Samsung recently unveiled a 200MP ISOCELL HP2 camera sensor with improved pixel technology to capture beyond pro-level photos in ultra-high resolution. Samsung Galaxy S23 Ultra smartphone is going to have this powerful camera sensor.
Through Samsung’s 200MP HP2 camera sensor, the company is pointing to a different tech tree this time around. It is most likely Galaxy S23 Ultra main camera CMOS with a size of 1/1.3 inch, 200MP 16 in 1, and a single pixel of 0.6μm. However, we’re not surprised by the size, let’s check other perspectives.
HP2 camera sensor adopts D-VTG (Dual vertical transfer gate) technology that greatly increases the full well capacity (FWC) of a single pixel. The well capacity is 33% higher than the previous generation ISOCELL. Most importantly, the smartphone’s CMOS with limited planar size, especially the HDR highlight suppression and the signal-to-noise ratio under sunlight.
The HP2’s D-VTG technology appeared in a paper published by Samsung in the Journal of Solid State Electronics. A transmission gate TG corresponding to the photodiode PD is changed into two, and the TG is placed vertically with respect to the PD, which is called a Dual VerticalTransfer Gate.
The benefit of D-VTG technology is that it not only has better conduction efficiency because TG controls the flow of electrons from PD to FD but it may well also increase the potential difference of charge, thereby increasing the maximum number of electrons that PD can carry.
Samsung designed a single pixel with 6,000 electrons, which has been increased to 9,000 electrons through the new technology, i.e., 1.5 times the maximum consumption, compared to the 33% improvement currently advertised by HP2.
Samsung HP2 200MP camera sensor is a full well of 10,000 electrons in a single pixel. It is important to note that the pixel size of A7R4/R5 is very large, and there are only 36,000 electrons in a full well.
Dual Slop Gain
Samsung introduced Dual Slope gain technology in the ISOCELL HP2 for the first time in Samsung’s image sensor lineup to take HDR for high-resolution images to the next level. In all pixels, it applies two different conversion values when light is converted into a digital signal.
The “two different conversion values” mentioned by Samsung are actually the readout capacity of the two FD areas, one larger and one smaller, and the smaller capacity corresponds to the higher conversion gain, the higher the conversion gain.
Conversion gain can be seen as slope, that is voltage converted into a unit of electron energy and the unit is mV/e-, which is the origin of slope in propagation.
It is actually mentioned that the FD field capacitor is switched to switch the conversion gain, but Samsung’s paper shows three capacitors playing a three-gain TCG. Sadly, in the current state, HP2 is not installed.
200 MP and Focus
The HP2 really is the right direction for mobile photography. It’s not that the HP2 has more pixels, but Samsung chose to grow the sensor vertically so it has a chance to compete with an inch, especially since the bottom half of the phone is usually equipped with a larger aperture.
HP2’s 200 million pixels is the equivalent of 4 in 1 5000w pixels, divided into four grids for Quad PD focusing, i.e. tilt and horizontal/vertical phase can be detected.
HP2 is the most balanced main camera CMOS at present. It should be considered a 50mp four-in-one, but each tiny pixel has its own focus function, as its DSG function can’t be turned on at 200mp. Meanwhile, reading speeds at 200mp are just 15fps, which is pretty low for a stack-heavy phone.
TSMC to produce 3nm Snapdragon 8 Gen 4: No talks w/ Samsung
Qualcomm may stick with TSMC for the 3nm Snapdragon 8 Gen 4 processor’s production next year. Earlier, the company was rumored to shift to a multi-foundry strategy, but it won’t happen next year. Samsung also has some 3nm clients, but no flagship product has been ordered.
According to China Times, Qualcomm and MediaTek plan to use the second-gen TSMC 3nm process to fabricate Snapdragon 8 Gen 4 and Dimensity 9400. The new N3E tech offers performance and efficiency improvements over the N3B, which is used in Apple’s A17 Pro.
Reports suggest that the Taiwanese chip manufacturer is producing 60,000 to 70,000 wafers per month using its 3nm process. This number is expected to rise to 100,000 wafers by the end of next year. As of now, 5% of its revenue comes from its 3nm process, which will rise to 10% next year.
Samsung to produce 5nm AI chips for Rebellions
South Korea-based AI fabless startup, Rebellions, is gearing up for mass production of data center-specific artificial intelligence (AI) chips using Samsung Foundry’s EUV (extreme ultraviolet ) process at the 5nm level.
BusinessKorea reports that Rebellions completed a mass production contract last week with SEMIFIVE for the Samsung 5-nm AI semiconductor ATOM, with mass production set to commence early next year.
ATOM is known for its industry-leading graphics processing unit (GPU) performance and up to 3.4 times greater energy efficiency compared to equivalent neural network processors (NPU).
Particularly notable is ATOM’s performance in the Global Benchmark, where it outperformed competitors like Nvidia and Qualcomm by 1.4 to 3 times, sparking significant interest.
Earlier this year, Rebellions produced prototypes of ATOM through Samsung Foundry’s “Multi-Project Wafer Service,” which prints various types of semiconductors on a single wafer.
Rebellions is also strengthening its collaboration with Samsung, co-developing the next-gen semiconductor Rebel, specialized for Large Language Models (LLM), which is being developed using Samsung Foundry’s 4-nm process.
MediaTek Vs Qualcomm: Dimensity 9300 giving tough competition to Snapdragon 8 Gen 3
Announced earlier this month, MediaTek’s Dimensity 9300 is giving tough competition to Qualcomm’s new Snapdragon 8 Gen 3 flagship processor. Vivo recently launched X100 boasting Dimensity 9300, while most of the upcoming Android flagships use Snapdragon 8 Gen 3.
It’s worth mentioning that the running score of Dimensity 9300 is slightly better than Qualcomm’s 8 Gen 3 SoC. The new MediaTek chipset is rivaling well against Qualcomm’s matured Snapdragon chipset portfolio due to its full-core CPU.
Tipster DCS notes that the Dimensity 9300 performs well in medium and high-load energy consumption. At the same time, it’s slightly better than Snapdragon 8 Gen 3 in high-load mobile games in the same scenario.
Genshin Impact 720p full high image quality is 59.6fps, consuming 4.7W. On the other side, the 8 Gen 3 does better in low-load games, while Dimensity 9300 isn’t far behind.
Fabricated on TSMC’s 3rd-gen 4nm+ process, the all-big core CPU design is the highlighting feature of the Dimensity 9300 processor. It features a prime Cotex-X4 core clocked at 3.25GHz alongside 3x Cortex-X4 cores @ 2.85GHz and 4x Cortex-A720 cores @ 2.0GHz.
// DCS / Weibo