Litavis SPAD image sensor moves photon processing on chip

A normal image sensor is a disciplined clerk. It counts light, hands over the receipts, and lets some other chip do the interrogation. Singular Photonics is trying something more ambitious with Litavis: move part of the photon bookkeeping into the place where the photons land. That is the hardware equivalent of putting the crime lab inside the getaway car, ridiculous until you remember how much energy and latency get burned moving data around.

This is a teardown without a screwdriver, because the architecture is the story. A SPAD sensor is already a strange little beast, a device built to notice individual photons instead of merely averaging a scene into brightness values. Litavis adds in-pixel processing, which means the pixel is no longer just a bucket. It is a tiny customs desk, asking each photon what business it has being here.

The pixel stopped being passive

Singular Photonics said on September 10, 2026, that Litavis is the world’s first single photon avalanche diode sensor using in-pixel processing to unify imaging, timing, histogramming, and photon statistics on one chip. EuropeSays describes the practical shift clearly: rather than sending large quantities of raw photon data to an external processor, Litavis performs part of the analysis inside the sensor itself. Singular Photonics also says the architecture combines photon counting imaging, programmable time gating, and advanced photon timing within a scalable SPAD array platform. In other words, the sensor is not just collecting light, it is sorting evidence at the scene.

That movement matters because data transfer is the sneaky tax collector in embedded imaging. EuropeSays reports that on-sensor analysis can reduce data transfer requirements and enable lower-latency, more power-efficient imaging systems. That is not a glamour spec, but it is the kind of spec that decides whether a robot sees in time to react or merely records a beautiful mistake. When photons are the raw material, hauling every event off-chip can turn the board into a tiny shipping port with a panic problem.

The buried spec changes the readout story

Daily.dev, summarizing The Quantum Insider, reports that Litavis offers 256 by 256 pixel photon counting imaging with picosecond resolution photon timing on a 64 by 64 macropixel grid. That is the line I would circle in red on the datasheet, then probably talk about too loudly at lunch. The resolution figure tells you the imaging canvas; the macropixel timing grid tells you where the sensor is doing the more specialized stopwatch work. It is not one generic camera pretending to be a lab instrument, it is a sensing fabric with different jobs mapped onto the silicon.

The same Daily.dev summary says Litavis supports software configurable modes including TCSPC, high dynamic range photon counting, and coincidence based detection. TCSPC, short for time correlated single photon counting, is the sort of acronym that arrives with a clipboard and refuses to leave, but the idea is simple enough: timing can be as valuable as intensity. Singular Photonics says Litavis can enable simultaneous intensity, timing, and histogramming modes. That is the jackpot combination for applications where the scene is dim, fast, depth related, or scientifically fussy.

Software configuration is the quiet power play

Technology Scotland describes Singular Photonics as an Edinburgh fabless semiconductor company and says Litavis is designed to capture and process far more information about light than conventional sensors. EuropeSays adds that the same hardware can be reconfigured for different imaging tasks without requiring a new sensor design. That is a big deal for product teams, because sensor selection usually feels like choosing a wrench size while the bolt is still being invented.

A programmable sensing architecture gives engineers more room to tune behavior after hardware decisions are already expensive. Singular Photonics frames Litavis as a software configurable sensing architecture, and that phrase deserves more respect than launch copy usually gets. Software configuration at the sensor level means parameters can be adapted closer to the physical event, before the host processor gets buried under photon receipts. It will not magically solve optics, illumination, packaging, calibration, or thermal design, because physics remains undefeated. But it can change where the system spends its bandwidth and power budget, which is where real products either behave elegantly or start bargaining with the heat sink.

Where this sensor could matter first

Technology Scotland says Litavis opens possibilities for robotics, medical imaging, scientific research, and other advanced imaging applications. Daily.dev also lists machine vision, depth sensing, spectroscopy, and quantum technologies among target applications. The common thread is not that these fields need prettier pictures. They need measurements where individual photons, arrival timing, or statistical patterns can carry useful information.

For robotics and machine vision, lower latency is the part to watch, according to EuropeSays. For scientific and medical imaging, the appeal is more about extracting useful signal from difficult light, especially where individual photon sensitivity matters. And for embedded designers, the lesson is broader than Litavis itself: the sensor is becoming a compute boundary, not a passive peripheral. The next interesting teardown question is not just what comes out of the image sensor, but what never has to leave it.