 Hello, my name is Ivan. I am an Applications Engineer with TST Microelectronics and today I want to talk about stability of offline bug converters with Viper Plus switchers. And I want to give you practical guidelines on how to verify your prototype, make sure there's nothing wrong with the switching activity, that it can work in the full input voltage range and that it can deliver the full output power that's designed to do. I will start with some basics and just a little bit of theory. I'm certainly not going into the details of the theory behind feedback analysis. Then I'm going to show you how to design a simple schematic. It's very quick and an easy way using our design suite. And then I'm going to go in the lab and give you some examples of a bug converter that has poor phase margin. It's not very stable in the way that it affects the switching activity. And I will correct that, fix the compensation and show you what the switching activity looks like. So this is a typical schematic of a bug converter that uses Viper Plus. It's fairly simple. On the input we have the rectifier and the EMI filter. The Viper itself or the MOSFET of the Viper is the switching device. This diode is a freewheeling diode which ensures recirculation of the current during the off time. L2 is the energy storage device during the off time and C9 is the output filtering capacitor. As far as feedback we sampled the output voltage onto C8 via this high voltage diode which is then divided down by the resistive divider R4 and R3. That voltage is fed into the feedback input which is compared to the reference inside the air amplifier. The output of the air amplifier is the camping or compensation where we connect the compensation network. We will focus on this compensation network and ensure with the right values for the components in it that the system is stable. That's what internally it looks like. Again the air amplifier is integrated. The feedback pin is one of the inputs of the air amplifier. The other input is the reference voltage which could be different depending on the type of Viper. It could be 1.2 volts or 3.3 volts. The camping is the output of the air amplifier where we would connect the compensation network. That's just a simplified schematic of what it looks like. Viin in this case is the output voltage that's sampled onto the capacitor divided down to the air amplifier. These components are external again. Vi out is the output of the air amplifier or the camping which we will always look at and make sure that the values of those components are right so that the system has sufficient phase margin and it works correctly throughout the full specification. Okay so as I said before I'm not going to go into the theory of feedback loop analysis and modeling. I'm just giving an example and this example is shown in a number of application notes that we have along with the demo boards but that's what it is. It's just an example and it only considers one type of switching mode. This continuous conduction mode. It is different for continuous conduction mode type of buck converter so bear that in mind. Essentially what we're trying to achieve is calculate a number of parameters which will then we use for calculating the components inside the compensator. So this is the the step-by-step procedure for designing the compensation network. And in the end with all these parameters that we calculated above we will use them to calculate the values of all these resistors and capacitors as part of the feedback loop so that we have a system that meets the specification that we set out to have. So that's just an example again. It will produce a closed loop system with cross over frequency of 2.1 kHz and face margin of 72 degrees. 70 degrees is usually recommended. The theoretical minimum is for a well dampened system is 60 degrees. We recommend higher at least 70 degrees so that component tolerance or varying ambient parameters like temperature will will not affect and push the system outside of the stable zone and make it unstable. But you don't have to go through this procedure you can pause it slide by slide and go through it calculate values for your specific example. But you don't have to do that because we have a tool designed sweet that you can go ahead design your system according to the specification that you have and have the components ready calculated for your full compensator. So this is what I'm going to talk about next. Okay that's how we get to design you can find it on the website but it's very easy to just type in eds.st.com you have to create an account so I'm just going to sign in and then when log in you can access some projects that you worked on before you have some examples there are a number of things that you can design with design suite. I'm going to navigate into converter ac to dc not isolated there's a few other topology options isolated but converter is not isolated obviously so you come back converter and then you have all 37 devices available right now that you can select from ideally you would go ahead first and type in your specification. I will make this one as close as possible to the demo board that I will show you later so it's a 5.6 volt output voltage current I think is 300 milliamp thermal power you will notice all devices are filtered out devices that don't really need the specification so what's left is those that you can choose from and I will select 5.115 and there's three options of this device h for high frequency 120 kHz l for low frequency 60 kHz and x for the lowest switching frequency of the device 30 kHz you can look at the data sheet there's some very brief specification of the device itself so I'll select this one it will show you a preview of the design again link to the data sheet and product folder on the AST website this is your specification it's a wide range input 85 to 65 and then you click on start design this is what the result would look like so your design is ready essentially that's the schematic with all critical values components that are important for the design to work input rectifier in rush current limiter fuse those are not necessarily always specified because they are not so critical it depends on the type of circuit that you're gonna be designing this one in everything else that's blue color text you can modify like the input by filter capacitors you can select different diodes different capacitors different inductors and so on so this is the compensation network and you can see 1.2 nano farad in parallel with 120q ohm and 47 nano farad this is the feedback resistive divider those can be modified so what else do we have here just very briefly we have a full build material we have a simulation I wouldn't call it simulation because it's static it will be affected by changing the values on some of the input parameters for example the input voltage and you'll see the duty cycle will change from 1.5 percent to 6.5 percent and so on different current will affect that result as well you have a net temperature and so on you have more information about the operating mode to the device here you can see the device went from continuous conduction mode to discontinues right now you can expand this view and look into more details if I increase the current to 300 milliamps it will work in this in continuous conduction mode now you can see the current does not reach zero the red trace is the current okay so the other things that we have efficiency plots again those will be affected by the input parameters losses distribution there is a part chart and depending on the input power output power you can see what is consuming most or where the losses actually go I see switching most of the time is a fairly big chunk of the losses because in this case it's working very high voltage the output diode as well and the last thing is the body plot for this input output specification or operating point in this case you will see that the crossover frequency is at 113 hertz and the phase margin is 74 degrees oops sorry if I lower the power to one third of what it was you can see those things change the crossover frequency obviously is not lower because the gain was lower so you start off declining at roughly 20 dB a decade at a lower starting point lower gain then declines from there so the phase margin will increase you can see it's always about 70 77 about 70 usually so let's see what happens if we change the capacitor to slightly different capacitor let's select one with a very low esr this is 50 m that's very very low esr you can't select it's the same capacitance value 7478 micro ferret then you get out of complete by the way these are the steps so you can customize your design following step by step but I'll just hit out of complete it will automatically modify the compensation network for me and you can see the parallel capacitor went from 1.2 to narrow ferret to 560 pico ferret right now those values also slightly changed so the two automatically calculated all these things for you if you were doing it manually using the equations that I showed before that will take another iteration of recalculating everything only because you changed the output capacity now and you changed the esr of the output capacitor so you went from one part number to another part number not really paying attention to this secondary parameter which is the esr of the capacitor so that happens often in real life probably wouldn't have changed the compensation components and it won't necessarily break the buck converter but it may push it just a little bit so in some conditions it will be unstable and not behave very well and maybe you will lose regulation at some point which is the worst case that that can happen so the output will basically have to be regulated and go down to zero at one point which essentially is a failure for the power supply also you will see if I increase the capacitance say I decided to use a milli ferret because I want to have a 1.2 milli ferret because I want to have very little recalculate the output on a complete different compensation network you can see this value here now is much larger so that's how the tool is actually helping you every time you choose to redesign the output filtering capacitor change the inductor modify the feedback divider always double check that the compensation network is similar or if it changes highly recommend you to go ahead and modify the compensation network as well you can always test it in in the lab on the bench and make sure one component was changed I will show you how and what to look for but if you change specifically output capacitors or change the load or reuse the same design for a different output voltage always make sure you test it on the bench or have a look at it in the design and then double check on bench the last thing I'll show you today is how going the lab and show you some scope plots that will highlight what improper or poor switching activity looks like for when when the buck is not stable and then I will show you when I fix it all the differences between the switching activity and what is the correct functionality of a buck converter and how switching activity is is different than when face margin was insufficient and it was working poorly all right so this is the demo board that I'm using today it's a buck converter based on viper 115x this is a 30 kHz switching frequency device the compensation network is right here you have a capacitor a resistor and a capacitor I hope you can see it probably not very well unfortunately the demo board is quite beat up that's my only one that I have for buck converter available at home and we have been working from home for the last couple weeks because of the virus unfortunately but it's functional and it works well one more thing that I did is I lifted one of the pins of the inductor to connect this wire which creates a current loop where I can hook a current probe and have a look at the current in the inductor also I'm going to connect this voltage probe on the same net this is the cathode of the output freewheeling diode and I'm going to be looking at the voltage on it as well I have another probe that's connected to the output wires which I will be using for the output voltage and that's about it so let me power it up and have a look at some waveforms so I modified the compensation network of the board so that it exaggerates the unstable behavior and you will see it in a second I'm just going to power up ramp up the input voltage slowly and you'll see it okay so that's about 90 volts right now and you can see switching activity the yellow trace is the voltage on the the cathode of the freewheeling device the blue one is the output voltage and the red trace is the current in the output inductor you can see the voltage is about 6.23 volts and the load is about 40 to 50 milliamps that's the minimum I can set in this electronic load right now but it's fine so it's working nothing really stands out as far as output regulation it should regulate at about 5.6 volts it's meant to have a downstream LDO or DC-DC converter but it's fine so if I increase the voltage you will notice the activity starts to change a little bit let me pause it you will see pulse to pulse there is a difference in the current peaks that would essentially mean that the duty cycle is slightly different between those pulses it's actually not slightly it's a lot different you can hopefully you can see it the duty cycle is longer this is shorter and this one is a little longer again so that's the first sign of instability and I'm going to start increasing the load and the behavior starts to be more obvious there you go so you can see these cycles very high current peaks lower switching frequency low current peaks higher switching frequency I suspect the lower switching frequency is probably the device going into burst mode this is about 30 kHz the device is not skipping pulses you can see this irregular behavior you wouldn't expect that if everything was normal you want to see above 30 kHz which is the switching frequency of the device and no change in the peak current and now you see this pattern which is actually it's not a petal if it's kind of random as well so the device goes into burst mode out of burst mode for this power level should always be out of burst mode and very likely in this continuous mode but you see this behavior which is a sign of instability I'll certainly recommend every time you evaluate your solution to always check the waveforms and have a look if this behavior is happening this is a sign of instability and next thing I'm going to do is go ahead and fix the compensation network I will increase the capacitor that's parallel to the comp then and you will see the improvement in behavior so I modified the compensation network of this board I increased the parallel capacitor on the campaign from a few couple picofarad to about four I think it's a 470 picofarad I know that's the right value for stability because I've experimented before that but you will see an obvious improvement in the switching activity so let me ramp up the input voltage there you go that's about 85 volts at the input right now we have 100 milliamp output current and immediately you will notice the peak current in the inductor cycle to cycle as the same level that's a sign that the duty cycle does not change cycle to cycle and a sign of stable operation if I increase the load you will see that the duty cycle is increasing at one point about 220 milliamp output current the converter will go into continuous conduction mode you will see the current never reaches zero nevertheless the peak cycle to cycle do not change like they did change before so the converter works in a very stable way I think the current limit for this device will be reached at about 400 milliamp so I'm loading it quite a bit behavior is stable I will increase the voltage and you'll see there is no change in the behavior it's all stable cycle to cycle no variation in the peak current that's what you want to see and I highly encourage everyone looking at switching more power supplies have a look at the main waveforms the device current the switching nets you want to make sure the duty cycle is stable if you don't change the load the duty cycle should not change you will see this variation in frequency here so it's measuring between 28 and 31 kilohertz roughly so this is the jitter of the viper and it's an intended function it improves EMI it's randomized slight change into the switching frequency of the device to improve EMI by spreading the spectrum of the immediate emissions as far as switching it's very stable throughout the input voltage range so I lowered it below 70 volts and I think I've hit the current limit behavior stable even at low output current and I'm just varying the input voltage so for low output current low power operation you will see the frequency I don't know if you can see it on the camera but this is really 15 kilohertz this means the device is working burst mode so it's skipping one cycle in this case by effectively that's lowering the frequency to half of what it normally is just because the load does not require higher switching frequency as I'm increasing the output current the device will move out of burst mode and start switching at the typical switching frequency of 30 kilohertz right now so I increase the slide slightly when it was unstable you saw the frequency was lower as if the device was working burst mode even when the output power was much higher and it was going in and out of burst mode well in this case even if I increase the the time scale you'll see there is no change in the in the frequency or change the current peak this slide peak here is just a noise it's a it's an artifact of the switching it exists every time when the the MOSFET is turned on this is due to this capacitive discharge between the drain source so there you go that's a stable operation I'm changing input voltage between minimum maximum and changing output current between minimum maximum there is no cycle to cycle variation like we saw before and the last test that I recommend you always do before you're 100% sure that the power supply is stable is run a step response in this case I set up the electronic load to change between minimum and maximum output current and let me pause for a second here this is changing every couple of seconds so it's a slow type of test you don't have to heavily try and load to do it you just make sure that you capture the event when the load is changing and analyze the behavior during the change also edit the green trace here which is looking at the AC component of the output voltage so here is the event of change the current at the output is changing between maximum and minimum as a result there is a slight jump of the output voltage but it settles down nicely same when the volt the current changes between minimum and maximum there's a slight dip down and it settles back to DC nicely what I mean by nicely there is no ringing there is no large overshoot it's a few couple hundred millivolts changing the voltage and we saw that that's the result of the load regulation of the converter and not a result of the dynamics of the loop if you saw some type of ringing at the output that would suggest that the face margin is insufficient or if the overshoot was too high maybe the bandwidth wasn't sufficient as well so that's a nice way to analyze stability in addition to what we did before looking at the cycle to cycle switching activity and also make sure that you're looking at the full input voltage range as you're running the test so I'm about 80 volts right now and I'm going to sweep it slowly making sure that there isn't any specific input voltage level at which we see some ringing or undesired behavior so I think it's looking pretty good I'm almost to 60 volts right now the range is 260 to 85 to 65 to 85 volts and there isn't anything that concerns me really at this point so I think it works good we saw the switching cycle to cycle was stable we see that the step response is good and the settling time is not too long and there isn't any huge overshoots so everything is within specification and this is going to work well for this application I think this is a good example all right that's everything that I wanted to show you here are the four points that I want to highlight as a takeaway from the video always use e-design I highly encourage you to use it it's very easy way to for you to create a schematic and it will be final most of the time at least it will give you the best approximation for all the component values that are best for your specification and it will save you time to go and work through all the theoretical equations to calculate values for the various components always make sure that you verify once you have a prototype verify the switching of the viper always probe the voltage of the freewheeling diode always look at the inductor current this will help you avoid things like saturation in the inductor poor switching activity high output voltage repo always look at the ac on the output voltage as well it's a very good idea to test you know input voltage cases not just minimum and maximum but also in the full range of the input voltage depending on the application and of course for all the voltage corners tested full output power and minimum output power as well you want to verify at every likely operating point that the switching mode power supply is working well and finally always perform a low step test always check the power up and power down sequence make sure that the startup of the viper is not too long there isn't any overshoot when you step the power from no load to maximum load these are the tests that will certainly exaggerate and show you if there is a problem with the design stability most of the time will show there and I highly encourage you that you you try and leave no stone unturned when you verify your power supply okay thanks a lot