The story behind SiC module packaging used in Tesla model 3 – Interview of APC and Boschman

SiC power device market is clearly entering into the rapid growth phase, in particular, driving by automotive applications, which is growth with a CAGR_2017-2023 of 73%, according to the lastest SiC report, Power SiC 2018: Materials, Devices and Applications. Indeed, EV/HEV applications are not only driven the market volume but also leading the technology progress; such as power module packaging technology analysed in the report Power Module Packaging 2018: Material Market and Technology Trends. As the pioneer on the EV market, Tesla’s move attracts vast attention. Its adoption of SiC MOSFETs in the Model 3 is one of the most noticeable news for power semiconductor as well as SiC community in year 2018. In this interview, we invite you to discover the story behind the SiC MOSFET module design used by Tesla, the technology and the difficulties associated.

Yole Développement’s analyst, Hong Lin, Senior Technology & Market Analyst specialised in Compound Semiconductors, recently had the opportunity to interview Marco Koelink, Business Development Manager at APC and Boschman Technologies.

Marco Koelink: APC was founded about eight years ago by Boschman Technologies. Initially, APC’s primary focus was package development and sample-making. Over the last three years, APC has begun diversifying into assembly/production of small – medium volumes for products with high added-value. We focus on MEMs, sensors, power devices and medical/optical products.

MK: APC developed the package for the STM power MOSFET in the power module for the Tesla Model 3. Back in 2014, Boschman Technologies was the first supplier to introduce an industrial sinter press to the market. Together with leading material suppliers (e.g. Alpha, Heraeus, and Kyocera), APC pioneered the pressurized sinter process that is applied in automotive power electronics. This made us the first choice to co-develop a package for the SiC MOSFET. Initial discussions began in 2015 and most of the prototype development was done in 2016. Fine-tuning for pre-production and release took a while, since a complete sinter production infrastructure at STMicroelectronics was non-existent and many processes were still very new!

MK: APC was instrumental in designing the transfer-molding and sinter process, as well as tooling development/production. APC also designed the complete back-end assembly flow and executed the entire assembly (sintering, soldering, wire-bonding, molding, trim & form). Moreover, APC built the first prototypes in-house and optimized all assembly processes for yield and transfer to production. APC’s subsidiary, Boschman Technologies, created all production tooling and installed the first fully-operational sinter machines at STMicroelectronics.

MK: The use of SiC was new, as was the use of the sinter die-attach technology – specifically for automotive power electronics. Most importantly, our module was 2 – 3 years ahead of the crowd. At the time, the reliability of a “SiC and sintering” combination was in question. It was believed to be “visionary”, but also “risky”.

MK: Time pressure was a huge challenge. Although it became quickly evident that the device could work, many of the process were completely uncharted. Nobody really knew what the “safe-process windows” were. In order to find consistently good performance under defined process variations, numerous tests were needed. 

When you’re under the gun time-wise, suppliers sometimes deliver wrong/defective materials. To avoid this pitfall, APC had to improvise in order to achieve the desired results. Once we finished the design, industrialization at STMicroelectronics was the next step. But they had no existing infrastructure for Ag-sinter equipment, nor any process experience in production. Thus the whole industrialization process was challenging. The technical obstacles pertained mainly to the sinter process. The Ag-bond line thickness must be homogeneous in thickness and density. Good performance means high pressure (which equals high Ag-density and thus high conductivity), but this increases the risk of die-crack. To avoid die-crack, the die must be positioned accurately and perpendicular to the substrate, and the pressure must be applied homogeneously, precisely, and exactly at the right position. This ensures high performance, high reliability, and high production yield. Fortunately, Boschman’s proprietary Dynamic Insert technology was the perfect choice, and the product is now in production at sufficiently high yields.

MK: Both SiC and GaN are technologies that enable high-power and high-frequency die performance. With traditional packaging technologies, the package actually becomes limited in performance – not the die. High thermal conductivity is necessary for allowing high power output at higher die junction temperatures and low inductance, and capacitance of all package structures is required for high-frequency performance. Typically you now see Ag-sintering for die-attach and thick aluminum wires, ribbon bonding, or sintered clip bonding to reduce parasitic influences. Also, the symmetric design of the high current flow (physical layout) inside the device can be optimized to further reduce parasitic inductance.

MK: SiC is far from mainstream yet. The majority of power devices are still in Si. Industrialization on the front-end is still ongoing: new fabs are built, larger wafer diameters introduced. Also, the know-how is not widespread yet for positioning these devices in the optimum electrical environment (~circuitry). On the packaging side, there is not much difference between SiC and, for instance, Si. The design requirements for SiC can be more stringent, but all processes used to optimize SiC can also be used to optimize IGBTs for example. But of course, SiC simply puts higher demands on all design aspects, so it does push the limits a bit more.

Initially, there was some concern that SiC could not withstand the necessary pressure of the sinter process (SiC is more prone to cracking than Si), but with the pressure applied accurately and homogeneously, that concern proved negligible. There is one area where a big difference exists: in the materials for transfer molding. In order to support SiC’s higher junction temperatures, more molding materials are required with a higher glass-temperature. These materials are currently in development by several different suppliers, but only a few are proven and released. Also, if you look at discrete vs. module, CTE mismatch becomes in general more of an issue for larger devices. Although large modules can be interesting economically, discrete devices are more reliable and provide scalability at system level.

MK: Yes, I perfectly agree! To date we have worked almost exclusively on power packages that are either solely designed for automotive, or for general industrial use with automotive as the main target. Other applications include power electronics for trains, inverters for wind-power generators, and high-power, high-efficiency, high-frequency converters.

MK: APC typically works on very similar projects, but whereas a few years ago only a handful of pioneers were using this technology, now everybody in the business is jumping on it. Automotive is the main driver, with Germany and China leading. From our perspective the USA is lagging, however recently we have received an increasing number of inquiries and some development projects have started. Maybe some activities in the USA are happening outside our scope. 
A development cycle is still typically 1.5 – 2 years until full production at a captive or OSAT location. If we start pre-production in-house, it can be 6 – 9 months faster.

MK: Several things are happening:

Interviewee:

Marco Koelink holds a PhD/MSc in Applied Physics from the University of Twente and an MBA from Tilburg University. He has diverse experience with a.o. Philips and NXP in semiconductors, display technology, solid-state lighting, medical, and industrial equipment. His previous roles include Development Manager for NXP RF Power; Director of Materials Analysis for Philips Research; and Market Intelligence Manager for Philips Lighting. As a Business Development Manager, Marco is currently responsible for marketing and sales activities at APC and Boschman Technologies.

Interviewer:

Dr. Hong Lin works at Yole Développement (Yole), as a Technology and Market Analyst, Compound Semiconductors within the Power & Wireless division since 2013. She is specialized in compound semiconductors and provides technical and economic analysis. Before joining Yole Développement, she worked as R&D engineer at Newstep Technologies. She was in charge of the development of cold cathodes by PECVD for visible and UV lamp applications based on nanotechnologies. She holds a Ph.D in Physics and Chemistry of materials.