Saturday, 14 March 2009

Our lovely system

Most FourThirds camera owners say: "this is simply the next invention which is being copied from an Olympus product" when Samsung announced its NX system. We are getting used to this, but we have to separate our feelings from facts even though it is trivial what happened.

The most sobering thing in connection with micro FourThirds (mFT) cameras is the lack of autofocus with most of the FourThirds (FT) lenses. Canon or Nikon who have not released any new contrast-AF specific lenses and where lens firmware update is not possible at your home could solve this problem even if they are very slow sometimes. Only FT, the most modern system could not do it.

The first question about Samsung's new cameras was: "will they use the K bayonet"? This question was perfectly pointless because the first announced feature of the NX system was the very small flange distance (similar to mFT). This automatically makes all of K lenses unusable directly on NX bodies because any type of focusing is impossible, so the bayonet type is indifferent. The primary consideration on the systems like this is the size, so they will reduce the bayonet diameter if it is possible.

But what about adaptor? Except SDM type ones, all Pentax AF lenses have mechanically coupled AF driving. If the adaptor contains a micromotor (or an axis which pass the torque to the lens if the motor will be built into the body again), Samsung will be able to eliminate any contrast-AF incompatibility stemming from the mechanical drive. Pentax released some quality DA lenses in the last years, like the pancakes for example, so their adaptability is a very relevant problem.

We have to accept that there are serious reasons behind the lack of AF when FT lenses are mounted on mFT cameras, but very sad to see that more and more parts of the FourThirds's main concept fall to the sand. I think Olympus does not put enough energy into compatibility and consistency. I hope that the announcement of the first Olympus mFT camera will be a pleasant surprise from this point of view.

Monday, 9 March 2009

640K...

...ought to be enough for anybody - Many people thought of this famous sentence when they have heared Akira Watanabe's opinion about megapxiel race. The biggest difference between two statement is that Gates told it as a general rule but Wantabe said: "...for covering most applications most customers need."

Is this statement a suicide from a marketing point of view? It really can be dangerous. It is unpredictable how customers will react to the permanent amount of megapixels + increasing image quality while other manufacturers continously increase the resoulution. Everybody knows that megapixels are the most attractive (and most misinterpreted) feature on the market.

Is it really imaginable that one of the most innovative manufacturer who fabricates scanning electron-microscopes suddenly discovers that their system is a dead end? They precisely knew the limits given by optical and phisical laws when they decided to launch the 4/3 system. I think that Wantabe was very honest. He did not say that more than twelve megapixels is simply unnecessary in photography. He talked about most customers and most applications, and I must agree with him. I used an Olympus E-3 during the last 12 months. This camera has some pesky imperfections but my biggest problem is to exploit the ten megapixels. The real-life themes are not test charts, not papers glued onto a flat surface. The perfect AF, ergonomy, camera speed and good noise characteristic worth much more than some extra megapixels over ten, and Olympus wants to focus on these features.

There are a couple of 60cm x 45cm prints on the wall of my living room and anybody who has seen them was pleased with quality. Of course, there are applications where relatively easy to hit the resoulution limits. Panorama-stiching is a solution for most landscapes in this case. I have some panorama photos which were made under 3-5 seconds without a tripod and they are nearly perfect from a technical point of view. By using this technique, and in exchange for some extra work, we can get rid of corner sharpness problems of wide-angle lenses. But such techniques are not panacea. Large group portraits and some professional work require fullframe and medium format cameras, there is nothnig to do.

The launch of the Canon 1D Mark II in 2004 was a milestone in professional digital photography. Many professionals have dropped their stereotypes about dSLR becasue it was the first which was widely used in studios, sports events and could substitute analog cameras in most cases. The announcement had the same effect in the amateur world. I can remember that mere indication of the camera type was enough to get the maximum points at most of photo-sharing sites. The downsized, 8 megapixel image of 4/3 cameras will have reached or exceeded the quality of the 1D Mark II in next 1-2 years, moreover it is happened already from some point of view.

Today, many 10-12 megapixel dSLR are used for professional applications and most of them are the flagships of their strict category. Are they really temporary solutions? Every photographer who bought a Nikon D3 did it because of a necessity until 16-20 megapixel version are announced? Or are the needs of users continously increasing? How can this be reconciled with the legends about the resolution of the films? High resolution films have been overtaken by the Nikon D3 and the Canon 1D Mark III by many years. So where is the logic in worrying around twelve megapixels? I think it is nowhere.

The evolution of highly integrated sensors has just begun so maybe the real number of megapixels will be indifferent in the future. The first two steps were the individual A/D converters for every coloumn to increase speed and precision and reduce noise (Sony Exmor), and on-chip exposure-bracketing to increase the dynamic range (Fuji EXR, Panasonic in the future). The integration drifts in a direction that every pixel or pixel-group will behave like a little computer, and the consequences can be astonishing: colour pixel binning (quasi three layer), sensor aided anti-shake/anti-motion-blur, auto-stich, high-quality upsampling, auxillary data recording for post processing (DOF, noise reduction) etc. The evolution is unforeseeable also for Olympus engineers.

I would write many pros and cons about this resolution limit. The limit in itself is not an advantage. Higher resolution images can be resized down to twelve megapixels anytime. The gaps between pixels or microlenses will exists forever so they will reduce the pixel fill ratio stronger in higher resolution sensors, but maybe the gap will be negligible in the future. Only the harmony of the whole system can be an advantage. The market will have the last word. If Olympus can attain their promises about image quality, reliability and services, and keep the price under that of competitive products, customers will be the winners. But holes in lens-palette must be filled irrespectively of megapixels. Olympus has to rethink its in-camera image processing and AA-filter strategy (not removing) to keep itself in the market. I have niggling opinion about this but it is pointless to talk about it without understanding signal processing basics. I am working on a new website which will help in it by my unique interactive, illustrative method.

Wednesday, 4 March 2009

There is truth, after all!

I was so disappointed some months ago by the Panasonic G1 announcement. I predicted that there will be a wider sensor in it to allow switching between aspect ratios without loss of pixels. But G1 got the same dimension sensor like any other 4/3 camera, so the switching translates to simple cropping of the image in this case.

Panasonic has announced the GH1 on the day before yesterday, the new version of the G1 which can record HD movies. The video capability is the most important feature for most of the people, but not for me. There is another very good news. GH1 has been fitted with a sensor similar to what I have predicted. To see the point a little bit better, look at the following figure (the real size of the sensor area is a little bigger than the official size due to technological reasons):


As you can see, the sensor is hardly wider than its conventional 4/3 counterpart but the effect of this difference is very important. Just imagine: You can take portraits with 3:4 aspect ratio in vertical mode, but for landscapes 16:9 is also available without significant loss of quality. Let us calculate the advantage:


There is another very interesting question: Which image size will be recorded in RAW mode? If RAW images will be cut to selected aspect ratio, there is no surprise, but if whole sensor data is recorded, we have to face another possibility and a related problem as well. Most of zoom lenses can project larger image circle than official at longer focal lengths. The difference between the official and the image circle required by wider formats is very small, so these lenses can exploit the whole 14 megapixel (13 megapixels are used) sensor at almost any point of their focal range. The lens hood of conventional 4/3 lenses may cause a little vignetting if they are used with adapter on GH1-like mFT cameras at 16:9, but this is a minor issue and one can take care of it.

I hope that we shall see the same concept in dSLR cameras soon and that Olympus will use the same sensor in mFT bodies!