1-inch 8K 18 stops of dynamic range

[Emanuel]

https://ymcinema.com/2025/04/15/omnivision-unleashes-a-game-changing-1‑inch-8k-18-stops-of-dr-sensor-for-smartphones/

https://ymcinema.com/2025/09/30/omnivision-ov50r-8k-sensor-action-cameras/

https://www.ovt.com/press-releases/omnivision-launches-ultra-high-dynamic-range-1-inch-image-sensor-for-movie-grade-video-capture-in-flagship-smartphones/

 

Breaking news: currently in mass production...

 

Imagine that reaching the mirroless/compact cameras segment, no less...

Why not? ; )

 

In a single line:

economies of scale.

 

The funny part of the equation is the smartphones market pushing the boundaries of this industry in every front, from iPhone sales to Android.

 

;- )

[ND64]

Same technology used in Xiaomi 17 Pro. It has more DR than G9m2, but the lens quality and the algorithm used to combine low gain and high gain are horrible and so the extra highlight stops are wasted. 

[Clark Nikolai]

6 hours ago, ND64 said:

 

 

I was looking at the Chinese characters and wondering what rice would have to do with lenses. I looked it up and 小米 is pronounced xiao mi. Hah!

[ND64]

2 hours ago, Clark Nikolai said:

I was looking at the Chinese characters and wondering what rice would have to do with lenses. I looked it up and 小米 is pronounced xiao mi. Hah!

For anyone who knows one or two things about imaging science, it was obvious we can't go beyond 16 stops with current combinations of tech, as DR is not just about pixel's saturation. You can save harsh highlights from clipping but you get an overall contrast level that doesn't match to that saved highlights, which is what exactly demonstrated here.

[Emanuel]

5 hours ago, ND64 said:

For anyone who knows one or two things about imaging science, it was obvious we can't go beyond 16 stops with current combinations of tech, as DR is not just about pixel's saturation. You can save harsh highlights from clipping but you get an overall contrast level that doesn't match to that saved highlights, which is what exactly demonstrated here.

For sure.

The “18 stops” claim comes from the manufacturer (sensor-level, special HDR mode). It’s plausible as a sensor specification under controlled conditions/multi-exposure HDR — but practical DR measured in real cameras has been substantially lower due to optical and pipeline factors.

Although the sensor itself has a “theoretical” 18 stops, many other factors (noise, analog-to-digital conversion, image pipeline, compression, amplifier architecture, pixel readout, linearity, saturation capability) limit the actual practical performance.

 

Manufacturer's claim = sensor under optimized conditions (two/more readings, dual-gain, on-sensor HDR, internal measurements). This isn't necessarily the DR you'll get in a final phone/camera.

Actual limits: optics (flare/veiling glare), amplifiers, pipeline noise, compression, RAW/JPEG processing — all of which reduce useful DR.

 

https://www.imatest.com/docs/full.html

Imatest/DxOMark measure the system, not just the sensor die (photodiodes (light-capturing pixels), transistors, amplifiers, readout lines, A/D converters, etc.). 

Lab values applied to cameras tend to show much lower useful DR on sensors of this size — hence the large difference between manufacturer claims and practical DR.

 

However, if this OmniVision sensor actually comes to fruition with 18 practical stops, that will be revolutionary for such small sensors — it would be worth watching for announcements and technical tests.

[PPNS]

i worked more days in construction this year than i have in the film and commercial industry. sorry but who cares about pushing boundaries and constant releases of new means when there’s less and less work to use the means on?

[Emanuel]

Got you : ) But. Saying something new doesn’t matter is like refusing to plant seeds because not every one will grow. Complaining about "the new" is like grumbling at a garden: some seeds fail, but a few surprise you with a nice tree. 

They also mean democratization today. Of course, this has led to a flood of mediocre content, but it also opens doors to those who otherwise wouldn’t even have the chance to try.

I finished film school twenty-two years ago. Still kickin’ ; ) not exactly 'cause of the old-school film I got from there. : P

[sanveer]

It doesn't look 18 stops. Also the shadows are insanely noisy.

Maybe they meant 10 stops and it got translated as 18 stops. 

Also, except the noise pattern, both pics seem to have similar dynamic range. 

[ND64]

LOFIC is inherently noisy, because its optimized for saturation. Its like having a sensor that its base ISO is ISO 25, but its max ISO is 400. Thus you need a second readout optimized for low noise to combine the two. And you can't get really good noise performance in the second one until you go up in size to at least 1 inch format. 

[PPNS]

19 hours ago, Emanuel said:

Got you : ) But. Saying something new doesn’t matter is like refusing to plant seeds because not every one will grow. Complaining about "the new" is like grumbling at a garden: some seeds fail, but a few surprise you with a nice tree. 

They also mean democratization today. Of course, this has led to a flood of mediocre content, but it also opens doors to those who otherwise wouldn’t even have the chance to try.

I finished film school twenty-two years ago. Still kickin’ ; ) not exactly 'cause of the old-school film I got from there. : P

I’m not complaining about the new, i’m fairly indifferent towards it. It won’t really affect my life very much either way. I just fail to see if this is an interesting topic compared to a labor market that keeps getting worse and whatever capital is trending towards now.
 

The complete devaluation of quality in arts (and even commerce fwiw) and the tendency to constantly cut corners by automating tasks, or even straight up skipping the entire process and straight up generating content from large data models has had an extremely negative impact on my life, as well as others. The fact that no one with any political power anywhere wants to regulate it and just sees it as a given is also bad i think.

[Emanuel]

The world is changing fast. The real challenge is adapting or we’re out of the game. Brexit happened nine years ago because of this regulation you mention, and Trump (as well as others like him) rose for similar or related reasons, at least partly (the phenomenon is complex!), I’d tend to dare myself to guess.

[Ilkka Nissila]

1 hour ago, Emanuel said:

The world is changing fast. The real challenge is adapting or we’re out of the game. Brexit happened nine years ago because of this regulation you mention, and Trump (as well as others like him) rose for similar or related reasons, at least partly (the phenomenon is complex!), I’d tend to dare myself to guess.

In an unregulated state, all the money will go to the owners of the AI built on stolen data (from creatives without compensation) and no working person will have money. It'll be like the 1920s again, and remember the tariffs then made the US depression spread worldwide, leading eventually to World War II. After which a period of relative decency began, until the 1980s where all the money more and more were given to the fewest of people, leading eventually to Brexit, Trump, the Russia-Ukraine war. All of these phenomena since the 1980s happened because the multimillionaires and billionaires want to have all the money and keep it too.

 

Adapting is the same as capitulation which makes working people the equivalent of slaves. All the money will go to the techno-oligarchs and their criminal politician friends.

 

The only way to solve the problem is to make AI models based on stolen data illegal and erase them or give due compensation to the creators of the original teaching data that was used to make the model, and tax billionaires so that they end up with only the money that a decent life requires. This would restore fairness and decency in society and good lives to ordinary people.