A current-limiting load switch

The SY6280 is a five-pin part in a SOT23-5 package that costs about the same as two polyfuses. Inside it there is a MOSFET, a current sense amplifier, a comparator, and a thermal sensor. It sits where the polyfuse went and does the same job in a completely different way.

Instead of getting hot until it stops conducting, it measures the current continuously and closes the MOSFET's gate down until the current is exactly what you told it to be. The fault is still a fault; it just never reaches the supply.

The one resistor

A resistor from the ISET pin to ground programs the limit:

I_LIM(A) = 6800 / R_SET(Ω)

6.8 kΩ gives 1 A. 3.9 kΩ gives 1.74 A. 47 kΩ gives 145 mA. The part covers about 130 mA to 2 A, and outside that range the formula stops being the thing in charge.

The tolerance is the part worth taking seriously. The datasheet's own test condition is R_SET = 6.8 kΩ, and it specifies the result as 0.75 A minimum, 1 A typical, 1.25 A maximum. A quarter either way is a correct part, not a bad one.

So the design rule is two-sided. Pick the resistor so that your real load still runs at the bottom of the band, and so that the supply still copes at the top of it. A board that draws 800 mA in bursts and a limit programmed at "1 A" is a board that will work on four units and brown out on the fifth.

What the supply sees

Press play and watch the current, not the clock.

The polyfuse lets the entire fault current through — whatever the wiring allows, often five or ten amps — for 150 milliseconds. The switch never lets more than its programmed limit through at all, from the first microsecond.

That difference is what decides whether a USB host shuts its port down, whether a wall adapter's own protection hiccups, and whether the 3.3 V regulator feeding the rest of your board browns out while the fault is being dealt with.

The sawtooth is correct

Holding 1 A across most of 5 V is about 5 W in a package designed to dissipate 0.4 W. The die reaches 130 °C, the part shuts down, it cools by 20 °C, and it tries again.

If you have ever plugged a shorted USB cable into a hub and watched its LED blink about once a second, you have watched this happening. It is called hiccup mode, and it is the only sensible response to a fault that has not gone away: keep the part alive, keep testing, recover the instant the short does.

The three things you get for free

An enable pin. EN is a logic input — above 2 V it is on, below 0.8 V it is off. A GPIO can now switch a whole peripheral's supply, which is the cheapest power saving there is. Do not leave it floating.

Reverse blocking. There is no body diode across this switch. When it is off and the output is pulled above the input — by another supply, or by a charged capacitor bank — nothing flows backwards.

Output discharge. At shutdown it pulls the output down through about 150 Ω, so the peripheral you just switched off is actually off rather than coasting on its own decoupling for half a second. That matters more than it sounds: a sensor that keeps 1.8 V on its rail does not reset when you power-cycle it.

Where it does not help

The SY6280 protects against too much current. It does not protect against a supply fitted backwards — only against the output being pulled above the input while the switch is off. For that you need something pointing in a direction, which is the whole of the next chapter.