Wuji Equipment Editor丨Liang Shuang
Factors such as recurring outbreaks of COVID-19 and fluctuations in exchange rate of US dollar and Japanese yen in 2022 will have a significant impact on camera industry. According to latest data from CIPA, camera market is gradually recovering. Interchangeable lens cameras will outperform 2021 from May 2022. Among interchangeable lens cameras, mirrorless cameras are already leading in terms of shipment volume, shipment volume and unit. The price of SLR cameras is several times higher, and mirrorless cameras have long become undisputed mainstream in market. Major manufacturers are comprehensively addressing mirrorless cameras in terms of product shape, while focusing more on mid-range and high-end value-added products. Compared to full-frame flagship competitors in 2021, “coffee position” has declined in 2022, and technology continues trend of using multi-layer CMOS, high-performance processors, and AI detection technology.
Truth be told, don't let photography stop at fever!
The latest CIPA data shows that camera market has gradually recovered in 2022
Part-frame stack sensor
In fact, non-full-frame small-sized CMOS sensor was applied very early on to multi-layer sensors, not to mention smartphones. Multilevel CMOS sensor. Interestingly, stacking technology directly passes 4/3 and APS-C and is used for first time at full frame. We see that difficulty of applying this technology is not in size of sensor, but rather depends on requirements of application (high speed) and cost.
In February 2022, Aozhixin launched OM-1 camera equipped with a BSI Live MOS multi-layer sensor, and in May, Fuji introduced X-H2S with a back-illuminated X-TransTM CMOS 5 HS multi-layer sensor. Among them, Fuji highlights that signal reading speed of this sensor is more than 4 times faster than previous generation. The most direct manifestation of "speed" is continuous shooting. OM-1 with continuous autofocus/exposure tracking can achieve 50fps continuous shooting and 120fps continuous shooting at AF/AE lock; The Fuji X-H2S can shoot continuously at 40fps with focus/exposure tracking.
OM-1 uses a multilayer sensor
In addition to continuous shooting (no black screen), use of high-speed multi-layer sensors will greatly improve focus, video frame rate, EVF frame rate, and jelly effect. We see that in 2022, from full frame to 1 inch, stack sensors have embraced flagship models of various frames, and "only fast, but not broken" has become secret of every company's pursuit of highest performance.
It is worth noting that foreign media resource image mentioned in an interview with Fuji development that X-H2S multi-level matrix does not have any cache, but only maximizes analog-to-digital converter (more than 4 times) and transmission chain (transmitted to channel processor reads). It can be seen that foot sensor is currently a custom product, and manufacturers can choose which function to place on second layer of stack sensor according to their own design. We can't just view various sensors of stack as a structure, but Eight Immortals cross sea to show off their special abilities.
Fuji places chip that processes and reads signal on back of sensor surface
High-performance processors further expand AI capabilities
Aozhixin OM-1 uses new TruePic X processor, its high-speed computing performance is about 3 times that of TruePic IX, which greatly promotes progress of features such as shortening processing time of shooting in high resolution mode. and real-time ND with improved slow shutter effects, etc. In addition, developer claims that deep learning processing power of new processor is 60 times higher than previous one, and there is still room for improvement in future firmware.
Fuji's "fifth generation" model is equipped with X-Processor 5 processor, which has twice processing speed of current model. In addition to significantly improving accuracy and speed of autofocus, AI-based subject detection autofocus can be achieved. As for video, processor supports several codecs, including Apple ProRes 422 HQ.
Fuji X-5 processor
We see that image processor of modern cameras is no longer solely an ISP. Due to continuous improvement in video performance, more codecs need to be integrated. In addition, to increase processing power of AI, an AI processor unit must be integrated into chip. Of course, there are also external AI chips, such as Sony α7R V, as a transitional solution to increase computing power. In future, processors will be highly integrated in "SoC" direction, forming an image processing microsystem that is more complex and powerful than before.
A wider range of artificial intelligence AI
As mentioned above, all new processors are expanding AI deep learning capabilities that are widely used in object detection for focus tracking by adding more objects. The Canon EOS R6 Mark II adds detection of horses, trains, planes and helicopters. In addition to racing cars, steam locomotives, motorcycles, airplanes, helicopters, trains, and birds that E-M1X can recognize, OM-1 recognizes animals such as cats and dogs. Fuji X-H2S, X-H2 and X-T5 "fifth generation" models can automatically detect and track many types of objects, in addition to detecting and automatically tracking human faces/eyes, they can also detect animals, birds, cars, motorcycles. , bikes, planes and trains.
The Canon EOS R6 Mark II has recently added many detection and recognition targets
It is worth noting that, in addition to BIONZ XR image processor, α7R V is also equipped with a deep learning AI smart chip, which is specially used for processing large amounts of data and has advanced object recognition functions. The chip can intelligently predict and process people's body position information, and improve camera's object recognition accuracy. At same time, AI intelligent recognition subject categories have also been further expanded: in addition to people, animals, and birds, up to 7 subject objects such as vehicles and insects have been added.
Sony adds insect detection
In addition, Sony has more AI applications than other manufacturers. The auto white balance function of α7R V is designed at same time as AI intelligent recognition. With an IR sensor on front of fuselage, AI intelligent chip also helps to achieve more accurate white balance in low-light (such as shade) environments.
Pixels have been updated and some remain same
In 2022, Sony α7R V and Leica M11 cameras couldn't beat 61MP full-frame record set by α7R IV, and Hasselblad X2D 100C is also on par with 100MP Fuji GFX100. A new sensor with a large number of pixels, which causes pixel peak of camera circle to stagnate.
Of course, not all new items are like this: Panasonic GH6 has 25.2 million pixels, and Fuji X-H2/X-T5 has 40.2 million pixels, which accordingly updated pixel records of cameras of corresponding format. Fuji has improved a lot, surpassing Canon's record of 32.5 million in one fell swoop.
In addition to "best students" that clean records, there are also some models with a certain degree of pixel enhancement. For example, Canon EOS R6 Mark II increased from 20.1 million to 24.2 million in previous generation. , which can be considered a catch-up level of mainstream. Another drawback. It is worth noting that Sony has launched Super 35mm FX30 camcorder, which is equipped with a back-illuminated APS-C sized sensor with a resolution of 26 million pixels. Compared to Fuji X-T4 and other models, this sensor is more likely to be a regular version of Bayer sensor. In future, Sony may release an α series camera (α7000?) with same sensor as FX3-α7S III.
Typically, people have ever higher demands on camera focus, detection, burst shooting, video, etc., and "quickly" has become driving force behind almost every feature. We can see that Canon EOS R6 Mark II's CMOS sensor is neither backlit nor complex, but it's not ambiguous in terms of speed and can reach ultra-high 40fps continuous shooting. On contrary, Sony's non-stacked sensor lost a little in "performance" - this gap began to appear even with advent of new Canon sensors, such as EOS M6 Mark II and EOS R5. Sony's iron law of "10fps is enough for a flagship" has caused sensor speed to stagnate, while at same time putting friends and sellers equipped with Sony sensors in a passive position. "Slow" not only affects continuous shooting index, but also focus, detection, jelly effect, video frame rate, etc. Of course, it's not enough for a product that creates images to just blindly be "stupid and fast". , white balance improvement, there are still many aspects of computational photography in smartphones that camera manufacturers can learn from, and Aozhixin is obviously at forefront of synthetic imaging.
Editor | Bi Chao