The King's Way: A Guide to Flyback Power Supplies and Transformer Design

I hesitated for a long time to discuss the topic of flyback power supply and transformer. Because the topic of flyback has been discussed a lot, and this topic has been discussed very thoroughly. There are also many articles summarizing the parameter design of flyback power supply. There are also enthusiastic netizens who have written their own software or spreadsheets to make the calculation foolproof according to the calculation process. But I also noticed that there are posts asking for help about problems in the flyback design process almost every day, so after thinking it over, I decided to bring up this topic again! I don’t know if I can write something new, but I will try my best to write it well. I don’t expect to be appreciated by experts, but I hope to be able to give some help to beginners.

Looking at the power supply market, no other topology is as popular as the flyback circuit , which shows that the flyback power supply has an irreplaceable position in power supply design. It is not an exaggeration to say that if you thoroughly understand the flyback power supply design, even if you don’t know anything about other topologies, it is not difficult to find a job with a monthly salary of 10K in the workplace.

outline

1. The flyback circuit is evolved from the buck-boost topology. Let's first analyze the working process of the buck-boost circuit.

Working sequence description:

At t0, Q1 is turned on, then D1 is subjected to reverse voltage and is cut off, and the inductor current increases linearly under the input voltage.

At t1, Q1 is turned off. Since the inductor current cannot change suddenly, the inductor current passes through D1 and charges C1. Under the action of the voltage across C1, the current decreases.

At time t2, Q1 is turned on and a new cycle begins.

From the waveform above, we can see that the current of inductor L1 does not reach zero during the entire working cycle. Therefore, this working mode is the continuous current CCM mode, also known as the incomplete energy transfer mode, because the energy stored in the inductor is not completely released.

From the working process, we can also know that the way of energy transfer in this topology is that when the MOS tube is turned on, energy is stored in the inductor, and when the MOS tube is turned off, the inductor releases energy to the output capacitor . The MOS tube does not directly transfer energy to the load. The entire energy transfer process is a process of storing first and then releasing. The output capacity of the entire circuit depends on the storage capacity of the inductor. We should also note that according to the direction of current flow, it can be judged that when the input and output are grounded, the output voltage is a negative voltage.

When the MOS tube is turned on, the inductor L1 bears the input voltage, and when the MOS is turned off, the inductor L1 bears the output voltage. Therefore, in steady state, the circuit must ensure that the inductor does not enter saturation, and must ensure the balance of the forward and reverse volt-second products borne by the inductor. Then:

Vin×(t1-t0)=Vout×(t2-t1). If the entire working period is T and the duty cycle is D, then: Vin×D=Vout×(1-D)

Then the relationship between output voltage and duty cycle is: Vout=Vin×D/(1-D)

At the same time, we pay attention to the voltage stress of the MOS tube and diode D1, both are Vin+Vout

In addition, because it is CCM mode, it can be seen from the current waveform that the diode has a reverse recovery problem. There is a current spike when the MOS is turned on.

The working mode above is the CCM mode with continuous current. Based on the original figure, the inductance is reduced to 80uH, and other parameters remain unchanged. The steady-state waveform of the simulation is as follows:

At t0, Q1 is turned on, and D1 is subjected to reverse voltage , which is cut off. The inductor current increases linearly from 0 under the action of the input voltage.

At t1, Q1 is turned off. Since the inductor current cannot change suddenly, the inductor current passes through D1 and charges C1. Under the action of the voltage across C1, the current decreases.

At t2, the inductor current and diode current drop to zero. D1 is cut off, and the junction capacitance and inductance of MOS begin to resonate. Therefore, it can be seen that the Vds voltage of MOS oscillates periodically.

At time t3, Q1 is turned on again and enters a new cycle.

In this working mode, because the inductor current will reach zero, it is a discontinuous current DCM mode. There is also a so-called complete energy transfer mode, because the energy stored in the inductor is completely transferred to the output. Since the diode also works in the DCM state, there is no reverse recovery problem. However, we should note that the peak current of the diode, inductor and MOS drain in the DCM mode is greater than that in the CCM mode above.

Note that the balance of the volt-second product in DCM is:

Vin×(t1-t0)=Vout(t2-t1)

It's just a matter of the waveform being reversed, just like if the probe and clip of an oscilloscope are reversed, the waveform will be reversed.

Look at the right side of the picture to see the specific definition of the waveform. Some waveforms are obtained by subtracting two points.

The waveform diagram should also be viewed in conjunction with the schematic diagram.

When the MOS is turned on, the diode D1 is subjected to reverse voltage, which is a negative voltage. When the MOS is turned off, the diode is turned on, and the forward voltage drop is very low. The reverse recovery of the diode is related to the movement of carriers in the PN junction when it is working. In DCM, because there is no current flowing through the diode, the internal carriers have completed the recombination process. Therefore, there is no reverse recovery problem. There will be a little reverse current, but it is caused by the junction capacitance.

There is a transition state between CCM and DCM mode, called CRM, which is critical mode. In this mode, MOS is turned on when the inductor current just drops to zero. This mode is the critical mode of transition from DCM to CCM. CCM will enter DCM mode when the load is light. CRM mode can avoid the reverse recovery problem of the diode. At the same time, it can also avoid the disadvantage of large current peak in deep DCM. To keep the circuit working in CRM mode all the time, variable frequency control method is required.

I also noticed that in DCM mode, after the inductor current drops to zero, the inductor will resonate with the MOS junction capacitance and discharge the MOS junction capacitance. So, is it possible to have a working mode that when the MOS junction capacitance is discharged to the lowest point, the MOS is turned on to enter the next cycle, so as to reduce the loss of MOS turning on? The answer is yes. This mode is called quasi-resonance, QR mode. It also requires variable frequency control. Whether it is PWM mode, CRM mode, or QR mode, there are now a variety of control ICs available for design.

2. We often say that the flyback circuit is derived from the buck-boost circuit. How did the buck-boost topology evolve into the flyback topology? Please see the following figure:

This is the basic buck-boost topology. Now we change the position of the MOS tube and the diode and move them to the bottom. The circuit structure becomes as follows. This circuit is completely equivalent to the above circuit.

Next, we disconnect the circuit from points A and B, and then connect a transformer to the disconnected points to get the following figure:

Why is the transformer connected here? Because in the buck-boost circuit , the bidirectional volt-second product on the inductor is equal, which will not cause the transformer to accumulate magnetic bias. We noticed that the transformer's primary and the inductor in the basic topology are in parallel, so the transformer's excitation inductance and this inductance can be combined into one. In addition, the transformer secondary output is adjusted to suit reading habits. The following figure is obtained:

This is the most typical isolated flyback circuit. Since the working process of the transformer is to store energy first and then release it, rather than just transferring energy, the essence of this transformer is a coupled inductor. Using this coupled inductor to transfer energy can not only achieve isolation between input and output, but also realize voltage conversion, rather than just relying on duty cycle to adjust the voltage.

Since this coupled inductor is not an ideal device, there is leakage inductance, and there is also stray inductance in the actual circuit. When the MOS is turned off, the energy in the leakage inductance and stray inductance will generate a very high voltage spike at the drain of the MOS, which will cause damage to the device. Therefore, we must process the leakage inductance energy, and the most common way is to add an RCD absorption circuit. Use C to temporarily store the leakage inductance energy, and use R to dissipate it.

Let's simulate the working process of the flyback circuit. When using coupled inductors for simulation, we need to know how to use coupled inductors in Saber. A simple way is to select an ideal linear transformer and set its inductance for simulation. Another way is to use the coupled inductor K model for simulation. The following figure is the circuit diagram we use for simulation. In order to let everyone see the setting of component parameters, I have displayed the key parameters of all components. In addition, because of the need for simulation, I have shared the input and output ground, and the actual circuit is of course isolated.

Careful friends may notice that the primary inductance of the transformer is 202uH, but only 200uH is involved in coupling, so 2uH is leakage inductance. The secondary is 50uH, with no leakage inductance. The inductance ratio of the transformer is 200:50, which means that the transformer turns ratio NP/NS=2:1. Set the transient scan, time 10ms, step length 10ns, and take a look at the waveform in steady state:

The following is a brief description of its working principle:

At t0, the MOS is turned on. The transformer primary current rises linearly under the input voltage , with a rising rate of Vin/l1. The transformer primary voltage is sensed to the secondary, and the rectifier diode is reversely cut off. The diode is subjected to a reverse voltage of Vin/(NP/NS)+Vout.

t1时刻，MOS关断。 变压器初级电流被强制关断。我们知道电感电流是不能突变的，而现在MOS要强制关断初级电流，那么初级电感就会在MOS关断过程中，在初级侧产生一个感应电动势。根据电磁感应定律，我们知道，这个感应电动势在原理图中是下正上负的。这个感应电动势通过变压器的绕组耦合到次级，由于次级的同名端和初级是反的。所以次级的感应电动势是上正下负。当次级的感应电动势达到输出电压时，次级整流二极管导通。初级电感在MOS开通时储存的能量，通过磁芯耦合到次级电感，然后通过次级线圈释放到次级输出 电容 中。在向输出电容中转移能量的过程中，由于次级输出电容容量很大，电压基本不变，所以次级电压被箝位在输出电压Vout，那么因为磁芯绕组电压是按匝数的比例关系，所以此时初级侧的电压也被箝位在Vout/(NS/NP)，这里为了简化分析，我们忽略了二极管的正向导通压降。

Now we introduce a very important concept, reflected voltage Vf. Reflected voltage Vf is the voltage reflected from the secondary output voltage to the primary winding according to the turns ratio of the primary and secondary windings when the secondary winding transfers energy to the output capacitor after secondary rectification. The value is: Vf=(Vout+Vd)/(NS/NP), where Vd is the forward conduction voltage drop of the diode . In this example, Vout is about 20V, Vd is about 1V, NP/NS=2, then the reflected voltage is about 42V. This can be confirmed from the waveform diagram. Then we can know from the schematic diagram that the voltage MOS withstands at this time is Vin+Vf.

也有朋友注意到了，在MOS关断的时候，Vds的波形显示，MOS上的电压远超过Vin+Vf！这是怎么回事呢？这是因为，我们的这个例子中，变压器的初级有漏感。漏感的能量是不会通过磁芯耦合到次级的。那么MOS关断过程中，漏感 电流 也是不能突变的。漏感的电流变化也会产生感应电动势，这个感应电动势因为无法被次级耦合而箝位，电压会冲的很高。那么为了避免MOS被电压击穿而损坏，所以我们在初级侧加了一个RCD吸收缓冲 电路 ，把漏感能量先储存在电容里，然后通过R消耗掉。当然，这个R不仅消耗漏感能量。因为在MOS关断时，所有绕组都共享磁芯中储存的能量。其实，留意看看，初级配上RCD吸收电路，和次级整流滤波后带一个 电阻 负载，电路结构完全是相同的。故而初级侧这时候也像一个输出绕组似的，只不过输出的电压是Vf，那么Vf也会在RCD吸收回路的R上产生功率。因此，初级侧的RCD吸收回路的R不要取值太小，以避免Vf在其上消耗过多的能量而降低效率。t3时刻，MOS再次开通，开始下一个周期。那么现在有一个问题。在一个工组周期中，我们看到，初级电感电流随着MOS的关断是被强制关断的。在MOS关断期间，初级电感电流为0，电流是不连续的。那么，是不是我们的这个电路是工作在DCM状态的呢？

In the flyback circuit, the judgment of CCM and DCM is not based on whether the primary current is continuous. Instead, it is based on the synthesis of the primary and secondary currents. As long as the primary and secondary currents are different and are zero, it is CCM mode. If there is a state where the primary and secondary currents are zero at the same time, it is DCM mode. The CRM transition mode is between the two.

So based on this, we can see from the waveform that when the MOS is turned on, the secondary current has not dropped to zero. When the MOS is turned on, the primary current does not start to rise from zero, so the circuit in this example works in CCM mode. We have said that the CCM mode is an incomplete energy transfer. In other words, the energy stored in the magnetic core is not completely released. But after entering the steady state, the newly stored energy is completely released to the secondary when the MOS is turned on in each cycle. Otherwise, the magnetic core will be saturated.

In the above circuit, if we increase the resistance of the output load and reduce the output current, the circuit operation mode can enter the DCM state. In order to keep the output voltage unchanged, the MOS drive duty cycle should be reduced a little. Other parameters remain unchanged.

Similarly, set the transient scan time to 10ms and the step length to 10ns to see the waveform in steady state:

At t0, MOS is turned on and the primary current increases linearly.

At t1, the MOS is turned off, and the primary induced electromotive force is coupled to the secondary to transfer energy to the output capacitor . The leakage inductance generates a voltage spike on the MOS . The output voltage is coupled through the winding and reflected to the primary according to the turns ratio. These are the same as in CCM mode. This state lasts until t2.

t2时刻，次级 二极管 电流，也就是次级电感电流降到了零。这意味着磁芯中的能量已经完全释放了。那么因为二管电流降到了零，二极管也就自动截止了，次级相当于开路状态，输出电压不再反射回初级了。由于此时MOS的Vds电压高于输入电压，所以在电压差的作用下，MOS的结电容和初级电感发生谐振。谐振电流给MOS的结电容放电。Vds电压开始下降，经过1/4之一个谐振周期后又开始上升。由于RCD箝位 电路 的存在，这个振荡是个阻尼振荡，幅度越来越小。

From t2 to t3, the transformer does not transfer energy to the output capacitor . The output is completely maintained by the output energy storage capacitor. At t3, the MOS is turned on again. Since the core energy has been completely released before, the inductor current is zero. Therefore, the primary current starts to rise from zero.

From the waveforms of CCM mode and DCM mode, we can see the difference between the two waveforms:

1. Transformer primary current, CCM mode is trapezoidal wave, while DCM mode is triangular wave.

2. The secondary rectifier tube current waveform is a trapezoidal wave in CCM mode and a triangular wave in DCM mode.

3. Vds waveform of MOS, CCM mode, Vds is always maintained at the platform of Vin+Vf before the next cycle is turned on. In DCM mode, Vds will drop from the platform of Vin+Vf and cause damped oscillation before the next cycle is turned on.

Therefore, as long as we have an oscilloscope, we can easily tell from the waveform whether the flyback power supply is working in CCM or DCM.

In addition, we can also get some meaningful hints from the working waveform of DCM.

For example, if we control the moment when the secondary winding current drops to zero and turn on the MOS to enter the next cycle, this can effectively utilize the duty cycle and reduce the primary current peak and RMS value.

This working mode is called CRM mode. It can be controlled by using IC with variable frequency current zero crossing detection, such as L6561MC34262.

Another way is that after the secondary current passes zero, the MOS junction capacitance and primary inductance resonate and discharge. If we let the MOS turn on when Vds drops to the lowest point, the energy loss caused by capacitive turn-on can be effectively reduced. This is the QR quasi-resonance mode mentioned earlier. There are many such control ICs now.