Adapt DC Voltage To Perform PWM DImming Of High-Brightness LEDs
High-brightness LEDs (HBLEDs) are making inroads into more traditional lighting applications that include a dc distribution system (for example, 24-V MR-16 track lights). HBLEDs are more efficient, and they have a potentially longer lifespan than do halogen or xenon lamps.
Because hysteretic controllers are inexpensive, simplify lighting designs, and require no compensation networks, they’re well-suited for driving HBLEDs. Hysteretic controllers usually have a pulsewidth- modulator (PWM) input that enables a pulse train of varying duty cycle to provide the dimming function. One problem, however, in converting a traditional lighting system is that many dimmers provide a 1- to 10-V dc signal rather than a PWM signal. Also, to increase the HBLEDs’ operating lifetimes, a controller should provide temperaturebased current foldback.
Converting a dc voltage to a PWM signal is easy. The PWM signal appears at the output of a comparator when you apply the dc voltage at one input and a triangle wave at the other. Headaches can arise, however, when trying to align the triangle wave with the control voltage. You need a linear relationship between duty cycle and control voltage, with a 0% duty cycle at the minimum control voltage and a 100% duty cycle at the maximum.
The circuit in Figure 1 includes the hysteretic controller, U1 (MAX16820); related power components; and a control circuit based on a quad op amp, U2 (LMX324). U1 drives five HBLEDs from a 24-V source, using only inductor L1, MOSFET Q1, and catch diode D1. A sense resistor (R1) sets the current to 0.5 A. U1 turns Q1 on whenever the current-sense voltage drops below 190 mV and turns Q1 off when that voltage exceeds 210 mV.
Hysteretic controllers have no clock and require no external compensation. U1 also provides a regulated 5 V to power the PWM conversion circuitry. Figure 2 illustrates the currentsense waveform corresponding to a small “ON” time in the PWM signal.
The difficulty in converting a control voltage to a PWM signal involves setting the triangle wave’s peak and valley voltages to closely match the corresponding maximum and minimum values of the control voltage (VCNTL). Two of U2’s op amps generate the triangle wave, which oscillates between an upper voltage level set by the R7-R8 divider and a lower voltage level set by the divider formed by R7 and R8 in parallel with R9. U2’s output is a 50% duty cycle, rail-to-rail square wave. Setting U2b+ equal to VCC/2 causes U2b’s output to integrate the square wave, producing a symmetrical and linear triangle wave. R10 and C4 set the operating frequency.
Achieving 0 V at the valley of the triangle wave is difficult, because U2b’s output has a worst-case minimum of 60 mV. We therefore chose a valley of 250 mV and a peak of 2 V. Because VCNTL ranges from 0 to 10 V, R12-R13 divides VCNTL by 5. This limits the reduced control voltage, VRED, to 2.0 V and thereby matches the triangle wave’s peak voltage.
U2d creates the PWM pulse train by comparing the triangle wave to VRED. The triangle-wave valley is 250 mV, so the PWM signal remains at 0% until VCNTL reaches 1.25 V. This action causes a small offset error that’s most pronounced at low values of VCNTL, but it also confers an advantage by guaranteeing an OFF position. Figure 3 shows how the triangle wave converts the divided control voltage into a pulse-width-modulated waveform.
Op-amp U2c provides the temperature-based current foldback. The R4-R5-R6 divider delivers 1.5 V to U2c’s noninverting input, which is almost a diode drop below the triangle wave’s peak (2 V). Thermistor R2 (a resistor with negative temperature coefficient) is nominally 100 k? at 25°C, but its value declines to 33 k? at 50°C. At that temperature the R2-R3 divider produces 1.5 V—a balance point at which U2c’s positive, negative, and output terminals are all at 1.5 V, and just about to pull VRED lower, via D2.
At 70°C, R2 drops to 15.5 k? and the op-amp output drops to 1.0 V, pulling VRED to about 1.6 V. This action achieves the desired current foldback by limiting the maximum duty cycle at 70°C to 80%. A simple change of resistor values allows the circuit to accept different VCNTL ranges and to have different temperature- foldback characteristics.