Self-oscillation principle and compensation solution of error amplifier in switching power supply IC

At present, with the widespread application of switching power supplies, the control IC plays an important role as the heart of the switching power supply. The control IC of the switching power supply generally contains an error amplifier, which is used to amplify the offset of the output voltage to control the action of the main switching circuit and achieve a regulated output. The error amplifier itself is an operational amplifier, and negative feedback is added in actual use. Due to the influence of external components and PCB and other factors, the error amplifier sometimes generates self-oscillation, which makes the switching power supply unable to work normally. The author analyzes the principle of self-oscillation when the error amplifier adds negative feedback, and designs an external compensation circuit using the UC3875 control IC as an example, and conducts experimental verification.

　　1 The principle of self-oscillation generated by the error amplifier

　　1.1 Causes of self-oscillation

　　The expression of the closed-loop gain G of the error amplifier after adding negative feedback is:

　　Where A is the open-loop gain, F is the feedback factor, and AF is the loop gain.

　　From the above formula, we can know that when 1+FA approaches 0, |G| =∞. This means that there will be waveform output even if there is no signal input, thus generating self-oscillation.

　　The gain and phase shift of the amplifier will change with frequency. When the frequency becomes higher or lower, the output signal and the feedback signal will produce additional phase shift. If the additional phase shift reaches ±180°, the feedback signal is in phase with the input signal, and the negative feedback becomes positive feedback. The feedback signal is strengthened. When the feedback signal is greater than the net input signal, there is a signal output even if the input signal is removed, so self-oscillation occurs.

　　Right now:

　　There are many natural poles inside an actual operational amplifier. The additional phase shift caused by them will cause the output phase to shift more than -180°, and when negative feedback is used, the amplifier will produce self-oscillation. Therefore, most operational amplifiers have compensation ports or are directly compensated internally for ease of use. These internally compensated operational amplifiers are generally compensated to have only one pole above 0 dB gain, and will not self-oscillate when used alone even as a unity-gain amplifier.

　　1.2 Determination of the stability of negative feedback amplifier circuit

　　The method to judge self-oscillation is to see whether it meets the phase condition. Only when the phase condition is met can self-oscillation occur. That is, if the loop gain |FA|≥1 when the additional phase shift φ=±180°, then the circuit will generate self-oscillation.

　　On the contrary, if the loop gain |FA| is less than 1 when φ=±180°, the circuit will not produce self-oscillation.

　　2 UC3875 error amplifier circuit

　　2.1 UC3875 Error Amplifier Circuit Structure

　　UC3875 is a phase-shifted full-bridge soft-switch controller produced by TI, which is widely used in high-power switching power supplies with ZVS and ZCS topologies. It contains an error amplifier, and the output voltage of the error amplifier output terminal is compared with the output voltage of the ramp generator to generate a phase-shift signal. Its AB and CD outputs can set the dead time respectively, which is very suitable for full-bridge resonant switching power supplies. The circuit connection of the error amplifier part of UC3875 used in this article is shown in Figure 1.

　　The non-inverting input of the error amplifier is connected to the reference voltage, the output is fed back to the reverse input through a 150 kΩ resistor, and the inverting input is connected to the output voltage sampling circuit through a 470 kΩ resistor.

　　When the switching power supply was debugged and its output was measured, it was found that the output was very unstable. Then the UC3875 control output terminals OUTA and OUTC were observed with an oscilloscope, as shown in Figure 2. It was found that the output phase-shifted signal had a large amplitude jitter, causing the switching power supply output to become unstable. Subsequently, when the output of the error amplifier was observed, it was found that the error amplifier had oscillations, and an unstable sinusoidal signal was generated at the output (Figure 3). Since the output of the error amplifier is compared with the output voltage of the ramp generator to generate a phase-shifted control signal, the output control signal of the UC3875 will have a large jitter.

　　2.2 Analysis of UC3875 Error Amplifier Oscillation Phenomenon

　　According to the data sheet of UC3875, its typical bandwidth and open-loop gain are 11 MHz and 90 dB respectively. The error amplifier of most control ICs has been internally phase compensated, and the compensation is such that even when the closed-loop gain is 0 dB (when the feedback is the largest), no oscillation will occur. However, in actual use, due to the influence of external components and other factors, new poles may be generated, causing the circuit to have an additional phase shift exceeding -180°, thus causing oscillation.

　　According to the previously observed self-excited oscillation waveform at the output of the error amplifier, its oscillation frequency is approximately around 50 kHz. At this frequency, the additional phase φ≥-180°, and its open-loop gain is greater than 0 dB. Based on these conditions, it can be estimated that the pole frequency generated by the external circuit should be around 5 kHz, and it is added to the frequency characteristic diagram of the gain and phase of the error amplifier to obtain Figure 4. Among them, P1 is the pole set during internal compensation, and P2 is the pole generated by the external circuit (the solid line in the figure represents the gain, and the dotted line represents the phase, the same as Figure 6).

2.3 Design of external compensation network

　　Since the zero point can produce a leading phase shift, it can offset the lagging phase shift produced by the pole. Therefore, if a compensation network is added to the circuit, setting a zero point will be able to offset the pole produced by the external circuit, thereby suppressing the self-oscillation of the amplifier. Since the error amplifier does not have a compensation port, the compensation network needs to be set externally. As shown in Figure 5, a capacitor Cf is connected in parallel across the feedback resistor Rf, thereby generating a zero point. By properly setting the frequency of this zero point, the additional phase shift produced by the new pole can be offset, so that the total phase shift does not exceed -180°. Because the estimated external pole frequency is 5 kHz, the zero point frequency should be set near 5 kHz.

　　According to the formula:

　　Substituting fz=5 kHz, we get Cf=212 pF.

　　Select Cf as 220 pF. Since the capacitor Cf is placed in the circuit, a new pole will be generated, and its frequency is:

　　Substituting the numerical values ​​into the above equation, we can get the new pole frequency of 1.5 MHz, which is equivalent to moving the external pole P2 to the position of P2′ as shown in Figure 7.

　　It can be seen from Figure 6 that although there are two poles above the gain of 0 dB, when the gain drops to 0 dB, the phase shift still does not exceed -180°, so the self-oscillation condition is destroyed and the circuit will not produce self-oscillation. At the same time, it can be seen from the figure that the bandwidth loss of the amplifier is very small when this method is used. However, according to formula (3), it can be seen that the frequency of the new pole is related to the gain of the amplifier. If the amplifier gain is too small, the compensation effect will be greatly affected because the pole moves too small to the high frequency. In particular, when used as a voltage follower (at this time, the amplifier output is directly connected to the inverting input terminal and the feedback resistance is zero), the frequency of the new pole will not move to the high frequency, and this circuit will have no effect at all. Due to the influence of various factors and estimation errors, the actual characteristic curve will have some gaps with the theory, so the set zero point needs to be adjusted through experiments (later experiments also confirmed this).

　　3 Experimental verification of external compensation network

　　The experimental circuit is connected as shown in Figure 5. Cf with capacitance values ​​of 22 pF, 100 pF, and 220 pF are connected to the circuit respectively, and the control output waveform of UC3875 is observed. Figure 7 shows the waveform when using 22 pF capacitance. In this circuit, since the zero point is set far away from the pole, the waveform jitter is slightly weakened, but its jitter amplitude is still large.

　　Figure 8 shows the waveform when using a 100 pF capacitor, and it can be seen that the jitter amplitude is greatly reduced. At this time, the zero frequency set in the circuit is relatively close to the pole position, which has already reflected the effect of oscillation suppression, but the output oscillation amplitude is still obvious.

　　When the capacitor is replaced with 220 pF, the jitter of the waveform basically disappears. The zero point in the circuit is near the pole position estimated above. By carefully observing the waveform on the oscilloscope, extremely weak jitter can still be found. This shows that the actual pole position is somewhat different from the previous estimated value. Therefore, when the actual circuit situation is not very clear, the compensation network parameters obtained by estimation need to be verified and debugged in actual experiments.

　　Considering the influence of various factors in practical applications and the estimated errors, it is necessary to maintain a certain margin when designing the compensation network. Therefore, Cf is selected as 470 pF. After it is connected to the circuit, the waveform of the output control of UC3875 is shown in Figure 9. The output waveform jitter has completely disappeared, and UC3875 has been working stably. After observing the output end of the error amplifier, it is found that its output has become a straight line. The oscillation of its output voltage has completely disappeared.

　　4 Conclusion

　　Although many common operational amplifiers and switching power supply control ICs have internal error amplifiers that are phase compensated, sometimes new poles are generated externally, making the circuit unstable. The method used by the author is to use a zero point to offset the new pole, so that it can work stably. This method basically does not lose the bandwidth of the operational amplifier and can achieve good results. There is a prerequisite for using this compensation method, that is, the amplifier needs to have a relatively large closed-loop gain, so that it can produce better results. In switching power supply applications, in order to obtain a stable output voltage, the closed-loop gain of the internal error amplifier is generally large, so this method is very suitable.