LM2596 DC-DC Converter Module Testing

A bench that measures the module, not the cables

Most bad measurements of a buck module come from the wire between terminal and meter. Load leads carry amps and drop volts; sense leads carry none. Take the output voltage from a separate pair on the module’s terminals, not the far end of the load cable, and the same at the input: 0.1 Ω of cable at 2 A costs 200 mV.

A resistor bank is a stable load with no loop of its own; an electronic load sweeps faster, but its loop runs in parallel with the module’s, so a few hundred microfarads at the load keeps that local. Measure current in the return leg with a shunt.

  PSU ──┬── A ──┬── MODULE ──┬── A ──┬── LOAD
        │       │  IN     OUT│       │  (resistors
       V (Vin)  │            V (Vout)│   or CC load)
        │       │  tab   L   │  D    │
        └───────┴──[TC]─[TC]─┴──[TC]─┘

Tape thermocouples to the regulator tab, inductor winding and catch diode, plus one in still air. An infrared thermometer aimed at a shiny TO-263 reads low: the emissivity is unknown.

Probing: why a clip on the output posts lies

The 10 cm ground lead on a passive probe is a loop. The switch node swings the input voltage in nanoseconds, and the loop couples that edge into the trace. The LM2596 datasheet says it plainly: a probe terminator soldered at the output capacitor gives a short ground and a far cleaner picture than the standard three-inch lead.

Clip onto an output post and you have the worst version of that loop, millimetres from the inductor. Many cheap modules use open-core bobbin inductors that leak flux along their axis; the datasheet notes that a nearby trace acts as a single-turn secondary, producing false scope readings and real instability alike. Use tip-and-barrel at the output capacitor instead: tip on the positive pad, exposed ground barrel on the negative pad.

A differential probe is better for numbers meant to be compared: it removes the chassis ground path entirely.

Ripple, switching frequency, and the shape of the waveform

Ripple is the inductor’s triangular current multiplied by the output capacitor’s ESR. The datasheet puts typical ripple at 0.5% to 3% of the output voltage and gives the equation: peak-to-peak inductor current is V_OUT(1 − D)/(L·f), with D equal to V_OUT/V_IN. For 33 µH at 150 kHz and 5 V out that is about 0.59 A at 12 V in and 0.80 A at 24 V; across 0.1 Ω of ESR, 60 to 80 mV.

Check the switching frequency at the switch node, not the output, where the fundamental sits under the ripple harmonics; the oscillator is specified from 127 kHz to 173 kHz. At light load the converter runs discontinuous and the ripple changes shape.

Efficiency across load and input voltage

Efficiency is V_OUT·I_OUT over V_IN·I_IN, all four read at once at the module’s terminals. Sweep 10, 25, 50, 75 and 100% of the printed rating at 12 V and 24 V, and plot both. The datasheet’s typical figure for the 5 V version is 80% at 12 V in, 3 A out: 3.75 W of heat.

Where that heat sits shifts with input voltage. Switch conduction scales as V_SAT·D·I_OUT, which at the datasheet’s 1.16 V saturation figure and 3 A is about 1.45 W on 12 V and 0.72 W on 24 V. The catch diode conducts for the complement of the duty cycle, so its loss scales as V_F·(1 − D)·I_OUT: about 0.87 W on 12 V and 1.19 W on 24 V at the 0.5 V the datasheet assumes. Raise the input and the regulator cools while the diode heats.

Heat: the inductor, the diode, and 50 °C/W of board

Thermal resistance is where a cheap module loses its argument. The TO-263 package is specified at 50 °C/W junction-to-ambient with half a square inch of copper, 30 °C/W with 2.5 square inches, and 20 °C/W double-sided with 3 square inches. Put 1.5 W into 50 °C/W and the junction is 75 °C above ambient against a 150 °C limit.

Run each load point fifteen to twenty minutes, not two, and record the rise. Winding temperature can be estimated from resistance cold and hot using R/R₀ = 1 + 0.00393(T − T₀); a 20% rise is roughly 50 °C of winding rise.

Why the printed current rating does not survive 24 V

The silk-screened number is a DC load current at one input voltage, usually 12 V, at one ambient. The regulator’s limit is on peak switch current: 3.6 A minimum, 4.5 A typical, 6.9 A maximum at 25 °C, and 3.4 A minimum over temperature. Since peak current is I_LOAD + ΔI/2 and ΔI grows as the duty cycle shrinks, usable DC current falls as the input rises: a 3 A load peaks at 3.29 A on 12 V and 3.40 A on 24 V, already the minimum limit cold.

The inductor usually gives up first, and saturation is a peak-current event. Modules in this class often carry an inductor rated close to the printed output current. The datasheet gives the sequence: the core saturates, inductance collapses, switch current rises within the cycle, the cycle-by-cycle limit trips, and DC output current falls. Nothing breaks; the module simply stops delivering the printed amps.

Record, per load point: V_IN, I_IN, V_OUT, I_OUT, efficiency, ripple in mV peak-to-peak, switching frequency, the three temperatures, ambient, and whether the module hit current limit. Ten load points, each with a dozen fields, say more than the label.

These figures are calculated from the LM2596 datasheet’s equations and typical values and describe how the part is specified to behave; no specific module was measured here. Which limit binds first on a given board — die, inductor, diode or copper — only a bench run on that board can say.

References

• Texas Instruments, LM2596 SIMPLE SWITCHER Power Converter 150-kHz 3-A Step-Down Voltage Regulator, datasheet SNVS124G, March 2023.
• Texas Instruments, Reducing noise on the output of a switching regulator, Analog Applications Journal SLYT740.
• Texas Instruments, How to measure ripple for better design outcomes, SSZTB25.