In general, the industry is focused on making devices and components smaller. Canon, however, has recently created the world's largest CMOS sensor.
The new sensor measures a whopping 20cm on each side. Strangely, Canon would have gone bigger—but, since the largest wafers available are ~30cm (and these are circular), the 20cm sided square sensor just barely fits onto the wafer.
Each pixel size on the sensor is 2.2µm which means that this sensor has a staggering 120 megapixels. Each pixel is incredibly sensitive, and the sensor can be used in lighting conditions as low as 0.3 lux (for comparison, the full moon has a lux of 0.1).
On the video side of things, Canon claims this sensor can shoot video at 60fps. This is mildly surprising since surely the size of the sensor will cause issues with delayed signals and synchronization.
The large sensor compared to the T6i DSLR
Understandably, Canon has not released too much information on the process mechanism, itself, but they have said that the sensor takes advantage of parallelism and some clever transfer method. From this, the sensor is probably split up into small cells that each have their own small processor which record data from the sensor and then send the data to a central processor or directly into a memory bank using DMA.
But synchronizing on this scale could be potentially tricky but there are a few ways in which this could be achieved. One method would involve synchronizing all the cells to a timer and then sensor packets sent by the individual cells could be time stamped. Then the central processor could arrange the image blocks in the correct order. Another solution could be that each cell has its own memory bank and continuously records images. Since each cell knows exactly where it is in the array when it receives a “send me your data now” signal, it can send information taken n frames ago where n accounts for the signal delay.
Defend the Earth!
This sensor clearly has many applications—but how could it defend the Earth? As it turns out, this sensor was mounted onto a telescope in a Japanese observatory at the University of Tokyo and was used to make the world's first recording of a meteor with an apparent magnitude of 10. For those who are not familiar with apparent magnitude in astronomy, a celestial object with a magnitude of 10 is incredibly faint to the point where its invisible to the naked eye, not visible with a pair of binoculars, and makes Neptune look bight!
One of the biggest issues with planetary defense is detecting small to mid-sized objects which are difficult to see. While some may think that large asteroids like those seen in films are the real threats, the reality is that they are far, few, and trivial to spot. Asteroids and meteors like the one that flew over Chelyabinsk are arguably more dangerous as there are many more of them and they are challenging (sometimes next to impossible) to detect with enough time to prepare or warn people in their way. If the asteroid over Russia had not disintegrated in the atmosphere and fell in a city, it would have the destructive capability of 500 kilotons of TNT. For comparison, the bomb that dropped on Hiroshima was only 20 kilotons.
The trail left by the Russian meteor. Image courtesy of Nikita Plekhanov [CC BY-SA 3.0]
This type of sensor could play a major role in detecting such faint objects before they become impossible to intercept. Meteors that would normally be too faint to detect can be spotted by digital equipment that can process the data to automatically determine if there is a threat to Earth.
The Cost of Big vs the Cost of Small
Transistor features now measure in the nanometers. For example, a 14nm gate can be as thin as 26 atoms across. Therefore, there is not much material left to play with to produce reliable devices and when devices contain billions of transistors the chances of a chip failing naturally increases (and thus the yield falls).
From one perspective, going smaller provides an increase in processing capability at the cost of yield. It turns out that increasing chip size comes at a cost. As the size of a chip increases, the yield also decreases because of errors. Errors that can cause issues when spanning out include spot problems (issues in the silicon crystal), and misalignment of layers.
A silicon wafer. Image courtesy of Peellden [CC BY-SA 3.0]
Despite the issues with large silicon chips, Canon's new sensor is going big. It's unlikely that Canon will publicly publish yield information—what's your experience with chip production? What issues come with small chips compared to large ones in today's manufacturing facilities? Share your thoughts in the comments below.
- CMOS Imaging Sensor by ams Aims to Be the Future for Machine Vision
- Exploring the Solar System with the Voyager Spacecraft’s Cameras, Polarimeters, and Magnetometers
- Drones on Mars? NASA Projects May Soon Use Drones for Space Exploration
- How NASA’s Raven Module Is Developing Machine Learning Algorithms for Autonomous Spacecraft Docking
- How Sensor Fusion Works