Analysis of the causes of howling in switching power supply testing

　　Anyone who has done development work has had this experience: testing a switching power supply or hearing the uninvited sound of high-voltage leakage or high-voltage arcing from a similar product. The sound may be loud or soft, or appear and disappear at times; the rhythm may be deep or harsh, or change erratically.

　　1. Transformer (Transformer) Poor varnish impregnation: including no varnish impregnation. Howling and causing spikes in the waveform, but generally the load capacity is normal. Special note: the greater the output power, the more severe the howling, and the performance of low power may not be obvious. I have had the experience of poor load in a 72W charger product, and found that there are strict requirements for the material of the magnetic core in this product. (The customer requirements for this product are relatively strict) In addition, when the transformer design is poor, it is also possible to vibrate and produce abnormal noise during operation.

　　2. PWM IC grounding routing error: Usually, the product will show that some parts can work normally, but some parts cannot carry load and may not be able to oscillate. Especially when using some low-power ICs, it is more likely to not work properly. I have used the SG6848 test board. Because I did not have a thorough understanding of the performance of the IC at the time, I hurriedly laid out based on experience. As a result, I could not do a wide voltage test during the experiment. Sad!

　　3. Improper routing of the optocoupler working current point: When the working current resistor of the optocoupler is connected before the secondary filter capacitor, there is a possibility of howling, especially when there is more load.

　　4. The grounding wire of the reference regulator IC TL431 is wrong: The grounding of the secondary reference regulator IC has similar requirements as the grounding of the primary IC, that is, it cannot be directly connected to the cold ground or hot ground of the transformer. If they are connected together, the load capacity will decrease and the howling sound will be proportional to the output power. The PCB in the previous article had such a mistake, which was later corrected after JACKY WANG pointed it out. When the output load is large and close to the power limit of the power supply, the switching transformer may enter an unstable state: the duty cycle of the switch tube in the previous cycle is too large, the conduction time is too long, and too much energy is transmitted through the high-frequency transformer; the energy storage inductor of the DC rectifier is not fully released in this cycle, and the PWM determines that no drive signal to turn on the switch tube is generated in the next cycle or the duty cycle is too small; the switch tube is in the cut-off state in the entire subsequent cycle, or the conduction time is too short; the energy storage inductor releases energy after more than one entire cycle, the output voltage drops, and the duty cycle of the switch tube in the next cycle will be large again... This cycle repeats, causing the transformer to vibrate at a lower frequency (regular intermittent full cut-off cycles or frequencies with drastic changes in duty cycle), emitting a lower frequency sound that can be heard by the human ear. At the same time, the output voltage fluctuation will also 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, it may even reduce the vibration frequency of the transformer originally working in the ultrasonic frequency band, enter the frequency range audible to the human ear, and emit a sharp high-frequency "whistle". At this time, the switch transformer is working in a serious overload state and may burn out at any time - this is the origin of the "screams" before many power supplies burn out. I believe some users have had similar experiences.

　　When the load is not loaded or very light, the switch tube may also have intermittent full cutoff cycles. The switch transformer also works in an overloaded state, which is also very dangerous. This problem can be solved by presetting a dummy load at the output end, but it still occasionally occurs in some "saving" or high-power power supplies. When there is no load or the load is too light, the back electromotive force generated by the transformer during operation cannot be absorbed well. In this way, the transformer will couple a lot of clutter signals to your 1.2 winding. This clutter signal includes many AC components with different spectra. There are also many low-frequency waves. When the low-frequency waves are consistent with the natural oscillation frequency of your transformer, the circuit will form low-frequency self-excitation. The magnetic core of the transformer will not make any sound. We know that the human hearing range is 20--20KHZ. So when we design the circuit, we generally add a frequency selection circuit to filter out the low-frequency components. From your schematic diagram, it is best to 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.