This is the third part of our five-part video series on common mistakes in flyback power supplies and how to fix them. In this video, we'll examine issues related to efficiency and thermal management. For our next common mistake, we are continuing to look at our 5 volt, 25 watt, quasi-resonant flyback example. 

Now we're looking at it from the thermal perspective. When we look at the board with the thermal camera, we notice that the synchronous rectifier, which is Q1, is getting very hot, around 85 degrees C. The first thing to suspect was synchronous rectifiers is that they're getting too hot because they're not turning on. 

If that were to happen, the power supply would still function, but all of the secondary current would run through the body diode of the synchronous rectifier, which has a much higher forward drop than if it were on. This would lead to higher power dissipation. However, looking at the gate and drain voltage wave forms of the synchronous rectifier, we see that it is indeed turning on, so that is not the issue. 

To further help debug the issue, we need more information about the synchronous rectifier effect. First the ARM is current and the FET is fairly high at around 10 amps RMS. 

Next we have some information from the FET data sheet. Noticed that the on resistance is only 3 and 1/2 mil-eons. So there should only be around 350 millowatts of conduction loss in the device. So do you know why it's getting so hot? The answer is in the table in the chart. 

Notice that the on resistance is specified with a 6 volt drive. And you can clearly see from the Miller region in the drive voltage in the chart around 5 volts. You can see that right here. With our 5 volt output providing VCC to the driver, we are not fully enhancing enhancing the fet. And on resistance as much larger than what is shown in the data sheet as a result. 

The solution is to replace the synchronous rectifier with a more appropriate one that has a lower turn on threshold. Additionally, in this case, we don't need such a high voltage rating for the synchronous FET. Replacing the original 60 volt FET with the 30 volt device shown on the right reduces the case temperature by over 15 degrees C. Notice that the on resistance is clearly specified for a 4.5 volt drive. 

For our next common mistake, we are looking at another quasi-resonant flyback. This time it is a 24 volt 36 watt output. Again, we are looking at a thermal issue. This time, it is the TVS clamping diode that is getting too hot. It is D2 in the schematic, and is getting to 96 degrees C. The efficiency is not too bad though, at 88.8%. But 96 degrees is just too hot, and we need to find a way to reduce this component temperature. 

The purpose of D2 is to absorb the energy and the leakage inductance of the transformer, and prevent over voltage spikes on the drain of the primary effect, Q1. This design has a 4 to 1 turns ratio from primary to secondary and 200 microhenrys of magnetizing inductance and transformer, and 4 microhenrys of leakage inductance. 

The voltage and current wave forms in Q1 are shown on the bottom right. Do you know why the clamping diode is getting so hot? Did you guess that the clamping voltage was too low? If so, you're correct. The difference between the clamping voltage of the TVS and the reflected output voltage is impressed across the leakage inductance when Q1 turns off. 

With a 4:1 turns ratio and 24 volt output, 96 volts is reflected back to the primary, leaving not much voltage to reset the leakage. This means it takes a longer time to reset the leakage energy, and during that leakage reset time, a portion of the magnetizing energy is also dissipated in the clamp. 

The solution is simple-- just adjust the clamp voltage higher. By replacing the 110 volt TVS with 150 volt TVS, the temperature is reduced to 71 degrees C, and the efficiency increases by 1%. Of course, you need to be careful not to over voltage the primary FET when selecting the clamp voltage. 

An RC network can be used in place of the TVS, which is cheaper, but also less accurate and can burn more power at very light loads when operated in a burst mode. Active clamp flybacks replace the clamping network with the second fet, and recycle the leakage energy for very high efficiency. But this requires a specialized controller. 

Our next problem deals with standby power loss, or loss when the power supply is operated at no load. Here we have a 5 volt 10 watt power supply. In order to meet the EU standard code of conduct specifications, the no load power consumption must be below 75 millowatts. The no load power consumption of this power supply ranges from around 150 millowatts at 80 volts AC input to 750 millowatts with the 265 volt AC input. You can see that in the chart on the bottom right. 

Take a moment to look over the schematic of this fixed frequency 100 kilohertz discontinuous flyback, and see if you can find a few things that we could change to improve the standby power consumption. You can press pause if you need more time. 

We have highlighted a few things here. First of all, the choice of controller is very important. Older, fixed frequency PWM controllers were developed before modern standby power regulations were put into place. Replacing the controller with one that has a variable frequency or a burst mode of operation at light loads will great greatly help manage frequency related losses like switching loss and gate drive loss. 

R1 and R2 are part of a startup network. But they are always present, pulling current from the high voltage input and burning power. At higher input voltages, they dominate the standby power loss. Look for controllers with very low startup currents which allow for very high startup resistor values, or better yet, look for a device that has an integrated startup circuit, which will totally eliminate these losses. 

The secondary regulating circuit can dissipate tens of millowatts of loss, which might not seem like much, but with a budget of only 75 millowatts, every millowatt counts. Controllers with primary side regulation eliminate the circuitry completely and help to achieve ultra low standby power consumption. 

Finally, the choice of the FET can also have an impact. Look for FETs with very low gate charge and low output capacitance. Here's a better solution for the same power requirements. We've replaced the older fixed frequency controller with a modern value switching flyback controller, the UCC28730. 

This device goes to a very low operating frequency at no load, as low as 30 Hertz. And this reduces the frequency related losses. It also has an integrated active startup circuit to eliminate losses in the startup circuitry. And it uses primary side regulation to eliminate the losses in the TL 431 circuit. Notice that the standby losses are now well below 20 millowatts.