Lists

Wednesday, August 14, 2024

EETimes article on imec

Full article: https://www.eetimes.eu/imec-getting-high-precision-sensors-to-market/

Imec: Getting High-Precision Sensors to Market

At the recent ITF World 2024, EE Times Europe talked with imec researchers to catch up on what they’re doing with high-precision sensors—and more importantly, how they make sure their innovations get into the hands of industrial players.

Imec develops sensors for cameras and displays, and it works with both light and ultrasound—for medical applications, for example. But the Leuven, Belgium–based research institute never takes technology to market itself. It either finds industrial partners—or when conditions are right, imec creates a spinoff. One way to understand how imec takes an idea from lab to fab and finds a way to get it to market is to zoom in on its approach with image sensors for cameras.

“We make image sensors that are at the beating heart of incredible cameras around the world,” said Paul Heremans, vice president of future CMOS devices and senior fellow at imec. “Our research starts with material selection and an overall new concept for sensors and goes all the way to development, engineering and low-volume manufacturing within imec’s pilot line.”

A good example is the Pharsighted E9-100S ultra-high-speed video camera, developed by Pharsighted LLC and marketed by Photron. The camera reaches 326,000 frames per second (full frame: 640 × 480 pixels) and up to 2,720,000 frames per second at a lower frame size (640 × 32 pixels), thanks to a high-speed image sensor developed and manufactured by imec.

Another example is an electron imager used in a cryo-transmission electron microscope (cryo-TEM) marketed by a U.S. company called Thermo Fisher. The instrument produces atomic resolution pictures of DNA strands and other complex molecules. These images help in the drug-discovery process by allowing researchers to understand the structure of the molecules they need to target.
Thermo Fisher uses direct electron detection imagers, developed by imec and built into the company’s Falcon direct electron detection imagers, each composed of 4K × 4K pixels. The pixels are very large to get to the ultimate sensitivity. Consequently, the chip is so large (5.7 × 5.7 cm) that only four fit on a 200-mm wafer.

A third example is hyperspectral imagers, with very special filters that detect many more colors than just red, green and blue (RGB). Hyperspectral imagers pick up tens or hundreds of spectral bands. They can achieve this level of performance because imec implements processing filters on each pixel.

“We can do that on almost any commercial imager and turn it into a hyperspectral camera,” Heremans said. “Our technology is used by plenty of customers with a range of applications—from surveillance to satellite-based Earth observation, from medical to agriculture and more.”

Spectricity

To bring some of its work on hyperspectral imagers to market, imec created a startup called Spectricity. “The whole idea is to bring this field of multispectral imaging or spectroscopy into cellphones or other high-volume products,” said Glenn Vandevoorde, CEO of Spectricity. “Our imagers can see things that are not visible to the human eye. Instead of just processing RGB data, which a traditional camera does, we take a complete spectral image, where each pixel contains 16 different color points—including near-infrared. And with that, you can detect different materials that look alike but are actually very different. Or you can do color correction on smartphones. Sometimes people look very different, depending on the ambient light. We can detect what kind of light is shining—and based on that, adjust the color.”
The first use case for cellphones is auto white balancing. When a picture is taken with a cellphone, sometimes the colors show up very differently from reality, because the camera doesn’t have an accurate white point, which is the set of values that make up the color white in an image. These values change under different conditions, which means they need to be calibrated often. All other colors are then adjusted based on the white point reference.

Traditional smartphone cameras cannot determine the ambient light accurately, so they cannot find the white point to serve as a viable reference. But the multispectral imager obtains the full spectral information of the ambient light and applies advanced AI algorithms to detect the white point, which leads to accurate auto white balancing and true color correction.

Spectricity said its sensor is being evaluated by seven out of the top eight smartphone manufacturers in the world for integration into phones. “By the end of this year, you will see several smartphone vendors launching the first phones with multispectral imagers inside,” Vandevoorde said.

While smartphones are the ultimate target for high volume, they are also very cost-competitive—and it takes a long time to introduce a new feature in a smartphone. Spectricity is targeting other smartphone applications but also applications for webcams, security cameras and in-cabin video cameras for cars. One category of use cases takes advantage of the ability of multispectral images to detect health conditions.

 

Spectricity’s spectral image sensor technology extends the paradigm of RGB color image sensors. Instead of red, green and blue filters on the pixels, many different spectral filters are deposited on the pixels, using wafer-scale, high-volume fabrication techniques. (Source: Spectricity)

 
Spectricity’s miniaturized spectral camera module, optimized for mobile devices.

“For example, you can accurately monitor how a person’s skin tone develops every day,” Vandevoorde said. “We can monitor blood flow in the skin, we can monitor moisture in the skin, we can detect melanoma and so on. These and many other things can be detected with these multispectral imagers.”
Spectricity has raised €28 million in funding since it was founded in 2018—and the startup has its own mass-production line at X-Fab, one of the company’s investors. “We have our machinery and our process installed there,” Vandevoorde said. “It’s now going through qualification—and by the end of the year, we’ll be ready for mass production to start shipping large volume to customers.” 

How imec finds the right trends to target
Spectricity is a good example of how imec spots a need and develops technology to meet that need. Spectroscopy, of course, is not new. It’s been around for decades, and researchers use it in labs to detect different materials and different gases. What’s new is that imec integrated spectroscopy onto CMOS technology and developed processes to produce it in high volumes for just a couple of dollars. Researchers worked on the idea for about 10 years—and once it was running on imec’s pilot line, the institute set up Spectricity to take it into mass production and develop applications around it. 

“We sniff around different trends,” said Xavier Rottenberg, scientific director and group leader of wave-based sensors and actuators at imec. “We’re in contact with a lot of players in the industry to get exposed to plenty of problems. Based on that, we develop a gut feeling. But gut feelings are dangerous, because it might be that you’re just hungry. However, with an educated gut feeling, sometimes your intuition is right.”

Once imec develops an idea in the lab, it takes the technology to its pilot line to develop a demonstrator. “We do proofs of concept to see how a device performs,” Rottenberg said. “Then we set up contacts in the ecosystem to form partnerships to bring the platform to a level where it can be mass-produced in an industrial fab.”

In some cases, an idea is too far out for partners to pick up for near-term profit. That’s when imec ventures out with a spinoff company, as it did with Spectricity.


2 comments:

  1. This is all very interesting work done at IMEC.
    There is a caveat however: IMEC does not only develop new technologies with / for image sensor solutions, but also likes to directly compete with the Flemish industry it is supposed to support with what IMEC thinks is novel, but in reality has become plain vanilla, albeit high tech, sensor technology development...

    ReplyDelete
  2. This is very exiting! In the Astrophotography community this would be something like a near holy grail, as this provides multi-color *per pixel* as opposed to the standard RGGB bayer matrix which only registers light assigned to the respective pixels and rejects the rest.
    What would actually make the full holy grail is a custom spectrum (RGB, Oiii, Sii, Hydrogen-Alpha/Beta, UV/IR). Depending on the Quantum Efficiency (what percentage of photons will actually be converted in the right signal) this would end the market for Mono sensors and the Filter market all together.

    So I'm VERY eager to learn how this actually performs!

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.