1. Why this regulator is everywhere

The AMS1117-3.3 is the default answer to "I need 3.3 V from 5 V" on a small board. It is a three-terminal low-dropout regulator in a SOT-223 package, rated around 1 A, with a fixed 3.3 V output and no external components beyond two capacitors.

That simplicity is why it appears in routers, development boards, sensors, adapters and industrial controllers. It is also why so many second sources exist: the pinout is trivial, the application circuit is published, and the market is enormous.

The catch is that the part is often used outside its thermal comfort zone. A designer sees "1 A" on the datasheet and a 12 V input, and assumes it will work. It will not, and the failure is a slow thermal one rather than an immediate fault.

2. What actually matters in an LDO swap

Most substitution failures come from the third and fourth rows, not the first. A part that is electrically compatible can oscillate with the wrong capacitor, or run 20 °C hotter because its thermal path is different.

3. The pin-compatible families

All of these use the same SOT-223 footprint: pin 1 is ground (or the adjust pin), pin 2 is the output, pin 3 is the input, and the tab is connected to the output. That is why they are interchangeable mechanically.

One important exception: parts like the SPX3819 are also 3.3 V LDOs, but they sit in a five-pin SOT-23 package with a different pinout. They are not substitutes for a SOT-223 board, even though they appear in the same category.

4. The thermal math that decides everything

A linear regulator dissipates the voltage difference multiplied by the current: (Vin − Vout) × I. That power has to leave through the package, and the SOT-223 is small.

Now apply a typical SOT-223 thermal resistance. With about 80 °C/W junction-to-ambient on a modest copper area, a 1.7 W dissipation raises the junction by roughly 136 °C above ambient. At 40 °C ambient that is 176 °C — far beyond the maximum rating.

Even 0.5 W is warm. This is why the "1 A" rating is misleading in practice: it assumes a large copper area and a small voltage drop. For 12 V to 3.3 V, the practical continuous current is a few hundred milliamps at best, and often less inside an enclosure.

Dropout is the other half of the thermal story. As the input falls toward the output voltage, the regulator leaves regulation and the output follows the input minus the dropout. On a 5 V rail that is rarely an issue; on a battery rail it is the whole design.

For a 3.7 V lithium cell, a 3.3 V LDO at 300 mA needs a dropout below 0.4 V, and it needs that dropout near the end of discharge. Most 1 A LDOs specify dropout at full rated current, where it is largest. Check the curve at your actual current, not the headline figure.

A worked example with copper area

Take a 5 V to 3.3 V rail at 0.5 A: dissipation is 0.85 W. With 80 °C/W the junction rises 68 °C above ambient — acceptable at 25 °C ambient, marginal at 50 °C inside an enclosure.

Adding copper changes the number quickly. A tab connected to a 25 mm × 25 mm pour on both layers, with a few thermal vias, can bring the effective resistance closer to 50 °C/W. The same 0.85 W then raises the junction by about 43 °C, which is comfortable.

The lesson is that the PCB is part of the regulator. Two boards using the same part can differ by 30 °C purely because of copper area, and a substitution that ignores the layout is only half a comparison.

A 24 V industrial rail makes the case obvious. Even 100 mA from 24 V to 3.3 V dissipates more than 2 W, which a SOT-223 cannot shed. Every such design needs a switching stage first, regardless of which LDO is used. The LDO then becomes a post-regulator for noise, not the main converter, and its dissipation drops by an order of magnitude.

5. When a linear regulator is the wrong answer

If the input is more than about twice the output, or the load is above a few hundred milliamps, a switching regulator is the correct choice. It is not a substitution — it is a different topology with an inductor, a different feedback network and a different EMI profile — but it is the only design that survives the thermal math.

A common architecture is a two-stage supply: a buck converter from 12 V to 5 V, then an LDO from 5 V to 3.3 V for a quiet analogue rail. That keeps the LDO in its thermal comfort zone and preserves low noise where it matters.

For a board that already has a 12 V rail and a 3.3 V LDO running hot, the cheapest fix is often a small buck module followed by the existing LDO, not a search for a "better" LDO.

Fixed versus adjustable

The fixed 3.3 V version is the simplest choice: two capacitors, no divider, no tolerance stack. The adjustable version replaces the ground pin with a feedback node and needs two resistors.

For a board that will be copied into several variants, the adjustable part often wins on total cost because one reel covers 1.2 V, 2.5 V and 3.3 V builds. The penalty is an extra node to route and a divider that must be recalculated — and the reference voltage is not identical across families.

A subtle trap: on some adjustable parts the reference is 1.25 V, on others it is different. A divider copied from a reference design can produce a 5–10% error if the reference voltage is not checked.

6. Where pin-compatible parts differ

Dropout. The minimum input voltage at full load differs between families, and it rises with current. A part with a 1.3 V dropout cannot regulate 3.3 V from a 4.2 V battery near the end of discharge.

Quiescent current. Some LDOs draw milliamps at no load; others draw microamps. For a battery product that sleeps most of the time, that difference dominates the energy budget.

Capacitor requirements. Older LDO designs need a minimum ESR and can oscillate with modern ceramic capacitors. Newer parts are stable with ceramics. Check the stability curve, not just the capacitance value.

Thermal resistance. The same package can have different die attach and lead frame construction. Two parts with identical footprints can run 10–20 °C apart at the same dissipation.

Noise, PSRR and analogue rails

For digital rails, noise is irrelevant. For analogue, RF and sensor rails, it is the specification that matters. An LDO's output noise and its ability to reject input ripple (PSRR) decide whether a 12-bit ADC sees a clean reference.

PSRR falls with frequency. A part that rejects 60 dB at 1 kHz may reject only 20 dB at 100 kHz, which is exactly where a switching pre-regulator puts its ripple. If the rail feeds a precision measurement, check the PSRR curve at the switching frequency of the upstream converter, not just at DC.

Output noise is usually specified in µV RMS over a bandwidth. Compare the bandwidth: a number quoted over 10 Hz–100 kHz is not comparable with one quoted over 10 Hz–1 MHz.

Capacitor selection and stability

LDO stability depends on the output capacitor's ESR and capacitance. Older designs require a minimum ESR — typically a tantalum or an electrolytic — and can oscillate with a pure ceramic capacitor. Newer parts are designed for ceramics and may require a minimum capacitance instead.

Two practical rules. First, read the stability curve in the datasheet, not just the recommended capacitor value. Second, keep the input capacitor close to the device; a long input trace adds inductance and can destabilise the loop even when the output side is correct.

If a substitution produces a rail that looks fine on a multimeter but has a low-frequency oscillation, the capacitor is the first suspect. Check with a scope at the output, AC coupled, over a few milliseconds.

7. A practical test plan

Parallel operation and current sharing

Linear regulators do not share current well. Two AMS1117s in parallel will not split a 1.5 A load evenly; the one with the slightly higher output voltage takes most of the current until thermal drift changes the balance.

If more current is needed, use a part rated for it or a switching converter. Adding a small ballast resistor in series with each output can force sharing, but it costs voltage headroom and stability margin — rarely worth it for a 3.3 V rail.

Reverse current and protection

If the output can be driven above the input — for example when a downstream rail stays alive after the input is removed — current can flow backwards through the regulator. Some LDOs tolerate this; others do not. Check the reverse-current specification if the board has multiple supplies.

A Schottky diode across input and output protects against reverse polarity but adds leakage and a small voltage drop. For battery products, that leakage can matter more than the protection is worth.

Quality differences between second sources

At this price point, second sources compete on process control as much as on design. The parameters that vary most are output voltage tolerance at temperature, dropout at full load, and quiescent current.

For a non-critical digital rail, those differences rarely matter. For a rail that feeds a sensor or a radio, they can. Ask for the temperature curves rather than the 25 °C typical values, and compare the worst case.

8. Cost and supply

These regulators cost cents, so price differences are small in absolute terms. The variables that matter are availability, packaging and traceability.

Second sources are common and usually acceptable for non-critical rails. For analogue or RF rails, stay with a part whose noise and PSRR you have characterised, because a cheaper regulator can degrade a measurement chain without failing any test.

Reel quantities and date codes matter more than unit price at this level. A full reel of a cheaper part can cost more than a partial reel of the original, and old stock needs a functional check before use.

Pre-regulators and modules

The most practical fix for a hot LDO is to reduce the voltage across it. A small buck converter from 12 V to 5 V, followed by the existing 3.3 V LDO, cuts the dissipation by more than half and keeps the quiet rail.

Ready-made buck modules are acceptable for prototypes but not for production: the inductor, capacitors and thermal design are unknown, and the module becomes a single-source component that nobody qualified. Design the pre-regulator onto the board once the architecture is settled.

When to keep the original

If the rail is quiet, the thermal margin is small, or the product is certified, the cheapest part is not worth the requalification. Keep the original in those cases and use the second source only as insurance, not as a production change.

9. FAQ

Is LM1117 a direct replacement for AMS1117-3.3?

They share the SOT-223 footprint and pinout. Dropout, current rating and capacitor requirements differ, so verify the load and the thermal design.

Can I run 1 A from 12 V to 3.3 V with this package?

No. That is about 8.7 W of dissipation, far beyond what a SOT-223 can handle. A switching regulator is required.

Why does my LDO oscillate with ceramic capacitors?

Some older LDO designs need a minimum ESR that ceramics do not provide. Choose a part specified as stable with ceramic output capacitors.

Does a lower dropout always mean better?

Usually, but it can come with higher quiescent current or different stability requirements. Match the parameter to the application.

Can I substitute a SOT-23-5 LDO on a SOT-223 board?

No. The package and pinout differ. It needs an adapter or a new layout.

What is the practical current limit?

It depends on the voltage drop and the copper area. For a 5 V to 3.3 V rail, a few hundred milliamps is typical before thermal limits dominate.

10. Sources

• TI LM1117 product page: ti.com
• Diodes AZ1117 documentation: diodes.com
• MaxLinear SPX3819 documentation: maxlinear.com