Power Ultra-Brights in Series
High-Power LED Series-String Power Supply
Max Carter
If your lighting or decorative application calls for more than one high-power LED (Luxeon, Cree, etc.) you might be interested in this power supply circuit. It will power up to 35 LEDs in series at currents up to 1.5 amp (see maximum current).
Why series?
Ease of active current regulation is the main reason. With LEDs connected in series, the same current flows in all LEDs in the string. One current regulator can serve them all. The series configuration also allows LEDs of different colors to be mixed within the string. With a voltage-adjustable power supply, a series string can be as flexible as a parallel configuration, allowing as few as one LED, up to the maximum number the supply can handle. Wire size is a secondary consideration. The interconnecting wire for a series string need be sized to handle only a fraction of an amp (typically), whereas the parallel configuration may require wire (and power supply) sized for tens of amps.
This DC power supply presently energizes a series string of 30 high-power LEDs on a decorative holiday star (see star). It was developed experimentally using (mostly) surplus parts. The design requirements were that it be adjustable from about 0 to 125 volts DC (up to 35 ultra-brights in series), be current regulated, be reasonably efficient and (to contain out-of-pocket costs) use parts on hand and surplus parts from my junk pile. The resulting supply is unconventional but does meet the all of the requirements. Notice that isolation is NOT one of the requirements. This power supply is not isolated from the power line.
Here's the as-constructed circuit, followed by a description.
*alternate current regulator ↓
Circuit description
Leviton dimmer
The function of the Leviton dimmer is not to dim the LED string. Its purpose is to set the operating voltage of the power supply. The dimmer uses the same principle as all dimmers (and motor speed controllers) made in the last 30-40 years: it varies the firing time - the "firing angle" - of a triac. (A triac can be thought of as two paralleled silicon control rectifiers of opposite polarity sharing a common gate.) The output of the dimmer, instead of being a steady sinewave, is a series of pulses at the line frequency but of varying width (on-time). When the dimmer is set to maximum the output pulses will have maximum on-time; the full AC sinewave will appear at the output. When set to minimum the pulses will have minimum on-time, or no on-time at all. Ordinarily the dimmer is used to power an incandescent lamp which serves as a "flywheel". The pulses are integrated into a steady brightness by the lamp's inherant inability to respond instantaneously to changes in voltage. The brightness depends on the width of the output pulses from the dimmer: wider pulses result in a brighter lamp, narrower pulses produce a dimmer lamp. (A motor speed controller works similarly. The mass of the motor's rotor performs the pulse integration, again by its inherent inability to respond instantaneously to changes in applied voltage.)
I found the Leviton dimmer in my junk pile.
Bridge rectifier
The function of the rectifier is to change the AC from the dimmer to pulsating DC to charge the 1500 uF capacitor. I liberated this one from a failed computer power supply.
1-ohm 10W resistor, 2 mH choke, NTC thermistor, 1500 uF capacitor and 100k resistor
These components form a second-order network that serves as the "flywheel" that integrates the pulses from the dimmer. The time constant of the LC network (inductor and capacitor) is a significant fraction of the period of the applied AC line voltage, resulting in a lag, or inability to respond instantaneously to changes in applied voltage. The result is an almost constant voltage on the terminals of the 1500 uF capacitor. Wider dimmer pulses produce a higher voltage, narrower pulses produce a lower voltage. The actual voltage is also somewhat dependent the load connected to the capacitor (the LED string). The voltage can be varied from near zero to about 130 volts DC.
The function of the NTC (negative temperature coefficient) thermistor is to limit component-damaging surge current when power is first applied. When cold the NTC has a resistance of around 100 ohms. When the NTC is hot (heated by the current passing through it) the resistance drops to a couple of ohms.
The 100k resistor is a bleeder. It slowly discharges the capacitor after power is removed. The slow bleed rate allows the NTC to fully cool by the time the capacitor is completely discharged.
The 1-ohm 10W resistor was found in my junk collection, as was the 2 mH choke (a high-current/low-resistance model); the 1500 uF capacitor came from a failed computer power supply; the NTC thermistor was ordered from Mouser Electronics for a couple of bucks; the 100k resistor was new from old stock.
About now you might be wondering why not just deliver the variable-width DC pulses directly to the LED string, let the LEDs pulse at the line frequency, and let the observer's eyes do the integration - again by the eye's inability to respond instantaneously. Why bother with the flywheel network? The answer is because I wanted to encorporate active current regulation. The non-linear conduction curve of an LED (even more so a string of them) does not lend itself to passive current regulation. I wanted to protect the LED string from the posibility of thermal runaway from a maladjusted dimmer (or whatever). Much easier to accomplish with steady DC.
LM317 linear voltage regulator, 3-ohm and 1-ohm 1W resistors
The LM317 acts as a current regulator. The 1-ohm and 3 ohm (R1) resistors sense the current through the LEDs. Higher current produces a higher voltage drop across the resistors. The sensing input (ADJ) of the LM317 is connected across the resistors and acts to keep the voltage drop, and thus the LED current, constant. Additionally, the 1-ohm resistor allows monitoring of the LED current. The combination of the two current sensing resistors produces a regulated current of 310 mA through the LEDs. The current can be changed by changing the value of R1 according the the formula,
R = (1.25/I) - 1,
where R is the value of R1 in ohms and I is the desired LED current in amps.
The LM317 and resistors came from new stock on hand. I found the heatsink for the LM317 in my junk.
back ↑
The power supply adjustment procedure starts below the photos.
Photos
The power supply is installed in a water-tight plastic box, also from the junk pile.
The power supply mounted.
Adjustment
Use caution when making adjustments. The circuit is not isolated from the power line. Potentially lethal voltages are present.
With load (LED string) connected and power not applied:
- Set dimmer to minimum (fully CCW).
- Set multimeter to read millivolts DC; connect multimeter probes to TP1 and TP2; power the circuit.
- Increase the dimmer setting (turn CW) slowly until no further increase in current is noted on the multimeter (indicates circuit is regulating). Note: The scale factor is 1 mV = 1 mA.
- Set multimeter to read volts DC; relocate multimeter probes to TP1 and TP3; verify voltage reading of 5-10 volts; re-adjust dimmer if necessary to meet this requirement.
- Remove multimeter probes.
- Verify all LEDs in string are lit.
