Thursday, 15 April 2010

Video killed the dSLR star

I have planned for a long time to write a post about video mode of dSLRs. The announcement of Panasonic micro4/3 type professional videocamera have given me a new impulse to do this. Let's start with a quote from me: "Probably all Micro FourThirds camera will be able to record movie but maybe a videocamera-system with interchangeable lenses is among the main goals of Panasonic. August 5. 2008." Of course, I smirked a lot when I read about AG-AF100 some days ago.

A first thought of many might be that this will be the ultimate video system. But we have to clearly define the category of this equipment to judge its applicability. There are two main categories in the professional videography market which we have to talk about: cameras equipped with three relatively small (1/3'' or 2/3'' type) sensors, and large, single sensor cameras (35mm Arri stantard type, close to APS-C size in photograpy).

trichroic beam splitter of a 3CCD videocamera
(move the cursor over the animation to rotate the prism)

Single sensors use a filter array over the pixels to determine the color of the light. Color filter passes only a small part of incoming light by dropping out the remaining part of it. The other method to sense the color is to split the light beam into its color components by a trichroic prism. It is a complex and precise optical system which uses three monochromatic sensors. This solution is expensive, requires lots of space and requires special optical design of the attached lens but is two to three times more efficient than single sensors with color filter arrays.

The introduction of 3CCD cameras was a milestone in the video world. Every low-light situation where the videographer has not chance to use a lot of artifical light, the three sensor setup is very useful. You might think that a large sensor is a substitute for it, but it is in fact not. The shallow depth of field of a large sensor is not acceptable in many cases and stopping down to achieve the required sharpness, you lose light and thus the advantage of the large sensor. The 1/3'' type sensors allow engineers to create lense systems with high zoom ratio (>10x) and fast aperture (>f/2). It is impossible to design comparable lenses for 4/3'' sensors. To sum up, the advantages of a large, single sensor (variable depth of field and good signal to noise ratio) can be exploited only if the situaton is allows it and these advantages are partially substitutable by fast lenses of multi-sensor cameras. On the other hand, the gain in light-efficiency with three sensors is an absolute profit. Unfortunately, the flange distance of micro 4/3 standard is not small enough to place three sensors into the camera.

So in what area are videocameras with large sensor usable? Movies, videoclips, in studio and most outdoor situations, the advantages can outweigh the disadvantages. But to record the most important indoor video-report of my life I would take a 3CMOS camera with me. The whole question will of course be moot if one of the manufacturers develops a multi-layer sensor with the same efficiency as the three sensor method. But there are only plans and patents about this idea, and the single presently available technology, Foveon, is in a lightyear away from this quality.

Video mode of dSLRs is often criticized for a typical flaw. In videocameras equipped with CCD the interline type is used instead of fullframe transfer type. Exposure and reading of the image is totally separated in interline CCDs. The photosites can not transmit the charge to nearby photosite, but there are CCD registers between every coloumn which are covered from light. At the end of the exposure, after the charge of photosites has been loaded into the masked CCD coloumns, a new exposure starts while the previous frame is being read. Thanks to this solution the exposure starts at same time on the whole surface of the sensor.

Interline CCD vs. CMOS

CMOS sensors are able to produce moving picture in itself because their pixels are individually addressable and readable. But a pixel can do only one thing at a time, so the reading electronics has to continously scan the sensor and start the exposure at the row last read. This results to an odd phenomenon called "rolling shutter effect". The name deriving from distortion produced by focal plane shutters at high shutter speed and on moving theme. Horizontally moving (relatively to optical axis) objects will be tilted, vertically moving objects will be compressed or lengthened.

As the amplifiers and switches in CMOS sensors, the masked coloumns in interline CCDs also reduces the efficient surface of pixels. The microlenses help to collect more light in both type of these devices. The rolling shutter effect is not accepptable in lots of applications, however manufacturing of cameras equipped with CMOS is cheeper and these sensors produce less noise due to high level of integration (on chip A/D, noise reduction etc.). By increasing frame rates and digitally combining these frames helps to make virtually invisible the imperfection of CMOS sensors. Panasonic will use this solution in AG-AF100 too.

Another important question is the lens strategy of Panasonic. The only lens which comparable for professional videocameras from point of view of AF drive is the 14-140 HD due to its linear motor, but its maximum aperture makes it a toy, rather than a professional equipment. I do not think that Panasonic wants to focus only professional segment because the micro 4/3 bayonet compatibility would be pointless, but it is a telling sign that the mockup was presented with a Zuiko Digital 14-35mm 1:2 mounted on it. Yes, the professional market will not accept lenses with f/5.6 maximum aperture, Panasonic has to clearly communicate that there will be more price and performance lens category to create the possibility of a smooth transition for advanced users. Maybe, photograpers will profit from this too thanks to new, fast and high quality prime lenses.

A welcome news in the announcement that AG-AF100 will support 1080p mode, and 50 or 60 frame per second at smaller resolution. I hope, we will meet this feature in Olympus cameras, because all of new Canon video-dSLRs can do it. I think the most positive impact of Panasonic's new system that it speeds up the development of the 4/3 size sensors, and this is what we have been waiting for since 2003.

1 comment:

Anonymous said...

"Unfortunately, the flange distance of micro 4/3 standard is not small enough to place three sensors into the camera."

this isnt true, there is more than enough room for a beamsplitter within a 4/3rds mount. There is such a thing as an RI (refractive index) for all translucent materials, pure glass has an RI of 1.5, which means that the lens register can be maintained within a prism system that is 1.5 times longer pathway.

And even if it wasnt so the beamsplitter stills fits with less than 1mm to spare