Transistor Module Switching Timing Control: The Methods That Actually Work

Switching timing is the invisible hand that decides whether your power stage runs clean or destroys itself. Get the timing wrong by even a few hundred nanoseconds, and you get shoot-through, voltage spikes, EMI nightmares, or modules that fail inside of a week. Every half-bridge, every full-bridge, every multi-level converter depends on precise timing between devices. This is not a "nice to have" — it is the foundation of reliable power electronics.

Why Timing Control Is the Hardest Part of Power Design

Most engineers focus on component selection and thermal management. Timing gets treated as a detail. It is not a detail. The moment one switch in a bridge turns off and the other turns on, there is a window — dead time — where both devices must be off. Too short, and you get shoot-through: both switches conduct simultaneously, the DC bus shorts through the bridge, and current spikes to destructive levels in microseconds. Too long, and the body diode of the opposing device conducts, which introduces reverse recovery current, extra loss, and voltage ringing that stresses both switches.

The timing also affects EMI. Fast dv/dt edges generate conducted and radiated noise that can corrupt nearby control circuits. Slow edges reduce EMI but increase switching loss. The sweet spot lies in a narrow band, and finding it requires more than a datasheet — it requires measurement and iteration.

Dead Time: The Most Critical Timing Parameter

What Dead Time Actually Does

Dead time is the intentional delay between turning off one switch and turning on the complementary switch in the same leg of a bridge. During this interval, both devices are off, and the load current freewheels through the body diodes. The duration must be long enough to guarantee that the first device has fully turned off before the second one starts turning on.

For IGBT modules, turn-off delay (td(off)) plus fall time (tf) typically totals 0.5 to 2 microseconds depending on the device and gate drive strength. Add a safety margin of 0.5 to 1 microseconds, and your minimum dead time is roughly 1 to 3 microseconds. For fast MOSFET modules, the numbers shrink — total turn-off time can be under 100 nanoseconds, so dead time can be as low as 200 to 500 nanoseconds.

The danger zone is temperature. At high junction temperature, IGBT turn-off slows down significantly — tail current extends, and the device takes longer to fully block voltage. A dead time that works at 25°C might be dangerously short at 125°C. Always size dead time for the worst-case temperature in your application.

Adaptive Dead Time vs Fixed Dead Time

Fixed dead time is simple: you pick a number and stick with it. It works fine for constant-frequency, steady-state operation. But in variable-frequency drives or resonant converters, the switching period changes, and a fixed dead time becomes either wasteful (too long at high frequency) or dangerous (too short at low frequency).

Adaptive dead time adjusts dynamically based on operating conditions. The driver monitors the actual turn-off waveform — specifically, when the collector-emitter voltage (for IGBTs) or drain-source voltage (for MOSFETs) rises above a threshold indicating the device has fully turned off. Only then does the driver enable the complementary switch. This method guarantees minimum safe dead time under all conditions, maximizing efficiency at high frequency while maintaining safety at low frequency.

The trade-off is complexity. Adaptive dead time requires a driver with desaturation detection or voltage-sensing capability. But for any serious industrial application, the extra cost of the driver is nothing compared to the cost of a destroyed module.

Turn-On and Turn-Off Sequencing in Bridge Topologies

The Half-Bridge Sequence

In a half-bridge, the high-side and low-side switches must never conduct at the same time. The basic sequence is: low-side on, high-side off. Then: low-side off, wait dead time, high-side on. Then: high-side off, wait dead time, low-side on. Repeat.

The critical moment is the transition from low-side on to high-side on. When the low-side turns off, the load current transfers to the low-side body diode. During dead time, this diode conducts. When the high-side turns on, it does not turn on into zero voltage — it turns on into the forward voltage of the body diode, roughly 0.7V to 1.5V. This is called hard commutation, and it generates a current spike equal to the bus voltage divided by the stray inductance in the loop. Minimize that inductance, or the spike exceeds the module's safe operating area.

The Full-Bridge Sequence

A full-bridge has four switches and two legs. Each leg follows the half-bridge sequence independently, but the two legs must be coordinated to produce the correct output voltage. In unipolar PWM, one leg switches at high frequency while the other switches at line frequency. In bipolar PWM, both legs switch at the same high frequency with complementary signals.

The timing challenge in bipolar PWM is cross-conduction between legs. If the timing between legs drifts, you can get a momentary short across the DC bus. Use a single PWM controller with built-in dead time generation rather than driving each leg from separate controllers. The controller guarantees that the two legs never overlap, regardless of temperature or supply voltage variation.

Soft Switching and Timing Optimization

Zero Voltage Switching (ZVS)

ZVS means the switch turns on when the voltage across it is already zero. This eliminates turn-on switching loss entirely and reduces EMI dramatically. To achieve ZVS, the timing must be precise: the resonant tank must complete its oscillation and bring the switch voltage to zero before the gate signal arrives. If the gate signal arrives too early, the switch turns on into a voltage and you lose the ZVS benefit. If it arrives too late, the body diode conducts for too long, increasing conduction loss.

The window for ZVS turn-on is typically 50 to 200 nanoseconds wide. This demands a driver with propagation delay jitter under 20 nanoseconds. Any jitter eats into your ZVS window and can cause hard switching at high frequency.

Zero Current Switching (ZCS)

ZCS means the switch turns off when the current through it is already zero. This eliminates turn-off switching loss and avoids the tail current problem in IGBTs. The timing here is controlled by the resonant inductor and capacitor values, not by the driver directly. But the driver still needs to respect the current zero-crossing point. Turning off too early means the switch interrupts current while it is still flowing, which generates a voltage spike. Turning off too late means the switch stays on after current has reversed, which creates reverse conduction loss.

In practice, ZCS timing is set by the resonant frequency and the PWM duty cycle. You calculate the expected current zero-crossing from the resonant period, then set the turn-off edge to land within ±50 nanoseconds of that point. Verify with a current probe on the oscilloscope — do not trust the calculation alone.

Timing Jitter and What Causes It

Propagation Delay Variation

Every gate driver has a propagation delay — the time between the input signal edge and the output signal edge. This delay is not constant. It varies with temperature, supply voltage, and the slew rate of the input signal. A driver might specify 100ns propagation delay at 25°C, but at 125°C it could drift to 150ns or more. That 50ns drift is enough to collapse your dead time margin and cause shoot-through.

The solution is to use a driver with matched propagation delay between channels. For a half-bridge driver, the high-side and low-side channels should have propagation delay matching within 10ns. This ensures that the dead time you set in the controller is the dead time the switches actually see, regardless of temperature.

Input Signal Slew Rate

A slow input signal edge confuses the driver. The input comparator inside the driver takes longer to switch, which adds jitter to the output. If your PWM controller outputs a 5V signal with a 100ns rise time, and the driver expects a 5ns rise time, you are introducing unpredictable delay. Use a controller with strong output drivers, or add a buffer stage between the controller and the gate driver. The goal is an input edge faster than 10ns for most modern drivers.

Measuring and Tuning Timing in the Real World

Do not rely on simulation alone. Simulators model ideal components. Your real board has parasitic inductance, uneven trace lengths, and component tolerances that shift timing by hundreds of nanoseconds.

Hook up a four-channel oscilloscope. Probe the gate signals of both switches in one leg, plus the collector or drain voltages. Trigger on the PWM input signal. Measure the actual dead time between the falling edge of one gate and the rising edge of the other. Compare it to your target. If it is off by more than 10%, adjust the dead time setting or shorten the gate traces.

Check the turn-on and turn-off edges on the Vce or Vds waveform. If you see voltage rising before the gate signal arrives, your dead time is too long. If you see both switches conducting simultaneously (voltage near zero on both at the same time), your dead time is too short. The oscilloscope does not lie — trust it over the simulation every time.

One more thing: always test timing at maximum junction temperature. Run the module until it hits thermal equilibrium, then capture the waveforms. The timing you measured at room temperature might be completely different at 125°C, and that is when failure happens.