Audible Noise in LED Drivers: Causes, Standards, and Practical Mitigation

An LED lighting installation should be silent. When it is not — when a low buzz or whine emanates from a driver in a quiet office, a library, or a bedroom — the cause is almost always the driver's magnetic or ceramic components vibrating at audible frequencies. Audible noise in LED drivers is a well-understood phenomenon with identifiable causes, measurable standards, and practical mitigation strategies. It is also a quality issue: drivers that emit unacceptable noise have typically cut corners in component selection or circuit design.

The primary source is the transformer. A flyback transformer stores and releases energy at the switching frequency, and the magnetic flux density in the core varies with load current. The ferrite core changes its physical dimensions slightly as field strength changes — a phenomenon called magnetostriction. At high switching frequencies — 65kHz and above — these dimensional changes are ultrasonic and inaudible. But when the driver operates in burst mode or quasi-resonant mode, the switching frequency drops into the audible range (20Hz to 20kHz), and magnetostriction becomes audible as a whine. The effect is strongest at light load, where many drivers switch to burst mode for efficiency, and at the 100Hz mains envelope, where energy delivery pulses at twice the line frequency modulate the audible component.

A second source is the ceramic capacitors used for output filtering. Class 2 dielectrics such as X7R and Y5V are ferroelectric materials with a piezoelectric effect: an applied voltage causes physical deformation, and a varying voltage causes mechanical vibration. The high-frequency output ripple, filtered by ceramics in parallel with the electrolytic bulk capacitor, can make these ceramics vibrate and couple their motion into the PCB, producing an audible tone. The effect is most pronounced when a ripple component falls in the audible range due to control loop interaction.

Inductors in the input EMI filter and the PFC stage can also contribute. The PFC stage, operating in critical conduction mode at the boundary between continuous and discontinuous conduction, has a varying switching frequency that can sweep through the audible range as input voltage and load vary — a common source of the characteristic 'singing' of a PFC-equipped driver under certain conditions.

The governing requirement is embedded in the European Ecodesign Regulation 2019/2020 for light sources and separate control gear, which includes a specific limit on audible noise from control gear, measured with a sound level meter at a specified distance in a quiet environment. Some national codes and procurement specifications impose stricter limits for hospitals, libraries, and residential buildings.

Mitigation begins with design. The most effective measure is keeping the switching frequency and its harmonics above the audible range under all operating conditions — avoiding deep burst-mode operation at light load, or shifting burst-mode frequency above 20kHz. The transformer can be designed with lower flux density (larger core) to reduce magnetostriction; softer ferrites with lower magnetostriction coefficients can be specified; ceramic capacitors can be replaced with film types or used in smaller values. Potting the driver or applying conformal coating damps the PCB and absorbs vibration.

For the installer, the practical advice is to listen before finalizing. Power the driver at the actual load and mounting position, and listen at one meter. A driver that buzzes on the bench will buzz in the ceiling. If a model is known to be noisy, swapping brands or choosing a higher-switching-frequency driver is often the simplest fix. Audible noise is not a safety issue — the driver functions correctly — but it undermines perceived quality and, in noise-sensitive spaces, is a legitimate reason for rejection.