Light Up That Burned-Out Bulb

by Norm Stanley

Beautiful flame-like discharges can be produced in burned-out incandescent light bulbs by applying a voltage high enough to ionize the low-pressure gas within the bulb, increasing its conductivity to the point where arcing occurs between the ends of the broken filament or between the filament supports.  Other spectacular effects can be produced if a high-frequency oscillatory current is used. 

Figure 1. Click image to enlarge

These effects are best seen with clear glass bulbs which, unfortunately, are now hard to find in larger sizes (60 watts and up).  They are still used to some extent for industrial lighting, although largely replaced by more efficient lamps, and for theatrical lighting.  Street lighting with series strings of incandescent lamps was once common and still may be found in some areas.  These usually contain a fuse to short out the lamp when the filament fails.  I don't know if the fuse is in the base or the envelope.  If the former it could be disabled by carefully removing the base; if the latter, no such luck.  Creative dumpster diving at likely locations should turn up some suitable bulbs.  Of course, frosted bulbs can be used, though obviously not the best choice.  Very old lamps were high vacuum rather than gas filled.  Characteristically these had a pointed tip at one end, produced when the bulb was sealed off from the vacuum line.  These are hard-to-find items, probably of interest as antiques, since I would guess production was phased out somewhere around the 1920s.  However applying high voltage to the bulb can produce a curious "Crookes tube" effect.

A neon sign or oil burner transformer is the best source of high voltage for these experiments.  These transformers are current limiting; when an arc is struck between the high voltage terminals (essentially a short circuit) the output drops from its rated voltage (6 to 15 kV) to a few hundred volts, sufficient to maintain the discharge without tripping the breaker on your AC circuit.  A warning, though: Don't be complacent about working around your transformer just because it's current limited.  Contact with the high voltage terminals can give you a bone-rattling shock, a nasty burn, or even kill you.  And remember HV can jump out and bite if you get too close to the terminal.  Obey the one-hand-in-pocket rule and stay clear of the high voltage wiring.  A variable transformer (e.g., Variac or Powerstat) in the primary circuit can be used to adjust the HV for the best effects (Figure 1).

Figure 2. Click image to enlarge

A porcelain lamp socket can be used to support the bulb.  If you can find one, an old-fashioned surface-mount socket from the days of knob-and-tube house wiring is the best.  Run wires from the transformer secondary to the socket terminals, preferably using high voltage insulated cable of the type used for connection to the cathode of a TV picture tube.  The socket should be set on a nonconductive surface.

Turn the Variac down to zero output and connect to the 120 VAC line.  Then turn up the Variac until an arc strikes between the broken filament ends or the support leads within the bulb.  Generally ionization occurs before the voltage is high enough to flash over between the contacts in the socket.  Advance the setting to obtain the best effect; current flow increases with higher settings to give a more spectacular discharge at the expense of the time it can be maintained.  Heat generated may melt down the supports or even soften the glass seals, rendering them conductive.  If the discharge takes place between the filament ends they will be heated to incandescence and glow brilliantly, obscuring the discharge between them  Figure 2  illustrates some of the effects I have seen.

Figure 3. Click image to enlarge

If a capacitor and spark gap are added to the secondary circuit (Figure 3) a high-frequency damped wave oscillatory current will be generated.  On each half-cycle of the 60 Hz AC from the  transformer the capacitor is charged up to the voltage at which the gap breaks down allowing a current pulse to flow to the bulb.  The pulse "rings down" exponentially to the point where the gap ceases to conduct, allowing the capacitor to recharge and repeat the cycle. 

The rating of the capacitor is not too critical.  Anything within the range of 0.002-0.02 =B5F at a DC working voltage (WVDC) of 15 kV or higher should do.  Oil-filled capacitors are obtainable from surplus dealers such as Fair Radio Sales.  Avoid GE Pyranol or other PCB-filled units.  Alternatively, you can build your own capacitor from window glass and aluminum foil: Obtain 21 sheets of 5" x 7" window glass.  Cut 20 pieces of foil 3" x 8", rounding the corners about 1/4".  This is to prevent sparking from sharp corners.  Dry the glass by warming in an oven.  Take a sheet of warm glass and coat the top side with shellac.  Lay a piece of foil on the glass, allowing a 1" margin on three sides and press into place with a photo print roller, thus leaving a 2" tab projecting from the 5" side.  Apply another coat of shellac and sandwich it between a second plate.  Proceed in this manner to stack up the plates with the foils projecting alternately right and left.  When done bind the stack together with electrical tape.  At each end press the  projecting foils together and punch a hole through their center, taking care not to tear them.  Use a 3/4" #8 machine screw, flat washers, and nuts to bind the foils together and attach a lead, then trim the edge of the bundle evenly.  To protect the stack it should be potted.  Place it in a suitably sized rectangular plastic pan, using blocks of dry wood or other insulator to support it off the bottom by about =BD".  Attach the leads to binding posts mounted on the long side of the pan.  Allow about three inches of space between them.  Finally pour in melted paraffin wax to cover the stack to a depth of about 1/2".  Several pounds will be needed, depending on the size of the pan.

When working with high voltage capacitors keep in mind that a capacitor may hold a charge after it's removed from the circuit.  For safe handling it should be discharged with a discharge fork or insulated screwdriver placed across the terminals.  It's a good idea to store capacitors with the terminals shorted by a wire.

Figure 4. Click image to enlarge

The effects produced when the bulb is energized vary according to the voltage and pulse rate generated by the circuit.  These depend on the time constant of the LC (inductance-capacitance) circuit and the width of the gap.  A wider gap results in a slower rate and higher peak voltage.  Construction of the spark gap is not critical and can be improvised from whatever materials are available; however if one arm is pivoted so that the gap can be varied, while energized, from contact to about one inch separation, the circuit can be more readily "tuned" for the most interesting effects.  Figure 4 is a sketch of the gap I used (a long time ago).  It was constructed from window opening hardware with wooden spools for insulating handles.  For safety I'd recommend longer handles than shown.

With the capacitor in the circuit noisy crackling blue sparks occur across the gap as the energy stored in the capacitor is suddenly dumped into the discharge on each half-cycle.  Without the capacitor the discharge is the whitish, humming, arc typical of 60 Hz AC.

Figure 5. Click image to enlarge

When using this LC circuit to illuminate a burned-out 75-watt tungsten bulb, I observed the following odd phenomena: With the gap opened too wide for a spark to pass, the tips of the broken filament glowed bluish, while a faint discharge was visible between them.  Nevertheless the discharge was sufficiently intense to cause one section of the filament to vibrate so that the glowing tip described a small circle.  In the dark this appeared as a ring of discrete points of light (Figure 5).  A faint brush discharge from the filament to the glass envelope was also present (Figure 6).

Figure 6. Click image to enlarge

I then closed the gap until sparks passed. Snapping blue sparks then occurred between the filament tips and also across the filament leads.  The entire filament simultaneously glowed dull red.  On progressively shortening the gap sparking ceased and the filament uniformly glowed more brilliantly, becoming white hot and reaching the brilliance one would expect from an intact bulb (Figure 5).  Further decreasing the gap caused a decrease in overall brilliance while the tips continued to glow white hot.  With the gap entirely closed the tips achieved maximum brilliance while the filament overall did not glow.  Figure 6 shows some of the other effects observed with different settings of the gap.  In some cases the arc was so intense as to melt down one of the filament supports, causing the arc to elongate

How do we explain these results?  My thought has been that the phenomenon is due to induction.  Passage of the high-voltage, high frequency, low  current through the bulb induces a low voltage, high current in the filament.  After all, it is a tiny coil.  At the outset, with the gap open, no oscillation occurs.  Brush discharge across the gap and within the bulb causes enough leakage current to pass to produce visible effects albeit with sufficient energy to set the filament to vibrating gently.  The 60 Hz frequency of the leakage current accounts for the stroboscopic effect seen.  At the point of voltage breakdown at the spark gap oscillation occurs in the LC circuit.  Voltage is high, resulting in flash-overs which consume most of the energy in the discharge.  With further shortening of the gap pulse rate increases and voltage decreases to the point where flash-over ceases.  More energy is thus available to heat the filament.  Decreasing the gap still further continues to lower the applied voltage and correspondingly the induced voltage across the filament.  Current in the filament is correspondingly reduced, resulting in decreased light output, with more of the available energy dissipated in the gap between the filament segments.  With the gap completely closed we're back to 60 Hz AC and inductive energy transfer to the filament becomes negligible.

Whew!  That's a pretty crude analysis, and I may be way off on what's going on here.  Perhaps some EE type can offer a more sound analysis.  I did these experiments over 60 years ago, and would now be interested in hearing from anyone who might try to reproduce them.