OAKLAND, Calif. — The startup NanoLambda has developed a $10 spectrometer-on-a-chip that's suitable for use on a wide range of consumer devices.
The Spectrum Sensor measures individual wavelengths and is accurate to 1 nm with 10nm resolution, allowing NanoLambda to selectively pass light through nanostructures to determine their makeup. The company used traditional wafer processes to create 5mm x 5 mm x 2mm thin nanofilter arrays. Nine hundred of these arrays can be stacked to create a millimeter-scale chip that can be embedded into various Internet of Things devices.
The spectrometer is embeddable in a variety of devices.
"Every material has its own spectral fingerprint, which means, when certain very flat white light shines on to the target material, its material absorbs light differently, making... specific patterns or shapes," NanoLambda CEO Bill Choi told us. "Each filter will detect different wavelengths. Then, collectively, the sensor can figure out" the material's characteristics.
Spectrometer technology has been used to analyze biochemicals for several decades, but it was previously too bulky and expensive to use at a consumer scale. NanoLambda's nanoarrays have brought the chip's price point down to $10. Choi expects the small size and price to allow for the technology to be embedded in everything from chopsticks to medical devices.
NanoLambda CEO Bill Choi.
Sweet red apples would exhibit a specific color that users could identify with ripeness, or a spectrometer could be used to identify the freshness of milk. Choi said consumers could monitor drinking water from kitchen faucets; certain colors would show whether water is contaminated or safe to drink. Requests for this particular kind of analysis have been commonplace in China, where people want to identify fake whiskey and wine.
"The chemical composition of real wine and fake wine will be different," he said. "You could add additional color, but the chemical fingerprint itself is different. When you shine light on to the target whiskey, the transmitted light will have a slighter different absorption pattern."
Each pixel detects a specific wavelength collectively to describe its
relationship to the spectrum.
(Image: TSensors Summit)
A spectrometer-on-a-chip could also conduct noninvasive medical monitoring. Choi has received suggestions for tongue monitoring, as well as monitoring change in organ color over time, though that app is invasive. He hopes his technology will become the basis of a health testing consortium.
"Fusion or multi-mode sensing abilities will definitely bring the solution to next-generation healthcare," he said. "I don't think one single sensor technology will solve all the problems. I think collaboration among companies is very important."
Those collaborations will also help solve NanoLambda's biggest challenge: creating a database of spectral fingerprints for comparing food freshness, organ color, or other biochemical materials.
"The sensor itself is useless without an algorithm," TSensors Summit chairman Janusz Bryzek said at a San Diego conference last month. "You are trading transistors and processing power for optics. It requires a lot of post-processing."
Choi used milk tests as an example of database collaboration:
If we want to have some kind of database for milk, you want to measure all kinds of milk's spectral fingerprints from different companies and different flavors. Of course, at the beginning you can start with a very baseline database from [your own experience], NanoLambda, or milk companies. As we move on, then maybe the consumers can participate by measuring the milk using this very simple device and uploading the data on to the system. Those won't be scientifically measured, but have some value. Who will get the database service later is open for question.
The fabless NanoLambda hopes to deliver the spectrometer and interface software IP to OEMs such as Samsung and Apple, which will build final devices. Choi could not speak about any partnerships, but he said the company will release a software developer kit in the first few months of 2015. He expects applications to be created in three months to one year after that.
The company's software runs on both Linux and Windows systems, he said.
— Jessica Lipsky, Associate Editor, EE Times