Pre-Switch Ups AI-based Inverter Efficiency

Article By : Maurizio Di Paolo Emilio

The CleanWave200 inverter now reaches an efficiency of 99.3%. The result is that the range of EVs using the inverter can be improved by up to 12%.

Pre-Switch says its 200-kW CleanWave200 inverter now reaches an efficiency of 99.3% (space-vector–modulated) at a switching frequency of 100 kHz with a flat profile as the load varies. The practical result is that the range of electric vehicles (EVs) using the inverter can be improved by up to 12%.

“We have a huge amount of data published as of today showing how we can achieve 99.3% with an accuracy of 0.01%,”  Bruce Renouard, CEO of Pre-Switch, told EE Times.

Leveraging its artificial-intelligence-based DC/AC, AC/DC soft-switching technology, Pre-Switch demonstrated how this was achieved by using only three discrete, low-cost 35-mΩ SiC FETs per switch location.

“We are primarily focused on silicon carbide, with the goal to virtually eliminate almost 100% of switching losses,” said Renouard. “And as a result, [by] limiting the switching losses, we can reduce the amount of silicon carbide needed per system by approximately 50%. The amount of SiC saved depends on the amount of switching losses the alternative system has, but it’s certainly a big chunk. And that’s a big cost saving.”

Bruce Renouard

The CleanWave200 inverter (Figure 1) offers fast switching frequencies that create a near-pure sine wave that makes electric motors efficient. The increased switching frequencies also reduce the size and cost of the DC link capacitors, in proportion to the increased switching speed,  and have the added benefit of enabling low-weight low inductance motors needed in aviation.

Power electronics needs AI
The Pre-Switch AI (artificial intelligence) solution allows users to migrate from costly, lossy, hard-switching implementations to efficient, soft-switching designs with a 10× higher switching frequency that produces a near-pure sine-wave output. The AI technology analyzes its parameters in real-time, making the necessary adjustments to the small resonant transistors, thus resulting in soft-switching even in difficult, changing environments. The Pre-Switch AI algorithm takes into account a range of parameters such as temperature, device degradation, changing input voltages, and abrupt current fluctuations.

Hard-switching simply forces the transistor to turn on and off by adding current or voltage to enable the modified states. Hard-switching is known to be very hardware-demanding on transistors, and it shortens their lifespan. The concept of soft-switching, on the other hand, uses an external circuit to avoid the overlapping of voltage and current waveforms when switching transistors.

Figure 1: CleanWaveTM evaluation system (top view). Click the image above to enlarge.

Inverters for EVs
In the automotive sector, research into the efficiency of EVs focuses on battery performance and the efficiency of the inverter and electric motor employed. Stringent automotive safety and quality standards are steering technological innovation to approaches that maximize the efficiency and autonomy of EVs while minimizing battery size and weight and reducing costs. AI is providing essential support in the push for EV autonomy and efficiency, including efforts to eliminate switching losses in order to ensure rapid transistor commutation.

Extending the range of an EV requires improving both motor and inverter efficiency know as drivetrain losses. Drivetrain losses dominate most EV losses up to about 50 mph, at which point wind resistance takes over. But drivetrain losses account for the largest share of all losses in EVs, so it is crucial to keep an eye on both the inverter and motor, with a trade-off between switching losses and higher motor efficiencies. Motor iron losses decrease as the switching frequency increases, but inverter losses increase.

Renouard pointed out that Silicon Carbide (SiC) helps the inverter at low power levels but that many EV inverters are still using SiC devices at lower switching frequencies -in the order of 10 kHz. However, increasing the switching frequency does not always solve the problem. Switching faster results in higher switching losses, which decreases the efficiency of the inverter.

Furthermore, Renouard said that if you want to try to hard-switch FETs faster and keep the inverter’s efficiency high, you need to add more FETs to reduce conduction losses in an attempt to compensate for the higher switching losses. This results in increased cost, and often the high dV/dt associated with fast switching frequencies requires thicker motor insulation and ceramic bearings to make the motors more robust. Pre-Switch addresses this challenge by incorporating AI into an FPGA that is used to precisely control the timing of the auxiliary resonant transistors, shown as S1 and S2 in Figure 2. The result is the virtual elimination of all switching losses in the main SiC transistors (Q1 and Q2).

Figure 2: Pre-Switch embeds AI into an FPGA, which precisely controls the timing of auxiliary resonant transistors S1 and S2. Click the image above to enlarge.

During each switching cycle, the timing of auxiliary resonant transistors S1 and S2 is adjusted to ensure that Q1 and Q2 have virtually zero switching losses. The algorithm calculates and minimizes dead time based on full knowledge of how and when each switch is transitioning. “Let’s look at Figure 3, which shows 20 switching cycles,” said Renouard. “At switch-on, the algorithm starts the learning process, and then at the fourth switching cycle, the first correction provided by the AI is made. In this case, a reduction in the resonant current of the inductor [shown in green] is observed. Moving on, the algorithm will adjust the inductor resonant current independently to ensure that it oscillates briefly above the load current [shown in blue]. All adjustments are fast enough to ensure accurate, smooth switching with any PWM input and can be used to create a perfect sine wave with a DC/AC inverter. The system also works perfectly in reverse.”

Figure 3: Switching cycles show power-up, the algorithm-learning process, and ongoing corrections for optimized soft switching. Click the image above to enlarge.

The CleanWave inverter evaluation system’s AI continuously adjusts the timing conditions within the switching system required to force a resonance to offset the current and voltage waveforms. This minimizes switching losses, enabling a step function in higher switching frequencies with improved inverter efficiencies.

CleanWave200 at 99.3% at 100 kHz
The published data plots system efficiency for switching speeds from 50 to 100 kHz, input voltages, power output, and current output, allowing system designers to compare Pre-Switch results with their requirements. Renouard highlighted that from an EV application perspective, the analysis allows optimization of efficiency requirements, with a net improvement of up to 12% more EV range from the same size battery (Figures 4–7). Going from 50 kHz up to 100 kHz, the efficiency is relatively constant, with a variation of about 0.2% (at 60 A). “What makes this particularly important [is that] these curves were all done at 800 V,” said Renouard.

Figure 4: Pre-Switch CleanWave200 system efficiency results. Click the image above to enlarge.


Figure 5: CleanWave efficiency at 100 kHz vs. amps per phase. Click the image above to enlarge.


Figure 6: CleanWave efficiency at 100 kHz vs. power. Click the image above to enlarge.


Figure 7: Pre-Switch switching frequencies extend EV range by 5% to 12%. Click the image above to enlarge.

“We are migrating toward a new system-on-a-chip FPGA that will enable us to run significantly faster than what we’re doing here, and we will soon have a new CleanWave200 update ready,” said Renouard.

Higher inverter frequency switching (Fsw) reduces motor losses. At 5 kHz, you have the worst case, as Renouard points out in Figure 8, for the Nissan Leaf’s silicon IGBT inverter. “If you go 20× faster, you can see that the sine wave is very clean, without any added output filters,” he said.

Figure 8: Higher inverter Fsw reduces motor losses. Click the image above to enlarge.

Renouard said that this system could also be applied to GaN solutions. While they have not yet been tested, the flexibility of the system makes it implementable. “SiC and GaN switches share some similar characteristics; both turn on very quickly and turn off very quickly,” he said. “We think that, actually, this silicon carbide algorithm would work largely as it is with GaN, and it will work with silicon IGBTs. But we will soon offer versions of the algorithm that are optimized for each switch technology ”

When considering motor benefits, switching inverters are always compromised in the application by minimizing switching frequencies in order to maintain high inverter efficiency. The result is a large amount of output ripple that is experienced by the motor, representing, in effect, inductive heating that takes place within the motor. This heat obviously has to be dissipated, which is another cost. Renouard pointed out that the near-perfect sinusoidal output provided by a 10× to 20× increase in switching frequency, resulting in much better motor efficiency and reducing the cooling required in the motor.

“When working at 100 kHz, motor efficiency is greatly improved at low torques and low to medium speed, which is where most of the driving is done,” said Renouard. “This is how we achieve the 5% to 12% increase in EV range by reducing switching losses.” The increase in engine efficiency also translates into savings in battery space as well as costs.

This article was originally published on EE Times.

Maurizio Di Paolo Emilio holds a Ph.D. in Physics and is a telecommunication engineer and journalist. He has worked on various international projects in the field of gravitational wave research. He collaborates with research institutions to design data acquisition and control systems for space applications. He is the author of several books published by Springer, as well as numerous scientific and technical publications on electronics design.

Leave a comment