How to control LM2596 buck-converter with microcontroller

What a buck module actually exposes

The LM2596 modules sold for a couple of dollars are a fixed-frequency (150 kHz) step-down regulator with an internal switch, an inductor, a Schottky diode and two capacitors. The feedback pin is held at 1.23 V, and the output voltage is set by the divider from Vout to that pin:

Vout = 1.23 V × (1 + R1 / R2)

On the module, one of those resistors is the trimpot. To control the output from a microcontroller you are not “sending a voltage” to the module — you are changing that divider ratio, either by replacing the pot with a digital potentiometer or by injecting current into the feedback node.

Option 1: a digital potentiometer in the divider

Replace the top resistor of the divider with a digital pot used as a rheostat (wiper tied to one end, the other end to the divider node). The output then follows the pot’s code.

• Voltage limit. Most digital pots specify a maximum terminal voltage of 5 V, and some only 3.3 V. The top resistor sees close to Vout, so a 12 V output will destroy the pot. Feed the pot from a low-voltage node or use a pot rated for the voltage — check the datasheet, do not assume.
• Resolution. A 100-step pot across the divider range gives about 1 % steps, which at 12 V is 120 mV per step. A 256-step part is better; a 32-step part is a toy.
• End-to-end resistance tolerance. Digital pots are specified at ±20 % or worse, so the absolute voltage will not match the calculation. Calibrate in firmware against a measured output, or accept a few percent error.
• Wiper current. Keep the divider current small (a few hundred microamps) — the wiper has a limited current rating and its resistance adds to the divider.

Option 2: PWM into the feedback node

This is the cheap trick: leave the divider alone, and inject a small current into the feedback pin through a resistor from a PWM pin, with an RC low-pass filter so the pin sees a steady voltage rather than a 490 Hz square wave.

MCU PWM ──[ R_f ]──┬── FB (1.23 V)
                   │
                 [ C ]  to ground
                   │
                  GND

The filtered PWM acts as a pull-up or pull-down on the feedback node and shifts the regulation point. Two effects make it awkward:

• It shifts, not sets. The relationship between duty cycle and output voltage is not the clean divider equation, and it drifts with the LM2596’s own reference tolerance. You end up calibrating a curve.
• Loop stability. The RC filter’s time constant interacts with the regulator’s compensation. Too little filtering and the output carries the PWM ripple; too much and the loop becomes sluggish and overshoots on load steps.

For a bench supply or a battery charger where you set a voltage and leave it, the digital pot is the cleaner design. For an experiment where you just need the output to move under software control, the PWM injection is fine — with the filter and a calibration table.

What the module will not do

• It cannot boost. Below about 1.5 V above the input, the LM2596 drops out of regulation. A 12 V input cannot produce 12 V out.
• It has no current limit you can set. The internal switch has cycle-by-cycle current limiting, but the module has no adjustable constant-current mode. Charging a battery or driving an LED string needs a different topology (a constant-current buck, or a current-sense loop).
• It is not a precision instrument. The reference is ±4 % over temperature on the plain LM2596, with load and line regulation to match. If you need 1 % you are using the wrong part.
• Its inductor and capacitors are the cheapest that work. Raising the output voltage raises the ripple; the stock 220 µH inductor and generic electrolytics are chosen for 5–12 V.

A practical recipe

1. Measure the module’s existing divider: with the pot set mid-travel, note Vout and confirm it matches 1.23 × (1 + R1/R2). That tells you the pot’s actual value, not the label.
2. Choose a digital pot with a terminal-voltage rating above your maximum Vout, 256 steps, and an SPI or I²C interface your MCU already uses.
3. Keep the feedback trace short and away from the switch node; the FB pin is high impedance and picks up switching noise.
4. Add a firmware calibration step: measure Vout at three codes and fit a straight line. Store the coefficients in EEPROM.
5. Never step the output while the load is sensitive — the LM2596’s transient response to a divider change is a visible glitch. Ramp the code in small increments and let the output settle.

References

• Texas Instruments, LM2596 SIMPLE SWITCHER power converter datasheet — 1.23 V feedback reference, switching frequency, current limit, dropout and component selection.
• Microchip, MCP4xxx digital potentiometer family datasheet — terminal voltage limits, wiper current, end-to-end resistance tolerance, rheostat configuration.
• Atmel/Microchip application notes on PWM-filtered DACs — ripple and settling trade-offs of an RC-filtered PWM into a high-impedance node.

The equations and limits above come from those documents. Specific module behaviour varies by clone; measure yours before trusting a calculated value.