Noise sources and repairs in flyback power supplies

　　Although switching power supplies operate at frequencies far above the human hearing range, they can still generate audible noise under certain load conditions. There are many possible sources for audible noise. The noise can be caused by design flaws, such as an oscillating output voltage, or by noisy components such as capacitors or transformers. In some cases, the high-pitched hum or hiss you hear may sound like a fan oscillating at an unusual frequency, or may be caused by the power supply being close to external EMI sources (fluorescent lighting or power strips).

　　This article will examine the most common sources of noise in flyback power supplies and describe possible solutions. All of the procedures described below can be accomplished using a programmable AC source or autotransformer and an electronic load. Keep in mind that in some cases the noise level produced by your power supply may be very low and if the power supply will be used in a sealed enclosure, then audible noise will not be a problem.

　　Possible noise sources

　　The most common source of noise in a flyback power supply is the noisy components. This noise is usually generated by ceramic capacitors or ferrite transformer cores. Noise in ceramic capacitors is usually caused by the reverse piezoelectric effect. When a voltage is applied to a dielectric structure, it induces mechanical stress or strain, causing the material to deform. When this material deforms, it displaces the surrounding air, which generates noise.

　　Because the reverse piezoelectric effect occurs when large voltage swings occur, designers can focus on finding ceramic capacitors that experience high dV/dt swings. In a typical power supply, these capacitors include snubber capacitors, clamp capacitors, and ceramic output capacitors. To quickly determine if a ceramic capacitor is generating noise, replace it with a metal film capacitor of the same capacitance value and the appropriate voltage rating. If the noise level drops, you have found the source of the noise in your circuit.

　　If the noise source is the clamp capacitor, you can replace it completely with a metal film capacitor, or try a ceramic capacitor with a different dielectric material. Another method is to change the clamp capacitor you are using, for example, replace it with a Zener clamp circuit. If the noise problem comes from the snubber capacitor, you can replace it with a metal film capacitor, or increase the value of the series resistor to reduce the dV/dt noise on the capacitor. You can also use a ceramic capacitor with a different dielectric to see if the noise can be reduced.

　　Figure 1: Methods for repairing noisy clamp capacitors

　　If the noise problem is with the ceramic output capacitors, there are many different strategies that can be tried. One approach is to try switching to electrolytic capacitors or capacitors with other dielectric materials. Alternatively, the problem capacitor can be replaced with multiple ceramic capacitors in parallel. The reduction in the size of each capacitor will reduce its surface area accordingly, thus changing the mechanical resonance of the capacitor.

　　Managing Transformer Core Noise

　　On the other hand, noise generated by transformer cores is usually caused by magnetostriction, which is similar to the reverse piezoelectric effect. Many ferromagnetic materials change shape when subjected to a magnetic field. As the magnetic field in the transformer core changes, these materials cause the core to physically vibrate. When the vibration frequency reaches the mechanical resonant frequency of the transformer, the vibration is amplified and causes louder audible noise. In AC electrical equipment (such as transformers using a 60Hz applied magnetic field), the maximum length change occurs twice per cycle, resulting in the familiar 120Hz noise.

　　If this problem occurs in your design, make sure it is not caused by improper design before you start troubleshooting. First, confirm that the input voltage and output load provided meet the design specifications. If the power supply is operating below the specified minimum input voltage, or above the specified output load, part of the AC cycle will lose regulation, which will cause the magnetic flux in the core to increase and generate noise.

　　If the input voltage and load are within specifications, next verify that the input bulk capacitors have the correct value. If the input capacitors are too small for the application, the DC bus voltage will drop significantly between AC refresh cycles, causing the input to lose regulation for part of the AC cycle.

　　Transformers contain many moving parts, such as coils, isolation tapes, and bobbins, which make them a common source of noise. Current in the coils generates electromagnetic fields, which generate forces that cause many transformer components to mechanically vibrate. The most effective way to reduce the physical movement of transformer components is to use bonding materials or varnishes. For example, impregnating the core with varnish is a widely used method to prevent the core from vibrating with the bobbin. Although there are many varnishing techniques available from suppliers, we recommend varnish impregnation rather than vacuum impregnation because vacuum impregnation significantly increases winding capacitance, which reduces efficiency and increases EMI.

　　If your design requires a long core transformer, another strategy you can use is to use a standard core length. Long core products, such as EEL and EERL transformers, have very low mechanical resonant frequencies. This low resonant frequency tends to increase audible noise. Using a standard core length with a higher resonant frequency can alleviate this problem. However, it is important to note that if you switch to a shorter standard core, you must use a larger core size to provide sufficient winding window area.

　　Processing pulsed beams

　　Pulse bunching is another potential source of noise. Pulse bunching occurs when the conduction current pulses in a design are bunched together and then skipped in greater numbers. Pulse bunching creates frequency components in the switching mode that are usually in the audible range. Pulse bunching is most common in power supplies that use on/off control.

　　To determine if this phenomenon exists in your design, disconnect the MOSFET drain trace and insert a current loop to monitor the switching pattern of the drain current. Detailed instructions for doing this can be found at www.powerint.com/sites/default/files/PIU-104_MeasuringDrainVoltageCurrent.pdf.

　　A current probe and an oscilloscope were used to capture a set of drain switching pulses over a wide time scale while the power supply was operating under normal load. The following figure compares the waveform showing pulse bunching with the waveform with normal switching pattern. If you see pulses similar to the one on the left – a large number of pulses in a row followed by two or more skipped pulses, you may have this problem in your design.

　　Figure 2: Comparison of circuit waveforms with pulsed beam current (left) and normal switching mode (right)

　　Typically, bunching indicates that the feedback circuit is too slow, causing the controller to respond slowly. To diagnose this problem, you can first confirm that all component values ​​in the feedback circuit match the values ​​specified in the design. One solution you can try is to use a D-type optocoupler in your design. D-type optocouplers have higher gain than standard optocouplers. Another strategy is to add a feedback loop speed-up circuit to shorten the response time. This circuit will ensure that the optocoupler transistor always operates in the active region, which prevents it from saturating and increases the response speed.

　　Figure 3: Feedback loop acceleration circuit

　　in conclusion

　　Although there are many sources of audio noise in flyback power supplies , the most common culprits are often ceramic capacitors or ferrite transformer cores. If you test and find significant noise in your power supply, you can try the strategies described in this article. In most cases, you can quickly find the faulty component and solve the noise problem.