Flashback: the second megapixel race
The first megapixel race started around 2004 (when camera phones first crossed the 1MP mark) and came to an end in 2013 or so with the arrival of the Nokia 808 PureView. Its 41MP camera would only be beaten in terms of resolution in 2018. During that period phones focused on other aspects of the camera rather than resolution. But now a second megapixel race is upon us.
The first one ended in part because common wisdom shifted from “more pixels is better” to “bigger pixels are better”. The arrival of computational photography also played a major role and allowed iPhones, Galaxys and Pixels to stick to a 12MP resolution for their main cameras for several years (in fact, Apple only now left 12MP behind).
These days things have swung back towards “more pixels is better”, although not completely – the current trend is actually a mix of the two approaches. We already have phones with 200MP sensors as well as phones with 1” sensors.
We will take a closer look at the two branches that have emerged, starting with the “more pixels is better” branch today and leaving the other one for next time.
The second megapixel race brought about the proliferation of 48MP sensors, which became quite popular on mid-rangers (with flagships still preferring larger sensor size to higher resolution). Let’s look at the Samsung ISOCELL GM1 from 2018, for example. It’s not huge at 1/2” optical format, but it had 0.8µm pixels, which grew to rather big 1.6µm with binning.
Here we have to take a small detour to talk about the Bayer filter. We wrote a detailed explanation back in the day, but Bayer – and Quad Bayer and so on – filters are central to today’s topic. The GM1 sensor used Tetrapixel technology, Samsung’s term for Quad Bayer. It looks like this, four neighboring pixels sharing the same colored filter square. This makes it natural to combine the four into a single output pixel (4-in-1 binning).
There are now sensors that cover 3×3 groups and even 4×4 groups of pixels with the same color filter, they use 9-in-1 and 16-in-1 binning, respectively. The 12MP resolution we mentioned earlier didn’t go away – 108MP and 200MP cameras still target 12MP as the final output after binning. That makes for a good default setting as you get enough resolution to zoom in but don’t have to juggle photos that take up tens of megabytes of storage.
Back to the megapixel race. As high resolution sensors started becoming the standard on mid-range phones, there was a push to keep costs low and that meant only one thing – smaller sensors.
Whereas the GM1 had 0.8µm pixels, the 48MP ISOCELL GM5 from 2020 dropped to 0.7µm, making it a 1/2.55” sensor. The JN1 from 2021 went even smaller with 0.64µm pixels, so despite its high 50MP resolution it only had a 1/2.76” optical format.
Samsung is not alone in using tiny pixels, for example, OmniVisions OV60A is a 60MP sensor, 1/2.8” optical format with 0.61µm pixels and a Quad Bayer filter. There are also larger sensors like the 1/1.34” OV64A, but again we’ll talk about those next time.
Alright, we’ve covered pixel sizes and Bayer filters, it’s time to cross the 100MP barrier. The first sensor to go beyond was the Samsung ISOCELL Bright HMX. Its full resolution was 12,032 x 9,024px and it had 0.8µm pixels, making for an optical format of 1/1.33”.
The first phone to use it was the Xiaomi Mi CC9 Pro (it was supposed to be the Mi Mix Alpha, but that was canceled). You can check out our hands-on review for camera samples. The phone defaulted to shooting in one quarter resolution, 27MP, with pixel binning.
Another 108MP 1/1.33” sensor is the HM3, which also has 0.8µm pixels and was used in the Galaxy S21 Ultra. However, this one does 9-in-1 binning, defaulting to 12MP resolution. Like with the 48MP sensors things may have started at 0.8µm, but quickly started going down – at 0.7µm and 108MP we have the likes of the 1/1.52” HM2, then at 0.64µm and 108MP resolution there is the HM6, a 1/1.67” sensor.
We already mentioned the JN1, another 0.64µm sensor. As you can probably tell, sensors can be grouped by pixel size. For example, Samsung built several sensors on its 0.7µm tech:
OmniVision has a pair of competing sensors. The OVB0B has 0.61µm pixels, the OVB0A matches the HP3 at 1/1.4” and 0.56µm.
200MP is as high as current smartphone cameras have gotten. However, Samsung is said to be working on sensors with up to 600MP resolution, so this isn’t the end of the road.
Before we wrap up we should quickly go over the advantages of having that many pixels. The first is obvious, bragging rights. We know that marketing departments love this, especially when they can slap a “First!” label on it.
But there are practical advantages too. Digital zoom has greatly benefited from this – the sensors that use pixel binning can usually do lossless digital zoom at the same factor (e.g. binning 2×2 pixels and zooming in 2x). Even when forced to do interpolation the end result is better because there are more pixels to work with.
Without a motorized lens this is the only way to achieve smooth zoom (e.g. in videos). Motorized zoom lenses that are tiny enough for modern smartphones are already on the market, though they are extremely rare.
Another cool use for high-resolution sensors is to treat them as two or even separate sensors. For example, half of the pixels can shoot at low ISO and the other half at high ISO, which can then be combined into a single image that has detail both in bright and dark areas. Staggered HDR does a similar trick except it varies the exposure time (low, medium and high).
That’s it for today, next time we will look at the other branch and we will track the growth of smartphone image sensors up to the 1″ mark. We’ve had 1″ phone cameras before, but the Panasonic CM1 makes even the Xiaomi 12S Ultra look slender.