All mobile phone vendors hope to repeat the success of Apple with a new Jesus phone. Sony Ericsson’s candidate is Satio (previous working name Idou) and they are positioning the integrated 12 megapixel camera as the “killer app” that will lift Satio above the other smartphone competitors when it hits the shelves this fall.
Unfortunately, I think they have made a serious design mistake in their choice of camera. For marketers that are used to the IT industry’s logic that “more is always better” it is a given that cramming as many pixels as possible into the smartphone is a desirable goal. Sony Ericsson’s motivation is that the Facebook generation wants an easy way to take high quality pictures. That is correct, but a 12 MP camera phone is not the answer.
The problem is that a sensor with a lot of megapixels diminishes the ability to take good pictures in low light which, for most users, is much more important than taking high resolution pictures with 4000 x 3000 pixels. Sony Ericsson (SEMC) will run into problems when disappointed buyers realize how limited the camera is in practical usage.
12 megapixels used in the tiny sensors in a camera phone is not the same as 12 MP used in larger DSLR cameras. The number of megapixels is not the only performance factor. How tightly each pixel is packed on the sensor (Mpix/cm²) is of equal importance. If too many pixels are crammed on a sensor that is less than a square centimeter (6 × 4.5 mm if Sony’s new sensor is used in the Satio) each pixel will be so tiny that the physical limitations of the number of photons that can hit this pixel will determine the capability to produce an image.
If the camera sensor is viewed as a football field filled with buckets, the photons can be viewed as a rain of billiard balls that fall down in the buckets. Each bucket is one pixel. Stronger light produces many billiard balls which increases the precision when the number of balls in each bucket is counted. Weak lighting conditions (few billiard balls falling down) might work if each bucket is large enough to at least catch a few balls. But if the same number of buckets is crammed into a handball field each bucket will only be the size of drinking glass and the errors (noise) will be much larger because the billiard balls are too few to fill up all the small glasses.
For pictures taken in direct sunlight in the middle of the day or with a strong flash, the lowest sensitivity (ISO 100) is sufficient. Under these lighting conditions a small sensor (the handball field) is almost on par with a large sensor (the football field). There are still differences in quality because the optics in a small sensor camera is always inferior to larger, high quality optics. Quality differences are also caused by the fact that the leading camera vendors (Nikon and Canon) have more experience with electronic image processing than new players such as SEMC.
But as soon as you take pictures in low light and have to increase sensor sensitivity to ISO 400, 800 (or even higher) the difference between a large and a small sensor with the same megapixel count becomes dramatic.
In web forums such as Nikonians and review sites like DPReview most pro photographers and photo nerds have been in agreement about this for a long time. They are generally skeptical about small point-and-shoot cameras since the image noise level becomes unacceptable indoors or with low lighting. They are also critical about the way the vendors try to compensate for mediocre image quality with exaggerated electronic post-processing, by saturating the colors and by increasing edge sharpness. The result is quite often pictures that look unnatural. The table below shows the differences in the sensor area and pixel density for a selection of digital cameras and camera phones.
Sensor format and pixel density
|Canon EOS 50D||15||22.3 × 14.9 = 332.3||4.5|
|Nikon D90||12||23.7 × 15.7 = 372||3.2|
|System cameras with full-frame sensor|
|Canon EOS 5D Mark II||21||36 × 24 = 864||2.4|
|Nikon D700||12||36 × 23.9 = 860||1.4|
|Nikon Coolpix P6000||13.5||7.40 × 5.55 = 41.1||32.8|
|Olympus Stylus 9000||12||6.13 × 4.60 = 28.2||42.6|
|Samsung HZ10W||10||6.13 × 4.60 = 28.2||35.4|
|Fuji FinePix F31fd||6||7.60 × 5.70 = 43.3||13.9|
|Sony Ericsson Satio||12||6 × 4.5 = 27||44|
|Sony Ericsson K850||5||5.22 × 3.91 = 20.4||24.5|
|Nokia N95||5||4.86 × 3.6 = 17.28||29|
Notice in the table above that the value for pixel density (megapixel per square centimeter) should be as low as possible. That figure is a measure of how densely the pixels are packed on the sensor and it is of almost equal importance as the number of megapixels. With a sensor area of 27 mm² Satio will have a sensor in the same range as the smallest point-and-shoot cameras. This will give Satio a pixel density of 44 megapixels per square centimeter which is very high.
Another observation from the table is that there are large differences in the sensor area between the camera types. The sensor in most DSLRs is around 350 mm². Compare this with the sensor in compact cameras which is between 28 and 40 mm². The sensor in a DSLR has an area that is twelve times larger than the smallest compact digital camera, thus providing twelve times higher capacity to collect light.
The first vendor to use a small sensor with fewer megapixels which is optimized for maximum low light performance will capture an empty market segment. There are no products on the market today to fill this latent demand. A few years ago Fuji released the Finepix F31fd camera, a 6 Mpix compact camera using a larger sensor with very good low light performance. It was even possible to take decent night pictures without a flash at ISO 1600 which no other compact camera had ever managed. When Fuji discontinued the model the camera gained cult status and the used prices on eBay have sometimes been close to double the new price.
Of course a company should not design their products for the nerd market. However, it is worthwhile to listen to advanced users. In this case they are right, high sensitivity in low light is relevant for everyone. Taking pictures indoors of your friends without being forced to use flash is perfect for the Facebook generation.
That the mass market customer believes “the more megapixels a camera has the better it is” is natural given that almost all marketing from the vendors has focused on megapixels. However, sooner or later the well-founded criticism of the megapixel obsession (e.g. from the New York Times’ technology editor) will bring awareness to the general public.
The only major advantage with a high megapixel sensor in a camera phone is that you can zoom in electronically without a disastrous loss of resolution. The electronic zoom only provides magnification from the center of the sensor if the lens has no moving parts. However, the marketers’ obsession with large zoom range (12x for Satio) is also criticized by those who actually use these cameras. Even if a camera has image stabilization is not enough to compensate for the fact that light sensitivity and resolution deteriorate significantly with electronic zooming. The New York Times points out that a good wide angle range is more important (so you can take pictures of all your friends around the dinner table) than useless telezooms.
A manufacturer that launches a camera (or cam phone) with a smaller zoom range and superior low light performance but with fewer megapixels can easily communicate user value. A marketing message could be built on comparing test pictures or copy text such as: “No flash at the wedding”, “Say goodbye to washed out flash images”, “Take pictures without anyone noticing”, etc.
If Sony Ericsson wants to reduce the pixel density my suggestion is that they develop a much more light sensitive sensor of the same size with a resolution of 2400 × 1800 pixels. This resolution will deliver 4.3 MPix pictures and the pixel density will be 15.9 Mpix/cm² which is much better than 44. A resolution of 2400 × 1800 pixels should be comparable to the resolution of a typical PC screen. Most screens have a lower resolution of 1680 × 1050 or 1920 × 1080, which means that the picture’s size has to be reduced in order to fit on the screen. Even larger 24 inch and 27 inch screens don’t have a higher resolution than 1920 × 1200. (Since most digital pictures are only displayed on a screen and never printed on paper, the higher resolution requirements for printing are not as relevant here.)
Sony Ericsson’s Satio as a case study
For Sony Ericsson’s sake one has to hope that the Satio will be well received by the market and that all the other advanced features are enough to convince the customers, in spite of the problems with the camera, to buy the smartphone. SEMC’s choice of camera is of general interest for product strategists in the tech sector and I will use it as case for an analysis of how the process went wrong (from my perspective as an outsider) and give some free advice to SEMC:
- Do not allow internal politics within the Sony Ericsson alliance to govern product development. If Sony was given the responsibility of product development for the camera just because they have their own camera division it was mistake. As a camera maker Sony is weak brand. For instance, their DSLR division is built on Sony’s 2006 acquisition of two budget brands (Minolta/Konica) that were too weak to survive the competition from Nikon and Canon. Just because Sony developed a 12 Mpix sensor that they adapted for camera phones does not mean that SEMC should be forced to use that sensor.
- That SEMC did not pay attention to photographers’ discussions on forums and in product reviews is a sign of weak competitive intelligence and consumer insight. Criticism of the megapixel race and clumsy image post-processing is readily available on the Internet. Even though SEMC has no intention of developing cameras they need to be competent buyers of the camera module.
- SEMC (or another vendor) should be brave enough to go against the group think in the industry. Unfortunately, most consumer camera manufacturers have been technology driven in the wrong way, for example in their obsession with zoom range and megapixels. (My impression is that SEMC is stuck in the same pattern. A few years ago I spoke with one of the senior executives at a conference about the downsides of the megapixel race. My concerns were completely dismissed.) In addition, the vendors have also had a Microsoft-style ambition to fill their products with 100s of unnecessary, silly functions such as smile detection. At the same time they have neglected to implement basic functions such as lossless rotation of JPG pictures. (The last time I checked this in 2007, several compact cameras degraded image quality if they were rotated 90 degrees.) By breaking conventions and getting it right it is possible to gain a strong competitive advantage. Just look at the iPhone.
- SEMC should have procured technologies from leading vendors with strong brands. Nokia uses Zeiss optics in their high end camera phones, which improves the image quality somewhat. SEMC could have done the same or partnered with Leica. In the built-in camera module Sony sells as an add-on to their new 12 MP sensor the lens is made of plastic. Plastic!
- Using the same argument, SEMC should have bought the software for noise reduction, color balance etc. from Nikon or Phase One (though Nikon would probably refuse to sell their technology to a company affiliated with Sony).
- SEMC should already prepare for an immediate launch of the next model ”Satio II” this fall if Satio fails on the market. It should use a 4.3 Mpix sensor designed for low light and use a smaller zoom range (3x).
This article has previously been published on my Swedish blog.