Why a boost converter fails when the Wi-Fi radio starts

The board runs fine on the bench supply. On a single lithium cell and a small boost converter it works until the Wi-Fi radio starts, then it resets. On a bad day the boost converter itself is dead, and the next one dies the same way. Swapping parts feels like progress and is not.

This is one of the most common power faults in battery products with a radio. The cause is almost never one part. It is a short, sharp current demand meeting a power stage that was sized for the average.

What the radio actually asks for

A Wi-Fi or BLE radio does not draw its datasheet average. It draws a burst, then nothing, then a burst. Espressif says this plainly in its hardware design guidelines: when an ESP32 is transmitting, there may be a sudden increase in current draw that can collapse the power rail, so it recommends a 10 µF capacitor on the radio supply rail alongside the smaller decoupling capacitors (Espressif ESP32 schematic checklist). The same guide asks for a 3.3 V supply that can deliver at least 500 mA when a single supply is used, and at least 10 µF where power enters the board.

Those are the chip vendor’s numbers for the chip’s own pins. They say nothing about whether the converter upstream can hold the rail while the burst lasts. That is your job, and it starts with the burst, not the average.

Measure it. Put a current probe or a shunt with a fast scope in the battery lead and trigger on the radio start. Look at the peak, the duration and the rise time. A multimeter or a slow logger will show you a comfortable number that the board never actually sees.

The usual suspects, in the order we check them

1. The inductor saturates

A boost converter stores energy in its inductor each cycle. Every inductor has a saturation current. Above it, inductance collapses and the current rises much faster than the controller expects.

TI’s guidance for boost converters is to pick an inductor with a saturation current rated higher than the peak current the application will see, and its TPS61230A datasheet recommends a margin of 20% to 30% above that peak (TI TPS61230A datasheet, inductor selection). It also warns that an inductor with insufficient saturation performance can lead to excessive peak current. That is a datasheet for one device, but the principle applies to the whole class.

Two details catch people:

• Peak current is not output current. With a low battery voltage and a 5 V output, the input current is much larger than the output current. Compute the peak at the lowest battery voltage you will allow, at the highest load burst, not at a fresh cell. As a rough shortcut, Monolithic Power’s application note estimates the peak as the output power divided by the minimum input voltage, scaled up for ripple and losses (MPS AN122). Use the converter datasheet’s own equation for the final number.
• Saturation curves differ. TI’s LM5123-Q1 note describes hard saturation, where inductance drops sharply, and soft saturation, where it rolls off slowly (TI SNVA991, LM5123-Q1). A hard-saturating part can push the current past the converter’s limit faster than the limit circuit reacts. Read the inductor’s current-versus-inductance curve, not only the single rated number, and note whether the rating is defined by a drop in inductance or by temperature rise. Vendors define both and they are not the same.

A substitute inductor with the same inductance and footprint is not the same part. If the layout allows a larger package with a higher saturation rating, that is often the cheapest fix on the list.

2. The input side rings or sags

A single cell has internal resistance, and a battery holder, a connector, a protection circuit and a load-sharing FET each add more. When the boost converter pulls a large current step, the input voltage dips. If the input also has a long lead and only a ceramic capacitor, the lead inductance and the ceramic can ring. The input can swing well beyond the dip you expected.

The checks:

• Probe the converter’s VIN pin at the pin, with a short ground spring, not a long clip lead. A long ground lead creates its own ringing and you will chase it.
• Look for undershoot below the converter’s minimum operating voltage or its undervoltage threshold. A converter that drops out and restarts under load produces exactly the “reset when Wi-Fi starts” symptom.
• If the input rings, add bulk capacitance with some damping, such as a polymer or electrolytic part in parallel with the ceramic. Check the converter datasheet’s own input capacitor guidance for your part, since it differs by family.

Tantalum bulk capacitors deserve a caution. Some designers use them to add capacitance cheaply, but they can fail badly from surge currents on low-impedance sources such as a battery. If you use one, check the manufacturer’s voltage derating guidance for that series.

3. The layout of the switching loop

The boost converter’s switching loop, formed by the inductor, the switch node, the output capacitor and the ground return, carries fast current edges. Loop area sets the voltage spikes. On a two-layer board with no continuous ground plane, spikes and ground bounce reach the feedback pin and the load.

What we look for on a board:

• The output capacitor sits right at the output and ground pins of the converter, as in the datasheet’s layout example, not a few millimeters away on a thin trace.
• The feedback divider takes its signal from after the output capacitors and stays away from the switch node.
• Ground is a solid pour under the converter, with the return path of the switching loop short and wide. Check that vias carrying that loop are not undersized or placed far from the capacitor.
• The radio’s decoupling capacitors are close to the module’s supply pins, as Espressif’s checklist asks.

A converter dying repeatedly on a board with spikes at the switch node is not unlucky. It is a layout finding.

4. Startup and load sequencing

Inrush is a separate event from the transmit burst. Output capacitance charging, the load turning on and the radio calibrating at boot can stack on one another. If the converter and the load come up together, the load may start while the rail is still below its operating level. Gate the load with an enable line or a supervisor, or let the converter’s soft-start finish first. Read how the specific converter behaves with a load present at startup, since it varies.

Fix order that saves time

Do these in order, and re-measure after each:

1. Scope the battery current and the converter VIN, VOUT and switch node during a Wi-Fi start, with proper probing.
2. Compute the true peak inductor current at the lowest battery voltage and highest load burst. Compare it against both the inductor’s saturation rating and the converter’s current limit.
3. Fix the inductor first. It is the smallest change with the largest effect.
4. Add or damp the input bulk capacitance if VIN dips or rings.
5. Fix the layout if the switch node or ground shows spikes.
6. Only then consider a different converter, or consider whether the radio should be on a separate regulated rail. If the product runs from one lithium cell and the radio needs 3.3 V, a buck-boost or a low-dropout regulator running directly from the cell may avoid a 5 V rail that only one peripheral needs. Check that the cell’s minimum voltage still supports that regulator.

Resist the urge to fit larger and larger capacitors without measuring. Large output capacitance raises the startup inrush and can make the problem worse. It can hide a saturating inductor for one revision and bring it back in production, where cell resistance, temperature and part tolerances all move.

Test it where it ships

A bench supply with a stiff output proves little. Test on the real cell, at the cold end of its range and at the lowest charge you intend to support, because cell resistance rises when the cell is cold or depleted. Run the radio through connect, transmit and reconnect cycles many times and log every reset. Our first power-up procedure covers current-limited bring-up, and why sleep current beats the datasheet covers the other side of the same power budget.

How we approach it

We start from a scope capture of the burst and work back through the inductor, the input, the layout and the startup sequence. If you send us the schematic, the layout files and a capture of the failure, we can tell you which of these four it is.