First generation Roadster

Before reviewing the Tesla inverter design, I have to mention AC Propulsion, a company that has left a strong mark on the history of electric vehicles. Founder Al Cocconi was involved in developing the first production electric car, the GM EV1. Still, after GM "killed" the EV1, Cocconi left S.C. to found AC Propulsion, which designed and built a small number of prototype electric cars, the T-Zero, which was intended for one-person use only. There was only one small door, making entry and exit difficult. The power section was powered by lead-acid batteries connected in series. The inverter had 4-6 IGBTs connected parallel to each arm's upper and lower bridges for 24 or 36 IGBTs.

After Tesla was founded, it licensed the powertrain technology from AC Propulsion, including the IGBT single-tube parallel technology used in the first-generation Roadster inverter. The royalty payments from AC Propulsion were not stopped until Tesla produced approximately 500 powertrain systems and changed the system control from analogue to digital.

But since then, multiple parallel tubes have become a central feature of Tesla's inverter design. This has supply chain considerations in addition to path-dependent reasons. In the first decade of this century, few mass-produced automotive-grade IGBT modules were available in the market, only Infineon HybridPACK1 and others. Still, they could not meet Tesla's requirements for power output. Although there were high-current versions of industrial modules, they were not designed for automobiles, their reliability, traceability, and form factor did not meet Tesla's requirements, and no manufacturer was willing to customize expensive automotive-grade power modules for Tesla.

The earlier PEM 185 used IGBT single tubes in a standard TO247 package, with each switch consisting of 14 IGBT single tubes connected in parallel, significantly increasing over the original AC Propulsion solution. 84 IGBTs were used in the inverter, including at least the Infineon 75A IGBT IKW75N60T.

In later versions, Tesla switched to IR's custom 600V 120A AUIRGPS4067D1, which also uses 14 parallel chips. This IGBT is in a TO-247 Plus package (also known as TP-247, Super-247), which eliminates the screw holes for fixing in the TO247 package so that a larger die size can be loaded to increase the output current.

However, both IGBTs are mounted the same way, with the IGBTs bending the pins (Trim and Form) and attaching them to the power PCB at 90 degrees. At the same time, the conductive collector on the back is applied to the heat sink through an insulating thermal paste coating, and then the entire IGBT power board is fixed to the heat sink with screws. The main failure modes of this mounting method are short-circuiting of the IGBT due to cracking of the insulating thermal conductive layer after long-term use and damage to the electrolytic capacitors.

Model S/X

The Model S, which went into production in 2012, significantly improved the powertrain. The inverter design was completely abandoned in favour of a three-dimensional configuration instead of the flat-lay approach of the previous generation. The Model X, which went into production in 2015, also follows the same design, which can be called the second-generation powertrain.

The second-generation Tesla powertrain is divided into two types: Large Drive Unit (LDU) and Small Drive Unit (SDU). The former is mainly used for the rear-wheel drive in the single-motor version of Model S/X and the dual-motor high-performance 4WD version. The latter is mainly used for front and rear-wheel drive in the dual-motor normal version and front-wheel drive in the dual-motor high-performance version.

As the name implies, the LDU is larger, cylindrical and has a higher output, while the SDU is the opposite. Although the two powertrains appear in the same model, the LDU was developed earlier than the SDU and was withdrawn from the market earlier, mainly because of cost and power density considerations.

LDU

The inverter in the LDU is constructed as a trigonal prism, with each phase or each half-bridge section occupying one side of the trigonal prism. The top and bottom of the tri-prism are the high-voltage DC input section and the high-voltage AC output section, respectively. On the DC input side are three other small triangular PCBs, the drive PCBs for each phase.

The LDU uses the same as the PEM, TO247 package IKW75N60T, but the amount is more. Each switch is 16 IGBT single-tube parallel, a total of 96 IGBTs. Although the IGBTs in the LDU still need to bend the pin, its connection with the bus copper and power PCB board is greatly optimized, and the area of the power PCB board used to reduce a lot. Because of this, half of the IGBTs in each half-bridge section (middle two rows) can be fixed with a bus bar, while the other half (outer two rows) needs to be fixed with two sets of clamps.

Regarding the design of the inverter in the LDU, I still have a few questions that remain to be clarified. One is why Tesla continues to use the lower current IKW75N60T instead of the newer and higher current AUIRGPS4067D1, and the other is that there are two versions of LDU, green PCB and red PCB. Is there any difference between them?

(top) just disassembled the inverter shell of the LDU (middle) inverter details, taken from the DC side and AC side, respectively (bottom) the details of the half-bridge section, you can see that there are 8 IGBT single tubes per row, and another 8 x 2 rows of IGBT single tubes hidden under the busbar copper row and long power PCB board.

SDU

The SDU also adopts a three-dimensional structure in the inverter. Still, the design approach is very different from the PEM and LDU, making the overall structure more compact, and the power density reaches 30kW/L and 33.3kW/kg, respectively.

First, the IGBT single tube is AUIRGPS4067D1, six pieces in parallel, totalling 36 pieces. Although single IGBT costs increase, the total cost is lower because of the reduced usage. However, according to the communication with Tesla engineers, the small number of parallel IGBTs requires higher chip consistency, which makes the actual design more difficult. Therefore, Tesla has added special specification binning requirements for IGBT single tubes, which brings considerable challenges to the back-end process of IGBT manufacturing and supply chain management.

Secondly, IGBT single tubes' layout and heat dissipation have been significantly changed. The IGBT tubes in each half-bridge upper and lower arms are fixed back-to-back to the heat sink by low-temperature welding and further strengthened by clamps, forming a sandwich-like structure. Compared with LDU, the three-dimensional structure between half-bridges and the three-dimensional structure of the upper and lower bridge arms within the half-bridge makes full use of space. In contrast, low-temperature welding makes better heat dissipation. Some semiconductor suppliers of double-sided water-cooled cooling modules also use a similar thermal design to improve power density. Again, the connection of the IGBT single tube is also very different from the past. SDU does not need a power board to connect IGBT single tube but adopts the inverted way to connect with the driver board. Therefore, it is no longer necessary to bend the IGBT single tube pins, which reduces the installation cost and avoids various troubles that may be caused by this (IGBTs may fail sporadically after bending the pins, which is difficult to determine the cause and often leads to mutual accusations between IGBT suppliers and system vendors). Then, by appropriately adjusting the length of the three pins of the single tube G/D/S, they are moderately connected to the driver board and the busbar copper row. Therefore, the design and manufacturing of IGBT pins also become important.

The emergence of SDU gives Tesla more stringent mechanical, electrical and manufacturability requirements for IGBT devices.

Model 3/Y 

The powertrain is more compact than its predecessor, especially in the inverter section. One reason is that the Tesla inverter has chosen to remove the cover from the previous generation and fit close to the gearbox, thus reducing the weight and size of the inverter compared to other companies' 3-in-1 electric drive systems. But more importantly, new power devices have been chosen for the new generation of inverters, and the overall design of the inverter has been changed.

While Tesla was still optimizing the design of the SDU, the core developers were already thinking about how the next-generation powertrain should be implemented. In particular, the TO247 and TO247 Plus packages used for the core device IGBT single tubes in the first two generations of systems and three designs no longer had much potential to further increase current specifications and improve performance. At the same time, although IGBT technology continues to progress, it brings more quantitative change rather than qualitative change. In summary, the IGBT single-tube is about to reach the performance bottleneck. With this in mind, Tesla is working with power semiconductor manufacturers to explore options for new power chips and with many advanced packaging technology companies to develop new packages. The result is the TPAK (Tesla Pack) module, a revolutionary advancement that includes the following points.

First, Tesla pioneered silicon carbide chips instead of IGBT chips in mass-produced electric vehicles. Although the module cost of TPAK SiC is high, it is in line with the trend of industrial upgrading and has obtained the data of large-scale field use of silicon carbide at least three years earlier than the competitors.

Second, the TPAK package adopts a single switch module (Single Switch Module) design between single-tube and conventional modules, which not only surpasses the limitations of output current, output power, and parasitic inductance brought by the previous single-tube package but also retains the flexibility of parallel connection of multiple tubes, which allows you to choose how many TPAK modules are needed in parallel according to different inverter power output requirements. And Tesla's experience in parallel connection of multiple tubes accumulated in the past ten years can continue to be used.

Third, the TPAK module uses sintering as the connection method inside and outside. Inside the module, the chip is connected to the DBC through a silver sintering layer instead of a solder layer. On the outside of the module, the TPAK base plate is also sintered to the heat sink instead of the thermal paste coating. The two together not only bring the system's thermal capability to a higher level, but the reliability of the TPAK itself, especially the number of power cycles, is also greatly improved. In addition, the improved thermal performance means that the same size chip can output more current at a limited junction temperature or output the same current with a smaller chip, achieving chip cost reduction.

Finally, the parasitic parameters of TPAK are very small, so it can be used as a general-purpose module to put silicon carbide chips, IGBT chips, and GaN chips. This makes it easy for suppliers to share the back-end production line to produce different TPAK modules and achieve cost reduction and production increase. At the same time, only one module package is considered for the inverter design, reusing the mechanical and thermal design and reducing the cost at the inverter system level. Thus, four of these TPAK SiC modules were connected in parallel to form the top or bottom bridge of the bridge arm, and the drain and source of the modules were connected to the bus bar by laser welding, making a total of 24 TPAK modules to form the first generation of Model 3/Y inverters.

Model S/X Plaid

The Model S Plaid and Model X Plaid were delivered to the public in the middle and end of last year, respectively, so there is not much disassembly analysis on the internet now. From what we can gather, the Model S/X Plaid continues to use the same inverter design as the Model 3/Y, and even the drive and control PCBs of the former inverter are marked "Model 3". The only visible change is the inclusion of a pyrotechnic actuator in the high-voltage section of the Plaid inverter, which immediately cuts off the connection to the motor in the event of a short circuit caused by the failure of the TPAK module.

At the system level, the Model S/X Plaid differs significantly from the Model 3/Y in that the rear drive of the Model S/X Plaid is a dual motor driven by two TPAK modules in separate inverters. In addition, the motors used in the Model S/X Plaid have been improved, especially in the rotor section with carbon fibre reinforcement.

Cybertruck and second-generation Roadster

Both models are still in internal development, so information is extremely limited. From the information revealed by Elon Musk in his tweet, the motor used in the second-generation Roadster will have a faster speed than the Model S/X Plaid. I speculate that Cybertruck and the second-generation Roadster will follow the TPAK module, only to choose a new, better-performance SiC chip inside. Expect Tesla to announce more news about these two models this year.

Previous: Classification of Electric Drive Systems for Electric Vehicles

Next: What is Automotive Electronics