Canon Inc. announced today that the company has developed the world's first1 single-photon avalanche diode (SPAD) image sensor with signal-amplifying pixels capable of capturing 1-megapixel images. SPAD image sensors are ideal for such applications as 2-dimensional cameras, which capture and develop still image and video in an extremely short span of time. These sensors also hold potential for use in 3-dimensional cameras due to their ability to obtain information about the distance between them and a subject as image data.
A SPAD sensor is a uniquely designed image sensor in which each pixel possesses an electronic element. When a single light particle, called a photon, reaches a pixel it is multiplied—as if creating an “avalanche”—that results in a single large electrical pulse. The ability to generate multiple electrons from a single photon provides such advantages as greater sensitivity during image capture and high precision distance measurement.
The SPAD image sensor developed by Canon overcomes the longstanding difficulties of achieving this effect with high pixel counts. By adopting new circuit technology, Canon's sensor uses a method known as photon counting to realize a digital image resolution of 1 megapixel. What's more, the sensor employs a global shutter that allows simultaneous control of exposure for every pixel. Exposure time can be shortened to as little as 3.8 nanoseconds2, making possible clear and distortion-free image capture. In addition, the sensor is capable of up to 24,000 frames per second (FPS) with 1 bit output, thus enabling slow-motion capture of fast movement within an extremely short time frame.
Thanks to its ability to capture fine details for the entirety of events and phenomena, this technology holds the potential for use in a wide variety of fields and applications including clear, safe and durable analysis of chemical reactions, natural phenomena including lightning strikes, falling objects, damage upon impact and other events that can't be observed with precision by the naked eye.
The sensor also features a high time resolution as precise as 100 picoseconds2, enabling it to determine the exact timing at which a photon reaches a pixel with ultra-high accuracy. Leveraging this functionality, the sensor is capable of Time of Flight distance measurement. What's more, with a high resolution of 1 megapixel and high-speed image capture, it is also able to accurately perform 3D distance measurements in situations where multiple subjects overlap—useful in such scenarios as a vehicle distance measurement for self-driving automobiles and grasping 3D spatial information for xR3 and similar devices.
Canon's development of a SPAD image sensor enables 3D cameras capable of recognizing depth information to achieve a resolution of 1 megapixel is expected to rapidly expand the use of such cameras as the “eyes” of high-performance robotic devices. Going forward, Canon will strive to anticipate the needs of industry by continuing to advance its innovative image sensor technology, further expand the possibilities of what is visible, spur evolution in science and industry through high-precision detection of information and contribute to the development of fields yet to be discovered.
- Among SPAD sensors. As of June 23, 2020. Based on Canon research.
- 1 nanosecond = one billionth of a second. 1 picosecond = 1 trillionth of a second.
- “xReality,” where x represents a variable for the unknown. Includes AR (augmented reality), MR (mixed reality), and VR (virtual reality).
*Canon* Here's 24,000fps
If you can capture all the images, after the fact you could move a sliding window around and adjust the moment the exposure starts down to the 1/24k of a second and alter the effective ISO after capture. In fact normal exposure terminology doesn't work when you're capturing single photon events over time. You'd be able to do something like bayer pattern demosaicing except temporally, which ought to enable some new tricks.
In a sense yes, it comes from their energy level. But assuming single point trigger on the avalanche the sensor cannot detect color.
Now I started thinking some foveon-style avalance photon detector where it triggers the single avalanche at different depths... hmm...
You can definitely use APDs as "geiger counters" for photons and count them up over time. Typically, you need to divide the count by some ginormous number, since even your average 60W light bulb emits 1.8 x 10^20 photos/second. However, you could use different factors to accommodate different light levels and in theory get very VERY good detection in low light.
Single-photon avalanche diode imagers in biophotonics: review and outlook
That is true. The gain is very high: The cell is a photo diode that functions as a current source with a current proportional (in best case but there is also noise) to the number of photons hitting the diode. A single photon can be measured since a so-called avalanche effect induces a large amount of electrons resulting in a measurable current.
The question is how small a diode and a necessary resistor can be made such that high integration i.e. high mp sensor is possible.
Since we ARE a chip manufacturer ourselves, we do this quite often where we use a 6x6 array of RGB colour photosites intermixed with luminance-specific photo avalanche diodes AND longwave and shortwave infrared and 400 nanometres ultraviolet sensitive photosites giving a FUSION IMAGE DETECTION DEVICE.
The arrays are 6x6 of 30 micron photosites combined into one virtual 180 micron-sized fusion pixel which allows us to offer simultaneous multi-spectral views at various wavelengths. We run that info through 2D-XY SOBEL EDGE detection to find and track our moving targets or provide full RGB colour even at midnight.
For now we are running 186 mm GaN and CMOS substrates so that means we are only at 1024 by 1024 fusion pixels. That good enough for most warfighter nightvision though! We will bring that up to 8192 by 8192 virtual pixels by changing the substrate dopants to increase sensitivity which means each photosite only needs to be around 8 microns. The next version sensor will be a VERY LARGE SENSOR at almost 400 mm by 400 mm but it WILL be DCI 8k by DCI 8k resolution for IR, UV and Optical wavelengths.
---
The basic meaning of my above statements is to outline that military-spec gear has used such technology for many years now and Canon is JUST NOW bring that technology to consumers AND THAT MEANS YOU will eventually get super high-sensitivity GaN/CMOS fusion sensors that can give you low-noise, nearly perfect RGB colour EVEN AT NIGHT !!!!
---
Based upon our OWN production costs and schedules, we estimate that our own GaN/CMOS fusion image sensor technology will come down to less than $750 per sensor chip in maybe 7 to 10 years and that Canon can bring it down to $250 per similar-style sensor chip using CMOS in maybe 5 years.
Canon won't be using GaN/CMOS fusion substrates like we do BUT they can get QUITE A LOT of sensitivity in those Canon-made sub-$250 image sensor chips which will QUITE OUTPERFORM the Exmor-series chip used in the Sony A7s2 camera which for now is the best consumer-grade near-full-RGB-nightvision chip around. Think ISO ONE MILLION sensitivities with LOW NOISE!
So again, ALMOST FULL COLOUR RGB NIGHVISION is just around the corner in about five years at $250 per sensor chip which can be used in RF-mount Canon cameras.
V