The first thing to know is that light diffraction is a very general phenomenon and quite natural. It’s been a long time already that scientists and engineers noticed that -on the one hand- light rays are slightly deflected while running through a very small orifice and -on the other hand- two light rays can interact with each other on the condition that they are have very paths nearly perfectly parallel. This is even one of the most significant and founding elements of quantum physics. When you combine both those phenomenons, you will notice that light going through a very small opening like camera lens diaphragm will produce not exactly the expected neat circle but a more irregular and circular shape that you can see in the illustrative figure that goes along this paragraph: Usually named an Airy disk or Airy disc.

As a matter of fact, the size of the Airy disc can be computed relatively easily and only depends on the opening of the diaphragm where light goes through.

As this is only happening at the microscopic scale, you could think that this has no importance. But there is an additional difficulty for the digital photographer: Silicon-based image sensors are really working at this scale. For example, here are the sizes of some of the common images sensors on the market:

• APS-C sensor (1:1.5 crop factor / Sony SLR) of 6 mega-pixels: 8 µm
• APS-C sensor (1:1.6 crop factor / Canon SLR) of 8 mega-pixels: 6.5 µm
• APS-C sensor (1:1.6 crop factor / Canon SLR) of 12 mega-pixels: 5.3 µm
• Full Frame sensor of 12 mega-pixels (Canon 5D): 8.5 µm
• Full Frame sensor of 24 mega-pixels (Sony Alpha 900): 6 µm
• 1/2,5″ sensor of 8.2 mega-pixels (HP Photosmart R937 point-and-shoot): 1.7 µm

As you can notice all cameras are not born equal from this point of view. For the same number of pixels, Full Frame sensors have notably larger pixels than APS-C cheaper SLR cameras. Point and shoot compact cameras are in a completely different range. And phone cameras are pushing the lower limit of technology for smaller sensors (and correspondingly small pixels).

When you compare the values from the table of Airy circle sizes and the size of the pixels of your photo camera, you easily understand that closing the diaphragm of your top-notch lens is not enough to produce super-crisp images out of your brand new 12-mega-pixel camera: a 12MP Canon SLR will hit the Airy disc size as soon as you close down to f/5.6/; The images do not immediately become bad, but your pixels are not used at their best. And, sometimes, you wonder why you don’t get better results than from the older camera still in your bag.

A good algorithm of image correction may compensate part of this problem (as this phenomenon is very predictable and easy to modelize with computers, it is relatively easy to counter-act with software), but there limits to what firmware can do in our cameras, too. The situation is simply desperate with the smallest sensors of small point-and-shoot cameras and phone cameras. More pixels bring them nothing except bigger images files on Flash cards.

On the contrary, camera manufacturers firmly aim at larger sensors (so-called Full Frame sensors) in order to compensate these issues. This is what is driving the current rush to Full Frame at Canon, Nikon, Sony and the others despite the still prohibitive cost of the huge chunks of silicon used at the heart of these photo cameras.

Of course, do not refrain from closing the diaphragm to improve the depth of field of your image, but keep in mind the magnitude order of the limit diaphragms of your camera in order not to be surprised by the appearance of some diffraction in some of your images.

Additional information:

• Diffraction & photography, at Cambridge in color,
• Airy disk, at Wikipedia.

The arrival of the Nikon D700 with a Full Frame sensor of a lower resolution than the high-end cameras like the Nikon D3 (or the Sony Alpha 900) had people murmuring that maybe (and I have to heavily insist on this maybe word) Nikon and Sony would share a 12-16 million-pixel Full Frame sensor. Nikon would launch a Nikon D700 and Sony would launch a similar Sony Alpha 800 sometimes nicknamed “flagcheap” (“flagship” designating the top-of-the-line Sony, the Alpha 900, “flag-cheap” is a word game on “cheap” for a cheaper camera).

It is impossible to tell if there is any truth to it. Assuredly, the lack of recent information about the Sony Alpha 900 (and most importantly its price target) had heads spinning and marketing laws or logic law being what they are, plugging a price hole in the Sony SLR camera range may become an urgent necessity. Sony could fill it with the Alpha 800. And it previously occured that Sony hid some products behind the visibility of another camera (see the example of the Alpha 300/350).

As you can notice, it is expected that the Sony Alpha 900 will be targetting low prices despite it’s 24MP sensor (or it will have a little Alpha 800 brother, while the Alpha 900 would shoot for a higher price point). But the current Canon EOS 5D seems to be an astonishingly good bargain (while stocks last).

Nikon Rumours affirmed last week that the Nikon D700 (or D700 FX) will be priced just under 3000 USD. It has been said that the Canon EOS 5D MkII should arrives around 3500 USD, but the Canon EOS 5D is already sold under 2000 USD now. And the Nikon D3 (already previous generation camera) is still near 5000 USD, except in some very small on-line shops whose reputation is a bit low.

Not taking into account the actual performance and exact features of each, I would say that Sony will be able to position itself around 2000 USD which is the price point that Canon is setting for all the others. But keep in mind that there is a large body of amateurs enthusiasts that are ready to pay much more than the usual 600-1500 USD price tag for a digital SLR, but the competition is heating to bring Full Frame capability to the masses.

Your guess is as good as mine, but I would say that the tendency is to prepare to a very low price point. 2000 USD will be considered a turning point (notice how Canon is letting the price of the EOS 5D plummet down). A camera willing to position itself as top-quality Full Frame will probably aim at $2500 (I think the price of the Canon EOS 5D MkII or a possible Canon EOS 6D could climb there, not even speaking of a possible Nikon D3x in 2009 with a stratospheric price), but a product aimed at a larger market would have to start around $2000 and preparing for a street price going under this limit (this should be the case of the Nikon D700 -the name is referring more to the D300 than to the D3/D3x pro family- and of the Sony Alpha 900).

And it’s a good question to ask, but the answer goes with some of the comments about sensor resolution. About the same people who noticed that resolution is not all that counts (despite what is said or implied by some of the major brands on this market) also understand that going Full Frame is a possibly very important decision for the attentive digital photographer.

WTF is Full Frame?

First you’d like to understand what we are speaking about – precisely. Most Digital Single Lens Reflex (D-SLR) cameras use a sensor whose size is about half of the size of a usual (old?) 35mm film frame. We call them APS-C-sized sensors because it’s not far from the (younger but still old?) APS-C film frame size. Since so many photographers have been trained in the world of 35mm films, everything tends to be compared to it.

Those small sensors are used by camera manufacturers because the smaller the silicon chip, the cheaper it goes. This is the main reason why phone cameras use even smaller sensors (a few millimeters wide for some of them).

Full Frame sensors are those ones that are nearly exactly the same size as a 35mm film: 24mm by 36mm sensitive surface. By silicon manufacturing standards this has long been too big to be reasonably manufactured in large quantities but progress being what it is in this industry (Moore’s law, you know?) it is now becoming feasible. So why not?

The pros of FF

The first one is that if you stick to a fixed number of pixels (say 10 millions of them on one single sensor), the wider the sensors, the larger the individual pixel sensitive element. This has several enormously interesting advantages.

Being big, the pixel is also much more sensitive (it can collect more photons at the same time). So, there is less need to do mathematics and algorithmics to compensate for low light situations. The sensor collects directly enough light and the image is less impacted by noise. Like when you push the volume up on a radio to hide the background noise of a conversation. Those big pixels naturally “push the volume up”.

And noise is a real problem. On very high ISOs and/or on low quality cameras (read “small sensors” like we find them in camera phones or even compact point-and-shoot cameras), you keep shooting images that are dull, covered with little colored specks or lack in details because the camera manufacturer forced on “noise cleaning algorithm” and this removed most details.

Parallely, if you pixel is large and collecting a lot of light photons, it is easier for the electronics to cut it in small pieces: you get better dynamic range. It means that you will find easily more details in the dark areas of the image and you will mostly avoid the blown-out white zones of some light areas. This is a good thing for your photos.

But there are also some other even more subtle but important issues at hand. Many photographers migrating from analog to digital SLR cameras have noticed that the viewfinder now looks like a tiny little tunnel: everything seems to be small and dark. Since the sensor image is smaller, it is sent back to the eye in a smaller size too. It is uncomfortable unless you invest in a camera with a top of the line viewfinder. But if the sensor is Full Frame, it is easier for the camera manufacturer. They don’t have to enlarge the image, it’s already big.

A much more technically difficult argument is something that is now cropping up with large pixel numbers. When we had 6 million pixels on an APS-C sensor, nobody noticed it, but with 10, 12, 14 millions pixels people start to notice that the increase of resolution does not always go with improved image quality. One of the reasons is linked to light diffraction when you close the diaphragm. The pixels are getting so small that the slight diffraction from a lens closed at f/22 or f/16 is becoming larger than a pixel (or at least perceptible there).

Manufacturers (and intensely attentive photographers) now understand that if they want to keep getting more pixels, they -first- have to buy extra-high-quality lenses (at a price) and -second- they must get bigger pixels (and bigger sensors). 20 million pixels on an APS-C sensor will probably never be significantly better than 10-12 millions of them. You have to go to Full Frame if you want more pixels! It essentially gives you more real pixels and an extended diaphragm range: in most cases, it is understood that around 12 millions pixel on an APS-C sensor, closing to more than f/16 is just breaking down your picture.

As a side note, this also explains why so many small cameras and phone cameras are producing really ugly images out of an astonishing number of pixels…

Questionable advantages

Another bane of small sensors is that they have so much Depth of Field (DOF). Everything seems to be in focus. This is good for a digital photo notebook like your camera phone. But if you want to go a little more on the artsy side, you may want to ensure that the subject is neatly focused and leaving the background in a warm fuzzy light. This is more difficult with APS-C sensors than with full frame sensors. However, you have to admit that this is a mixed advantage. For example, macro photography is always lacking in depth of field. APS-C has been bringing a little advantage here.

Also, APS-C sensors being smaller, they tend to crop the image. So, the standard 50mm lens you used on the old analog camera looks like it is giving you the same image as your old 75mm or 80mm short tele lens when mounted on a Digital SLR camera. This is what is called the crop factor. Depending on your camera (or, more precisely, its sensor), it gives you a small conversion factor (1.6 for most Canon EOS, 1.5 for most Sony and Nikon). This is a nightmare for people in love with wide angle lenses and wide panoramas (it’s becoming more and more difficult to get a wide angle lens: super wide 16mm now looks like a wide 24mm). Many nature landscape photographers suffer from that change. However, being a wildlife photographer myself, I admit that this can be good too: My collection of tele-lenses got a boost going from Minolta analog to Sony digital; The latest addition, a 400mm f/4.5, is nearly exactly equivalent on my Sony Alpha 700 to a 600mm f/4.5 on the old Minolta Maxxum 9xi but it is about half the weight and half the price). Not everybody can be happy with that.

The dark side of the Full Frame

But there are also very bad aspects of these large Full Frame sensors. First, we already said that electronics manufacturing is utterly sensitive to surface. The larger a chip, the more expensive. And not only because of the added material. Defects on a silicon chip tend to be spread quite evenly on its surface. The larger it is, the more defects you have, the more difficult it becomes to manufacture, the more parts you have to scrap (if your chip is twice bigger, a single defect will kill it, but on a smaller one, you would get one good chip and one bad one). Price increase has long been so large that it is not mentionable. Now, things are a bit easier, but… it’s expensive.

Also to mention, bigger sensors mean bigger cameras (your phone camera is small because its photo sensor is small). Bigger means heavier. There is a definite difference between a 300g camera body and 800g camera body. Keep it for more than an hour while shooting! Some photographers may not really care (I often hold a camera behind a lens that is more than 2kg -I previously mentioned a 400mm lens of 2.9kg- and I’ve got used to it but many people would not even think about it. The first time I bought a second-hand Sigma 400mm/5.6 the previous owner was selling it because he had a back ache prohibiting to use it).

But you could also say that you don’t really care if the image is good. And here comes the size of the lens again. To cover the full surface of a Full Frame sensor you need a lens that is large enough for that. It must also be of perfect quality to the extreme corner of the image. But if you get the light in all corners you also have to ensure that the light rays are not suffering from too many angled paths through the lens. If you do not make sure of this during the design, it’s easy to get a correct lens for APS-C, but a pitifully bad one for Full Frame. Quite often, these are difficult goals to reach and you get vignetting (the corners of the image are a little darker than the center) and soft corners (zones that are not as crisp as the center of the image). The larger the sensor, the more difficult to make a good lens for it (again). You go to the pro-quality lenses. Again, there is a price hike and added weight on the scale.

So what?

In the end, the choice to go Full Frame is currently reserved to prosumers or top-flight amateurs or even pro photographers. But we have to keep in mind that the trend will be going there just to keep the quality while other factors are climbing. Don’t believe that electronics progress will remove all limits (diffraction is an optical phenomenon we have to live with). Not everybody will have to go there, but Full Frame is there to stay.

You may not want to invest in Full Frame, but you must take that into account when you buy a lens. Many cheap lenses are only able to serve APS-C sensors. While this was a sensible investment (pay less for the same service), remember that a good lens is supposed to live much longer than your camera (digital cameras have gone the way of computers: 2-year obsolescence is the standard). This is why so many Full Frame (even second-hand) lenses are keeping a high market price. This is why Sony and Carl Zeiss still only manufacture lenses able to work on both APS-C and FF.

When I buy a tele lens, I tend to pay a hefty price on second-hand hardware (more than 2000 Euros on the old second-hand Minolta APO G High-Speed 400mm f/4.5 I got from eBay). I intend to keep that much longer than the Sony Alpha 700 and the Konica-Minolta Dynax/Maxxum 7D I use as SLR bodies behind it. If (and I say if) I go and buy the future Sony Alpha 900 Full Frame camera which is supposed to be available some time this year, possibly at the end of the Summer 2008, all my lenses will still be compatible with this camera body.

So, even if you don’t care about the added quality, think about the rest of your investment in the context of Full Frame sensors.

According to Amateur Photographer, Samsung is preparing a Full Frame D-SLR camera to compete with Canon EOS 5D (or future 5D Mk II), with Nikon D3, with future Sony Alpha 900. We know that Samsung has a very good technology (shared with Pentax) and that the high-end market is calling for Full Frame sensors.

Nothing is certain, and we will probably have to wait until 2009, but things are moving fast in Korea too.

Taking into account the existence of the Nikon D3 and its impressive set of features, Canon sure has to prepare a top-of-the-line DSLR with a superb Full-Frame sensor, with outstanding image quality. Of course, this is a normal (but difficult) target for the leader and the company that drove us into the Full-Frame world. There, we found: large pixel counts, exceptionally low noise levels, very high sensitivity and a normal 1-to-1 conversion factor for lenses. This last item may not always be a positive (for example, long tele-lens lovers appreciate that a smaller APS-C sensor transforms your common 300mm lens into a loveable 450mm optical gun), but the rest is a boon for photo quality. And there, as the Canon EOS 5D then the Nikon D3 proved, there is a giant step to be taken from more common DSLR.

So, it should be no surprise that photo-related forums are dripping with more or less reliable comments about what the Canon EOS 5D Mk II will be. Possibly, we will see specifications around the following:

• 15 to 16 MP sensor (good but not oustanding, in order to hit hard on image quality; Enough to top the 12MP of Nikon D3) with 14-bit A/D conversion (the de facto standard of Canon image quality)
• ISO 12800 (boost up to ISO 25600) that should give Nikon a run for their money
• All-new 29-point TTL CMOS AF sensor (needed to compete with the impressive 51-point AF sensor of Nikon)
• 6 fps continuous shooting (for more than 50 frames)
• Lens fine adjustment setups for 14 lenses

The rest should be more predictable and does not request a crystal ball (sensor dust cleaning, improved 1MP 3″ LCD screen, LiveView inspired by cheaper DSLR models, weather sealing that was lacking from the EOS 5D). But there are two important elements left out of this picture:

• The price target would be around $3500 (to be compared to the $5500 of the Nikon D3), definitely strong positive argument for Canon even if this is still targetting pros and rich amateurs;
• The announcement would be made in April 2008 (may be, not as a Fool’s Day joke) for availability before mid-June.

These last key arguments could prove critical to offer a very good position to Canon. They had the first Full-Frame DLSR camera (the EOS 5D). Now, Nikon is trying to grab the market with the Nikon D3 since last November. Canon has to hit hard in order to keep the upper hand and they cannot wait for some pro photographers to switch (even if this category of customers cannot jump out of the train each and every year like some enthusiast consumers). Moreover, we know that Sony will be soon arriving with a new Full-Frame DSLR based on a 24MP sensor presented a few months ago (rumoured to be named Sony Alpha 900).

It’s time for Canon to shoot a star. The Canon EOS 5D history needs a sequel to become a real Full-Frame saga. The Canon EOS 5D Mk II could be this long-awaited beacon for the highest end of the Canon DSLR line.

Sources: slo-foto forums, DPreview forums.

• Fact: Last year, Sony showed a mockup of a future Digital SLR camera tagged as “flagship”. Rumour: The enormous size of the pentaprism of the viewfinder shown lets people wonder whether Sony will make it a camera with a big Full Frame (after Canon and then Nikon).
• Rumour: The high-end “flagship” camera should be named Sony Alpha 900 to keep the naming conventions of the pro-line of late Minolta brand.
• Fact: Sony announces today that they have developped a Full-Frame sensor of 24.81MP resolution with a set of very good features for superlative image quality. The chip will be shipping in industrial quantities before the end of 2008.

The logical conclusion: Sony will use this new sensor in its flagship camera. It will be named Alpha 900. It will have 24,9 mega-pixel resolution, continuous shooting at 6 or 6.3 frames per second, 12 bit pixel management, optional capture of a window leading to lower resolution upon request (maybe with even faster continuous shooting). The Sony Alpha 900 will be available in 2008 (using the full production of the sensor).

The only thing that seems less certain seems to be the name, because this one relies only on hearsay.

Sources for the sensor specification: PhotographyBlog, DPreview.

From the look of it, I would draw a few conclusions and comments (only valid if the image is a real scoop).

• Sony has cleaned up the original design a little from what was shown last March. Specially the pentaprism, while still large, looks less rough and unfinished.
• No weather-sealing visible between the body and the lens, but there could be some level of weather protection for the body buttons and such.
• Upper LCD screen, probably in addition to the one on the back of the body. This was a feature disappearing from most cameras (think high-end customers, pro photographers in sight).
• Most of the interface elements have already shown on the Sony Alpha 700 (incl. the battery grip)

Anyway, it’s probably a fake model (look at the chamber closed by a black plastic, and the alignment of the top LCD with the rest of the body looks a bit low). But it could well be a reasonably good demonstration of the current state of the external design in the Sony labs.