Prior to our upcoming Workshop in Detroit on 9-10 April, which will include participation from Obsidian Sensors Company, we had the opportunity to meet with CEO Dr. John Hong. Obsidian’s offices are in San Diego, California. Obsidian Sensors use uncooled microbolometers on glass.
Hong earned his PhD in electrical engineering from Caltech, and his B.S. in electrical engineering from MIT. He has over 35 years’ experience in R&D and product development involving optics, MEMS, and semiconductors.
Prior to co-founding Obsidian Sensors, he was the General Manager of Qualcomm’s MEMS technologies, and a VP of engineering with Corporate R&D. He was Chief Technologist for astrophysics with JPL/NASA. He is a fellow of Optica (formerly OSA).
DVN: Could you tell us more about the history of Obsidian Sensors?
Dr John Hong: We formed the company in 2017 as a spinoff from Qualcomm, our first investor. Some of the know-how comes from the work we did at Qualcomm in the development of low-power reflective displays. That was quite unique because the manufacturing required the fabrication of MEMS devices on a glass sheet at a flat panel display fab. When the development was terminated, the executive team allowed us to explore microbolometers as another application. A successful proof of concept led to a seed investment followed by the typical steps in a startup.
DVN: Who are the main investors in Obsidian Sensors?
J.H.: Our current investors are Qualcomm, Innolux (our manufacturing partner), Hyundai Motor Company, Hyundai Mobis, STK Walden (a VC), Himax Technologies, and some individuals.
DVN: What types of thermal sensors do you develop?
J.H.: We make uncooled microbolometers, which are infrared sensors that operate in the longwave infrared band (LWIR, 8-14µm). Unlike shorter-wavelength infrared bands like SWIR and NIR, LWIR imaging does not need illumination and directly uses the radiation emitted by all warm objects. Like any imaging modality, resolution (number of pixels) is a very important parameter for users. We are currently shipping VGA (640 × 480) resolution cameras, and have a roadmap to offer higher resolution products including SVGA (800 × 600) later this year and SXGA (1,280 × 1,024) next year. Our unique approach in device design allows us to scale up the resolution much faster and much more economically than our competition.
DVN: What are the major markets or industries targeted by Obsidian Sensors?
J.H.: We serve both commercial and defense markets as the microbolometer is truly dual-use technology. Thermal imaging with microbolometers is a standout whenever day/night imaging is required without reliance on illumination systems and when viewing conditions are challenging like seeing through smoke or blinding glare. Automotive usage can become the highest volume sector driven by ADAS, autonomous driving, and in-cabin applications. Security/surveillance and industrial usage represent very large opportunities with a lot of potential customers eager to use thermal imaging for the value it brings. More recently, there has been an explosion in demand from drones and robotic platforms, both commercial and military.
DVN: How can your technology contribute to the safety of standard and autonomous vehicles in comparison with standard cameras, radars, and lidars?
J.H.: Visible cameras, radar systems and lidar collectively have a blind spot. And that is what thermal imaging fills in, so it is complementary to existing sensor suites. Being able to image objects at a long distance in total darkness, in varying, challenging conditions such as fog, smoke, headlight glare, or sun glare helps to address important corner cases which are at the center of FMVSS 127 and related standards.
DVN: What opportunities and growth factors do you see for the thermal cameras or the automotive market?
J.H.: Right now, the thermal imaging industry seems focused on finding some sort of middle ground between cost and resolution. The pioneers of the automotive thermal imaging actually advanced the resolution to VGA but there is talk of QVGA (¼ the pixels) trying to meet perceived cost targets. The FPA resolution affects imaging in two related factors: the field of view and range (how far can you see that there is a person on the road?). For the same field of view, VGA doubles the range of a QVGA camera. Range is important because it relates to stopping distances that increase with increasing speed. But so is field of view. Because of this we should not regress back to lower resolution but move it forward with a sensible roadmap that keeps pace with improvements in both visible and radar sensing.
DVN: How do you see the market price evolving for thermal sensors?
J.H.: The cost of cameras is dominated by the focal plane array cost, hence the desire to cap the resolution for overall cost control. By adopting our unique manufacturing approach with patented device designs, it is possible to get the resolution you need at the price you want. That is our message to our OEM customers who in turn need to deliver driving machines to drivers who demand safety, performance and affordability in what they buy or lease. The parameters involved in driving are complex. Field of view, range, detectability, false positives are all directly related to safety outcomes.
DVN: When will you have validated products ready to go to production (B & C samples)?
J.H.: We are shipping product to support real needs in robotics and consumer markets today. We can support automotive projects with SOP consistent with FMVSS 127 timelines.
DVN: What are the advantages of your technology regarding competitors?
J.H.: Although we offer both focal plane array and camera products, it is important to note that at the core, we are a sensor manufacturer. We can promote market leading FPA products as well as to customize them quit easily to exploit application particulars. Here is an important point. We make microbolometers in a way wholly different and fundamentally advantaged in both cost and scale when compared with everyone else in the industry. We make glass based microbolometers with a proprietary fabrication process implemented in a standard flat panel glass fab. There are other aspects to this I will describe in the DVN session. Per unit area, costs to build electronics and MEMS on glass is over an order of magnitude lower than a comparable manufacturing strategy based on silicon wafers. Moreover, in terms of capacity, glass production is capable of several orders of magnitude higher volume manufacturing output. This is the core value we bring. High resolution, high performance thermal imaging at a fraction of the cost and with much higher volume capacity.
DVN: Do you have projects or pre-developments with tier-1s or automakers?
J.H.: Compal, Quanta, Mobis. We also have had a project with HATCI, the North American Hyundai division.
DVN: How do you ensure the quality and reliability of your sensors?
J.H.: We can pass the same qualification standards as other microbolometer sensors. Also, we do not have wire bonds in our focal plane module due to our unique architecture. There are several other distinct advantages baked into our unique approach that cover both front end (glass) production and back end (packaging).
DVN: Do you sell the entire camera module or just the sensor?
J.H.: We sell both at this point. When the volumes become larger, it is more efficient for us to handle the sensor module and work with integrators and camera makers.
DVN: Does Obsidian provide a software stack with the sensor?
J.H.: We have an ISP with minimal latency and tone mapping that can be tailored for human vision, machine vision or radiometry.
DVN: Where are the sensors manufactured?
J.H.: We describe our manufacturing process in two stages, the front end and the back end. The front end is the manufacture of the glass-based device and that is done today in Taiwan by our manufacturing partner, Innolux—Taiwan’s largest LCD manufacturer, and an important partner for us. Last year, we announced a new front end manufacturing partnership to focus on a 12µm SXGA product launch in 2026, with JDI in Japan. The back end of the manufacturing process involves vacuum encapsulation, ROIC attachment and other packaging steps, all done in our facility in San Diego.
