Some suggestions for feedback loop design and tuning

zbz0529

Some suggestions for feedback loop design and tuning [Copy link]

First, read the principle of automatic control (classic part) and understand the concept of zero and pole, because the power supply is a typical small phase angle system under the condition of small signal. What does it mean to understand? Then, when you see a power supply, you can write out the zero and pole immediately after seeing the feedback part. For the fixed circuit topology and control form (voltage or current), the zero and pole of the PWM part and the filtering part have corresponding formulas written out. If you want detailed mathematical calculations, go to the lead and lag compensation part of the automatic control, but this kind of calculation is generally not too accurate, but it can be used as a starting point for debugging. The final detection is generally done with an electronic load for dynamic load addition and reduction experiments (special instruments are very expensive). The general power supply sets the current change rate to 1A/uS or 5, 10A/uS. 50-100% load change, look at the voltage change, if the voltage returns to the steady-state value very slowly, it means that the phase margin is too large. If it oscillates for more than 2 cycles to return to the steady-state value, the phase margin is generally only 20 or 30 degrees, which is too small. If it is about one cycle, the phase margin is generally 50-60 degrees, which is just right. Of course, if the power supply itself is oscillating, the frequency of the oscillation is the crossover frequency of your loop, that is, the bandwidth, which means that the phase shift has reached 360 degrees at this frequency. The solution is either to reduce the bandwidth: increase the compensation capacitor value, or increase the resistance value of the feedback voltage divider. When changing these values does not work, look at other aspects of the loop, that is, add a zero point. For example, when TL431 is used as feedback, when the compensation capacitor is added very large and it does not work, you should actually add compensation (RC) to the branch in series with the optocoupler. This adds a low-frequency zero point and a high-frequency pole. The high-frequency pole is not within the loop bandwidth due to its high frequency and has no effect on the loop.
There are too many to list. What I want to say is that as long as you know the concept of zero and pole clearly, the loop problem is actually very simple (of course, it takes a lot of time to apply it correctly to the power supply. Because I didn’t have a teacher, I studied it for nearly a year). You can also use related software to simulate, but it is not easy, because the model is difficult to establish accurately. For example, if the voltage-type control counterattack (CCM working mode) is added to the TL431 with only one compensation capacitor, the result of PSPICE simulation is basically unstable, but in reality most power supplies are stable. How to explain it? The reason is that the output filter part is not actually a strict second-order system. Due to the winding resistance, high-frequency impedance, diode resistance, capacitor resistance, and especially the secondary loss, it is equivalent to a larger resistance, so the two poles will not overlap (second-order system), it becomes a series connection of two first-order systems with different frequencies, so its phase shift does not change dramatically. With the influence of other zero points, the phase will not reach 360 degrees. This is when using PSPCE simulation, a large resistance, such as 1 ohm, must be added to the diode or capacitor to get the correct result. Just a few words, I hope it will be helpful to everyone.