[TEXAS INSTRUMENTS JINGLE] Welcome to part 2 of our video series on "Common Mistakes in Flyback Power Supplies and How to Fix Them." In this video, we'll examine some common issues that can cause premature shutdown and then also look at some feedback issues. 

Here, we have a 5 volt, 25 watt quasiresonant flyback powered from a 120 volt AC 60 Hertz input. This supply is intermittently shutting down when operated at maximum load. To help debug the situation, we have probed the circuit and captured these waveforms. On channel 1, we have the voltage on VCC for the controller. On channel 2, we have the switch node voltage, which is the voltage on the drain of the primary FET, Q2. Channel 3 shows the voltage on the current sense pin of the controller. Finally, channel 4 shows the voltage on the QR pin of the controller, which is used to sense the voltage on the auxillary winding of the transformer through the divisor of R4 and R7. 

When debugging shut down issues, it's good to start with a datasheet and understanding what are all the possible shutdown mechanisms for the controller that you're using. Here on the bottom right, we have a list of the three possible shutdown mechanisms for this particular controller. As we discussed in part 1 of our discussion, if the VCC voltage drops below the turn-off level, in this case 7.5 volts, the controller could turn off. The VCC voltage on the scope image shown here is around 15 volts, so that's obviously not the source of the problem. Current limit protection can also cause a shutdown. If the voltage on the current sense pin exceeds the 500 millivolt level listed in the table, then it could shut down. However, the blue trace is safely below this level. Finally, the QR pin can trigger a shutdown event if it exceeds the over-voltage threshold, which is listed in the table. 

Superimposing the minimum 2.85 volt over-voltage level on top of the QR pin voltage reveals the source of the problem. The wave form is barely touching this threshold at several points. This explains the intermittent behavior of the shutdown. To fix the problem, we need to filter this QR signal a little bit more and lower the level to prevent any kind of false triggering that might trip this overvoltage protection. Basically, we need to adjust the values of R4, R7, and C3 in the schematic. Here, we show the wave form after adjusting these components. Notice that we are safely away from the minimum overvoltage protection trip level and the false shutdown is no longer occurring. 

Our second fall protection problem example is a 12 volt output flyback converter using a different controller. But just like the previous example, this one is exhibiting a sporadic shutdown behavior. The pink wave form on the left shows the voltage wave form on the auxiliary winding. THe over-voltage protection is set at around 24 volts, so this is not the cause of the problem here. The pink wave form on the right shows the voltage on the current sense pin, and the table at bottom right shows information from the datasheet about the current limit protection. Notice that there are two different threshold levels listed for the current sense voltage. What do you think is causing this shutdown issue? 

The maximum current sense voltage threshold of 750 millivolts limits the voltage of each current pulse from exceeding that level. If it does exceed that 750 millivolts, then it's going to terminate the gate drive for that particular pulse, but it doesn't shut down the converter. The next switching cycle is still allowed to proceed as usual. The second threshold is the overcurrent threshold. It has a minimum value of 1.35 volts and is used to protect from fault conditions. If the current sense voltage ever exceeds this level, the controller will initiate a complete shutdown of the power supply. By superimposing this level on our scope image, we see that the trailing edge spike clearly exceeds this level and is the source of the shut down. This spike is likely caused by currents induced by the high dv/dt on the drain of the FET during turnoff. Slowing down the turnoff at the FET would likely solve this problem. 

To finish up our discussion on shut down mechanisms, here are a few tips and tricks. Use an RC filter between the current sense resistor and the current sense pin of the controller to filter out high frequency spikes and prevent false overcurrent protection triggers. Typically, the corner frequency of this filter should be at least 10 times the switching frequency of the converter. As our examples have highlighted, the key to debugging shut down issues is understanding the controller. It's really important to read the datasheet. Every controller is different, so you need to read the datasheet for your particular controller. Find all the possible shutdown mechanisms for your controller and then rule them out, one-by-one. 

Our next common mistake deals with a regulation issue. This is our 5 volt, 25 watt power supply from our over-voltage shutdown example. Remember that we fixed the OVP issue by adjusting R4, R7, and C3. Now, this particular product has been deployed and been in the field for several months. And now there's a high return rate from that field of around 2%. All of these returned units were working fine initially, but then when they were operated at maximum load for a long time in a hot environment, the power supply would shut down. Troubleshooting the problem, we have plotted the output voltage of the bad units versus ambient temperature along with the plot of one good unit. Notice the output voltage of the bad units increases with increasing temperature, while the good unit does not. When the output voltage rises above around 5.9 volts, the bad units shut down because of OVP. What could be causing the bad units to lose regulation as they get hot? 

Something is preventing the correct feedback information from getting to the controller. In this case, it's the optocoupler. The value of R11 is too large and preventing the optocoupler from being able to send the air signal across the isolation boundary. To understand why this happened and how to prevent it, let's walk through the proper way to pick an optocoupler and size the pull-up resistor. The first step is to find the maximum current needed into the collector of the optocoupler. In order to do this, you must know what's inside the controller, so you need to study the datasheet. Here, we show a portion of the internal block diagram of the LM5023 used in this design. The comp pin must be driven down to 0.8 volts in order to control the PWM comparator to a 0% duty cycle. With a 42k resistor connected from an internal 5 volts source, the optocoupler must be able to pull at least 100 microamps in order to pull the comp pin to 0.8 volts. 

The second step is to determine how much current is available to drive the optocoupler on the secondary side. The TL431 can pull its output voltage down to 2.5 volts. The optocoupler will have a 1 volt drop across its internal LED. With a 5 volt output, this leaves 1.5 volts across the pull-up resistor R11. With a 5k pull-up resistor, this limits the forward optic coupler current to 300 micrograms. 

Finally, calculate the current transfer ratio, or CTR, that is needed and compare it to the optocoupler curves in the datasheet. Dividing 100 microamps by 300 microamps gives us a 0.33 CRT which is needed in order to maintain regulation. The CTR curves from the datasheet are normalized to 5 milliamps forward current and are shown here. And the red dot represents our needed CTR. Notice that we are right on the edge of what's available. As the temperature increases, the CTR curves decrease. This explains the temperature-related regulation issue. If we simply reduce R11 to 2k, we increase the available current to 750 microamps and reduce the required CTR to 0.13, which is safely below the curves. 

Not only do we need to account for temperature, but also tolerance. Most optocouplers come in a wide range of CTR tolerances, so be careful when selecting a part. Also, CTR will degrade with life. 

Also, be careful not to reduce R11 too much, because this will increase the small signal gain of the feedback loop and make it more difficult to compensate. If you want a little higher voltage to work with across R11, consider using the TLV version of the 431, which has a minimum voltage of 1.25 volts instead of 2.5 volts.