AMS1117-3.3: Pinout, Real-World Limits, and Free KiCad Files

Flip over an ESP32 dev board, an older Arduino clone, or almost any hobby-grade module with a 5V input and a 3.3V brain, and the odds are good you will find the same four-lead SOT-223 package marked 1117. The AMS1117-3.3 earned that ubiquity honestly: it costs almost nothing, it delivers up to 1A, its fixed 3.3V output needs no adjustment network, and its failure modes are so well known they are practically documented folklore.

That last part is the reason this guide exists. The AMS1117 is a great part inside a fairly narrow envelope, and nearly every complaint about it, boards that reset on Wi-Fi transmit, regulators that run scalding hot, outputs that oscillate, comes from stepping outside that envelope rather than from bad parts. The envelope has three walls: dropout, capacitors, and heat.

On the fixed-voltage AMS1117 variants in SOT-223, pin 1 is ground, pin 2 is the output, pin 3 is the input, and the large tab is electrically tied to the output, not ground. That tab detail matters twice: it means the copper pour you use to cool the part is a VOUT pour, and it means a probe slip between the tab and a nearby ground zone shorts your rail. The adjustable variant repurposes pin 1 as the adjust pin, so double-check which variant a reference schematic actually uses before copying it.

If you would rather not transcribe any of that by hand, the verified KiCad files linked at the end of this guide carry the correct pin mapping and the SOT-223 land pattern, taken from parts imported and used in real designs.

A linear regulator needs its input to sit above its output by a margin called dropout, and the AMS1117's dropout is roughly 1.1V typical at heavy load, up to about 1.3V worst case at the full 1A. Add that to the 3.3V output and the arithmetic is strict: the AMS1117-3.3 wants about 4.4V or more at its input to guarantee regulation under load.

That makes 5V USB input its natural habitat, with comfortable margin. It also explains the most common AMS1117 misapplication on the internet: powering it from a single 3.7V lithium cell. A LiPo's 3.7V is a nominal number that sags toward 3.3V and below across the discharge curve and dips further under load spikes, so the regulator spends most of the battery's life below dropout, passing the sag straight through to your supposedly regulated rail. The symptom is a board that works on the bench with a full battery and browns out in the field. For battery designs, use a genuinely low-dropout, low-quiescent LDO or a buck converter; the AMS1117 also draws around 5mA of ground-pin current at idle, which is a real tax on standby life all by itself.

The AMS1117 is an older LDO architecture that relies on its output capacitor's characteristics for stability. The datasheet asks for a 22uF tantalum on the output, and a similar bulk capacitor on the input is standard practice. Countless dev boards substitute multilayer ceramics and mostly get away with it, but that is a statistical observation, not a design rule, and clone-to-clone variation among 1117-marked parts makes it worse.

The conservative recipe is simple and cheap: a 10uF or larger capacitor on the input, a 22uF capacitor on the output as the datasheet asks, both placed close to the package, plus a small 100nF ceramic at the load end of the rail. If your output shows tens of millivolts of high-frequency wiggle for no apparent reason, suspect the output capacitor before anything else; that is what marginal LDO stability looks like on a scope.

Every volt you drop across a linear regulator times every amp you pass through it becomes heat in that little package: P equals (Vin minus Vout) times Iload. At 5V in, 3.3V out, and 500mA, that is 0.85W, warm but manageable. At the full rated 1A it is 1.7W, which a SOT-223 can only survive with a generous copper pour under the output tab doing the work of a heatsink. Feed it 12V and ask for 500mA and you are at 4.35W, which is not a design, it is a thermal shutdown demonstration.

The practical rules: keep (Vin minus Vout) times current under about 1W if you want an unremarkable, long-lived regulator on ordinary copper, budget serious pour area as you approach 1.5 to 2W, and switch to a buck converter beyond that or whenever input voltage is high. The AMS1117 protects itself with current limiting and thermal shutdown, but a regulator cycling in and out of thermal shutdown looks exactly like a mystery software bug from the firmware side.

The AMS1117 is precisely the kind of part that should never cost you library time, and now it does not have to. Verified KiCad bundles, schematic symbol with the correct pin map, SOT-223 footprint with the datasheet land pattern, and a 3D model, are free to download for the AMS1117-3.3 at protoflow.ai/library/parts/C6186 and for its 5V sibling, the AMS1117-5.0, at protoflow.ai/library/parts/C6187. Both pages render the symbol and footprint so you can eyeball them before downloading, and both files came through the same pipeline engineers use to import parts into working designs.

Drop the symbol and footprint into your KiCad libraries, or open the part directly in ProtoFlow and let the schematic start with the regulator, its capacitors, and its ESP32 already connected. Either way, the hour you used to spend transcribing a land pattern from page nine of a datasheet is better spent on the parts of your board that are actually yours.