Researchers at Cambridge have built a convolutional optical spectrometer on a silicon nitride chip that costs roughly ten dollars to produce. By applying the convolution theorem directly in the optical domain using repeating components like micro-ring resonators, the device analyzes near-infrared light across the 1200-1700 nanometer range. In tests, the centimeter-scale sensor achieved high classification accuracy on materials and measured biomarkers including blood glucose and lactate in real time.
Shrinking spectrometers usually forces a trade-off between size, resolution, and computational overhead, so seeing benchtop-level precision on a chip is notable. While the noninvasive glucose tracking results from a single participant are promising, they remain preliminary and far from a validated clinical product. It is a neat application of mathematical restructuring to hardware limits, but actual wearable utility will depend heavily on how well it handles real-world skin variation and motion blur over time.
more: https://www.sciencedaily.com/releases/2026/10/261022081312.htm
Leave a Reply