DC-Output, Constant Wattage 12 Volt Mercury Ballast!

Warm up a mercury vapor lamp in 45 seconds!

Please read this article in its entirety before constructing this.

CAUTION - The following material gets quite technical at times, the device below makes high voltages that can kill you, and if things go wrong it is easy to make lots of smoke, custom and/or homebrew ferrite core inductors are necessary, this circuit can be abusive to mercury lamps, mercury lamps should be in suitable enclosed fixtures in case they explode. I recommend this as a project only for those with prior electronic project and electronic repair experience.

Corrections, additions, and questions to (don@donklipstein.com). DO NOT use this with metal halide or sodium lamps, since these lamps do not like DC and most metal halide lamps are less tolerant of constant-wattage warmup than mercury lamps.

Below is a schematic for a DC output, nearly constant wattage circuit to power mercury vapor lamps from 12 to 14 volts DC. The overvoltage detection/handling circuit is shown separately below the main portion of this schematic for clarity, but must be included. See the component descriptions following the schematics for values and descriptions.

See other instructions following the component descriptions. It is important to test for proper operation and proper power output.

This electronic ballast will give a mercury vapor lamp nearly constant wattage despite variations in the voltage across the lamp. This will give the lamp excessive current during warmup, when the voltage across the lamp is less. An advantage of this is accelerated warmup of the lamp. A disadvantage is overheating of the electrodes, especially the positive electrode. Expect some significant discoloration around the positive electrode, which will partially evaporate after several minutes to an hour of use fully warmed up. Some permanent discoloration will remain.

I have found this discoloration quite severe with the Philips 40/50 watt mercury lamp, but quite tolerable with 100 and 175 watt lamps. Please beware that after the positive electrode has been abused a few times with constant wattage warmup, it is likely to never work well in the future as a negative electrode.

  B+                                          B+
   ^                                           ^
   |                                           |
D1 V                                           >
   -                                           > Rsense
   |                                           >
   |    +--------------------------------------+             To Overvoltage
   |    |                B+                    )         +-->Sense Circuit
   >    >                 ^                    ) L1      |
R1a>    >R1b   330k       >                    )         |
   >    >  +--VVVV------+ >                    |   Dout  |   L2
   |    |  |            | >2.2k                +--->|----+--UUUU----+
   |    |  |      |\    +-+  __________        |         |          |
   +----|--+------|+ \    | |          |      /         === Cf      |
   |    |         |    >--+-|2         |   | |D          |          |
   |    +---------|- /    | |    555   |   | |   Q1      |          |
   >    >         |/      +-|6        3|___| |S          |        (lamp)
R1c>    >R1d       339      |__________|      \          |          |
   >    >                                      |         |          |
   |    |                                     Gnd       Gnd        Gnd
  Gnd  Gnd

  |                   |
--+-- is connected; --|-- is not connected.
  |                   |

555 Connections not shown:
Pin 1 - Gnd.
Pin 8 - B+.
Pin 4 - Connect a .1 uF capacitor from this pin to Gnd.
Pin 7 - Do not connect to anything.
Pin 5 - Connect as directed below in the overvoltage sensing circuit.

Not shown: Power supply bypass capacitors. I recomend at least .1 uF across
the power pins of each IC, as well as at least 1 uF per output watt of
tantalum plus 10 uF per watt of electrolytic capacitor(s) across the power
supply, as close to the sense resistor and mosfet source(s) as possible.

Overvoltage sensing circuit:
   |    +--------------------------->From output filter capacitor Cf
   >    |
1K >    >
   |    >470K (determines voltage limit) (560k at your risk for more voltage)
   V    >            470k (hysteresis)
D3 -    |    +-------VVVV-+
   |    |    |    |\      |
   +----|----+----|+ \    |
   |    |         |    >--+--->To Pin 5 of the 555
   |    +---------|- /
   |    |         |/
Z1 -    >          339
   ^    > 10K
   |    >
   |    |
  Gnd  Gnd

Important component descriptions:

555 - I recommend the National Semiconductor LM555. Digi-Key sells this. Other 555's often do not work as well in ultrasonic circuits where a response lag around a microsecond can change things - at best, this can affect the power regulation.

339 - This is a 339 comparator. This is a quad comparator, only two of the four comparator sections are used. Other comparators will probably work if they can source and sink _at least_ 5 milliamps and preferably 10. Please note that the 339 is an "open-collector" type, requiring the pullup resistors (2.2K) from their outputs to B+. Most other comparators do not need this. I do NOT recommend op-amps, since they will respond more slowly.

Rsense - use approx. .06 ohm for a 100 watt mercury lamp. Change this inversely with lamp wattage for other wattage lamps. Wirewound is OK - I have tried this, and the inductance is not enough to mess things up too badly. Nichrome wire is OK - but solders with some difficulty if bright and shiny and free of oxide (scrape any off), and not at all where oxide is present. Nichrome can be made solderable by first wetting with a brazing rod (this typically requires a torch - electrically heated nichrome wire gets heatsunk by brazing rods and will not melt them).

D1 - Any ordinary rectifier diode.

D3 - I recommend putting an LED here to keep the voltage at the comparator noninverting input well below B+ should the supply voltage drop really low for any reason. Otherwise, comparators may act strangely.

Dout - This must be a high speed type, and should be rated to continuously handle the current obtained from dividing the lamp wattage by 12 volts. Fast diodes don't handle the peak current of the output pulses from L1 as well as ordinary rectifiers do.

Z1 - 6.2V zener diode.

Cf - If this circuit operates at several kilohertz or higher, I recommend approx. .1 to .15 microfarad multiplied by the lamp wattage in watts. For a 100 watt lamp, use 10 to 15 uF.

L2 - Resonant frequency between L2 and Cf should be a fraction of the operating frequency of the main circuit, generally near a kilohertz, maybe as low as several hundred Hz. This indicates a few millihenries for a 100 watt lamp. This inductor should not saturate at twice the normal lamp current. I also recommend having an ohm or two of resistance in series with the lamp. The inductor should be ferrite core to get the eddy current losses down while having a reasonable size and weight and number of turns.
Alternatively, you can use a resistor with a value of 800 ohms divided by the lamp wattage in watts. This would be 8 ohms for a 100 watt lamp. This would give a gentler warmup that is not hard on the lamp's electrodes, but will waste at least 4 percent of your power when the lamp is warmed up. This resistor can get as much as 60 percent of the output power during warmup.

L1 - Must not saturate with 1.33 times the ratio of lamp wattage to supply voltage, plus allowance for losses and errors and tolerances. Ontime will be typically 2/3 of (inductance * average current / supply voltage). Average current is the ratio of wattage to supply voltage, plus a bit more for losses, and will be close enough to .09 amp for each watt of lamp wattage at 12-13 volts. That means the inductance in microhenries would be 180-200 times the desired ontime in microseconds, divided by the lamp wattage in watts. I recommend an ontime around 30-40 microseconds. This would result in an inductance in microhenries of 6000 to 8000, divided by the lamp wattage in watts, for 12-13 volts. This inductor must have a ferrite core.

A few guidelines formulas for home-winding inductors on ferrite cores:

1. Flyback transformer cores are OK.

2. Maximum peak magnetic flux should not exceed 4,000 Gauss - preferably not exceed 3,000 Gauss. Even less if the core is made of "3B7" ferrite. Magnetic flux in gauss is one tenth of 4pi times the current in amps times the number of turns divided by the effective air gap in centimeters. Or, that's 4pi * N * amps / effective gap in millimeters. Effective gap is close enough to twice the physical air gap, since in most cores the magnetic flux flows through two gaps. The separation between two core halves will be the half-gap. If you want to figure the effective gap more exactly (generally not necessary), add to the efective gap the (correction 7/15/2013) *reciprocal of* ratio of the total route length of an average line of flux to the permeability of the core material (typically 1/X-ing a couple thousand).

3. Inductance in nanohenries is 4pi times the square of the number of turns, times the cross section area of the center leg of the core (cross section of the core for toroids and C-C cores such as flyback transformer cores) in square centimeters, divided by the effective gap in centimeters.

In the 100 watt version I built and tested, L1 consists of 20 turns of 12-gauge wire on a ferrite E-E core with a center leg cross section area of approx. 1.8 square centimeters. The half-gap is approx. .8 millimeter, so the effective gap is .16 centimeter. This would make the inductance 56 microhenries (I did not actually measure this). With the 12 amp peak current, the magnetic flux would be nearly 1900 Gauss - nowhere near the limits of 3B7, 3C8, nor any common flyback transformer ferrite. The theoretical ontime with 12 volts across this inductor during the on-time and current increasing from 6 to 12 amps would be 28 microseconds. My actual ontime is longer.

Q1 - I recommend a paralled bank of IRF730 or IRF740 power MOSFETs. Use maybe 16 IRF730's or 8 IRF740's in parallel for a 100 watt mercury lamp. Since a 555 comfortably drives half of this at ultrasonic frequencies (I don't know yet about more), I recommend one 555 per 50 watts of mercury lamp. Connect all 555's in parallel, that is, connect all corresponding pins together - with the exception of Pin 3. Each Pin 3 drives a separate sub-bank of four IRF740's or eight IRF730's. Connect all drains together and all sources together. Connect all gates together within each sub-bank.

Please use some heat sinking. The IRF730 MOSFETs will be dissipating nearly half a watt each (twice this for IRF740's), which can make them get quite warm if they are close together without a heat sink. Their "on" resistance increases with temperature, which may cause "thermal runaway" if they get too warm.

R1a, R1b, R1c, R1d - These should be 10K 1 percent resistors. Do not put these anywhere where any heat source would heat them too unequally. Moderate heating is OK if all four of these are heated nearly equally.

Important Testing Information:

When you are ready to operate this circuit, use a 120 volt incandescent lamp instead of a mercury lamp. The incandescent lamp should have a rated wattage 1.5 to 3 times that of the mercury lamp. Apply power - the lamp should glow. It should glow much more brightly than it does with 12-14 volts, but dimmer than it does with 120 volts.

So far, so good? Monitor the voltage across the sense resistor (Rsense, the really low value one) with an oscilloscope. The waveform should be a triangle wave and the voltage across this resistor should have a minimum near .4 volt and a maximum near .8 volt. The average should be close to the voltage across the diode D1 that is in series with R1a.
If the voltage across Rsense is too low, add a small amount of extra resistance in series with R1a. If this voltage is too high, add a small amount of extra resistance in series with R1b.
If the average voltage is OK but the voltage does not swing enough (stays far within the range of 2/3 to 4/3 the average), replace the 330k resistor with one of a lower value. If this voltage swings far outside the range of 2/3 to 4/3 of the average, replace this resistor with a higher value one. If the waveform's shallower slopes (increasing current) get steeper as current increases, then the inductor is saturating. You may be able to fix this by replacing the 330k resistor with a higher value one to decrease the current range. Otherwise, you need to increase the gap in L1 or otherwise rebuild it.

If you feel comfortable enough with this circuit to do so, you can change the 330k resistor to adjust the oscillation frequency or to reduce switching losses or for whatever other purpose. Just be sure the inductor is not saturating and the minimum voltage across the sense resistor is not too low - if the minimum voltage hits zero without the "on" cycle being restarted, you get nothing out.

Is everything checking out OK at this point?

If so, then use a 120 volt incandescent lamp (or combination of lamps in parallel) of rated wattage equal to that of the mercury lamp that you will use. Apply power. The lamp should get rated power. You can measure the voltage across it with a DC voltmeter. Be sure the brightness is close to that of the same lamp getting 120 volts, in case anything strange is going on.
If the lamp is being overpowered, add a small resistance in series with R1b.
If the lamp is being underpowered, add a small resistance in series with R1a.
Major output power adjustments would need a different sense resistor - change this inversely with the desired change in wattage.
Verify that the voltage waveform across the sense resistor remains good after adjusting the output power.

Is everything still checking out OK at this point?

Now for a test of the overvoltage handling circuit: With power off, add a 100K resistor in parallel with the 470K (or 560K) one noted as determining the voltage limit. This is to set the voltage limit to less than the lamp normally receives, in order to force the overvoltage handling circuit to actuate. Apply power again. The incandescent lamp should be receiving about 60 volts or so, and a buzzing noise is likely.
If the overvoltage handling circuit works to limit the voltage applied to the lamp to roughly 60 volts, remove the lamp and verify that the voltage continues to stay close to 60 volts. A "ticking" sound is likely due to the ballast working intermittently at a low duty cycle. The voltage will normally be non-constant - rising quickly during a "tick" and falling slowly between "ticks". This may show up better with an analog voltmeter than with a digital one.
If things are still good, remove the 100K resistor that was just added in parallel with the 470K (or 560K) resistor that determines the voltage limit. Operate this electronic ballast with no load at all, while monitoring the voltage across the output filter capacitor. There should be a bleeder resistor around a megohm or so across the output for this. The voltage should quickly rise to 300 volts or so. If it goes much higher (anywhere near or over 400 volts), immediately shut this circuit down. Check the circuit again with an incandescent lamp to see if it got damaged by the overvoltage. Replace any damaged parts and make sure what you had working is working. Then repair the overvoltage handling circuit.
If the overvoltage handling circuit is working properly, the output voltage should rise quickly to approx. 300 volts, then the circuit should automatically shut down, then restart after the output voltage bleeds down a couple percent.
If repairs were necessary, be sure no mosfets heat up unusually during operation.

Is everything still checking out OK at this point?

If things are OK at this point, then you can use a mercury vapor lamp. It is preferred with some mercury lamps to make the shell of the base positive and the tip contact negative. If the socket has a white wire and a black one, make the white one positive and the black one negative. If you have a brass screw and a silver-colored screw, make the silver-colored one positive and the brass one negative. Polarity may be important for continued reliable operation of the starting electrode inside the arc tube of the mercury lamp. Making the electrode closer to the starting electrode positive may cause the lamp to deteriorate in a way that impairs starting.

Please note that full normal lamp life is unlikely since whichever electrode in the arc tube is positive will probably be trashed (for usage as a negative electrode) by just a few constant-wattage warmups.

Please note that wattage here will vary roughly proportionately with supply voltage. By "constant wattage", I am referring to wattage not varying much with the voltage across the lamp.

Back up to my Electronic Ballast Top Page.
Back up to my Home Page.

Written by Don Klipstein.

Please read my Copyright and authorship info.
Please read my Disclaimer.