2023 Author: Bryan Walter | [email protected]. Last modified: 2023-05-21 22:24
Nikon has developed an image sensor that can control exposure levels for different areas of the frame during a shot. It consists of two layers: on top are pixels, conventionally divided into small groups, and below these groups are control circuits that select the required exposure for each group. This design allows for a dynamic range of 134 decibels. The description of the new matrix is published on the Japanese Nikon website.
The dynamic range of the matrix shows how strongly the pixels in one image can differ in illumination without losing details, that is, without becoming completely white or black (or rather, illuminated at the matrix noise level). High dynamic range is important in scenes with strong shadows and highlights, and in general photography, low range can be used as an artistic trick, but in other applications it can be very annoying. For example, cameras of an unmanned vehicle at the exit of a tunnel may encounter an extremely large difference in brightness, and the control algorithms of the car may not notice an obstacle in front of it.
There are two main solutions to the low dynamic range problem: software and hardware. The software consists in the fact that the camera takes two or more images with different exposures, and then creates one based on them, in which the light and dark areas are taken from different frames. Hardware solutions are less common and differ in implementation. For example, until the early 2010s, Fujifilm produced cameras with sensors in which there were two types of pixels of different sizes, designed for light and dark areas.
Nikon has unveiled a new hardware-based HDR sensor design. The die block consists of two chips, one above the other. Above is a photomatrix responsible for image acquisition. Its resolution is 4224 by 4224 pixels, which are combined into blocks of 16 by 16 pixels. The main feature of the new matrix is that during shooting, exposure can be controlled not only for the entire array of pixels as a whole, but also for each individual block.
The second chip is located on the lower level. In it, under each two adjacent blocks of pixels, there is a block with control electronics: 16 analog-to-digital converters and pixel control blocks. They receive signals from the pixels of the upper layer, convert them into digital form, and, based on this, select the exposure for the 16 by 16 pixel block, taking into account the global requested exposure of the image.
The developers claim that when shooting at 60 frames per second, the dynamic range of the resulting frames is 134 decibels or more than 22 "stops". The matrix also supports shooting at a rate of up to 1000 frames per second, but in this mode, the dynamic range is reduced to 110 decibels.
An example of shooting without (the first two frames) and with an extended dynamic range
The matrix has a square shape and a diagonal of one inch. The pixels in it have a size of 2.7 micrometers, and the elements are made according to a 65-nanometer process technology. The cost and date of the start of production of the new matrix are still unknown.
In addition to classical photo matrices, there are also event-based ones. They do not, as such, capture a frame in which the camera simultaneously polls the pixels: instead, the pixels work asynchronously and fire only when the brightness has changed by a threshold value. One of the advantages of such cameras is precisely the large dynamic range. And recent advances in algorithms for processing data from such cameras have made it possible to implement high-speed and color shooting in them.