Flyback Diode and Relay Coil Turn-Off: Where the Current Goes

The transistor that drives a relay coil fails at the moment the coil is switched off, not while it is running. With the coil energised the current is steady and the device is fully enhanced, so it is dissipating almost nothing. The stress arrives in the microseconds after the drive is removed, when the inductance of the coil insists that the current continue and the device is no longer holding the path open.

That is the whole reason turn-off protection exists, and it is also why inductive load switching punishes a more highly rated transistor as readily as a smaller one. The energy has to go somewhere. The design question is where, and how quickly.

What the Coil Does at Turn-Off

Current through a coil builds a magnetic field, and that field stores energy. When the driver turns off, the inductance opposes the change in current, and to do so it generates a voltage in the direction that would keep the current flowing. If the original path is interrupted and no alternative exists, that voltage rises until something gives way: the drain to source rating of the transistor, a nearby input, or the supply itself.

The faster the interruption, the higher the voltage needs to be. A slow driver that spends time in the linear region lets the voltage rise more gradually, which is why a device that survives one board layout and fails on another is usually telling a story about stray inductance rather than about gain.

What the Flyback Diode Actually Does

A diode placed across the coil, oriented so that it is reverse biased while the coil is powered, becomes forward biased the moment the polarity of the coil reverses. The coil current then circulates through the coil and the diode rather than through the switch, and it decays gradually in that local loop.

The energy is not removed, it is redirected. That distinction explains two practical consequences. First, the diode has to withstand a pulse of current comparable to the coil current, repeated at every actuation, so its pulse rating and thermal behaviour matter more than the package size. Second, the diode does not need to be large in voltage; what it needs is a low forward drop so the loop current decays in a controlled way through a component that can carry it.

Why a Gentle Clamp Releases Slowly

A plain silicon diode clamps the coil voltage to roughly a volt, which is kind to the switch and slow for the relay. The current decays with a time constant set by the coil inductance divided by the loop resistance, and with a low clamp voltage the decay takes longer. A relay that must drop out quickly can stay held long enough to miss the intended timing, and in a safety path that delay is a functional problem rather than a nuisance.

The usual alternatives trade clamp voltage for speed: a diode in series with a resistor, a Zener clamp, or a transient suppressor. Each of these allows the coil voltage to rise further, which pulls the current down faster, and each of them asks more of the switching device. Any one of them is defensible if the whole loop is checked together: the voltage the coil is allowed to see, the rating of the switch, the release time required, the emissions the faster edge produces, and the energy the clamp component can absorb. Removing the protection to gain speed is not one of the options.

<img src="https://www.gopcba.com/wp-content/uploads/2026/08/PCB-Fabrication-Process.jpg.webp" alt="turn-off voltage spike at the relay driver” />

Layout Decides Whether the Clamp Works

A protection scheme that is correct on the schematic can still fail on the board. The freewheeling path should be short and should enclose the coil and the clamp component, because the parasitic inductance of a long route through several vias is exactly what generates the extra spike at the switch. Keeping that loop area small is worth more than adding a larger diode. If the diode sits at the far end of the connector harness while the coil sits at the near end, the parasitic inductance of the intervening path is still in series with the switching node.

The coil loop also needs to be kept away from sensitive networks, particularly reset lines, oscillator circuits and analogue references. A ground symbol that is shared on the schematic does not mean the two points are at the same potential during a turn-off transient, and a reset pin that references a lifted local ground will behave as if it were deliberately triggered.

A Four Point Check for a Relay Driver

Confirm the coil voltage and current against the driver rating using the actual relay and transistor datasheets rather than a family specification. Draw the complete current loop for both states, energised and released, and check the orientation of every diode in that loop, because a layout review that stops at the schematic symbol will miss the stray inductance that sets the peak, because a reversed flyback diode conducts continuously and turns the driver into a heater.

Then measure, using the methods a test plan would specify for a switching node. Capture the switch node voltage, the coil current and the disturbance on the supply at the same time, with a probe arrangement whose ground lead is as short as the measurement allows. Finally, treat the release time, the temperature rise, the emissions and the effect of repeated actuation as the acceptance criteria, rather than the amplitude of a single spike. A device that survives a hundred operations is not evidence for a specification that requires a million, and the difference only appears when the test is repeated and recorded. Components that behave differently between batches, in this case the relay and the diode, are the usual source of that spread.

Three Beliefs Worth Retiring

The first is that interrupting the switch also interrupts the current, which is true only for an ideal switch and an ideal wire. The second is that the presence of a flyback diode is sufficient, when orientation, pulse capability and placement each have to be right. The third is that a larger diode is a better one, which ignores reverse recovery, the added capacitance and the effect on release speed.

FAQ

Does the coil need a flyback diode if the driver has an avalanche rating? It can survive without one, but the energy is then dissipated in the device on every actuation, which is a thermal decision rather than a free option.

Can a resistor in series with the diode be used? Yes, and it is a common way to raise the clamp voltage and shorten the release time, provided the switch rating accommodates the higher peak.

Why does the fault appear only on some units? Because it depends on parasitic inductance and on the coil batch, both of which vary between assemblies.

Should the diode be placed at the connector? Place it where it encloses the smallest loop with the coil, and keep the path back to the driver as short as the layout permits.

Summary

Protecting a relay driver is a question about where the stored energy goes and how fast. Give the coil a controlled freewheeling path close to the coil, choose the clamp for the release time the function requires, check the switch against the voltage that clamp allows, and keep the loop small enough that its own inductance does not undo the design. Then measure the switch node and the coil together, repeat the test, and record the result. The failure mode is not mysterious once the current path is drawn. The same discipline applies to any board that switches a coil, as described in our industrial control boards overview.