---
title: "What do a chip fuse, a polyfuse and an eFuse do when a USB port shorts?"
topic: "Fuses"
answer: "Simulated on a 5 V port rated for 0.5 A, the chip fuse opens in 0.82 ms and stays open. The polyfuse trips in 78 ms, then keeps passing 160 mA for as long as the short stays. The eFuse cuts off in 0.63 µs, then comes back on and holds up to 0.5 A until something switches it off."
url: "https://nearfieldnotes.com/notes/fuses/"
site: "Near Field Notes"
author: "Michael Walker"
note: 37
family: "Power"
published: "2026-11-13"
revised: "2026-11-13"
revision: 1
video: "https://nearfieldnotes.com/videos/fuses/"
---

Note 37 · [Power](https://nearfieldnotes.com/families/power/) · Rev 1 · [Watch the video · 3:08](https://nearfieldnotes.com/videos/fuses/)

Fuses

# What do a chip fuse, a polyfuse and an eFuse do when a USB port shorts?

**Short answer:** Simulated on a 5 V port rated for 0.5 A, the chip fuse opens in 0.82 ms and stays open. The polyfuse trips in 78 ms, then keeps passing 160 mA for as long as the short stays. The eFuse cuts off in 0.63 µs, then comes back on and holds up to 0.5 A until something switches it off.

A tripped polyfuse is a heater running on the fault’s own current, and the colder the room, the more it lets through.

All three are sold as 0.5 A parts. On each one, 0.5 A means something different.

**Note 37’s opening figure: One dead short, three parts.** The current into a short at the end of a 1 m cable

![Note 37’s opening figure: One dead short, three parts](https://nearfieldnotes.com/notes/fuses/fig-opening.png)

*What the figure shows:* A dead short at the end of a 1 m USB cable on a 5 V port, and the current each of three 0.5 A parts lets into it, on log scales of current and of time from 0.1 microseconds to 5 minutes, simulated. The eFuse cuts off in 0.63 microseconds at 3.9 A, stays off for 30 microseconds, then comes back on and holds 0.5 A. The chip fuse carries 4.7 A and opens in 0.82 milliseconds, leaving at most 0.5 milliamps. The polyfuse carries 5.6 A, trips in 78 milliseconds and keeps passing 160 milliamps.

Simulated from models fitted to the datasheets. The chip fuse and polyfuse models leave out the cable’s inductance, so they start at full current. The chart ends with all three together, five minutes in.

## A 5 V USB port, rated for 0.5 A

*1 m cable, 28 AWG*

Take a board with a USB-A port that it powers from its own 5 V rail. A USB 2.0 port promises a device 500 mA, and the USB spec makes the host protect its ports against over-current.

One part sits in the port’s 5 V line, between the rail and the connector. Its job is to stop a fault in whatever gets plugged in.

**Fig. 37a: The board, with a polyfuse in the 5 V line.**

![Fig. 37a: The board, with a polyfuse in the 5 V line](https://nearfieldnotes.com/notes/fuses/fig-37a.png)

*What the figure shows:* Figure 37a: the 36 by 24 mm port board rendered from above. A 6-pin header on the left brings in 5 V. Its VBUS trace runs through a polyfuse in an 1812 case and past a 120 microfarad capacitor to a right-angle USB-A receptacle on the right, with a cable's plug in it. The data lines and ground run straight from the header to the receptacle.

36 × 24 mm. The board is an example drawn for this page, and its other two versions swap in the chip fuse or the eFuse.

As a circuit, the rail and its copper add 50 mΩ and the port’s contacts 60 mΩ. A 1 m cable with 28 AWG power wires adds 0.43 Ω and 0.6 µH.

With nothing in the way, a dead short at the far end of the cable would draw 9.2 A.

Two faults go at that far end. One is a dead short. The other is a device that pulls 750 mA, half again what the port allows.

**Fig. 37b: The same port, as a circuit.**

![Fig. 37b: The same port, as a circuit](https://nearfieldnotes.com/notes/fuses/fig-37b.png)

*What the figure shows:* Figure 37b: the port as a circuit. A 5 V source with 50 milliohms of source and board copper feeds the 0.5 A protection part, then 60 milliohms of port and plug contacts and a 1 m cable of 0.43 ohms and 0.6 microhenries. At the cable's far end is the fault: a 10 milliohm short, or a device drawing 0.75 A.

The 0.5 A part is one of the three in Fig. 37c. The 0.75 A device holds its current down to 2 V, and below that acts as a resistor.

## Three parts, each sold as 0.5 A

*1206, 1812 and 2 × 2 mm*

A chip fuse is a thin metal film on a ceramic body. Enough current melts the film, and the fuse is open for good. This one is Bourns’ SF-1206F050.

A polyfuse, or [PPTC](https://nearfieldnotes.com/glossary/#g-polyfuse), is a slab of plastic filled with carbon, between two metal foils. It trips when it gets hot and resets when it cools. Bourns’ MF-MSMF050 is the one in the Arduino Uno’s USB power line, though Bourns now marks it not recommended for new designs.

An eFuse, short for [electronic fuse](https://nearfieldnotes.com/glossary/#g-efuse), is a switch that measures its own current. One resistor sets its limit, and one capacitor sets how long it lets an overload run. TI’s TPS259474L is the latch-off version: after a long overload it switches off and stays off.

**Fig. 37c: The three parts, to one scale.**

![Fig. 37c: The three parts, to one scale](https://nearfieldnotes.com/notes/fuses/fig-37c.png)

*What the figure shows:* Figure 37c: the three parts rendered side by side on a strip of board, to one scale. From the left: the chip fuse, a white 1206 body with a gray top marked F; the polyfuse, a larger black 1812 block with gold end terminations; the eFuse, a black 2 by 2 mm package with small pads; and beside it two 0402 parts, a resistor and a capacitor.

Rendered. The two small parts beside the eFuse are the resistor that sets its limit and the capacitor that sets its timer.

**The numbers**

| Part | Model | Size | What it is | Resistance in the line |
|---|---|---|---|---|
| Chip fuse | Bourns SF-1206F050 | 1206, 3.2 × 1.6 mm | A thin metal film on ceramic that melts | 0.52 Ω typical |
| Polyfuse | Bourns MF-MSMF050 | 1812, 4.5 × 3.2 mm | Carbon-filled polymer between two metal foils | 0.15 Ω new, up to 1.0 Ω after a trip |
| eFuse | TI TPS259474L | 2 × 2 mm, plus a resistor and a capacitor | A switch that senses its current, with a timer and a temperature sensor | 28 mΩ typical |

## Why a tripped polyfuse still conducts

*160 mA at 23 °C*

Cold, the carbon grains in the plastic touch one another and form chains, and current flows along them.

Near 125 °C the plastic melts and swells. The chains pull apart and the resistance climbs by decades. That is how Bel Fuse describes it, and Raychem’s guide puts the switch at 125 °C for most of its parts.

But it only stays tripped while it stays hot, and its only heat is the fault’s current flowing through it. So it settles where that current keeps it just hot enough.

**Fig. 37d: Inside the polyfuse.**

![Fig. 37d: Inside the polyfuse](https://nearfieldnotes.com/notes/fuses/fig-37d.png)

*What the figure shows:* Figure 37d: the polyfuse rendered with its front cut away. A black slab of carbon-filled polymer lies between two thin copper foils, one along its top and one along its bottom, each joined to one of the gold terminations wrapped over the two ends.

Rendered with the front cut away. The polymer is drawn as one dark slab. Its carbon grains are far too small to show.

Bourns rates this part to give off 0.8 W once tripped, in a 23 °C room. In the simulation the tripped part settles near 123 °C and gives off 0.79 W to the board and the air.

`I = P/V`

where I: current it still passes (A); P: power it must give off to stay hot (W); V: voltage across it (V).

With the short at the far end, nearly all of the 5 V sits across the polyfuse, 4.9 V. Giving off 0.79 W at that voltage takes 160 mA. Its resistance becomes whatever makes that work, 31 Ω here.

A colder room carries more heat away, so the part has to pass more current to stay hot.

**Fig. 37e: What a tripped polyfuse passes, by room temperature.**

![Fig. 37e: What a tripped polyfuse passes, by room temperature](https://nearfieldnotes.com/notes/fuses/fig-37e.png)

*What the figure shows:* Figure 37e: the current a tripped polyfuse keeps passing into a dead short on the 5 V port, by the room's temperature, simulated. At 0 degrees C, 197 milliamps, At 23 degrees C, 160 milliamps, At 50 degrees C, 117 milliamps, At 85 degrees C, 63 milliamps. The colder the room, the more it passes.

Simulated, five minutes into the dead short on the 5 V port.

**The numbers**

Five minutes into a dead short, typical part (0.35 Ω cold). From the heat model, not a bench.

| Room | Current | Power in the polyfuse | Its temperature |
|---|---|---|---|
| 0 °C | 197 mA | 0.96 W | 123 °C |
| 23 °C | 160 mA | 0.79 W | 123 °C |
| 50 °C | 117 mA | 0.58 W | 124 °C |
| 85 °C | 63 mA | 0.31 W | 125 °C |

Tripped polyfuse

### 160 mA for as long as the short stays

**Field marks.** Five minutes into a dead short the port still delivers 160 mA. The polyfuse sits near 123 °C and gives off 0.79 W, for as long as the short and the supply last.

**Look-alike.** An open fuse. Nothing past it works, and the fault looks cleared. But current still flows, and the part stays hot.

**Range.** The room matters more than the part: 197 mA at 0 °C, 117 mA at 50 °C and 63 mA at 85 °C. A part at its lowest or highest resistance passes 158 to 162 mA.

**Habitat.** USB ports on development boards and computers, and other low-voltage rails. This one is rated for 15 V at most.

## A dead short at the end of the cable

*9.2 A with nothing in the way*

The chip fuse carries 4.7 A and opens in 0.82 ms. Bourns says it then measures at least 10 kΩ, so at most 0.5 mA gets through.

The polyfuse carries 5.6 A and takes 78 ms to trip, then settles to 160 mA over the next few seconds.

That is a little more current for 95 times as long. So before it trips, the polyfuse lets through 141 times the chip fuse’s [I²t](https://nearfieldnotes.com/glossary/#g-i2t), the current squared over time that heats the cable and the copper.

**Fig. 37f: The chip fuse and the polyfuse in the short.**

![Fig. 37f: The chip fuse and the polyfuse in the short](https://nearfieldnotes.com/notes/fuses/fig-37f.png)

*What the figure shows:* Figure 37f: the current into the dead short through the chip fuse and through the polyfuse, from 10 microseconds to 10 seconds on a log time scale, simulated. With nothing in the way the short would draw 9.2 A. The chip fuse starts at 4.7 A, sags to 3.2 A as its film heats, and opens at 0.82 milliseconds. The polyfuse starts at 5.6 A, sags to 2.7 A and trips at 78 milliseconds, then falls to 160 milliamps over the next few seconds. Under it, on a log scale, the current squared times time each lets through before it acts: 0.012 A squared seconds for the chip fuse and 1.67 for the polyfuse, 141 times as much.

Simulated. Both start at full current, since their models leave out the cable’s inductance. Their own resistance, 0.51 Ω and 0.35 Ω, keeps them under 9.2 A. Below, what each lets through before it acts.

**The numbers**

Both models start at full current: they leave out the cable’s inductance.

| Part | Current at first | Acts at | Current then | I²t until then | Current at 10 s |
|---|---|---|---|---|---|
| Chip fuse | 4.72 A | 0.82 ms | 3.22 A | 0.01187 A²s | 0.5 mA |
| Polyfuse | 5.58 A | 78 ms | 2.74 A | 1.672 A²s | 161.3 mA |

The eFuse watches its current. It reaches the trip level, twice its limit, 128 ns into the short, and its switch is off 0.63 µs in.

Then the eFuse turns itself back on. TI’s design waits 30 µs and switches on again in current limit, so the port recovers quickly if the fault was brief.

If the short is still there, it stays in current limit. TI’s table of faults lists a short as “Circuit Breaker followed by Current Limit”, not as one that latches off.

In the simulation it holds 0.5 A. TI says the limit drops lower when the output is under 1.9 V, without giving a number, and here the output sits near 0.25 V. So 0.5 A is the most it passes.

**Fig. 37g: The eFuse in the short.**

![Fig. 37g: The eFuse in the short](https://nearfieldnotes.com/notes/fuses/fig-37g.png)

*What the figure shows:* Figure 37g: the current through the eFuse in the dead short, against time from 10 nanoseconds to 1 millisecond on a log scale, simulated from TI's datasheet. The current climbs through the cable's inductance and crosses the eFuse's trip level of 1.00 A at 128 nanoseconds. The switch is fully off 0.63 microseconds in, at 3.93 A. It stays off for 30 microseconds, then turns back on and holds the current at 0.5 A.

From a model built on TI’s datasheet. The current climbs through the cable’s inductance until the trip, the switch is off 500 ns later, and it comes back on 30 µs after that. The climb back to 0.5 A follows TI’s figure, about 100 µs.

Holding 0.5 A with nearly all of the 5 V across it puts 2.4 W into a 2 × 2 mm package.

The eFuse [shuts itself down](https://nearfieldnotes.com/glossary/#g-thermal-shutdown) at 154 °C. How hot it gets depends on the board around it. TI gives two figures: 41.7 °C/W on a four-layer board with eight thermal vias under the part, and 74.5 °C/W on one with none.

The board here puts eight vias in the copper at the eFuse’s input and output pins.

**Fig. 37h: The eFuse version of the board.**

![Fig. 37h: The eFuse version of the board](https://nearfieldnotes.com/notes/fuses/fig-37h.png)

*What the figure shows:* Figure 37h: a rendered close-up of the eFuse version of the board. The eFuse's input pin sits on a copper pour with four vias and its output pin on another with four more. A 6.65 kilohm resistor sets its current limit and a 10 nanofarad capacitor sets its timer. Two thin traces, EN and PG, run from the header to its pins, so a processor can switch it and read its power-good flag.

Rendered. EN and PG run to the header, so a processor can switch the port and read its power-good flag. TI’s thermal figure is for vias right under the part. These sit in the pin copper beside it.

**Fig. 37i: Where the eFuse’s junction heads in the short.**

![Fig. 37i: Where the eFuse’s junction heads in the short](https://nearfieldnotes.com/notes/fuses/fig-37i.png)

*What the figure shows:* Figure 37i: the junction temperature the eFuse heads for while it holds 0.5 A into the short, 2.36 W, in a 23 degree room, on TI's two boards. With 8 thermal vias, 41.7 degrees C per watt, it settles at 122 degrees, 108 to 135 across the limit's tolerance, below its 154 degree shutdown, so it stays on. With no vias, 74.5 degrees C per watt, it would reach 199 degrees, so it reaches 154 and shuts down.

The room plus 2.4 W times each board’s figure, in a 23 °C room. The bars start at the room’s temperature. The whiskers span the limit’s tolerance.

**The numbers**

Room 23 °C. Power in the eFuse is the limit times the voltage left across it: 2.03, 2.36 and 2.69 W. Both boards are TI’s, from its datasheet’s thermal table.

| Board | Junction to room | Limit 0.425 A | 0.500 A | 0.575 A |
|---|---|---|---|---|
| 8 thermal vias | 41.7 °C/W | 108 °C | **122 °C** | 135 °C |
| No vias | 74.5 °C/W | 174 °C | **199 °C** | 224 °C |

eFuse in a short

### 122 °C, short of its 154 °C shutdown

**Field marks.** With the eight vias it settles at 122 °C. It stays on, passing up to 0.5 A into the short, until something switches it off.

**Look-alike.** A part that has done its job. Nothing past it works, but it is still giving off up to 2.4 W.

**Range.** Across its limit’s tolerance, 0.425 A to 0.575 A, it settles at 108 °C to 135 °C. The top of that is past the 125 °C TI characterizes it to. At the typical limit it would reach shutdown only in a room above 55 °C.

**Habitat.** Without the vias it heads for 199 °C, so it reaches 154 °C and shuts down, and this version stays off until EN or the power is cycled. The better the board cools it, the longer it holds the short.

Holding 0.5 A into the short is no more than the port may deliver to a working device, so the cable and whatever shorted are not overstressed. The cost is the eFuse’s heat, and the rail, or a battery, feeding a dead port for as long as the short lasts.

So in a dead short it needs something else to switch it off. The auto-retry version, the TPS259474A, would hold on the same way. It differs only after a shutdown, and with the vias there is none.

That something is usually a processor reading the eFuse’s PG pin. TI says PG goes low for faults “except overcurrent”, and in a short it follows only the voltage on the PGTH pin, set by two resistors.

TI lets that divider hang from the input or the output. From the input, it never sees the short, so the firmware never hears about it.

From the output, as in TI’s example circuit, a healthy port puts 1.32 V on PGTH, past the 1.20 V that raises PG. In the short PG drops, and the processor can pull EN low to switch the port off.

**Fig. 37j: Where the PGTH divider hangs, in the short.**

![Fig. 37j: Where the PGTH divider hangs, in the short](https://nearfieldnotes.com/notes/fuses/fig-37j.png)

*What the figure shows:* Figure 37j: the eFuse's PGTH divider, 47 kilohms over 16.9 kilohms, hung from two places during the dead short. From the input, which stays at 4.97 V, PGTH sits at 1.32 V, above the 1.09 V that PG falls below, so PG stays high. From the output, which the short holds at 0.25 V, PGTH sits at 0.07 V, below 1.09 V, so PG goes low.

The voltages are the simulated short’s. The two resistors are this page’s pick for a 5 V port.

**The numbers**

TI’s typical thresholds: PG rises past 1.20 V on PGTH and falls below 1.09 V. The divider is 47 kΩ over 16.9 kΩ, picked for this 5 V port.

| Divider from | PGTH, port healthy | PGTH, port shorted | PG in the short |
|---|---|---|---|
| Input | 1.32 V | 1.32 V | Stays high |
| Output | 1.32 V | 0.07 V | Goes low |

## A 750 mA device

*Half again the port’s 500 mA*

The eFuse lets it run for 8.4 ms: the time its timer current, 1.8 µA, takes to pull 1.51 V off the 10 nF capacitor, and 2 µs more to open. Then the TPS259474L switches off and stays off until EN or the power is cycled. The auto-retry version, the TPS259474A, would instead try again 110 ms after each trip.

The polyfuse sits between its [hold current](https://nearfieldnotes.com/glossary/#g-hold-current), 0.5 A, and its [trip current](https://nearfieldnotes.com/glossary/#g-trip-current), 1.0 A, where it may do either. Its resistance decides. Bourns allows 0.15 Ω new and up to 1.0 Ω an hour after a trip or soldering, and this page takes 0.35 Ω between them.

A part at 0.15 or 0.35 Ω warms to 35 °C or 55 °C and carries the device for the whole hour. One at its 1.0 Ω limit trips at 14 s and then passes 174 mA.

A warm room tips it. At 40 °C the 1.0 Ω part trips at 8.9 s, and at 60 °C the 0.35 Ω part still holds on, at 110 °C.

Bourns’ curve for the chip fuse stops at 0.95 A, 100 s. The part meets UL 248-14, under which a fuse must open within an hour at 135% of its rating, 0.675 A here. So at 750 mA it opens within the hour, and the datasheet gives no closer time.

**Fig. 37k: 750 mA for an hour.**

![Fig. 37k: 750 mA for an hour](https://nearfieldnotes.com/notes/fuses/fig-37k.png)

*What the figure shows:* Figure 37k: a device drawing 0.75 A for an hour, simulated, on a log time scale from 1 millisecond. Top, the current: the eFuse lets it run for 8.4 milliseconds, then switches off for good. A polyfuse at 0.15 or 0.35 ohms carries the 0.75 A for the whole hour. One at its 1.0 ohm limit trips at 14 seconds and then passes 174 milliamps. Bottom, the polyfuse's temperature: it settles at 35 degrees C at 0.15 ohms and 55 at 0.35 ohms, and at 1.0 ohm it climbs to its trip and holds 121.

Simulated in a 23 °C room. The chip fuse is left out: its datasheet has no time for 750 mA, so its model has nothing to go on.

**The numbers**

A 750 mA device for an hour. The polyfuse’s resistance is its cold value, before any trip.

| Part | Trips | Current after an hour | Polyfuse temperature after an hour |
|---|---|---|---|
| 0.15 Ω, room 23 °C | never | 750 mA | 35 °C |
| 0.35 Ω, room 23 °C | never | 750 mA | 55 °C |
| 1.0 Ω, room 23 °C | 14 s | 174 mA | 121 °C |
| 0.35 Ω, room 40 °C | never | 750 mA | 79 °C |
| 0.35 Ω, room 50 °C | never | 750 mA | 94 °C |
| 0.35 Ω, room 60 °C | never | 750 mA | 110 °C |
| 1.0 Ω, room 40 °C | 8.9 s | 141 mA | 122 °C |
| eFuse, any room | 8.4 ms | 0 | – |

## What 0.5 A means on each part

*Three different promises*

On the chip fuse, 0.5 A is a test point. Bourns’ part carries it for at least 4 hours and opens within a minute at twice that. Fuse makers advise running one at no more than 75% of its rating at room temperature, 0.375 A here.

On the polyfuse, it is the hold current: what the part is sure to carry at 23 °C without tripping. It is sure to trip only at 1.0 A.

On the eFuse, it is a set point. A 6.65 kΩ resistor sets it, and TI’s table gives 0.425 A to 0.575 A for any junction from −40 to 125 °C.

Near their rating the chip fuse and the polyfuse take seconds to minutes. The eFuse takes 8.4 ms anywhere between its limit and twice its limit, and half a microsecond above that.

At 1 A, twice the rating, the chip fuse opens in 16 s and the polyfuse trips in 100 s. Higher up, the chip fuse pulls far ahead.

**Fig. 37l: How fast each part acts.**

![Fig. 37l: How fast each part acts](https://nearfieldnotes.com/notes/fuses/fig-37l.png)

*What the figure shows:* Figure 37l: how long each part takes to act, against current, on log scales. The chip fuse's curve, from Bourns' datasheet, runs from 100 seconds at 0.95 A to 0.44 milliseconds at 5 A and stops there. The polyfuse's typical trip curve runs from 100 seconds at 1 A to 20 milliseconds at 8 A. The eFuse never acts below its 0.5 A limit, acts at 8.4 milliseconds between that and 1.00 A, and within 500 nanoseconds above. The heat models' times sit on the two datasheet curves as dots. A line marks 750 milliamps, where neither curve reaches.

Lines: Bourns’ two curves and TI’s timing for the eFuse. Dots: the two fuse models at the same currents, a typical miss of 10% for the chip fuse and 5% for the polyfuse.

**The numbers**

The chip fuse’s times are Bourns’ average melting times; the polyfuse’s, its typical times to trip at 23 °C, for a part of 0.35 Ω. Both read off the datasheets’ graphs.

| Part | Current | Datasheet | Heat model |
|---|---|---|---|
| Chip fuse | 5 A | 0.44 ms | 0.45 ms |
| Chip fuse | 3.46 A | 1 ms | 1 ms |
| Chip fuse | 2 A | 5 ms | 4.8 ms |
| Chip fuse | 1.73 A | 10 ms | 9.7 ms |
| Chip fuse | 1.34 A | 100 ms | 81 ms |
| Chip fuse | 1.22 A | 1 s | 1.1 s |
| Chip fuse | 1.03 A | 10 s | 11 s |
| Chip fuse | 0.95 A | 100 s | 100 s |
| Polyfuse | 1 A | 100 s | 100 s |
| Polyfuse | 1.5 A | 8 s | 7.6 s |
| Polyfuse | 2 A | 2.5 s | 2.7 s |
| Polyfuse | 2.5 A | 1 s | 0.99 s |
| Polyfuse | 4.25 A | 100 ms | 96 ms |
| Polyfuse | 8 A | 20 ms | 21 ms |

Heat moves the polyfuse’s rating most. It trips by getting hot, so a warm room leaves it less to do. At 85 °C it holds only 0.29 A, where Bourns’ curve takes the chip fuse to 0.46 A.

Its resistance also costs voltage. At 0.5 A a part at 1.0 Ω loses 0.5 V from the port. The chip fuse, at Bourns’ typical 0.517 Ω, loses 0.26 V, and the eFuse 14 mV.

The chip fuse has one more number that counts: it can [interrupt 50 A at 63 V](https://nearfieldnotes.com/glossary/#g-interrupt-rating). Past that it isn’t rated to clear safely.

**Fig. 37m: What the polyfuse holds, by room temperature.**

![Fig. 37m: What the polyfuse holds, by room temperature](https://nearfieldnotes.com/notes/fuses/fig-37m.png)

*What the figure shows:* Figure 37m: the polyfuse's hold current, the most it is sure to carry without tripping, against the temperature around it, from Bourns' derating table. It falls from 0.77 A at -40, 0.68 A at -20, 0.59 A at 0, 0.50 A at 23, 0.44 A at 40, 0.40 A at 50, 0.37 A at 60, 0.33 A at 70, 0.29 A at 85 degrees C.

From Bourns’ derating table for the MF-MSMF050, in still air.

**The numbers**

The model’s column is the most a part at its 1.0 Ω limit carries for 30 minutes without tripping, the datasheet’s hold test. It runs above Bourns’ table, by 0.01 to 0.09 A.

| Room | Hold, Bourns | Heat model, 1.0 Ω part |
|---|---|---|
| −40 °C | 0.77 A | 0.78 A |
| −20 °C | 0.68 A | 0.72 A |
| 0 °C | 0.59 A | 0.66 A |
| 23 °C | 0.50 A | 0.59 A |
| 40 °C | 0.44 A | 0.53 A |
| 50 °C | 0.40 A | 0.49 A |
| 60 °C | 0.37 A | 0.45 A |
| 70 °C | 0.33 A | 0.40 A |
| 85 °C | 0.29 A | 0.32 A |

## Side by side

*The same port, the same faults*

**Figure: Three 0.5 A parts on a 5 V port.**

Fig. 37n

|  | Chip fuse | Polyfuse | eFuse |
|---|---|---|---|
| What 0.5 A means | Carries it 4 hours; opens within 60 s at 1 A | Holds it at 23 °C; sure to trip at 1 A | Its limit, set by one resistor: 0.425 to 0.575 A |
| Voltage lost at 0.5 A | 0.26 V typical; 0.19 to 0.32 V within ±25% | 75 mV new at 0.15 Ω, up to 0.5 V after a trip | 14 mV typical, 22.5 mV at most |
| Dead short | Opens in 0.82 ms | Trips in 78 ms | Off in 0.63 µs |
| After the short | 0.5 mA at most, for good | 160 mA while the short stays | Up to 0.5 A until switched off |
| 750 mA device | No published time; within an hour under UL 248-14 | Trips at 14 s at 1.0 Ω; never at 0.15 or 0.35 Ω | Off at 8.4 ms, stays off |
| Resets | No: replace it | Itself, once the fault is gone and it cools | On a toggle of EN or power |
| Tells anyone | No | No | PG goes low, if PGTH hangs from the output |
| Most voltage | 63 V | 15 V | 23 V |
| Its 0.5 A in an 85 °C room | 0.46 A, 91% on Bourns’ curve | 0.29 A held | 0.425 to 0.575 A, TI’s limits for any junction from −40 to 125 °C |

From the simulations above and the three datasheets. The eFuse’s PG pin reports a short only if its PGTH divider hangs from the output, as in Fig. 37j.

## Which part where?

*Two questions*

The choice comes down to whether a fault must stay cut, and whether anything is watching the port.

Key to protecting a 5 V output

### Which part goes in the line?

1. 1a The fault must stay cut, as at a battery or a supply that can deliver far more than the wiring can carry Chip fuse, rated to interrupt it
2. 1b A USB host port, or an output that should come back by itself Go to 2
3. 2a A processor can watch it eFuse, PG to the processor, PGTH from the output
4. 2b Nothing watches it Polyfuse, sized for the warmest room

USB 2.0 asks a host’s protection to remove or reduce power and to reset without anyone touching it, and names polymeric PTCs and solid-state switches as examples. That leaves a chip fuse out of a host port. On the latch-off TPS259474L, that reset is firmware toggling EN. The TPS259474A retries by itself. The board in Fig. 37a ends at 2b, and its eFuse version in Fig. 37h at 2a.

#### Conditions

- **Simulated:** Every current, time and temperature on this page comes from models fitted to the parts’ datasheets. None was measured on a bench.
- **Port:** A stiff 5.0 V rail, 50 mΩ of source and board copper, 60 mΩ of port and plug contacts, and a 1 m cable with 28 AWG power wires: 0.213 Ω per meter each way and 0.6 µH. Room 23 °C unless a figure says otherwise.
- **Faults:** A 10 mΩ short at the cable’s far end, run for 5 minutes. A device drawing 0.75 A that acts as a resistor once its voltage falls below 2 V, run for an hour.
- **Chip fuse:** Bourns SF-1206F050. A heat model with three steps, the film and two layers of body, fitted to Bourns’ average melting times from 0.44 ms at 5 A to 100 s at 0.95 A. It opens when the film reaches a set temperature. Its I²t at 5 A comes out at 0.0112 A²s, against Bourns’ 0.011 A²s. Once open it is 10 kΩ, the least Bourns allows.
- **Polyfuse:** Bourns MF-MSMF050. A heat model with two steps, the part and the board under it, and a resistance that climbs slowly and then by decades past 118 °C. Fitted to Bourns’ typical trip times, from 100 s at 1 A to 20 ms at 8 A, and to its 0.8 W tripped, taken at 125 °C. Cold, 0.35 Ω unless a figure says 0.15 or 1.0 Ω.
- **eFuse:** TI TPS259474L with 6.65 kΩ on its ILM pin and 10 nF on ITIMER. A model of its events from TI’s datasheet: a trip at 2.01 times the limit, off 500 ns later, and back on in current limit 30 µs after that. Between the limit and twice it, the timer runs out after 1.51 V at 1.8 µA, then 2 µs to open. Its junction is the room plus the power times TI’s figure for the board.
- **Render:** Blender. The board is an example drawn for this page.

Fitted to datasheets. Nothing on this page has been measured on a bench.

#### Limits

- Bourns gives no typical resistance for the polyfuse, so 0.35 Ω is an assumption between its 0.15 Ω minimum and 1.0 Ω maximum. In the short it hardly matters: 158 to 162 mA across that range.
- Below 1.9 V on its output, TI says the eFuse’s limit drops lower, without a number. The page holds it at 0.5 A, so its 0.5 A and 2.4 W in the short are the most they could be.
- The voltages lost at 0.5 A use the datasheets’ resistances, which Bourns measures cold. Both fuses run warmer at 0.5 A and read a little higher.
- The chip fuse’s model has no data below 0.95 A, where Bourns’ curve stops. Its 750 mA case rests on UL 248-14 alone.
- The chip fuse and polyfuse models are resistors, with no cable inductance. In their first microsecond the real current would still be climbing, as the eFuse’s does.
- The eFuse’s temperatures use TI’s two board figures, which TI simulated for its own boards. A real board will differ, and so will vias beside the part rather than under it.
- Near the hold current the polyfuse model is least sure. Bourns’ hold table sits 0.01 to 0.09 A under the model’s 1.0 Ω part, so a real part could trip in the 750 mA case where the model’s doesn’t.
- Bourns marks the MF-MSMF050 not recommended for new designs. Its 16 V version, the MF-MSMF050/16X, can measure up to 1.3 Ω after a trip.

#### Files

- sim/fuses.py The port, both heat models and their fits, the eFuse’s events and every fault
- render/scene.py The board and the three parts in Blender

#### Sources

- [Bourns SF-1206F series datasheet, rev. J, 2021 (SF-1206F050): typical resistance, melting times, I²t, the 4-hour and 1-minute tests, the 50 A interrupting rating, the resistance once open and the thermal derating curve](https://www.bourns.com/docs/Product-Datasheets/SF-1206F.pdf)
- [Bourns MF-MSMF series datasheet, rev. BE (MF-MSMF050): hold and trip currents, resistance, trip times, 0.8 W tripped and the derating table](https://www.bourns.com/docs/Product-Datasheets/mfmsmf.pdf)
- [TI TPS25947 datasheet, SLVSFC9C, 2026 (TPS259474L): the on-resistance, the limit and its accuracy, the fast trip, the timer, the table of faults, thermal shutdown, the auto-retry version and its 110 ms retry, the thermal figures, PG and its table, and the PGTH divider in the example circuit](https://www.ti.com/lit/ds/symlink/tps25947.pdf)
- [Universal Serial Bus Specification, Revision 2.0, section 7.2.1.2.1: over-current protection on a host port](https://www.usb.org/sites/default/files/usb_20_20240604.zip)
- [Bel Fuse 0ZCJ series datasheet, 2025: how a polyfuse trips, and holds itself near 125 °C](https://www.belfuse.com/resources/datasheets/circuitprotection/ds-cp-0zcj-series.pdf)
- [Tyco Electronics Raychem, “Fundamentals of PolySwitch Overcurrent and Overtemperature Devices”, 2008: the switch at 125 °C](https://www.electrokit.com/upload/product/41004/41004351/circuit-protection-psw-fundamentals.pdf)
- [Littelfuse 1206L series datasheet, 2011: R1max, measured an hour after a trip or soldering](https://media.digikey.com/pdf/Data%20Sheets/Littelfuse%20PDFs/1206L_Series.pdf)
- [Littelfuse Fuseology Design Guide, 2022: UL 248-14’s 135% test, and running a fuse at 75%](https://info.littelfuse.com/hubfs/Fuseology%20Design%20Guide.pdf)
- [Arduino Uno Rev3 schematic: the MF-MSMF050 in its USB power line](https://www.arduino.cc/en/uploads/Main/Arduino_Uno_Rev3-schematic.pdf)

What should the next note be? [Ask for one](https://nearfieldnotes.com/#ask-for-a-note)

---

Source: https://nearfieldnotes.com/notes/fuses/

Cite as: Michael Walker, “Fuses”, Near Field Notes, Note 37, Rev 1, 13 Nov 2026. https://nearfieldnotes.com/notes/fuses/

Figures: free to reuse under CC BY 4.0 (https://nearfieldnotes.com/license/). Each one's image is linked under its name. Credit Michael Walker, Near Field Notes, and link the figure: this address plus #fig- and its number, like https://nearfieldnotes.com/notes/fuses/#fig-37a.
