Cornell EV Team Design
Most modern electronics use power converters to operate properly. Learn how this Cornell Electric Vehicles (EV) project team member designed several highly efficient converters to power the team’s competition EV. The design blends efficiency with environmental friendliness.
I have been designing and building DC-DC switching power converters for my project team, Cornell Electric Vehicles. The student-led team builds a highly efficient battery electric vehicle each year to compete in the Shell Eco-Marathon (Figure 1). In this competition, teams compete against one another to build an electric vehicle that can complete a course, using the least amount of energy. This competition is not a race; it’s about being the most efficient.
I am part of the Electrical sub-team of Cornell Electric Vehicles, and we are responsible for designing, building, and testing many of the electronics that are placed on the vehicle. Some of these include a battery management system, a motor controller, a data acquisition system, and an automation system. In addition to the electronics we build, our team’s software system also has cameras and computers that are mounted on the vehicle.
These components are used to perform computer vision and motion planning as our team progresses toward a fully autonomous vehicle. All of these items have to be powered by a 24V lithium-ion polymer (LiPo) battery. However, none of our electronics actually operates at 24V, only the motor. Instead, they use a range of voltages, mainly 19V, 12V, 5V, and 3.3V, and all require different amounts of current.
Since our competition is efficiency focused, it makes sense to use a “switching power converter” (see the discussion in the next section). Although we could simply buy this product, there are two main reasons for designing our own. First, we pride ourselves on designing and building our car and its components ourselves, because it provides an exciting challenge. Beyond the competition, our main goal as a team is to learn and gain experience in our fields of interest.
Second, most off-the-shelf power converters have a wide range of input voltages and have a tunable output voltage. However, we have a specific input voltage from our 24V battery, and our electronics components, to which the battery connects, all have specific power requirements. We can therefore obtain higher efficiency by designing power converters for these specific needs. By tailoring them, we have been able to obtain efficiencies of more than 95% and reduced the size and weight of our converters.
WHAT POWER CONVERTERS DO
First off, let’s examine what power converters do. Power converters convert voltage up or down, while ideally keeping the total input power equal to the total output power. In the simplest form, a power converter is just a voltage divider, which is a simple, compact, and inexpensive way to obtain a lower voltage. However, a voltage divider or linear regulator becomes highly inefficient as the desired voltage drop increases. This is because we are only using the voltage across one of our resistors. All the current through our other resistor becomes wasted power.
A solution to this issue is “switching converters.” These converters use transistors as switches to connect the output to either the input voltage or ground. This is equivalent to pulse-width modulation (PWM), and results in an output voltage that is a square wave with a desired average value. However, we do not want an output voltage that looks like a square wave
We want to obtain a constant output. To accomplish this, we can add capacitors and inductors, which charge and discharge. This results in a constant output voltage equal to the average voltage when using PWM. Without dissipative components, such as resistors, we have now efficiently converted an input voltage down to desired output.
INITIAL DESIGN CHOICES
I designed three of the power converters (12V, 3.3V, and 5V) used on our vehicle. Before starting any of the converters, I considered several design choices. The overall goal of this project was to efficiently convert our 24V battery down to a lower value for electrical components. As already mentioned, a resistive converter, such as a linear regulator, could have been used. However, this is highly inefficient, and so switching converters were used instead.
The next decision made was the type of switching converter to use. There are many types, all of which have different topologies, resulting in different types of operations. All of our electronics required lower voltages, so a buck converter was selected. It is a common, simple, but highly effective design, so it would be the best choice for our lower-power electronics.
For our higher-power electronics, we considered using a flyback converter. This topology utilizes a transformer, which is helpful for isolating electronics from the battery in case of a malfunction. However, we decided not to use this, because selecting a core with which to make our own transformer is an arduous process that we did not have time to complete.
Another consideration was whether to implement the entire converter from scratch, or to use a regulator. A regulator is an IC that helps to implement feedback control for the converter. Although implementing the entire design ourselves would have been a great challenge, we decided to use a regulator on each of our converters. This helped lower the overall cost of PCBs, decreased their area, and ensured we finished before our competition. Time was an important factor, because all the converters had to be designed, manufactured and tested in approximately 7 months, during which time I was a full-time student.
The first converter I designed is a 24V-to-12V buck converter, which can be used to power the car’s horn or one of the motors used by the electrical automation system. This is a very high-power device, since our horn requires up to 6A of current. Even though the horn is high power, it is not used often so it contributes very little to the total energy consumption. The horn is not delicate, and therefore can handle much larger ripple current and voltage. It will work fine if the output of the converter is not completely steady.
For this design, I selected the Texas Instruments (TI) LM5085 regulator IC, which is a high-efficiency PFET controller for the converter. It has a wide range of input voltages from 4.5V to 75V, and a switching frequency of up to 1MHz. I set the switching frequency to approximately 700kHz, because a higher switching frequency generally helps to improve the efficiency. For all three power converters, I utilized their regulators’ datasheets quite heavily. The regulators are complex ICs, and their datasheets are meant to help designers by providing equations used to calculate component values.
I began by selecting the required PFET (P-type MOSFET), because I needed information from the selected part’s datasheet for later calculations. I found almost all the components, including the MOSFET on the Digi-Key Electronics website. In addition, most of the components are surface mounted, because they are small and easy to solder to the PCB that I am designing. I selected the PFET by looking at components that satisfied the 6A requirement and the maximum voltage rating of our vehicle’s batteries (around 28V). Beyond these requirements, I also looked at component cost, size, and availability. I selected the PFET to have low on resistance and low gate charge of 30mΩ and 12nC, respectively.
Next, I selected the switching frequency of the buck converter. In general, a high switching frequency increases the efficiency of a converter, because it allows you to use smaller capacitors. The buck IC has a programmable switching frequency between 100KHz and 1MHz, so I chose a higher value, around 700KHz. I could then use equations from my datasheet to select the appropriate-sized resistor to obtain this switching frequency.
SELECTING THE INDUCTOR
My next step was to select the inductor, whose value mainly controls the ripple current’s amplitude. The larger the inductor, the smaller the ripple current I expect to see through it. Ideally, the current in the inductor should not fall to zero, to maintain current conduction mode (CCM). If the current does fall to zero, the converter enters discontinuous conduction mode (DCM), which is not ideal.
I selected a maximum inductor ripple current of 2.7A. While this sounds like a huge ripple current, it is important to remember that the output current ripple will be less than this, due to additional filtering, and that it is meant to power our vehicle’s horn—which has an extremely high tolerance for ripple. I ended up picking an 8.6µH inductor that is rated for handling up to 17A. The maximum current was an important factor for selecting an inductor, because the horn draws around 6A. Therefore, it is imperative that all components on the main current path are capable of handling more than 6A, to allow some margin.
Another main component is the output capacitor, which provides filtering. The value of this capacitor should be directly proportional to the ripple current through the inductor mentioned earlier, and inversely proportional to the desired output ripple voltage and the switching frequency. Once again, since output voltage ripple is not critical, I selected a slightly smaller-valued capacitor of 4.7µF that was rated for our maximum input voltage.
An input capacitor is also used to help eliminate voltage noise. I selected a 15µF aluminum polymer capacitor. The main factor in selecting the input capacitor is the desired input voltage ripple. The larger the allowed ripple, the smaller the input capacitor can be. Finally, I selected a diode for the buck converter. A Schottky diode was recommended, and I chose one that was rated for the maximum input voltage and the maximum output current.
A schematic of the components is shown in Figure 2. The PCB software, Altium, was used for this task, because it allows creation of a layout based on the schematic (Figure 3). The schematic shows the electrical connections between components, whereas the layout shows where the components are physically placed on the board.
I designed this converter in the fall of 2020, while I was just starting my power electronics course and had not yet learned the best practices for the layout. As a result, my board was slightly larger than my later designs, for its number of components. The final dimensions were 2” × 1.6”. After fully populating the PCB, I tested it and confirmed an output of 12V when given a 24V input.
My next test, honking our car’s horn, not only was successful but also startled quite a few people, since the test was performed indoors. This test demonstrated that appropriate current was provided to the car’s horn.
The second converter I designed was a 5V-to-3.3V buck converter, intended to take a 5V input from another low-power converter (not the one discussed in the following section). This 3.3V converter is meant to power microcontrollers (MCUs) on the PCBs we designed. These MCUs are very low power, but require an input with very little voltage and current ripple. To minimize ripple, the 24V from the battery was stepped down in two stages: 24V to 5V, and then 5V to 3.3V.
A TPS64200DBVR buck regulator from TI was used for this power converter, because it was supposed to operate with extremely high efficiency at very low output current, which is perfect for an MCU. The challenging portion of this design was that the 5V board that provided input to the converter could output up to 1A of current.
To ensure that the power converter can operate for any input from this board, there needs to be a power balance. This means the total input power to the converter must equal the total power output. So, if the maximum input values are 5V and 1A, the total power is 5W, which must equal the total output power. Since the converter is outputting 3.3V, the total maximum output current it must handle is 5/3.3= 1.5A. In other words, the converter must be able to handle 1.5A of current, even though most applications require a maximum of only 150mA—a factor of 10 less.
The first part of component selection was similar to that for the 12V converter. I selected resistors to be placed along the feedback route that set the desired output voltage. The next step was the MOSFET selection. As for the previous converter design, this also required a PFET, but based on the recommendations in the regulator’s datasheet, a new MOSFET was used. The recommended specifications were a gate-source threshold voltage far less than 1.8V, and both the gate and drain-source must be able to handle the full supply voltage of 5V. The Si2323DS PFET (Vishay Siliconix) was selected, because it met these requirements. Choosing the diode was not difficult. The datasheet recommended a Schottky diode, so I selected one that could handle the circuit’s maximum current of 1.5A.
The next component to select was the inductor. The datasheet for the buck regulator recommends an inductor between 4.7 and 47µH. I needed a larger inductor to help minimize ripple current, which can affect delicate items, such as MCUs. The main issue, as mentioned, is that the converter needs to handle up to 1.5A, and in general, the more current you handle, the larger your ripple current will be.
Even though components such as MCUs never approach 1.5A draw, I also didn’t want large current ripple if this were used to power another device in the future. I therefore determined that a 56µH inductor would be appropriate. (This is the next standard-valued inductor after 47µH.) This limits the current ripple to 40.5mA for a device that draws the maximum 1.5A.
For the input capacitor, I was able to select much smaller values. The other 24V-to-5V power converter that connects to this has a very low output voltage ripple of 100mV, meaning that the board’s capacitors do not have to do as much work, and smaller components could be used. I calculated a value of 1µF, but the datasheet recommended a minimum value of 10µF. I therefore chose a small, 10µF ceramic capacitor for the input capacitor.
The last major component to select was the output capacitor. Unfortunately, this had to be significantly larger than the input capacitor. Switching converters are highly efficient, but one side effect of switching is noise. Opening and closing MOSFETs quickly introduces noise into the signal, requiring more filtering. I wanted to obtain a peak-to-peak output voltage ripple of 100mV. However, obtaining this for the maximum 1.5A load would require a 1.5mF output capacitor. This is enormous and unrealistic. Instead, I selected a 470mF capacitor, which is still quite large. This is not required for expected small loads, but becomes necessary if the 3.3V converter were used for a higher-power device in the future.
I added all these components to the schematic (Figure 4), and then created the layout (Figure 5). This time I did a much better job, by reducing the lengths of the paths that had high current or high-frequency signals. I achieved this by utilizing both layers of the PCB, allowing components to be mounted on the front and back, and connected with vias. Mounting components on both sides means they can be placed back-to-back. The overall size was 0.6″ × 0.7″.
After soldering the board, I began testing. I had some trouble with the diode, which seemed to occasionally break internally as I was soldering it. Despite this, I was able to get this converter to work after some careful soldering. As before, I made sure it could output the appropriate 3.3V, and then used it to power the MCU on our team’s motor controller board.
The members of the software team are currently working on making our car autonomous, so having cameras mounted on the vehicle will be essential for detecting objects around it. The cameras are normally powered via USB 3.0. Instead of directly powering the cameras, we powered a large USB hub to which all the cameras could connect directly. This avoided having to hack together multiple power and ground wires to each camera.
The last converter I designed was a high-current 24V-to-5V buck converter, to power the cameras used for computer vision. Unlike the 5V converter used to power the 3.3V buck converter I designed, this one is meant for larger currents. The previous one had a maximum output of 1A.
Each camera draws a maximum of 0.9A, and our team currently has four cameras, giving a total maximum of 3.6A. Rounding this up to 4.5A gives us a large safety net, but also allows us the option to add one more camera to the car, if needed. If more than five cameras are needed, we can then populate another 5V buck converter, and buy another USB hub.
An LM5166 regulator (from TI) was used for this power converter. This regulator required some extra support circuitry on the board, which made the final product slightly larger, but is supposed to operate at high efficiency with the expected load current of 2A to 4.5A. It also had excellent documentation. I began by selecting a middle-ranged switching frequency of 250kHz, which is controlled by a timing resistor. Using the documentation, I calculated a value of 12.5kΩ, and then selected a standard valued resistor of 12.4kΩ.
After this, I selected a 12µH inductor. As before, the inductor must be able to handle the maximum current plus the maximum ripple. Since operation is at a higher current than the 3.3V buck converter, a higher ripple current is expected.
In addition, our team’s cameras are not as delicate as the MCUs on our other boards. We could therefore use a standard-valued inductor with a lower value than previously selected. After considering this and using equations from the datasheet, the value of 12µH was obtained.
I decided to try a different approach to both the input and output capacitors on this board. Previously, I was generally selecting large-valued aluminum or electrolytic capacitors. One downside of these is that they have large equivalent series resistance (ESR).
In real-world design, capacitors are not ideal and have some associated resistance, resulting in extra losses. However, smaller capacitors, such as surface-mounted ceramic capacitors, have far less ESR. Instead of using one large capacitor on the input and output, I put several smaller capacitors in parallel, which adds their capacitances and results in a larger, desired capacitance. A larger value is obtained by using many smaller capacitors.
Although there were many other minor components, the last major one was the MOSFETs. As specified on the controller’s datasheet, the most important consideration in this case was the gate charge. Ideally, the MOFETs gate charge is as small as possible to help minimize losses. I selected MOSFETs with only 13nC of gate charge, which is quite small, considering the large amounts of current they are expected to handle. The schematic and layout for the 5V buck converter are shown in Figure 6 and Figure 7, respectively.
COVID-19 presented some unique challenges while I was finishing my designs. Online classes limited access to my project team’s normal laboratory, making it difficult to test boards and solder components. However, I was able to circumvent this by setting up a small work station in my apartment, where I could work safely using my own soldering supplies.
Also, as a result of the chip shortage, many of the controller integrated circuits and MOSFETs were out of stock. This meant that to make extra copies of my circuit boards, I had to go back and find similar components that met the necessary specifications. Although this was a tedious task, I was able to find in-stock replacement components.
This was a challenging project that helped me explore my interests in power electronics. I was able to help improve my project team’s electric vehicle by creating custom devices to power everything on our car more efficiently (Figure 8). Despite challenges and setbacks, I successfully completed all three converter designs, and they are now being used on our vehicle.
For detailed article references and additional resources go to:
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • MAY 2022 #382 – Get a PDF of the issueSponsor this Article
Circuit Cellar's editorial team comprises professional engineers, technical editors, and digital media specialists. You can reach the Editorial Department at firstname.lastname@example.org, @circuitcellar, and facebook.com/circuitcellar