Optimizing EV Inverter Design Through Advanced Waveform Analysis

Introduction

In recent years, the global push toward decarbonization has accelerated the transition from internal combustion engine vehicles to electric vehicles (EVs).
EV motor control is conducted by inverters, which convert direct current (DC) from the battery into alternating current (AC) to drive the motor. This conversion process incurs energy loss, which is a significant contributor to overall energy inefficiency in EVs. As a result, the design of the inverter plays a critical role in the vehicle’s energy utilization. Various techniques have been developed to reduce inverter energy losses, with one of the most effective being the use of high-speed switching devices such as silicon carbide (SiC) and gallium nitride (GaN). These technologies have already been implemented in automotive vehicles and are expected to see broader adoption across other industries that rely on electric motors.

Challenges

The overall energy loss of an inverter is best measured using a power analyzer. However, to specifically assess the switching loss of power devices, an oscilloscope is necessary due to the fast rise times of these devices. A common method for determining switching loss is to use the oscilloscope’s cursor function, though this process can be time-consuming. High-speed SiC and GaN devices generate significant voltage surges during signal rise times, which can potentially damage the components and surrounding circuits. Therefore, it is essential to verify that the signals remain within specified limits. Additionally, since inverters use multiple switching devices, another challenge is the need to monitor numerous test points during operation.
Furthermore, switching control instructions are typically communicated via serial bus protocols like CAN, which complicates the analysis of command timing and control system synchronization. This complexity makes it difficult to accurately assess the interaction between communication commands and switching device operations.

Solutions

In this material, we will present solutions that leverage the advanced features of the DLM5000HD High-Definition Oscilloscope for typical EV inverter evaluations.

• Switching Loss Measurement
• Surge Measurement
• Evaluation of Inverter Operations
• Measurements Related to CAN Signals

Switching Loss Measurement

In general, the switching loss of a power device is influenced by two key phases: the turn ON/OFF phases and the conduction phase. During the turn ON/OFF phases, the loss is calculated by multiplying the voltage and current. During the conduction phase, power loss is determined using the current along with constants such as the ON resistance (RDS(on)) and the saturation voltage (VCE(sat)). These complex calculations are available in the power analysis function of the DLM series oscilloscopes.

Figure 1. Switching loss overview

For example, if the device under test is a MOSFET, the switching loss can be easily calculated as power (in watts) and energy (in joules or watt-hours) by inputting the voltage, current levels, and ON resistance values. This approach allows for precise determination of losses during these switching phases :

• Turn ON loss
• Conduction loss
• Turn OFF loss

Figure 2. Determining switching loss by focusing on one cycle

The DLM series oscilloscopes can simultaneously zoom in on two separate locations. This enables detailed examination of both the turn ON and turn OFF phases of a target cycle. Figure 3 illustrates how this can be used to assess ringing and noise conditions.

Additionally, the cycle statistics function allows users to calculate losses for each cycle and display the results in a list format. Figure 3 demonstrates this function applied to the waveform shown in Figure 2. In this example, four types of parameters are computed for each cycle and listed in order. By selecting any entry in the list, the corresponding waveform is displayed, facilitating easy review of specific cycles. If an abnormality is detected in the results, this function helps quickly identify and examine the related waveform.

Figure 3. Cyclic statistical measurement function

The maximum record length is an important factor to consider when using these functions. For example, a motor rotating at 1000 rpms it completes a full rotation every 60 ms. A high sample rate of 2.5 GS/s is necessary to adequately observe fast rise times. To observe a full rotation at this sample rate a time per division of 10 ms/div is needed, which corresponds to 250 million data points of memory depth. Many oscilloscopes are ill suited for this application as they have lower memory specifications, whereas the DLM5000HD boasts 500 million points across all channels.

Moreover, with an inverter carrier frequency of 10 kHz and a motor with 4 poles, there are 1200 waveform cycles within 60 ms. When analyzing four parameters for each cycle, this totals 4800 parameters. The DLM series can manage and conduct up to 100,000 cycle by cycle parameters measurements, and these parameters can be displayed as histograms or trends.

Surge Measurement

As SiC and GaN devices operate at high speeds, significant voltage surges can occur between the drain and source due to the inductance of the device package and the wiring inductance of peripheral circuits. The inductance of the cable connecting the inverter to the motor can cause high voltage surges on the motor side, potentially leading to insulation breakdown of the motor windings. This effect is particularly pronounced with high-speed devices like SiC. Therefore, measuring motor surge voltage is essential for evaluating motor performance and ensuring reliable development and manufacturing.

When measuring a surge that reaches 1000 V using a voltage range of 250 V/div, an oscilloscope with 8-bit vertical resolution will have a minimum resolution of 10 V, since 1 division corresponds to 25 least significant bits (LSBs) of the A/D converter. Many users might find this 10 V resolution insufficient. The DLM5000HD, which features a 12-bit ADC, provides 16 times better minimum resolution, making it possible to observe the ringing effects more clearly.

Figure 4. Comparison of waveforms with different vertical resolution

It is important to note that the vertical axis accuracy of oscilloscopes generally applies to DC levels and may be less precise compared to power analyzers. When using values obtained via cursors or automatic measurements, it is advisable to apply averaging, to account for potential errors and variations.

Figure 5. Motor Surge Overview

It is crucial to verify that surges do not exceed the device’s maximum rated voltage. When measuring maximum surge values, it’s common practice to set the trigger level near the surge peak with the trigger mode set to normal to capture the highest point of the waveform. Utilizing an oscilloscope’s history function and the statistical measurement function can streamline this process.

The DLM series history function stores captured waveforms for future reference. The DLM5000HD can capture and store up to 200,000 history waveforms with a record length of 1.25 k points.

Figure 6. History Function

Additionally, the DLM series provides flexibility in waveform capture by allowing users to specify the number of captures. If concerns about dead time arise, the N single trigger mode minimizes this to 1 μs or less.

Furthermore, the history statistics function computes and displays statistical values—such as maximum, minimum, average, and standard deviation—based on all waveforms stored in history memory. It also offers a list display showing the maximum value for each waveform.

Figure 7. Statistical measurement display

Evaluation of Inverter Operations

Inverter operation typically involves measuring multiple points. For example, in the three-phase MOSFET inverter shown in Figure 7, there are 12 measurement points when assessing the voltage between each gate and source, and each drain and source of the six MOSFETs. Additionally, observing output current across the three systems adds 15 more measurement points.

Figure 8. Example of measurement points for inverters

With a standard four-channel oscilloscope, you would typically either perform multiple measurements or use multiple oscilloscopes to cover all channels simultaneously in a master/slave configuration. The first approach is inefficient, while the second method introduces time skew between the master triggering the slave instrument, making post analysis de-skew necessary.
Yokogawa addresses these challenges with its innovative DLMsync, which offers true synchronized measurements of master/slave configurations with no time skew. Setting up DLMsync is straightforward: simply connect two devices with a dedicated cable and select the appropriate option from the menu. This setup ensures synchronization at the samplingclock level with high precision, achieving an accuracy of ±50 ps.

Figure 9. DLMsync function

Measurement Related to CAN Signals

A vehicle’s Electronic Control Unit (ECU) is responsible for coordinating information between driver controls, sensors and actuators, including the tractions motor. The ECU achieves this by utilizing serial bus protocols such as Controller Area Network (CAN). An example of this is how the ECU converts accelerator controls and speed into inverter control signals.

The DLM series oscilloscopes are well suited for testing ECU signals through its CAN triggering and analysis functions. These features allow for precise triggering on CAN signal contents (ID/data) or detection of communication errors, improving timing analysis and troubleshooting. The serial bus auto setup simplifies the time-consuming process of configuring serial bus settings with a single button press. These oscilloscopes can decode and display CAN data overlaid on the waveforms and in list format.

Figure 10. CAN trigger analysis function