Load Switch and eFuse Circuit Design

Almost every board has at least one rail that has to be turned on and off under software control, or protected against a fault on a downstream module. A load switch does the first, an eFuse does the second, and the two are often confused because they share a topology. The difference lies in what happens when something goes wrong.

Where a Load Switch Is Used

A load switch is a controlled MOSFET in series with a rail, with an enable input and a gate drive circuit that ramps the device on. It replaces a discrete transistor and a handful of resistors with a single device that has a defined on resistance, a defined turn on behaviour and, usually, protection against the common mistakes of discrete design.

Typical uses include powering down an unused peripheral to save current, sequencing a rail so that it appears after another, and isolating a module that can be removed while the system runs. In each case the switch is expected to pass the load current with little loss and to be transparent to the circuit when it is on.

The switch is not a protection device in the ordinary sense. It may have a current limit, but its purpose is to control the rail rather than to survive a short circuit indefinitely. Using a load switch where a real fault has to be tolerated is a common cause of a device that fails after a few field events.

Load Switch versus eFuse

An eFuse is a load switch with a defined fault response. It measures the current through a sense element, limits it to a programmed value, and takes action when the limit persists: it may latch off, it may retry after a delay, or it may regulate the current for a defined period before shutting down.

The extra function brings extra requirements. An eFuse needs a sense resistor or an internal sense element with a defined tolerance, a way to set the current limit, and a fault output that the controller can read. The device also has to survive the energy it dissipates while limiting, which is the parameter that most often decides whether the design is adequate.

Choosing between them follows from the consequence of a fault. A rail feeding a connector that a user can short needs an eFuse. A rail feeding a soldered down device that cannot be misused is well served by a load switch, provided the upstream supply has its own protection.

On Resistance and Voltage Drop

The on resistance of the switch appears directly in series with the load, and its voltage drop reduces the rail that the load sees. A hundred milliohms at one ampere costs a hundred millivolts, which is significant on a one volt rail and negligible on a twelve volt one. The resistance also rises with temperature, by roughly forty percent over a hundred degrees for a typical device.

The power dissipated is the product of the resistance and the square of the current, so the loss grows quickly with current. At three amperes a hundred milliohms dissipates nine tenths of a watt, which is more than a small package can remove without copper area to help it. The resistance and the thermal design have to be chosen together.

The tolerance of the on resistance matters where the rail has a specification. A device specified at a maximum resistance over temperature allows the drop to be calculated at the worst case, while a design based on the typical value will produce a rail that falls out of specification in a warm enclosure.

Inrush Current and Slew Rate

Every load has capacitance, and charging it through a switch that turns on in nanoseconds produces a current limited only by the capacitor and the resistance of the path. inrush current therefore has to be controlled, either by slowing the gate drive so that the switch passes through its linear region gradually, or by limiting the current actively.

A slow turn on reduces the inrush but increases the energy dissipated in the switch, because the device spends longer with voltage across it and current through it. The energy is close to the energy stored in the load capacitance, and it has to fall inside the safe operating area of the device. A slew rate controlled switch translates that requirement into a defined number that can be checked.

Where the load capacitance is large, a switch with an adjustable slew rate is worth the additional pin. Fixed slew rate devices are convenient but force the capacitance to fit the device rather than the other way round, and a load that is larger than expected produces a current pulse that trips the limit during start up.

Current Limit, Fault Reporting and Retry

current limit is programmed with a resistor or with a register, and it has to sit above the maximum load current with margin for tolerance and temperature while staying below the rating of the connector and of the traces. The margin is set by the accuracy of the limit, which includes the tolerance of the sense element and of the internal reference.

When the limit is reached, the device has to decide what to do. A constant current response holds the load at the limit and dissipates the difference as heat, which is acceptable for a brief overload and destructive if it continues. A latch off removes the load until it is reset by the controller, which is the safest response for a fault that should not repeat. A retry after a delay suits a load that can draw a legitimate surge, such as a motor at start up.

The fault output is what makes the behaviour visible to the system. An open drain fault pin connected to the controller allows the firmware to log the event and to decide whether to retry, which is far better than a device that silently shuts down. The pin should be pulled up to a rail that is present throughout the fault condition.

gate drive for the internal switch is handled by the device, but the external components around the enable pin still matter. A slow rising enable signal can leave the device in an ambiguous state, and a long trace on that pin picks up noise that turns the rail on and off. A small filter at the enable pin is cheap insurance.

Layout and Thermal Design

The switch carries the whole load current, so the traces on both sides should be sized for it and the sense element should be placed in the current path with a four terminal connection where accuracy matters. The drop across the traces adds to the drop across the switch, and together they define the rail that the load actually sees.

The thermal path is part of the electrical design. Copper under the device and vias into the plane give the heat somewhere to go, and the connection that carries the current is usually the same one that carries the heat. The reservoir capacitor at the input of the load and the switch itself should be close together, since the general placement rules described in bulk capacitor placement apply to the input side as well.

The return path should be direct. A load current that returns through the sensitive ground of the module it feeds produces a voltage offset that varies with the load, and the resulting behaviour is a measurement that changes when the rail is switched. Keeping the return separate and joining it at one point is the same discipline used for any rail that supports both power and signal, as described in power integrity.

Verification and Common Faults

Verify the turn on behaviour with a current probe on the rail and a scope on the load voltage. The trace should show a controlled ramp with the current staying inside the limit, followed by a transition to fully enhanced operation. A step in the current waveform indicates that the load capacitance is larger than the slew rate allows, or that the loop is not behaving as designed.

Test the fault response deliberately by shorting the output. The device should limit the current, report the fault and behave according to its configuration, and it should survive the event at the maximum input voltage and the maximum ambient temperature. A test at nominal conditions proves very little about the energy the device has to absorb.

A rail that sags when the switch turns on, and recovers afterwards, has an inrush problem rather than a switch problem. Increasing the slew time or adding capacitance at the input of the load spreads the charge over a longer period, and the general layout rules for a switching stage, described in switching regulator layout, apply to the loop formed by the input capacitor and the switch.

FAQ

Can a load switch replace a fuse? Only if it has a defined current limit and a defined response to a persistent fault. A plain load switch will be destroyed by a short circuit.

Why does my rail sag when the load switch turns on? The inrush current is charging the load capacitance faster than the supply can deliver. Slow the turn on or add input capacitance.

How do I choose the current limit? Above the maximum load current with margin for tolerance and temperature, and below the rating of the connector and the traces. The margin follows from the accuracy of the limit.