This circuit drives a resistive heater element with a low-frequency, pulse-widthmodulation (PWM) voltage source, providing heat output that's directly and linearly proportional to the duty cycle of the drive signal. The circuit's low-power and low-frequency (approximately 1-kHz) drive contributes little noise to the system, especially if the driving circuit uses a generic Darlington transistor producing relatively slow (3-µs) rise and fall times. In addition, little, if any, voltage spiking is observed.
You can implement the circuit with a protected integrated-circuit transistor, like the LM395, to provide overload protection on the heater drive line. However, the circuit is then limited by the IC's voltage and current ratings. Additionally, the LM395's behavior under some overload conditions may not be entirely satisfactory, especially if the "on" time interval becomes dissipative due to the IC's internal current limiting.
Adding a small handful of parts, though, creates a better-behaved pulseby-pulse current-limited driver (Fig. 1). The enhanced circuit detects the emitter current from the Darlington (X1) and triggers a composite pnp-npn latch should the current exceed the VBE threshold of Q1. When triggered, the latch diverts the Darlington's base drive coming through the 2.2k resistor (R1). A small capacitor (C1) stabilizes the trip-point of the latch in the presence of noise.
The circuit permits full output at any duty cycle, as long as the resistive heating element's resistance exceeds a minimum value. Below that value of load resistance, the latch allows only a minuscule spike of current on each "on" transition (Fig. 2). The circuit is quite sensitive to the critical resistance value and is, of course, self-recovering as long as the PWM input is present. (If the PWM achieves 100% "on" time, the latch won't reset until the next pulse occurs.) No significant heatsinking is required.
The circuit can be scaled to different voltage and current levels. However, it may not be appropriate for power levels much higher than 25 W, due to the increasing magnitude of the spike current that appears at short circuit. It's also not a good choice for driving an incandescent lamp, unless the cold resistance of the lamp is high enough to allow the PWM to unlatch and deliver current.