A switched-mode power supply (SMPS) converts electrical power using high-frequency switching transistors and energy storage elements—specifically inductors and capacitors—rather than dissipating excess voltage as heat. For a standard 24V-to-5V, 3A industrial load, an SMPS operates at 85% to 95% efficiency. By contrast, a linear regulator handling the same drop wastes nearly 80% of its input power as heat, making switching topologies mandatory for voltage drops exceeding 2V or continuous currents above 100mA.
Linear vs. Switched-Mode Power Supply: The Headroom and Efficiency Math
When deciding between a linear regulator and a switched-mode power supply for a specific load, the decision hinges on dropout voltage, headroom, and thermal limits. Let us run the math on a 24V nominal input stepping down to a 5V output at 3A (15W output power).
The Linear Regulator Penalty
Take a heavy-duty linear regulator like the LT1083. It requires a minimum dropout voltage (headroom) of about 1.5V to maintain regulation. With 24V in and 5V out, the voltage drop across the pass transistor is 19V.
- Power Dissipated: $P_{diss} = V_{drop} \times I_{load} = 19V \times 3A = 57W$.
- Efficiency: $\eta = V_{out} / V_{in} = 5 / 24 = 20.8\%$.
Dissipating 57W requires a massive, expensive extruded aluminum heatsink and forced air cooling. It is entirely impractical for sealed DIN-rail enclosures or battery-powered systems.
The SMPS Advantage
A synchronous buck converter (like the Texas Instruments LMR33630) steps the voltage down by rapidly switching a MOSFET and storing energy in an inductor.
- Efficiency: Typically 90% at this load.
- Input Power: $15W / 0.90 = 16.66W$.
- Power Dissipated: $16.66W - 15W = 1.66W$.
The heat generated drops from 57W to 1.66W, easily managed by the IC's exposed thermal pad and standard PCB copper pours.
Power Supply Topology Comparison
| Topology | Typical Efficiency | Heat Profile | EMI / Noise | Cost & Complexity |
|---|---|---|---|---|
| Linear (LDO) | 20% - 60% | Extreme (proportional to V-drop) | Ultra-low (no switching) | Lowest ($0.50 - $2.00) |
| Buck (Step-Down) | 85% - 96% | Low (mostly conduction/switching losses) | Moderate (switching node ringing) | Low-Medium ($1.50 - $4.00) |
| Boost (Step-Up) | 80% - 94% | Moderate (higher peak currents) | Moderate-High (right-half-plane zero) | Low-Medium ($1.50 - $4.50) |
| Flyback (Isolated) | 75% - 88% | Moderate (transformer core losses) | High (leakage inductance spikes) | High ($5.00 - $15.00+) |
For more on selecting the right architecture, review the Texas Instruments Non-Isolated DC/DC Regulator Overview or Analog Devices' DC-DC Switching Regulator portfolio.
Buck Converter Design Example: 24V to 5V at 3A
Let us design a practical buck converter for a 24V industrial sensor hub. We will use the TI LMR33630, a 36V-max, 3A synchronous buck regulator with integrated FETs, chosen for its small VQFN footprint and high efficiency.
Design Specifications
- Input Voltage ($V_{in}$): 24V nominal (18V to 32V operating range)
- Output Voltage ($V_{out}$): 5.0V
- Max Output Current ($I_{out}$): 3.0A
- Switching Frequency ($f_{sw}$): 400 kHz (selected to keep the inductor physically small while avoiding AM radio band interference)
Inductor Selection and Math
The inductor dictates your ripple current. We target a 30% ripple current ratio ($\Delta I_L = 0.9A$). First, calculate the duty cycle ($D$):
$D = V_{out} / V_{in} = 5 / 24 = 0.208$
Next, calculate the required inductance ($L$):
$L = \frac{V_{out} \times (1 - D)}{f_{sw} \times \Delta I_L} = \frac{5 \times (1 - 0.208)}{400,000 \times 0.9} = 11 \mu H$
Component Pick: Select a standard 10 $\mu$H shielded ferrite inductor rated for at least 4.5A saturation current (e.g., Würth Elektronik 7443552100). Shielded inductors are mandatory here to prevent magnetic flux from coupling into nearby high-impedance analog sensor traces.
Output Capacitor and Feedback Network
To minimize output voltage ripple, avoid electrolytic capacitors. Use three 47$\mu$F X7R ceramic capacitors (e.g., Murata GRM32 series) in parallel. This drops the effective Equivalent Series Resistance (ESR) to roughly 3m$\Omega$.
For the feedback resistors, the LMR33630 uses a 0.8V internal reference. Setting $R_{bottom}$ to 10k$\Omega$:
$R_{top} = R_{bottom} \times (\frac{V_{out}}{0.8} - 1) = 10,000 \times (\frac{5}{0.8} - 1) = 52,500\Omega$
Component Pick: Use a standard 1% tolerance 52.3k$\Omega$ resistor for $R_{top}$.
Input Protection, Ripple, and Thermal Derating
A schematic is only half the battle. Real-world 24V systems are electrically hostile environments, and thermal constraints dictate your physical enclosure design.
Input Range and Protection Circuitry
A '24V nominal' battery or industrial supply will frequently see 18V during heavy motor cranking and can spike to 35V or higher during inductive load dump events. Because the LMR33630 has an absolute maximum rating of 38V, you must protect the input.
The Fix: Place a bidirectional TVS diode (e.g., Littelfuse SMAJ33A) on the input to clamp load dumps. Follow this with an ideal diode controller (like the TI LM74700) or a P-channel MOSFET for reverse-polarity protection, which drops only millivolts compared to the 0.6V lost in a standard Schottky diode.
Ripple and Noise Expectations
With our 0.9A inductor ripple and 3m$\Omega$ ceramic ESR, the calculated low-frequency output ripple is:
$\Delta V_{out} = \Delta I_L \times ESR = 0.9A \times 0.003\Omega = 2.7mV$
However, this ignores high-frequency switching noise. The fast $dv/dt$ of the switching node coupling through parasitic trace inductance will create 50mV to 100mV high-frequency spikes at 400 kHz. If your 5V rail powers a 12-bit ADC or an RF module, this noise will corrupt your readings. Add a small RC snubber (e.g., 10$\Omega$ + 1nF) directly across the low-side FET, or place a ferrite bead and a secondary LC filter stage on the 5V output before it reaches sensitive analog loads.
Thermal and Derating Note
The LMR33630 in a VQFN package has a junction-to-ambient thermal resistance ($\theta_{JA}$) of roughly 35°C/W on a standard 4-layer JEDEC test board. At 1.66W dissipation, the junction temperature rise is $1.66W \times 35°C/W = 58°C$. At a 25°C room ambient, the silicon sits at a safe 83°C.
The Derating Reality: If you mount this PCB inside a sealed, unventilated IP67 enclosure sitting in direct sunlight, the internal ambient temperature can easily reach 65°C. Your junction temperature jumps to 123°C, dangerously close to the 150°C thermal shutdown threshold. In this scenario, you must derate the maximum continuous load by 20% (limiting it to 2.4A) or expand your top-layer copper pour to act as a heatsink, effectively dropping the $\theta_{JA}$ to 20°C/W.
Switched-Mode Power Supply FAQ
Why does my switched-mode power supply whine or buzz under light loads?
This audio noise is caused by Pulse Frequency Modulation (PFM) or 'burst mode' operation. To maintain high efficiency at light loads (e.g., drawing 5mA), the controller skips switching cycles, dropping the effective switching frequency into the 20Hz to 20kHz human hearing range. The ceramic capacitors exhibit piezoelectric effects, physically vibrating and acting as tiny speakers. To fix this, either force the controller into Pulse Width Modulation (PWM) mode via a dedicated MODE pin (sacrificing light-load efficiency) or use a regulator with a fixed switching frequency strictly above 20 kHz.
Can I parallel two switched-mode power supplies for double the current?
No, you cannot simply wire the outputs of two standard buck converters in parallel. Because their feedback loops and internal voltage references have slight manufacturing tolerances, the supply with the marginally higher output voltage will attempt to source 100% of the load current, potentially triggering its overcurrent protection while the second supply sits idle. If you need 6A, buy a 6A regulator, or use a multi-phase controller. If you must parallel them, you need modules with an active 'current share' pin or you must add ballast (droop) resistors in series with each output, which ruins load regulation and wastes power.
How do I measure switched-mode power supply ripple correctly?
Most engineers measure SMPS ripple incorrectly by using the standard 6-inch oscilloscope ground alligator clip. That long ground wire acts as an antenna, picking up radiated EMI from the switching node and displaying 200mV of 'noise' that isn't actually on the power rail. To measure true ripple, remove the ground clip and use a 'tip-and-barrel' coaxial probe adapter, or solder a short 24-gauge pigtail ground wire directly from the probe ground ring to the ground pad of the output capacitor. According to Tektronix measurement guidelines, keeping the measurement loop area as close to zero as possible is the only way to see the actual millivolt-level ripple on the DC rail.
