Within this course, you will investigate the behavior of a buck converter through simulation and real measurements. Simulations will be carried out using LTSpice, which you can download at Analog Devices. For the measurements, you will receive our DIY power electronics evaluation board, equipped with three separately configurable half-bridges. The hardware is controlled using a Simulink experiment.
One main goal of the lab course is to train the usage of measurement tools such as oscilloscopes, current clamps, differential voltage probes, and more.

The problem often arises that electronic assemblies such as microcontrollers, sensors, small actuators or motors have to be supplied with 5V from a higher-level voltage source, e.g. 20V. Since linear regulators, which are nothing more than variable series resistors, generate excessive losses, buck converters are often used. With the use of modern semiconductors, efficiencies of > 98% can be achieved.
For our first experiment, we design a buck converter with the following properties:

Input voltage range: 15V … 24V, nominal 20V
Regulated output voltage: 5V
Rated output power: 25W

Unfortunately, a simulation is often very similar to installing real hardware: soldering everything together and applying voltage in the hope that it will work is utopian. Or in a simulation: parameterize all components, connect them and press the run button usually doesn't work
This usually doesn't work even if we only use ideal components. We have to gradually start approaching the matter and build up the simulation step by step. And each step must confirm the results we expected at the beginning. If this is not the case, the simulation should be reduced or simplified again until the result meets our expectations in order to achieve a trustworthy result. Don't let a simulation lie to you. It is just as important as the simulation itself to check whether the results are plausible. If, for example, currents > 100A or voltages in the MV range occur in the simulation, you should not trust the result under any circumstances.

For the simulation/calculation please use the following parameters:
Switching frequency: Last two digits of your Matrikelnummer in kHz ((if the value is less than 20, it is doubled until it is greater)
Inductor: Last three digits of your Matrikelnummer in µH (if the value is less than 40, it is doubled until it is greater)
Capacitor: Use a 330µF electrolyte capacitor.

Did you know? With LTSpice, it is very easy to calculate the RMS and average values for solutions in the time domain. Use Ctrl + left-click on the signal in the result view to see the calculated values.

Note: Within LTSpice you have to configure the switch parameters first:

Diodes always incur power loss, primarily caused by the voltage drop across the device. Thus, the freewheeling diode is often replaced by a second power switch, leading to the commonly used half-bridge topology. This is a clever solution, especially when using MOSFETs as switches. MOSFETs always come with an intrinsic diode, known as the body diode. Therefore, even if the MOSFET is not activated, the body diode can serve as the freewheeling diode. The body diode, of course, exhibits the same poor power loss characteristics as the previously used diode. Therefore, we activate the MOSFET to allow the current to flow through the MOSFET instead of the diode.The rule to avoid short circuits is as follows: whenever the high-side MOSFET is turned off, the low-side MOSFET is turned on. However, this introduces another problem. Since the switching speed of the devices is not infinite, a certain amount of delay is required before turning on the low-side MOSFET after the high-side MOSFET is turned off, and vice versa. This delay is referred to as dead time. With ideal switches, we do not need any deadtime. However, you can equip the PWM generator with a deadtime function, as we will need it later

Note: Dead time can be generated by shifting the comparator values with a negative or positive offset. In the following example, Qn is the inverse PWM signal of Q, with variable dead time.

LTSpice, like any network simulation program, is capable of accounting for the real behavior of MOSFETs. However, as we know, the parasitic elements of almost all semiconductors are strongly dependent on temperature. For MOSFETs, parasitics such as the parasitic capacitance are even dependent on the current voltage value. Nevertheless, having a simple, realistic behavioral model is better than having none at all. The models can normally be downloaded from the manufacturer's website. Here irf3205s.zip you can find the model for your IRF320 MOSFET.
Datasheet: irf320.pdf

To determine the required deadtime you can delete the load and increase the deadtime as long as there are short circuit pulses. Due to the parasitic mosfet capacities which are continuously charged and discharged, a remaining current can be seen even with sufficient dead time. In reality, however, this is not as large as in our almost ideal simulation environment, due to additional damping lead resistances or the internal resistance of the source.

In a simulation, it is very easy to generate an isolated voltage source that is independent of the common ground. In reality, generating isolated sources requires significant effort, as transformers are typically needed to decouple the grounding. The problem arises with the gate supply of the high-side MOSFET, as the drain potential continuously switches between the reference ground and the input voltage. A commonly used solution is the use of a so-called bootstrap circuit. Put simply, the circuit charges a capacitor which is then connected to the mosfet independently of the ground potential. Integrated circuits such as the IRS2890DS driver we use perform this task for us and can control one half-bridge at a time. Gate drivers are often equipped with additional features. Our driver can detect overcurrents and if necessary, will turn off all MOSFETs for safety. Since overcurrents in a simulation tool will not damage your PC or notebook, you can ignore this feature in the simulation. Pull up the RFE output and connect ITRIP to ground.
Datasheet: infineon-irs2890ds-ds-v01_00-en.pdf

As a rule of thumb, the bootstrap capacitor is selected in the range of the input capacitance Ciss of the MOSFET to be driven. However, many developers start with a 100nF capacitor and check the function in the design. During the turn on phase the voltage at the gate should be as constant as possible. What do you think?
Is a 100nF capacitor enough? Gate series resistors are used either to reduce the switching speed of the mosfets or to dampen unwanted oscillations. A value between 1 and 10 ohms is often used as the starting value.

Now we have a pretty good simulation of the buck converter to calculate the operation point and get a feeling for the losses. Simulation however ist getting slower and slower. To get even closer to the real behavior of ourbuck converter we have to consider the parasitic elements of the components.

• For the inductor we can add the series resistance given within the Datasheet: 74437429203101_100u.pdf
• The frequency behavior of the capcaitor can be calculated from the measured Z(f) plot z_f_330uf.png. Many manufacturers, such as Würth, offer the data for download on their homepage. Würth
• All PCB traces that carry a high current must be included in the overall consideration. The following values can be assumed as a rule of thumb. Inductance 1µH/m for supply lines or 10nH/cm for conductor tracks.The resistance can be estimated as 0,5mOhm/cm for a 35µm track with 1cm width.

It might now be clear that the final output voltage does not depend solely on the duty cycle. In a real system, many parameters influence its behavior. This is one of the reasons why we need a control loop, which we will implement later in the lab.

When we think of reactive currents or reactive power, we usually think of AC voltage systems but not of a step-down converter. But let's imagine the following scenario: The load resistor increases, or the load is lost— for example, because you turned it off or due to a broken connection. The DC-DC converter still attempts to regulate the output voltage to 5V with a constant duty cycle. In the half-bridge configuration, the high-side MOSFET will charge the output capacitor, while the low-side MOSFET discharges the capacitor within the same cycle. This creates a circulating current between the output capacitor and the power supply. You can easily demonstrate this behavior with a simulation. However, in the lab, it might be dangerous. Why? During the discharge of the capacitor, the circuit behaves like a boost converter. If the input voltage source cannot handle negative currents, it might get damaged. In many cases, however, the reactive current generates losses, and the bulk capacitor at the input of the buck converter can absorb the reverse current.

For a quick introduction to the world of power electronics, we have developed the DIY Power Electronic PCB. It contains three half-bridges that can be controlled independently of each other. The PWM signals are generated by an ESP32 microcontroller whose frequency and duty cycle can be adjusted. The driver circuit and the mosfets correspond to the components you have already simulated.

The PCB was originally not only developed for the power electronics lecture but also to drive different motor applications. Therefore, the PCB contains more features than are necessary for your measurements.

Download: schematic_diy_power_pcb.pdf

To investigate the behavior of the buck converter, we can connect different inductor and capacitor values. They can be individually connected using the connection cables.

You can choose between the following parts from Würth Electronic:

• Inductor: 74435584700
• Inductor: 74437429203101
• Capacitor: 860241081001
• Capacitor: 860010581025

Please take a moment to answer the following questions for yourself:

• Compare the size of the capacitors in relation to the maximum allowed DC voltage and the capacitance.
• Wow, what a huge tolerance! Do you think every capacitor has such high tolerance values?
• Why is the maximum allowed ripple current a function of the frequency?
• What is the main parameter for the capacitor's lifetime?

The power PCB is controlled with the help of a Simulink experiment. The microcontroller exchanges data cyclically with the PC via the serial interface.

- Connection, COM- Port, Start, Stop

Although we designed our DIY power PCB hardware with many safety features, such as reverse polarity protection, fixed minimal deadtime, and overcurrent protection, it is, of course, still possible to easily destroy any power electronic circuit.
For every electronic circuit, there are two main golden rules:

Especially the second rule is mandatory for power electronic circuits to avoid serious damage through overvoltages or electric arcs!

Safety warnings:

Therefore follow the procedure for each test setup:

The most important task of the microcontroller is to generate the PWM signals. Thus, we first start and check the signals.

MOSFETs with an SMD package such as D2Pack are commonly used. Even for measurements, they are suitable as we can directly connect to the pins. However, please be very careful when connecting the probes to avoid creating a short circuit. And very importantly: Turn off the power supply during any changes. A single short circuit can cause significant damage to the microcontroller or the driver IC.

Here you can find information about how to use differential voltage probes and our current clamp.

Starting conditions for the next measurement:
Frequency: 5kHz
Duty: 10%

The result should look something like this:

Take a screenshot of all four measurements for your report. Label the signals and the delay times. To work with stable signals on the oscilloscope, it is generally a good idea to use a PWM signal as the trigger source.

Now is the time for us to let a current flow. In the first step, we measure the current flow with a purely resistive load, before we connect an additional inductance. In the last step, a capacitor smoothes the output voltage of our buck converter. Our first experiment is calculating the power consumtion of an ohmic load.

Starting conditions:
Frequency: 5kHz
Load resistor: 20Ohm
Duty: Ramp up to 50%

At the latest, when we look at the voltage curve, it becomes clear that we can regulate the energy flow via the duty cycle but not the maximum output voltage. This means that even if we set the average value to 5V via the duty cycle for the power supply of a microcontroller, this type of voltage curve would certainly destroy it.
For this reason, we have to smooth the voltage curve. The first idea is to use a capacitor. However, due to the conservation of energy, the voltage on a capacitor must not jump. If we tried to do this, a very high current would flow into or out of the capacitor with every switching operation.
This leaves only a inductor that can be used as a storage element. So let's smooth out the voltage.

Starting conditions:
Frequency: 5kHz
Load resistor: 20Ohm
Inductor: 47µH
Duty: Ramp up to 50%

If we ramp up the duty cycle to the target value, you may already be able to hear it. The inductance oscillates with the clock frequency, which can produce a rather unpleasant beeping noise.

Okay, it's still not ideal. If you want, you can increase the inductance value, adjust the frequency (max. 40 kHz), and observe whether you can further reduce the ripple current.

It is not possible to use a capacitor directly at the bridge output, but can of course use a capacitor after the inductor. Let's try this out.
To smooth the output voltage, we need capacitance values in the mF range. Only electrolytic capacitors are suitable for this purpose. These are very sensitive to polarity reversal. If this happens, the lab will probably be over for all participants that day

In the next task, we will try to smooth the output voltage and reduce the output current ripple using capacitors. For a higher resolution of the ripple, you can filter out the average value of a signal by switching to AC coupling on the oscilloscope. As a result, a capacitor is connected in series with the signal at the oscilloscope input, blocking the average value. It is now possible to fully zoom out the remaining ripple and determine it using the cursor.

Your buck converter is almost ready. Now it is a good time to compare the measurement results with the simulation and calculations.
Please use the following parameters:
Switching Frequency: 20kHz
Inductor: 100µF
Capacitor: 4700µF
Load: 25W, 5V, 5A
Input voltage: 20V

Voltage ripple due to the remaining current ripple might not be a problem, as we can cancel the voltage ripple with capacitors. But there is a big problem:

Consider your DC/DC converter has nothing to do because the load is turned off — for example, any device that is in standby or low energy consumption mode. The output voltage is still set by the duty cycle to 5V. If you look at your formula to calculate the current ripple, nothing really changes. Thus, the current ripple through the inductor is still the same as calculated before. Even if we turn off the load, the remaining current ripple is heating up our components.

The transition from the point at which current flows back out of the capacitor occurs from the so-called critical current mode. Here, the lower reversal point of the current is set on the zero line. If the load becomes even smaller, the current in the capacitor becomes negative and flows back to the source as reactive current.

Reactive currents are not always bad. They can be used to eliminate the reverse recovery of the body diode. If we exactly meet the CRM mode, there is even no switching loss because we turn on the low-side MOSFET at zero current. If it is possible to generate a small amount of negative current, we even have no switching losses at the high-side MOSFET, as it operates in zero-voltage switching. This type of power electronic circuit is called a resonant converter, and it is now commonly used in many applications. The theory behind resonant switching converters is part of the Master's course Power Electronic Systems.„

The reduction of the reverse recovery effect can be seen by measuring the current through the MOSFETs with a Rogowski coil. If you don't have one, it is also possible to look at the drain-source voltage of the low-side MOSFET.

Congratulations. Now you are ready and prepared for the exam!