Where does DC switching power audible noise come from, and how can it be reduced or eliminated?
It is a common belief that multi-layer ceramic capacitors (MLCCs) or DC power supply circuits cause audible noise, but that’s not true. The noise is generated by the PCB, not by the components.
The noise of the components, as well as the influence of the circuit board size and its mounting, are examined step-by-step throughout this article, using three evaluation boards to demonstrate noise sources (Figure 1).
Figure 1 The MPQ4590, a 640V, non-isolated regulator with up to a 400 mA output current (left); the MPQ4316, a 45V, 6 A, low-IQ, synchronous step-down converter with frequency spread spectrum (center); and the MPQ4572, a 60V, 2 A, high-efficiency, fully integrated, synchronous buck converter (right) demonstrate noise sources for this article. Source: MPS
How does noise arise on the PCB, and which components of a DC power supply circuit are responsible? When the voltage on an MLCC changes due to the piezoelectric effect, the geometry of that capacitor changes, which results in a vibration (Figure 2).
Figure 2 MLCC vibrations are caused by voltage changes from the piezoelectric effect. Source: MPS
A voltage change in an MLCC generates a vibration stimulus. Vibrations are easily audible in the speech-sensitive frequency range (0.1 to 7 kHz). The vibration is transferred to the PCB via the solder joints, then the PCB emits an audible noise comparable to a loudspeaker membrane.
Figure 3 shows the typical components in a DC power supply circuit. MLCCs and the dimensions of the PCB are key to audible noise, as other components make no noise.
Figure 3 The MLCC is one of the the typical components in a DC power supply circuit that generates a vibration stimulus, and the PCB is the noise source. Source: MPS
Not all MLCCs behave in the same way. Only high-capacity Class II and III MLCCs have the piezoelectric effect. Other types of capacitors, molded inductors, resistors, and ICs do not show any change in geometry under a load. This means other components are insignificant sources of noise.
Table 1 Component classification in audible and inaudible systems
|MLCC Class I
|MLCC Class II, II X7R, X5R,
|MLCC Class II, II Interposer Type, Metal Strip||Electrolytic Tantalum Organic Capacitors||Switching Inductance (Molded)||Ferrite Beads, Resistors, DC/DC Converters|
DC power supply in FCCM or AAM
A DC power supply circuit operating in forced continuous conduction mode (FCCM) only produces audible noise within the speech-sensitive audio frequency range (e.g. GSM pulses or other periodic loads). A high DC switching power frequency is not audible.
When a DC power supply circuit operates in advanced asynchronous mode (AAM), light-load mode switching frequencies can be in the lower kHz range below 20 kHz. AAM switching frequencies are not fixed frequencies; they are random, which reduces audibility. AAM is only active under light-load currents, where there is generally no strong stimulus, and therefore seldom noise.
Comparing three mechanical systems
Audible noise on a PCB is created the same way that sound is generated on a stringed instrument (Figure 4). This theory is described in further detail below:
Measuring PCB noise with a microphone
Acoustic noise and the resonance frequency of a DC power supply circuit and PCB mount can be measured with a microphone and a small object that provides a Dirac impulse stimulus. A good choice is a condenser microphone, which is less sensitive against the magnetic field of the MLCC than a dynamic microphone.
A stick made of hard plastic or a plastic tweezer can be used as a simple mechanical stethoscope to make an audible noise easier to hear. Metallic objects make a louder noise, which can help search for points with a higher vibration amplitude (Figure 5).
Figure 5 Measure audible noise with this setup. Source: MPS
A comparison of the powered and unpowered microphone measurement shows that the PCB resonance frequency is exactly the same (Figure 6).
In the powered condition, the PCB is excited by an electrical signal. A 250 Hz load step causes the MLCC to vibrate, which excites the PCB at the 3900 Hz resonance frequency. In the unpowered condition, the PCB is excited by a mechanical shock, and a short push with the plastic stick causes the PCB to vibrate mechanically at the 3900 Hz resonance frequency.
The type of excitation, whether mechanical or electrical, has no influence on the resonance frequency of the PCB. The mechanical shock test can show the acoustic behavior of a test PCB, which behaves similarly to the later series PCB, as long as the dimensions and attachment points are comparable.
Measure PCB noise with a turntable and microphone
If a piezoelectric accelerometer is not available, a turntable is a simple alternative that can measure the exact horizontal vibration on the diamond. If a moving magnet or moving coil cartridge are the only unpowered measurements, the magnetic field of the capacitor current disturbs the signal. For a powered measurement, a crystal cartridge is a better choice to measure vibrations. While the microphone measures the integral, the cartridge or piezoelectric accelerometer measures a defined point (Figure 7).
Figure 7 A turntable can be used as an alternate set-up to measure horizontal vibration on the PCB. Source: MPS
Microphones show a second hammer touch and the mechanic bounce during hammer impact (Figure 8). The large cartridge amplitude shows the horizontal movement of the PCB and the cartridge with the tonearm. The PCB here is supported on two sides, and is free above the rubber mate of the turntable.
Table 2 lists different resonance frequencies under different conditions.
Table 2 Resonance frequency vs. PCB size
|PCB size||Condition||Resonance frequency|
|4×4.5 cm||Pressed with force lying on turntable rubber mate||5690 Hz
|4×4.5 cm||Lying on turntable rubber mate||5058 Hz
|4×4.5 cm||Two sides supported
|9×9 cm||Lying on turntable rubber mate
|9×9 cm||EVQ4590 free lying
|6×6 cm||EVQ4316 free lying
|9×9 cm||Two sides supported
During the practical design, a mechanical model of a PCB in a preliminary design status can be used for the first measurements. Mount the PCB in the housing before measuring the resonance frequency, and measure both in combination.
PCB vibration transfer functionality
Calculate the fast Fourier transformation (FFT) of the load currents (Figure 9), and compare these values with the resonance frequency from a PCB model. Check whether a calculated frequency reaches the PCB resonance frequency.
A PCB has a vibration transfer function, which approximately corresponds to a mechanical second-order resonance system. It consists of a mass and spring constant, defined by PCB size and stiffness (Figure 10).
Figure 10 This plot shows a simplified PCB vibration transfer function. Source: MPS
Superimpose the FFT with the PCB vibration transfer function, then check for overlapping frequencies with the PCB resonance. Consider the mechanical design and ensure that large vibration amplitudes cannot reach the area of the resonance frequency.
Reduce noise for a DC power supply circuit
Around the area of the PCB resonance frequency, vibrations are clearly audible. Avoid overlapping vibration frequencies and resonance frequency. For most PCBs, it is not possible to change the electrical excitation, but the PCB can be changed in the following ways to avoid acoustic noise:
A change in the voltage on an MLCC causes a change in its geometry due to the piezoelectric effect, resulting in a mechanical movement. This vibration generated in the MLCC is transferred to the PCB via the solder joints, which can amplify it audibly, similar to a speaker membrane. The frequency components of the vibration, the dimensions of the PCB, its mass, spring constants, and the type of installation determine whether an audible noise is generated.
When developing a DC PCB mount, take care to attach the circuit board to many distributed mounting points to increase the resonance frequency. Fastening with vibration-damping materials dampens the quality of the resonance frequency. Avoid vibration frequencies that can excite the PCB’s resonance frequency. Hardware developers should consider whether audible noise on a circuit board is distracting, such as on a phone or monitor in a quiet environment.
The frequency spectrum to be expected in the MLCC caused by the electrical load profile must be determined, and the resonance behavior of the planned, assembled PCB must be estimated. With this knowledge, the mechanics of the DC power supply circuit and PCB design can be optimized in advance. The methods described in this article can help engineers estimate whether acoustic problems are likely, and save multiple developments of PCBs.
Ralf Ohmberger is a Senior Applications Engineer at Monolithic Power Systems.
This article was originally published on EDN.