Does sensor size matter?
Does sensor size matter?
There are many interchangable-lens camera systems out there. Choosing one is a difficult decision, since these are expensive gear, and the user tends to "marry" with the system he/she chose. A "divorce" is always possible, but every divorce has a financial penalty, so it is understandable that everybody wants to make it right on the first try.
Some systems:
Everyone fears investing in a system and end up orphaned. Among the above mentioned systems, the last three have an uncertain future, even though A-mount and Four Thirds offer a migration path to E-mount and Micro Four Thirds, respectively.
There has been gossip that Samsung would abandon the MX system — in fact it disappeared from the Samsung brand store here in my town — this would mean a total loss for the respective users... Fact is, sooner or later there will be some market consolidatoin and the list above will shrink to half its size in ten years.
Or perhaps the modus operandi of photographers should change; clinging to a system is an obsolete behavior? Perhaps it would be better to buy whatever is best today, and stop worrying about the future. But this is another discussion.
The systems above may be grouped in four sensor sizes:
In the case of point-and-shoot cameras, whose lenses cannot be interchanged, the decision is not as loaded because it is restricted to the cost of one camera, that will last around four years until it breaks or the user is fed up of it. Most simple cameras use sensors smaller than 1". If one can pay the price, he/she should look for the biggest sensor that fits in the budget (1" or bigger) when buying a point-and-shoot.
There is a huge discussion regarding the best sensor format for a interchangable-lens system. In no other momemt of the history of photography so much ink has been expended as it is in the discussion full-frame vs. APS-C vs. Four Thirds (fortunately, most of the spent "ink" is digital).
As a general rule, the bigger the sensor, the better the image quality. This is beyond discussion because it is a result of laws of physics. The best thing would be those enormous bellows cameras with an 8"x10" sensor inside. A big sensor has a number or general advantages:
In spite of how much small-format proponents insist that sensor size does not matter... it does matter.
A smaller sensor also has its advantages, but most of them are pragmatic:
A terrible collateral effect of digital photography is to render larger formats unaffordable. While a film sheet has a price linearly proportional to its area, the cost of a digital sensor grows exponentially. The silicon chips are made in batches, on "wafers" or circular slices of pure silicon, around 12" in diameter. A certain number of random defects plagues every wafer, and some chips of the batch won't work. The bigger the chip, the higher the chance of getting one of these defects, and smaller the net chip production, or "yield", per wafer.
Suppose that converting a wafer in chips costs $5,000 per wafer, and each wafer has 10 defects in average. If 100 chips fit in one wafer, the yield is at least 90% (90 good chips per wafer), costing $56 each. If only 15 chips fit in one wafer, the yield is only 5 good chips, priced $1,000 each. A chip that is just six times bigger has a 18-fold cost.
A coincidence of destiny is, the photographic world had standardized around 35mm, a relatively small size of film. It is feasible to manufacture digital sensors this size, even though they are still expensive. If photography had settled around medium or large format, the digital photography would take much more time to catch on; or a change of paradigm would have to happen.
Around the 1990s, a full-frame sensor still cost something like US$ 30,000. At first, the pro digital cameras settled on APS-C format, which is around 33% smaller in each dimension. The first DSLRs cost "only" US$ 5,000.
The size difference is most often expressed as a crop factor (1.5x in case of APS-C) to simplify the comparison with full-frame, as well as the relationship between focal length and field of view. A 50mm lens in full-frame offers approximately the same view as a 35mm lens in APS-C (35 x 1.5 = 52.5). This kind of conversion was and is important because Nikon and Canon kept some degree of compatibility of full-frame lenses in APS-C cameras.
In time, the price of sensors fell down, and the best DSLRs went back up to full-frame. However, the APS-C models are still available, since the price difference is still high. A full-frame sensor costs 20 times more, in spite of having just twice the area. A full-frame DSLR is still too expensive for impulse purchases — which is how so many people "marry" into a a system. Finally, there is a whole ecosystem of lenses for APS-C formats (Nikon DX and Canon EF-S).
And the discussion goes on: APS-C or full-frame?
As mentioned before, a bigger sensor is always better. But the difference between full-frame and APS-C is not that big as the specialized media makes it. The central point of the debate is the loss of some power of controlling depth-of-field, that is, the capability of generating a very blurry background (bokeh!).
For a given field of view, the aperture should be divided by the crop factor to estimate the impact on depth-of-field. For example, consider a 50mm f/2.8 with full-frame sensor. In order to get the same shallow depth-of-field with an APS-C sensor, while keeping the field of view, it takes a 35mm f/1.8 lens (which is approximately 50mm divided by 1.5, and f/2.8 divided by 1.5).
This is a sensible diference; for example, there are no practical means to replicate the looks of an f/1.4 lens on APS-C cameras, since it would take an f/0.9 lens. But, in practice, most photographers don't possess an f/1.4 lens.
Unless the full-frame sensors get really cheap, to the point that every cheap camera gets one, I think that APS-C is a good compromise between price and quality.
Manufacturers like Samsung, Fuji and Canon must agree with me, since they have created their mirrorless systems around APS-C sensors. And creating a new system is not a walk in the park; it is necessary to develop dozens of lenses, and sell millions of them to recover capital and R&D costs (and a system must sell many million units to achieve any degree of respectability).
Sony adopted the shotgun approach, offering no less than four systems: the combinatorial explosion between two mountings ("A" and "E") and two sensor sizes (APS-C and full-frame). As I write this, everyone is paying attention to the A7R II camera, which is mirrorless, E-mount and full-frame.
And what about the systems with even smaller sensors, like Four Thirds and Nikon 1?
Nikon 1 has a crop factor of 2.7x, quite a strange choice for a fairly recent (2011) system. I don't believe it is viable, since the depth-of-field control is severely limited. Getting the looks of a 50mm f/2.8 full-frame in Nikon 1 would take a 18mm f/1.0 lens.
If it were the only digital camera of the world, perhaps the aestethic standards would give in (as the current aesthetic tastes are rooted on cinema, in my humble opinion). But it is not the case. Adapting DSLR lenses on Nikon 1 cameras is pretty useless, unless the user seeks an extreme telephoto (a 200mm lens gets the field of view of 540mm).
(Note that aesthetic tastes may indeed change with time, and they might be changing right now. With so many people making selfies and pictures using cell phones, it may well happen that this generation grows accustomed with uncontrolled depth-of-field, and people might begin to prefer cameras with smaller sensors, that better mimic the cell-phone-picture looks.)
The Nikon 1 cameras are shock-full of nice features e.g. burst rates of 60 frames per second and fast autofocus. Features like that could, just by themselves, carve a market niche. I won't deny that Nikon 1 gadgets are beautiful, as well. On the other hand, these things increase the angst about how good would be an hypothetical big-sensor Nikon mirrorless.
The Micro Four Thirds system is more intriguing. The 2.0x crop factor is high, but not too far from APS-C. A novel thing on 4/3 is being a standard shared by many manufacturers, including aftermarket device makers. This is something new in photography and increases the odds, because the system can serve many different markets. Even 4/3-compatible drones can be found these days. (Drone cameras must be light and drone photography is one of these game-changing things.)
My biggest fear about any current system: maybe tomorrow the technology changes abruptly, full-frame or even bigger format sensors become dirt-cheap, and whoever made a big investment on lenses will carry the bag. It is improbable that a change like this would happen so fast — the semiconductor industry is struggling with wafer sizes and chip yields for decades. (If it does happen, there is a bright side: we can all migrate to medium-format or large-format!)
The 4/3 cameras are also great pieces of technology, since they belong to a system thought for digital photography from the ground up. Even a pro DSLR looks like a Soviet-era device when compared to a 4/3 mirrorless. But again, this comparison makes people crave for mirrorless with bigger sensors. The 4/3 cameras are so good that they are not cheap at all, yet the sensor is rather small.
In spite of the technological advances, I still think that any system should use the biggest sensor that is still affordable. And the sensor that fits on budget right now is the APS-C. In this sensor size, we have the Nikon DX and Canon EF-S systems that need no introduction. The hipsters and early-adopters have other options.
If one privileges consistency, the Canon EF-M system makes the most sense. It has a new lens mount, but the EF and EF-S lenses are treated as first-class citizens. The camera kit bundles the adapter, while other brands charge several hundred dollars for a similar adapter. The EOS-M cameras are not ambitious, to the point that they are not found worldwide yet (each model is released just on one or two continents). Canon is not being agressive at all on mirrorless market, but at least it brandishes a roadmap. And this roadmap is based on the APS-C sensor.
The Fuji X and Samsung NX system cameras are prettier and more ambitious than Canon's. I had an X-E2 and an NX-200 in my hands, and both enchanted me. They are attractive enough to inspire impulsive purchases, and they are (as usual) loaded with technology, much more than current DSLRs. (I would tell the same about Nikon 1 and Olympus, if it weren't by the sensor size. Among all cameras, the Olympus ones are the most beautiful, IMHO.) Technological gimmicks don't take better pictures, but these things make life happier, which is important as well. Unfortunately there has been gossip about Samsung quitting the market. At least the Fuji system seems to be doing very well.
The biggest question mark is Nikon, that could launch a mirrorless system any day, hopefully with a good degree of compatibility with Nikon F mount. Personally, I would use the Nikon F mount, period; the extra 20mm of flange depth would not hurt the general camera size that much. The half-hearted Canon approach makes everybody sit on hands, waiting to see what Nikon will do next. This depresses the market as a whole, since (I guess) many people are waiting instead of shopping.