2010-01-04
Rope Light Controller Failure
The commercial Rope Light used in the Christmas Star project stopped working suddenly. The usual checks failed to revive it or indicate a reason for the failure so the sealed controller box was opened. Opening it was quite challenging, eventually I simply smashed the heck out of the box glue joints with a big hammer. Unfortunately this rather brute-force approach applied forces and accelerations to the internal components that were beyond their limits. The phenolic PCB was smashed and the heavier components torn off the board.
Even before completely gaining access it was immediately obvious by the smell that something rather catastrophic had occurred. A small amount of black fluid leaked out as well, indicating the likely cause for the failure was rain water penetrating the "outdoor" control unit. Upon inspection of the PCB there was extensive charring around the mains ingress point and foul-smelling, greasy pyrolysis products coated everything inside the box. Cleaning with water didn't shift the blackening, but 2-propanol made short work of it. The full horror of the failure was then evident; an arcing fault had occurred directly across the mains carbonising the phenolic board and spraying copper everywhere. The fault occurred before the internal fuse and did not open it, rather it appears to have burnt itself open, as the 30 Amp circuit breaker to the mains circuit in question did not open. It was probably only its containment in the sealed unit that prevented a much more violent event and resulting fire.
Despite my rather kinetic opening technique it was relatively easy to trace out the circuit and study it. The unit is a phase-angle controller, using SCRs to switch the bridge-rectified mains into the two circuits of the rope light. The controller logic is a small 10-pin SIP daughterboard with a die packaged directly to it covered in the familiar black encapsulation. The controller runs off a 5 volt rail derived directly from the rectified mains via a high-wattage dropper resistor, a 5v1 zener diode and a 220uF electrolytic capacitor. There is also a loading resistor across the rail, which I assume is to stabilise it? The SCRs are MCR100-6s, 800 mA, 400 V units. Their gates are driven directly from pins on the controller card, suggesting it has internal current limiting. There is a fairly substantial EMC filter on the mains side, with a common-mode choke and filter capacitors across the mains. There is also a MOV suppressor. The active feeds the controller a phase reference through a 200 K resistor and a 100 pF capacitor to LPF it.
The controller daughter card had some minor damage to its track-work thanks to my aggressive access method, but being a metric pitch required a little effort to test. I wired it up to a "standard" 0.1-inch header with the wiring pencil and plugged it into a solderless breadboard.
Once supplied with 5 volts, its timing resistor, and a zero-crossing reference I quickly determined it was a 4-channel unit (with the rope light only utilising channels 0 and 2) and was undamaged by the fault. The clock is about 115 kHz and varies quite dramatically with the timing resistor. It appears to be matched to the mains frequency, the zero-crossings of the reference synchronising its internal state so the control signals are phase-synchronous to the mains waveform. With smaller timing resistors it will happily generate several gating cycles per mains half-cycle, but it realigns upon the next zero-crossing. Larger timing resistor values compress the available phase-control range. The standard value 220 K is a little too large for 50 Hz, the supplied value in the original circuit (200 K) is also slightly too large, but wastes less of the waveform tail. The 100 Hz frequency of the full-wave rectified mains makes the fading quite smooth and flicker free, probably even with LEDs. It would not surprise me if this particular device is used with the typical 24 VAC LED lights as well. The gate drive signals are high for the entire conduction time, so they could drive non-thyristor switches too. There are a series of transients just after the leading edge of the drive lines going high, their purpose, if any is unknown? They exist into resistive loads but appear inversely proportional to current delivered (from a slightly higher impedance than the main logic level itself).
The drive outputs source enough current to directly drive LEDs, so I plugged some in. Not sure how much the outputs can sink, I didn't want to push my luck with what was clearly a working unit.
Now the choice becomes, do I rebuild the controller, returning the rope light star back to service? Or do I simply hard-wire the rope light straight to the mains (through a fuse) and have a static star? Another alternative is to chuck out the rope light and just buy a new one in December... Surely the time saved would be far more than the $22 price tag of the rope light - but that is never why I build things for myself anyway. I think I've had enough of blinken light projects for a little while, time to build some RF stuff again! Might get around to rebuilding it eventually though.
6 comments.
Parent article: Rope Light Christmas Star.