

On reassembly I noted how close the board mounting screws are to the  phase and neutral terminations to the printed circuit board.  While these screws are embedded in plastic bosses and therefore unlikely to short to anything, they are right on the lower track spacing limit for the applied voltage with an uncoated board.

Figure 7.  Phase Track Close to Mounting Screw

In completing this repair I have not had to resort to any live circuit measurements.  In retrospect, if I had set about the repair with an isolating transformer and oscilloscope then I would probably have identified the problem a little earlier, and without having to de-solder any components, but at the expense of having to consider what to measure and what to expect.  This would have been the very next step, had it been necessary.

I have also avoided using the ‘modern repair technique’ of changing out components   until the circuit worked.  Not only is this expensive and time-consuming but you can actually end up making the problem worse.  If you replace a damaged component with a new one, but the damaged part was a consequential failure, then you will end up damaging the new component and no closer to fixing the actual problem.

Figure 8.  Lamp Back in Service

So my desk lamp will live another day.  And I now have three appropriately sized spare tubes in reserve as well as an excellent understanding of the circuit configuration.  Job done.