Quick answer
These three parts all sit in the power path, but they answer different questions. A load switch turns a rail on and off with controlled inrush and almost no quiescent current — it does not limit current. An eFuse adds current limiting, overvoltage protection and fault reporting, at a higher cost and complexity. An ideal diode blocks reverse current and ORs two supplies with a low drop. Choose by the fault you must handle, not by the block diagram: on/off and inrush → load switch; overcurrent and overvoltage → eFuse; reverse or two-source ORing → ideal diode.
Executive takeaways
- A load switch is a FET with a driver. At 2 A and 60 mΩ it dissipates about 0.24 W; at 85 °C the 100 mΩ maximum makes it 0.4 W. That heat, not the current rating, is often the limit.
- Inrush is the load switch's real job. A 100 µF capacitor on a 5 V rail needs at least 167 µs of controlled rise time to keep inrush below 3 A.
- An eFuse replaces a fuse plus a supervisor. The TPS25947 gives 0.5–6 A adjustable limit, 5 µs overvoltage response and reverse-current blocking in one part.
- An ideal diode beats a Schottky on drop. At 2 A, a 0.4 V Schottky burns 0.8 W; a 79 mΩ ideal diode burns about 0.32 W and blocks reverse current.
1. Three jobs in the same place
All three parts connect between a source and a load, and all three can be drawn as a box in the power path. The difference is what happens when something goes wrong. A load switch only does what its name says: it switches. An eFuse watches current and voltage and acts on them. An ideal diode watches direction and blocks the wrong way.
| Function | Load switch | eFuse | Ideal diode |
|---|---|---|---|
| On/off control | Yes | Yes | Limited (enable) |
| Inrush control | Yes (slew rate) | Yes (timer) | No |
| Current limit | No | Yes, adjustable | No |
| Overvoltage protection | No | Yes (OVLO/OVC) | No |
| Reverse blocking | Body diode only | Yes (back-to-back) | Yes |
| Fault flag | No | Yes | No |
| Typical RON | 20–100 mΩ | 20–80 mΩ | 20–100 mΩ |
2. Load switch: the on/off and inrush part
A load switch is a P- or N-channel FET with a gate driver, an enable pin and often a controlled rise time. Its virtues are low quiescent current and a clean, repeatable turn-on. The TPS22916, for example, is a 1–5.5 V, 2 A part with 60 mΩ typical ON-resistance at 5 V, 0.5 µA ON-state quiescent current and 10 nA shutdown current; the C version slows the turn-on specifically to limit inrush.
The design numbers are straightforward. Conduction loss is I²R; at 2 A and 60 mΩ that is 0.24 W. Inrush is set by the load capacitance and the rise time: IINRUSH ≈ C × dV/dt. A 100 µF load on a 5 V rail with a 1 ms rise time draws about 0.5 A; with a 100 µs rise time it draws 5 A. Choose the part's slew rate for the capacitance, not the other way round.
What a load switch cannot do is protect against a short on its output. If the load fails short, the FET sees the full source current limited only by the source impedance; it will either survive briefly or fail. That is the case for an eFuse or a fuse downstream.
3. eFuse: current limit plus supervision
An eFuse is a load switch with a sense element, a comparator and a timer. It limits current to an adjustable value, blanks short peaks so inrush does not trip it, and can disconnect on overvoltage. The TPS25947 adds true reverse-current blocking with back-to-back FETs, a 0.5–6 A adjustable limit with ±10% accuracy above 1 A, an overvoltage clamp with 5 µs response, and an overvoltage lockout with 1.2 µs response.
The price is complexity: a current-limit threshold, a timer, an overvoltage option and often a fault flag to route to the controller. The benefit is that a shorted load is handled by the part itself, not by the source or the harness. For a 5 V, 2 A port, a 2.5 A limit with a timer that allows 5 A for a few milliseconds covers both inrush and normal load.
4. Ideal diode: direction, not magnitude
An ideal diode is a FET with a controller that emulates a diode: it conducts in one direction with a small drop and blocks in the other. It is the part you use when two supplies must be ORed, or when a connector can be back-fed. The LM66100, for example, is a 1.5–5.5 V, 1.5 A ideal diode with 79 mΩ typical ON-resistance at 5 V and reverse standoff to −6 V, plus input polarity protection.
The comparison that matters is with a Schottky. At 2 A, a 0.4 V Schottky dissipates 0.8 W and needs a heatsink or a larger package; a 79 mΩ ideal diode dissipates about 0.32 W and blocks reverse current when the source collapses. The ideal diode also avoids the Schottky's reverse leakage at high temperature, which matters in battery-backed designs.
5. Worked example: a 5 V, 2 A port with 100 µF
A 5 V port must power a 2 A load with 100 µF of input capacitance, keep inrush below 3 A, and survive a shorted load without damaging the connector.
- Inrush: with 100 µF and 5 V, staying below 3 A requires a rise time above 167 µs. A load switch with a slow turn-on (or an eFuse timer) meets this; a fast load switch does not.
- Short circuit: a load switch alone leaves the FET to survive the source current. An eFuse with a 2.5 A limit handles the short and reports a fault.
- Conduction loss: at 2 A, 60 mΩ is 0.24 W; at 85 °C, 100 mΩ is 0.4 W. Both are acceptable in a small package with copper, but they are not zero.
- Reverse: if the port can be back-fed, add an ideal diode or choose an eFuse with reverse blocking.
The honest answer for this port is an eFuse, not a load switch: the short-circuit case is the deciding requirement. A load switch is the right part only when the load is known-good and the source is already protected.
6. Inrush arithmetic
The inrush number is easy to calculate and easy to get wrong. For a load capacitance C charged to V in a time t, the average current is C × V / t. With a controlled rise time the peak is close to that average; without it, the peak is set by the source impedance and the FET RON, which can be tens of amps.
| Rise time (100 µF, 5 V) | Average inrush | Suits |
|---|---|---|
| 100 µs | 5.0 A | Strong source, eFuse with 5 A window |
| 200 µs | 2.5 A | eFuse with 2.5 A limit |
| 500 µs | 1.0 A | Small adapter, load switch with slew control |
| 1 ms | 0.5 A | Weak source, battery-backed rail |
Pick the rise time from the source and the protection, then choose a part that can produce it. A load switch with a fixed slow turn-on may already match; an eFuse with a timer gives more freedom but needs the timer set correctly.
7. Thermal budget
The pass FET loss is I²RON, and RON rises with temperature. The TPS22916 is 60 mΩ typical at 5 V but 100 mΩ maximum at 85 °C; at 2 A that is 0.24 W versus 0.4 W. In a small package with a few hundred square millimetres of copper, 0.4 W can be a 30–40 °C rise — enough to matter inside an enclosure.
The eFuse has the same conduction loss plus its own quiescent current, and the ideal diode has the same RON physics. None of these parts is a zero-loss switch; the question is whether the heat is acceptable at the maximum ambient, which is a layout and package decision as much as a silicon one.
8. Fault handling and recovery
The three parts behave differently after a fault. A load switch simply stays on if the enable is asserted; the fault is whatever the load does. An eFuse either limits current until the fault clears or latches off, depending on the part and configuration, and usually provides a fault flag. An ideal diode does nothing about a fault; it only blocks reverse current.
That difference drives the system design. If the product must recover automatically, an eFuse with auto-retry or a PPTC is preferable; if a fault must be reported and latched, an eFuse with a flag is the right choice; if the fault is upstream, the load switch is enough.
9. ORing two supplies
Two ideal diodes, one per source, make an ORing circuit with low loss and no reverse feed. The alternative, two Schottkys, is cheaper but loses more voltage and leaks more at temperature. The ideal-diode controllers also handle the transition when one source drops, which a simple diode does with a larger voltage excursion.
Check the reverse standoff and the transient behaviour: the LM66100's −6 V standoff is fine for a 5 V rail, but a 12 V rail needs a part rated for it. Also check the enable and status pins if the system must know which source is active.
10. Reverse blocking: body diode and back-to-back FETs
A single pass FET always has a body diode, so it conducts in reverse even when off. A load switch therefore does not block reverse current; if the output is back-fed, current flows through the body diode and the "off" state leaks. An eFuse with back-to-back FETs, or an ideal diode, blocks both directions because there is no forward-biased body diode in the path.
That is the physical reason the three parts differ, not just a feature list. If the application can be back-fed — a USB port, a battery connector, a board that can be powered from a programmer — reverse blocking must be explicit. If it cannot, a single-FET load switch is simpler and cheaper.
11. Who should not use which
- Do not use a load switch as a current limiter. It has no current sense and will not protect a shorted load.
- Do not use an eFuse where a load switch's 0.5 µA quiescent current matters. The supervision circuitry costs current; battery products may prefer the switch.
- Do not use an ideal diode for inrush control. It has no soft-start; a large load capacitance will draw whatever the source allows.
- Do not use a Schottky where reverse blocking matters. It leaks in reverse and its drop rises at low temperature.
12. Five mistakes
1. Ignoring RON at temperature. A 60 mΩ part is 100 mΩ at 85 °C; the loss rises by 67%.
2. Choosing the load switch for inrush and expecting current limit. They are different functions.
3. Undersizing the eFuse timer. The load's inrush profile, not its average current, sets the timer.
4. Using an ideal diode without checking reverse standoff. A 5 V part on a 12 V rail fails on the first transient.
5. Forgetting the fault flag. An eFuse that trips silently looks like a dead board; route the flag to the controller.
None of these choices is permanent: a board that adds a hot-plug connector or a battery backup later may need to move from a load switch to an eFuse or add an ideal diode. Leave the footprint and the fault flag for that possibility.
13. Checklist
- Required function: on/off, inrush, current limit, overvoltage, reverse blocking.
- Maximum load current and RON at the maximum temperature.
- Load capacitance and the inrush target.
- Short-circuit behaviour required by the system.
- Reverse or back-feed possibility.
- Quiescent and shutdown current budget.
- Fault reporting and recovery behaviour.
- Package thermal path and copper area.
- Threshold coordination with the upstream protection.
- Verification: inrush, short and reverse tests at temperature.
One number summarises the choice: the fault current the part must survive. A load switch is chosen by RON and inrush; an eFuse by its current limit and response; an ideal diode by reverse standoff and drop. Write the fault current next to the part number before signing off the schematic.
FAQ
What happens if the eFuse trips during a firmware update?
The board loses power unless the eFuse auto-retries. For products that must not drop during an update, use auto-retry or a load switch with a separate supervisor.
Does a load switch block reverse current when off?
No. Its body diode conducts in reverse; only a back-to-back arrangement or an ideal diode blocks both directions.
Can I put a load switch and an eFuse in series?
Yes, but it is rarely useful: the eFuse already switches and protects. Two pass FETs add loss and cost; use the load switch only if it serves a separate branch.
Why does the ideal diode get hot at high current?
It is still a FET with RON. At 1.5 A and 79 mΩ the loss is about 0.18 W; the advantage over a Schottky is the lower drop, not zero loss.
Can a load switch replace an eFuse?
Only if the load cannot short and the source is already protected. A load switch has no current limit, so a shorted output is limited only by the source.
Why does my load switch get hot?
Conduction loss is I²R, and RON rises with temperature. Check the maximum RON at your ambient, not the 25 °C typical value.
Do I need an ideal diode if I already have a Schottky?
If efficiency or reverse leakage matters, yes. A Schottky's 0.4 V drop at 2 A is 0.8 W; an ideal diode is a fraction of that and blocks reverse current.
What is the difference between OVLO and OVC?
Overvoltage lockout disconnects the load above a threshold; overvoltage clamp holds the output at a limit and lets the load continue. The choice depends on whether the load can tolerate the clamp voltage.
Can I use an eFuse for hot-plug?
Yes, with a blanking timer set for the inrush profile. That is one of its main advantages over a load switch.
Glossary
- RON
- ON-resistance of the pass FET; sets conduction loss.
- Inrush
- Current that charges load capacitance at turn-on.
- OVLO
- Overvoltage lockout; disconnects above a threshold.
- OVC
- Overvoltage clamp; limits the output voltage.
- ORing
- Combining two supplies so either can power the load.
Sources
- Texas Instruments TPS22916 load switch datasheet — 1–5.5 V, 2 A, 60 mΩ, controlled rise time.
- Texas Instruments TPS25947 eFuse datasheet — 0.5–6 A limit, OVLO/OVC, reverse blocking.
- Texas Instruments LM66100 ideal diode datasheet — 1.5–5.5 V, 1.5 A, 79 mΩ, −6 V reverse standoff.
Choosing a power-path part?
Browse our power switches and power modules, or send the rail, load and fault requirements for a recommendation.