Power Noise in Wireless Nodes: LDO vs DC-DC

Key Takeaways

• Power noise on a radio supply rail shows up as phase noise, reduced sensitivity or unstable ADC readings.
• Switching converters are efficient but emit ripple at the switching frequency plus broadband edge energy.
• A linear regulator adds no switching artefacts and rejects ripple, but turns the difference into heat.
• Cascading a DC-DC with an LDO keeps efficiency high and gives the radio a quiet rail.
• Decoupling capacitors work below their self-resonance; above it they stop behaving as capacitors.

1. What Is Power Noise?

Power noise is any unwanted voltage or current variation on a supply rail that reaches a sensitive circuit and degrades its performance.

The phrase covers effects that behave nothing alike, so two engineers can report a noisy rail and mean different things.

• Ripple: periodic variation left on the output by a switching converter, at the switching frequency and its harmonics.
• Switching noise: broadband energy produced by the fast edges of the switch node, which has no single frequency and resists notch filtering.
• Transient droop: the short sag when the load steps, for example when a radio keys up, set by loop bandwidth and capacitance.
• Intrinsic regulator noise: thermal and flicker noise from the internal reference and error amplifier, present even on a battery rail.

2. How Supply Noise Is Created and How It Reaches the Radio

Supply noise is generated inside the converter and couples into the radio along several paths at once.

1. The switching edge creates broadband energy. A buck switch node turns on and off in a few nanoseconds, so its edge spectrum reaches into the hundreds of megahertz and overlaps the receive band. A converter can pass every DC measurement and still lift a receiver noise floor.
2. Inductor current ripple becomes output ripple. The inductor current ramps every cycle and the output capacitor absorbs the difference, leaving triangular ripple at the switching frequency, typically 300 kHz to 2 MHz, plus harmonics.
3. Harmonics land on the channel. A harmonic inside the channel bandwidth or on an image frequency appears as a discrete spur rather than a raised floor.
4. Ground bounce adds common-impedance coupling. Switching current returns through the ground plane, and parasitic inductance turns it into a small voltage that moves the radio's ground reference. The symptom is noise only when another part of the node draws current.
5. Supply pushing shifts the oscillator. Every oscillator has some supply sensitivity, and a voltage-controlled oscillator has more than a crystal. Supply noise modulates the carrier and converts to phase noise.
6. Decoupling and layout decide the outcome. A module rated for a wide 2.3-5.5 V supply can still be quiet, because local decoupling and the ground return set the noise reaching the chip.

3. What Is an LDO?

A low-dropout regulator (LDO) is a linear regulator that holds a stable output voltage using a pass transistor across a small voltage headroom.

It has no switching stage, so its noise behaviour differs in kind from a converter's, and its weakness is thermal.

• Linear pass element: a series transistor drops the excess voltage continuously. With no switching node there is no ripple and no edge energy to couple into a radio.
• Noise set by the reference and error amplifier: low-noise parts are quoted in tens of microvolts RMS, and lower output noise usually costs more quiescent current.
• Power supply rejection (PSRR): often 60 dB or more below 10 kHz, rolling off as frequency rises. Above a few hundred kilohertz the LDO behaves much like a series impedance.
• Dropout and thermal limits: dropout is the smallest input-to-output difference that keeps regulation, a few hundred millivolts at moderate current, and dissipated power is (Vin minus Vout) times load current.
• Quiescent current trade-off: battery nodes want quiescent current in the microamp region, but the lowest-quiescent-current parts have a slower loop and weaker rejection.

4. LDO vs DC-DC: What Is the Difference?

The comparison is not about which regulator is better but about which cost each imposes.

The decision rule: use a DC-DC converter when the input-to-output difference or the load current is large, and an LDO after it when the rail feeds a radio or an ADC.

5. Power Noise Parameters and What They Mean in the Field

Each parameter below is where a supply design holds up or fails in the field.

• Switching frequency and its harmonics: typically 300 kHz to 2 MHz, with harmonics far above. A spur in one channel only usually points here.
• Output ripple: tens of millivolts peak-to-peak at full load. Growth with load current means an undersized capacitor or inductor.
• PSRR at the band of interest: 60 dB or more below 10 kHz, falling afterwards. Read the curve at 1 MHz, not the headline value.
• Dropout voltage and thermal rise: a few hundred millivolts at moderate current, and the package must dissipate (Vin minus Vout) times load current. Too little headroom sags the rail on transmit.
• Load transient response: recovery in a few microseconds. A slow loop shows as droop on a transmit burst, which looks like a protocol fault.
• Decoupling network and self-resonance: a 100 nF capacitor is self-resonant in the low tens of megahertz. Above that, parasitic inductance dominates.
• Ground return path: shared returns develop millivolts at high di/dt. The symptom is a noise floor that rises only while transmitting.
• Wide supply range: modules rated 2.3-5.5 V tolerate a partly discharged battery. That window is tolerance, not filtering.
• Sleep current in the microamp region: a 2 uA sleep current keeps a battery node alive for years; a high quiescent current erases that.

6. When to Use an LDO and When Not To

Good fit for an LDO

• A rail feeding a radio, a low-noise amplifier or an ADC reference.
• Battery nodes that sleep for long periods and need microamp quiescent current.
• Post-regulation after a DC-DC converter, where headroom is small.

Good fit for a DC-DC converter

• A large step such as 12 V down to 3.3 V, where a linear stage wastes more than the load uses.
• Loads of hundreds of milliamps and above, where heat, not noise, limits the design.
• Wide-input industrial rails, where the radio is a small part of the load.

Poor fit for an LDO

• High current with a large voltage difference, where the thermal rise is unavoidable.
• Inputs that fall close to the target output, leaving no dropout headroom.

Poor fit for a DC-DC converter

• A direct feed into a sensitive radio or ADC with no further filtering.
• Layouts with no room for a tight switch-node loop.

7. Real-World Applications

In industrial sensing, metering and battery-powered telemetry, the usual pattern is a switching converter at the front of the chain and a quiet linear stage at the radio, which absorbs the wide input and the heat.

Ebyte covers both ends of that chain. The AM31-12W12V is an ultra-small AC and DC dual-purpose switching power supply module with an 85-450 V or 120-630 Vdc input and a 12 V, 1000 mA output, described with ultra-low ripple, high efficiency, safety isolation and compliance with IEC60950, EN60950 and UL60950. The AM31-12W05V delivers 5 V at 2400 mA from 85-264 Vac or 100-370 Vdc, and the AM31-12W24V gives 24 V at 500 mA from the same input range in a 38.2 x 25.2 x 23 mm plug-in package. Downstream stages still need local decoupling.

At the radio end, the module sets how much supply tolerance is needed. The E22-400T22D-V2 is a Semtech SX1268 based LoRa UART module running 410.125-493.125 MHz with a 433.125 MHz default at 22 dBm up to 5 km line-of-sight, with TTL UART levels compatible with 3.3 V and 5 V MCU IO, a wide 2.3-5.5 V supply and 2 uA sleep current. The E103-W03 is a serial-to-Wi-Fi module based on the TI CC3220R SoC, covering 2.412-2.472 GHz IEEE 802.11 b/g/n at 18 dBm and 180 m with 12 uA hibernate current.

One detail is underestimated: the supply and the antenna share one ground return, which carries both switching and RF current.

8. FAQ: Power Noise in Practice

Q1: What does power noise do to a radio's sensitivity?

Power noise raises the noise floor and widens the carrier. Noise coupled into the oscillator converts to phase noise, so weak signals are harder to separate. At system level this appears as reduced range or intermittent packet loss.

Q2: Is an LDO quieter than a DC-DC converter?

Yes, for the frequencies that matter to a radio. An LDO has no switching edges, and its output noise is set by the internal reference and error amplifier, so it produces no switching harmonics. A DC-DC converter leaves broadband switch-node energy.

Q3: Should I use a DC-DC converter followed by an LDO?

Often, yes, when the input-to-output difference is large. The converter handles the voltage step efficiently and the LDO removes the remaining ripple and switching residue. Size the LDO for the full load current and budget its dropout headroom.

Q4: My node works on the bench but loses range from a switching supply. What should I check?

Start with the rail, because the switching converter is the usual culprit.

• Range drops only while transmitting -> check switching harmonics against the receive band.
• Noise floor raised on every channel -> check decoupling at the module supply pin.
• Sensitivity varies with load -> check output ripple at full load.

Q5: How do I choose decoupling capacitors for a radio module?

Choose by self-resonance, not by capacitance alone. A capacitor stops decoupling above its self-resonant frequency, where parasitic inductance dominates. Combine a few values, for example 10 uF, 100 nF and 10 pF near the supply pin.

9. Practical Checklist Before You Choose a Regulator

1. Estimate the headroom before the noise. If the voltage difference times the load current exceeds what the package dissipates, a linear stage is not a candidate.
2. Cascade when efficiency and quietness are both required. A converter for the voltage step plus an LDO for the radio rail is the usual answer.
3. Compare switching harmonics with the receive band. Check every harmonic up to a few hundred megahertz against the channel.
4. Treat decoupling as a network, not one capacitor. Choose values by self-resonance and place them at the module supply pin.
5. Verify with the radio transmitting on the assembled board. Measurements taken with the transmitter idle miss common-impedance coupling.