The audio noise generally refers to the audio signal generated by the switching power supply itself during operation, and can be heard by the human ear at an audio frequency of 20-20 kHz. When the oscillation frequency of the electronic and magnetic components is within the hearing range of the human ear, an audible signal is generated. This phenomenon has been known in the early days of power conversion research. Transformers operating at 50 and 60 Hz power frequencies often produce annoying ac noise. If the load is modulated with an audio component, a switching power converter operating at a constant ultrasonic frequency will also produce audible noise.
At low power levels, the audio signal is usually independent of the converter. However, designers may wish to reduce the acoustic emissions of their circuits. In low-power AC-DC converters, the core sheets of a 50 or 60 Hz transformer are soldered together to reduce the AC noise to an acceptable level. Similar techniques are used for ferrite transformers in high frequency switching converters.
In the past, advanced audio engineering equipment was used to study the acoustic radiation of switching power supplies. This device can measure absolute sound pressure level and sound spectrum very accurately, but human perception of sound is very subjective. It's hard to say how much sound can be heard, and it's harder to determine how much sound in a particular application would be considered unbearable noise.
Acoustic radiation is similar to electromagnetic radiation, but there is no universal benchmark for measuring hearing tolerance. Therefore, the designer can deal with the problems related to audible noise according to the following guidelines, and reduce the sound radiation of the product.
Power supply audio noise generation and suppression method
Â
One: audible noise generated by the transformer
In most flyback converter applications, the transformer is the primary source of audio noise. The noise generated by the first transformer prototype on the test board is often surprising. Using well-known appropriate structural techniques will essentially eliminate noise without adding extra cost. Pay attention to the repeatability of the finished product performance when assembling the prototype transformer.
There are some mechanisms that generate transformer noise, each of which produces a mechanical displacement that produces a sound. These mechanisms include:
Relative motion—The attraction between the two parts of the core moves it, pressing the medium that separates it.
Impact—If the surfaces of the two cores are in contact, they move in response to flux excitation, causing them to collide or scratch.
Bending—The cracks that exist only in the middle leg of the core of the EE or EI structure can cause the various parts of the core to follow the direction of attraction.
Magnetostrictive—the size of the core material varies with the flux density. The rate of change of normal power ferrite is less than 1 ppm.
Skeleton Movement—The displacement of the magnetic chip can be transmitted and amplified through the skeleton.
Coil Movement—The current in the coil creates the attractive and repulsive forces that move these wires.
The mobile sources work together to form a complex mechanical system that produces strong resonances at one or several frequencies within the human ear's hearing range. The commonly used structure of offline flyback converters below 10W generally produces resonances from 10 kHz to 20 kHz. When the fundamental frequency of the magnetic flux excitation or its harmonics passes through the mechanical resonance region, the movement emits a sound. The designer should change the load throughout to verify the audio noise, especially if dynamic loading is required.
The magnitude of the noise generated by these mechanisms is determined by the different locations in which they are located. Fortunately, designers can apply simple structural techniques to effectively attenuate the audible noise generated by various mechanisms.
The following is a brief explanation of common methods that can effectively attenuate the audible noise generated by various mechanisms.
Firstly, the transformer should be uniformly impregnated, so that the inherent gap between the coil and the coil, between the coil and the skeleton, between the skeleton and the magnetic core can be effectively filled, and the possibility of displacement of the movable member can be reduced, and the magnetic component and the line can be re-required if necessary. The plate contact surface is filled with white glue or sprayed with three anti-paint to further reduce the space of mechanical vibration and effectively reduce noise.
Try to reduce the peak magnetic flux density as far as possible, and fully consider the saturation magnetic flux density at high temperature, leaving enough margin to prevent the working curve from entering the nonlinear region, which can effectively reduce the audible noise of the transformer. Decreasing the density from 3000 Gauss to 2000 Gauss reduces the noise emitted by 5dB to 15dB.
The conditions allow the use of soft magnetic materials such as amorphous and ultrafine crystal alloys, their magnetic uniformity is much better than that of general ferrite, and the magnetostrictive effect tends to zero, so it is not sensitive to stress.
Two: audible noise generated by the capacitor
All insulating materials are deformed under the pressure of the electric field. This electrostrictive effect is proportional to the square of the electric field strength. Some insulating media also exhibit a piezoelectric effect, a linear displacement proportional to the strength of the electric field. Piezoelectric effects are often the primary means by which capacitors produce noise.
Non-linear insulating materials in inexpensive small ceramic capacitors typically contain a large proportion of barium titanate, which produces a piezoelectric effect at normal operating temperatures. Thus, these components produce more noise than the capacitance of the linear insulating component. Among switching power supplies, the capacitance in the clamp circuit with the largest voltage offset is most likely to produce audible noise.
Generally, in order to suppress electromagnetic interference and reduce the voltage stress of the device, the switching power supply generally adopts an absorbing circuit such as RC or RCD. The absorbing capacitor often uses a high-voltage ceramic capacitor, and the high-voltage ceramic capacitor is made of a nonlinear dielectric material such as barium titanate. The telescopic effect is obvious. Under the action of the periodic peak voltage, the dielectric is continuously deformed to generate audible noise.
General solution for capacitive noise
The solution is to replace the high-voltage ceramic capacitor used in the absorption circuit with a polyester film capacitor with a small electrostrictive effect, so that the noise generated by the capacitor can be substantially eliminated.
To determine if the ceramic capacitor is the main source of noise, it can be replaced with a capacitor of a different insulator. Film capacitors are a good value for money. However, it should be noted that the replacement can withstand repeated spike currents and voltage stresses.
Another price-competitive option is to replace the RCD clamp circuit with a Zener clamp circuit. The price of the Zener clamp is comparable to that of the RCD clamp, but takes up much less space and is more efficient.
Three: audible noise generated by circuit oscillation
When the power supply has a break-type oscillation during the working process, it will cause intermittent vibration of the coil core. When the oscillation frequency is close to the natural oscillation frequency around the transformer, it is easy to cause resonance phenomenon. At this time, the human ear can hear. To the audio noise.
There are many reasons for the oscillation of the circuit. The following is a brief explanation:
1: improper PCB design
A) The power high current ground wire and the control loop ground wire share the same trace. Since the PCB copper clad wire is not an ideal conductor, it can always be equivalent to an inductor or a resistor. When the power current flows through the PCB shared with the signal control loop. Line, voltage drop on the PCB, especially when multi-point grounding is used, because the nodes of the control circuit are dispersed in different positions, the voltage drop caused by the power current superimposes the disturbance on the control circuit, causing the circuit to emit noise. Point grounding can be improved.
B) The chip VCC power supply is too long, or is too close to the high dt/di high current trace. This problem can generally be improved by adding a 104 ceramic decoupling capacitor close to the chip VCC pin.
C) The grounding fault of the reference regulator ICTL431, the grounding of the same secondary reference regulator IC and the grounding of the primary IC have similar requirements, that is, they cannot be directly connected to the cold geothermal ground of the transformer. If connected together, the load capacity is reduced and the howling is proportional to the output power. When the output load is large and close to the power supply limit, the switching transformer may enter an unstable state: the switching cycle of the previous cycle is too large, the conduction time is too long, and excessive energy is transmitted through the high-frequency transformer. The energy storage inductor of DC rectification is not fully released during this period. It is judged by PWM that there is no driving signal or duty cycle that makes the switch tube conduct in the next cycle; the switch tube is off state in the whole cycle afterwards. , or the on-time is too short; the energy storage inductor is released after more than one full cycle of energy, the output voltage drops, and the duty cycle in the next cycle of the switch tube is large again... so that the transformer generates a lower frequency ( The vibration of a regular intermittent full cut-off cycle or a frequency of drastic changes in the duty cycle emits a lower frequency sound that the human ear can hear.
At the same time, the output voltage fluctuations will increase compared to normal operation. When the number of intermittent full cut-off cycles per unit time reaches a considerable proportion of the total number of cycles, the vibration frequency of the transformer originally working in the ultrasonic frequency band may be lowered, and the frequency range of the human ear may be heard, and a sharp high-frequency "whistle" is issued. call". At this time, the switching transformer works in a severe overload state, and there is always the possibility of burning - this is the origin of many "screams" before the power supply burns. I believe that some users have had similar experiences.
When the load is very low, or the load is very light, the switch tube may also have an intermittent full cut-off period. The switching transformer also works in an overload state, which is also very dangerous. This problem can be solved by presetting the dummy load at the output, but it still happens occasionally in some "saving" or high-power power supplies. When the load is not loaded or the load is too light, the back EMF generated by the transformer during operation is not well absorbed. This way the transformer will couple a lot of clutter signals to your 1.2 windings. This clutter signal includes the AC components of many different spectra. There are also many low-frequency waves. When the low-frequency wave is consistent with the natural oscillation frequency of your transformer, the circuit will form a low-frequency self-excitation. The core of the transformer does not make a sound. We know that the human hearing range is 20-20KHZ. So when we design the circuit, we usually add a frequency selection loop. To filter out low frequency components. From your schematic, you'd better add a bandpass circuit to the feedback loop to prevent low frequency self-excitation. Or you can make your switching power supply a fixed frequency.
Some other points to note about PCB traces:
The line must be as short as possible, and away from the MOSFET drain trace to prevent noise coupling, signal ground independently, as far as possible from the power ground. Optocoupler ground, Vcc ground, Y capacitively separated, feedback foot capacitance as close as possible to the IC.
Place the power supply in parallel with the ground. Keep sensitive and high-frequency traces away from high-distance power traces.
Widen the power and ground traces to reduce the impedance between the power and ground lines.
Minimize loop area consisting of drain, clamp, and transformer
Minimize loop area consisting of secondary windings, output diodes, and output filter capacitors
Increase the distance between traces to reduce capacitive coupling crosstalk.
2: Improper feedback design
For example, if the bandwidth setting is too wide and the phase margin is insufficient, the solution can try to push the bandwidth down. Some designs are designed to improve the transient response. If the bandwidth is too wide, the printing of high-frequency interference will be weakened. Blindly increasing the bandwidth is not possible. Take it.
High-power switching power supply short-circuit whistling
I believe that everyone has encountered this situation, the switching power supply suddenly shorts the power supply after full load test, sometimes it will hear the power supply whistling; or when setting the current protection, when the current is debugged to a certain position, there will be Howling, the sound of its howling is swaying, and it is very annoying. The main reasons are as follows:
When the output load is large and close to the power supply limit, the switching transformer may enter an unstable state: the switching cycle of the previous cycle is too large, the conduction time is too long, and excessive energy is transmitted through the high-frequency transformer. The energy storage inductor of DC rectification is not fully released during this period. It is judged by PWM that there is no driving signal or duty cycle which makes the switch tube conduct in the next cycle; the switch tube is cut off in the whole cycle afterwards. The state, or the on-time is too short; the energy storage inductor is released after more than one full cycle of energy, the output voltage drops, and the duty cycle in the next cycle of the switch tube is large again... so that the transformer generates a lower frequency. The vibration of a regular intermittent full cut-off cycle or a frequency with a sharp change in duty cycle produces a lower frequency sound that the human ear can hear.
At the same time, the output voltage fluctuation will increase compared with normal operation. When the number of intermittent full cut-off cycles per unit time reaches a considerable proportion of the total number of cycles, the vibration frequency of the transformer originally working in the ultrasonic frequency band may be lowered, entering a frequency range audible to the human ear, and a sharp high-frequency "whistling" call". At this time, the switching transformer works in a severe overload state, and there is a possibility of burning at all times - this is the origin of many "screams" before the power supply burns. I believe that some users have had similar experiences. No-load, or when the load is very light, the switch tube may also have an intermittent full cut-off period. The switching transformer also works in an overload state, which is also very dangerous.
For this problem, it can be solved by presetting the dummy load at the output, but it still happens occasionally in some "saving" or high-power power supplies. When the load is not loaded or the load is too light, the back EMF generated by the transformer during operation cannot be absorbed well. This way the transformer will couple a lot of clutter signals to your 1.2 windings. This clutter signal includes the AC components of many different spectra. There are also many low-frequency waves. When the low-frequency wave is consistent with the natural oscillation frequency of your transformer, the circuit will form a low-frequency self-excitation. The core of the transformer will not make a sound. We know that the human hearing range is 20-20KHZ. Therefore, when we design the circuit, we usually add a frequency selection loop. To filter out low frequency components. From your schematic, you'd better add a bandpass circuit to the feedback loop to prevent low frequency self-excitation. Or you can make your switching power supply a fixed frequency.
Audio noise generated by step load
Some switching power supplies produce audible noise when the load is tested throughout. For example, in the communication industry's test standard for switching power supplies, the dynamic load is defined as a period of 1 ms, a slope of 0.1 A/s, and a step of 25%-50% - 25% and 75% - 50% - 75%. Jump load, taking the forward converter as an example, the output inductor current is composed of the output ripple current and the step current. The frequency of the ripple current is the same as the operating frequency of the open source power supply. Generally, no audio noise is generated, and the step is The period of the current is consistent with the period of a given step load. When the output capacitance is small and the step current dt/di rate of change is too high, the method of audible noise generation is also to increase the output capacitance due to the internal volume of the power supply. Limitation, the output capacitance is generally not very large. At this time, it is also possible to try to delay the reaction time of the loop, and accordingly reduce the current change rate, thereby exerting a certain suppression effect. However, it should be noted that delaying the reaction time of the loop will make the overshoot or drop of the output voltage much larger, which is also a problem that needs to be considered.
Jinhu Weibao Trading Co., Ltd , https://www.weibaoxd.com