The operating mode has been selected and the IC is in place, but the high-frequency noise from a switching power supply may still exceed the limit. The solution is to add a final “filter gate” at the output. A passive second-order LC filter can attenuate the noise by approximately 47\,\text{dB} before it reaches the load. This article explains where the noise comes from, how to build the filter, how to calculate its component values, and how to verify its performance.
Where Does Switching Noise Come From?
Switching regulators naturally produce high-frequency noise for three reasons:
- PCB parasitic inductance and capacitance interact with rapidly changing voltage and current;
- the voltage and current change abruptly when the switch transitions;
- the switch node rings.
These high-frequency components can reach the load through conducted or radiated paths. For sensitive loads such as RF circuits, precision ADCs, and sensors, even a few millivolts of ripple can cause problems.
Recall Figure 1 from Part 1. It shows the output ripple before filtering—approximately 9\,\text{mV}, with the spikes mainly caused by the switching transitions. It is reproduced below for comparison:
Figure 1: Ripple and spikes at the LT8330 output before filtering. The simulated ripple is approximately 9\,\text{mV} (image sourced from Analog Devices).
Option 1: Can an LDO Be Used After the Boost Converter?
The most direct approach is to place a low-dropout linear regulator (LDO) after the boost converter and use its high PSRR to suppress ripple. This is not suitable for this example:
- With a 34\,\text{V} output, high-voltage LDO options are limited and dropout loss is significant.
- The 2\,\text{MHz} switching frequency is higher than the loop bandwidth of most LDOs, so PSRR falls sharply at this frequency.
Conclusion: An LDO is not an efficient or practical solution here.
Option 2: A Second-Order LC Output Filter
A more suitable approach is to add a second-order LC filter at the output. It requires no active devices—only an inductor, several capacitors, and a damping resistor.
Figure 2: Schematic of the additional output filter designed for the LT8330 (image sourced from Analog Devices).
Component-selection guidelines:
- Choose a series inductor L_1 in the range of 100\,\text{nH} to 500\,\text{nH}, with as little DCR as possible to avoid additional voltage drop.
- C_1 and L_1 determine the cutoff frequency.
- Choose C_2 as four to five times C_1. Together with the damping resistor R_1, it forms a damping network that suppresses LC resonance.
How Is the Cutoff Frequency Calculated?
For this example, choose L_1=0.3\,\mu\text{H} and C_1=4.7\,\mu\text{F}. The cutoff frequency is:
Whether this frequency is appropriate can be judged against three reference points:
- It is higher than the regulator loop bandwidth, approximately 13\,\text{kHz} in this example, so it does not interfere with loop stability.
- It is lower than the switching frequency, 2\,\text{MHz}, so it provides sufficient attenuation of switching ripple.
- As a rule of thumb, place it at least one decade above the loop bandwidth and one decade below the switching frequency.
How Is the Damping Resistor Calculated?
An LC filter has an impedance peak at its resonant frequency. Connecting C_2 and R_1 in parallel with C_1 suppresses this peak:
Technical tip: The damping network consumes a small amount of energy and introduces a slight voltage drop. Compared with the noise reduction it provides, this is usually acceptable.
Simulation Verification 1: Frequency Response
The Bode plot provides an intuitive view of how much noise the filter blocks:
Figure 3: Simulated frequency response of the additional filter (image sourced from Analog Devices).
At the 2\,\text{MHz} switching frequency, the filter provides approximately 47\,\text{dB} of attenuation. What does 47\,\text{dB} mean? The ripple amplitude is reduced to approximately 1/223 of its original value.
Simulation Verification 2: Time-Domain Output Ripple
Finally, examine the time-domain waveform after adding the filter and compare it with Figure 1:
Figure 4: Simulated LT8330 output ripple after adding the filter (image sourced from Analog Devices).
Comparing Figures 1 and 4, the high-frequency spikes are significantly suppressed and the output becomes clean and flat. This is the value of a second-order LC filter in a noise-sensitive application.
Important Reminder: Keep the Filter Outside the Control Loop
The filter is connected after the feedback-sense point, on the load side. In other words, the feedback network senses the voltage before the filter rather than the “clean” filtered output. Why?
- If the filter is placed inside the control loop, the LC resonance introduces additional phase shift into the loop gain and may compromise stability.
- When placed outside the loop, the filter only attenuates output noise and does not affect loop compensation.
- The tradeoff is a small load-regulation loss caused by the filter’s DC resistance and the inductor DCR, which is usually minor.
Summary
- A second-order LC filter is an economical and effective solution for suppressing high-frequency switching noise.
- The 134\,\text{kHz} cutoff frequency lies between the loop bandwidth (13\,\text{kHz}) and the switching frequency (2\,\text{MHz}).
- The damping network, R_1=0.25\,\Omega and C_2=22\,\mu\text{F}, suppresses the LC resonant peak.
- The simulation shows approximately 47\,\text{dB} of attenuation at 2\,\text{MHz}.
Key Takeaways
- A second-order LC filter is economical and effective; this example achieves approximately 47\,\text{dB} of attenuation at 2\,\text{MHz}.
- Set the cutoff frequency to 134\,\text{kHz}, between the loop bandwidth (13\,\text{kHz}) and switching frequency (2\,\text{MHz}).
- The damping network, R_1=0.25\,\Omega and C_2=22\,\mu\text{F}, suppresses the LC resonant peak.
- The filter must be placed after the feedback-sense point, outside the control loop, to avoid compromising stability.
Source: This article references the Analog Devices technical article Boost Converter Design for Sensitive Loads at Light Load.
Related Products
- LT8336 | 40 V, 2.5 A low-IQ synchronous boost converter
- LT8330 | 60 V, 1 A low-IQ boost/SEPIC converter
Additional References
More Content
- Boost Converter Design: Part 1 — The Light-Load Challenge: Choosing Between Efficiency and Noise
- Boost Converter Design: Part 2 — Design Goals and Input Current / Duty Cycle Calculations
- Boost Converter Design: Part 3 — Selecting the Inductor: Larger or Smaller?
- Boost Converter Design: Part 4 — IC Selection and Light-Load Operating Modes
- Boost Converter Design: Part 5 — The Final Gate: An Output LC Filter
- Boost Converter Design: Part 6 — Reviewing the Seven-Step Design Process