Tag Archives: optical sensor

Why high resolution cameras don’t need an AA filter

17 Dec

Disclaimer: Unfiltered truth ahead.

Digital cameras used to have a problem, which was that their lenses were so sharp that images would show moire or aliasing if there wasn’t an additional anti-aliasing or low-pass filter placed directly in front of the sensor to add a small amount of blur. Side effects of this filter include a small reduction in light transmission and a slight distortion of the colour distribution.

As camera sensors got to around 16MP APS-C or around 40 MP full frame, it became apparent this problem wasn’t so much of a problem any more. Glass quality generally wasn’t good enough to still cause moire and other aliasing at this tight packing of Bayer array and photosites. The lenses themselves served as AA filters, and throwing in an low pass filter on the sensor only made images blurrier, without any particular benefit. So some smart executive or engineer at Nikon said, “let’s ditch the thing and see if our customers like it”. And the customers came, and they saw that it was good, and they bought it.

Around the same time, some Pentax execs likewise came together and said unto each other, “let there be unwavering sharpness”. But knowing full wele that the Nikon gods had spread their bets across two models, one with and one withoute AA filter, the good sirs of Pentax did likewyse, and thus there came to be the K-5 II and IIs, made in the D800E’s image.

So there you have it: the story of why you can buy a D800E or K-5 IIs without regret. And how, as an engineer, you can do something for very good reasons and completely fail to communicate this to the customers. Probably because upper management told you to keep your trap shut about the sharpness of the lenses. And the fact that the Ricoh GR and Fujifilm X100S take sharper pics than your monster-tea-bagging SLR.

So there you have it: the reason you can leave that K-3 AA simulation switched off, save battery power, and never worry. Woohoo!

Advertisements

Q: Who is the most collaborative camera manufacturer?

13 Sep

A: As of June this year, the answer would have to be Panasonic. They are working with Olympus on the Micro Four Thirds standard, with Leica on compact and bridge cameras, and now with Fujifilm on sensor technology. An honorable mention goes to Sony, who provide their sensors to a number of other companies, including Pentax, Nikon, and Olympus, but do not integrate others’ technology into their own products, with the exception of Zeiss lenses. Sony has also been allowing Hasselblad to rebadge some of its cameras as fashion accessories, in a move that many believe is ill-fated on Hasselblad’s part but carries little risk to Sony.

On the flip side, the title of least collaborative company may go to Canon, and there’s not much to say about that, really. Among other less collaborative companies, we find Samsung, Pentax, and Nikon. Samsung uses Schneider-Kreuznach lenses in some of its cameras, and has previously shared sensor technology with Pentax, prior to the latter’s move to Sony. Nikon has recently switched its sensor supplier from Sony to Toshiba, for the D7100. Pentax also apparently takes Tamron lens designs and rebadges them as its own.

The four most influential image sensor technologies into 2013

5 Feb

As Panasonic once again proved with its introduction of the micro colour splitter that it is the less established companies that are more innovative in the photography world, I thought I’d take a look at what I believe may be the four most influential sensor technologies for the first half of this decade. I’m going to focus here on technologies that have actually made an impact in the consumer space, or are being promised to do so shortly – in no particular order (you are assumed to have heard of moire and the Bayer filter):

  • Foveon X3 or “stacked RGB”: This is the first of two proposed solutions for the avoidance of moire in digital photographs. The idea is to sense red, green and blue at the same site, rather than at sites next to each other. Similarly to sea water, differential colour absorption in silicon means that blue penetrates the least and red the most deeply (in sea water, colours are the other way around). Therefore blue can only be “harvested” close to the surface of a piece of silicon, whereas any deep absorption is sure to stem from red. If the photosite is vertically divided into three zones whose excitement can be read independently, the relative amounts of each colour coming in can be determined by subtraction among the three measured values. As moire results from the lateral displacement of photosites relative to each other (as in the Bayer array used on other sensors), moire is prevented in Foveon sensors. However, they can still suffer from aliasing. Foveon-type sensors are currently the sole domain of Foveon, Inc., and used in Sigma cameras. However, Sony has submitted patent applications for Foveon-like technologies and may eventually develop a similar sensor.
  • X-Trans: This is a simple modification of the Bayer array, such that the placement of red, green and blue photosites within any square containing four photosites, is different from the arrangement in any adjacent such square. Instead of a 2×2 repeating unit, the X-Trans sensor has a 6×6 repeating unit (or a 3×6, repeating with a lateral displacement of 3, or a 3×3 unit with rotation through 90 degrees in either direction, depending on your point of view), thus decreasing the regularity in arrangement that causes moire. Fujifilm started using the X-Trans technology in its “boutique” or enthusiast camera models, starting with the X-Pro1. Because this requires new demosaicing strategies, Fujifilm’s own in-camera JPEG engine has so far produced better output than desktop RAW processing software. Remember also that Fujifilm had previously established itself as an innovator in sensor technology when creating the EXR sensor and associated innovations, which also featured a modified array of red, green and blue sites. In this sensor, photosites are diagonally arranged in subdivided groups of one colour, so that they can either be combined to give greater sensitivity at stable noise ratios, exposed at different sensitivites to give a within-exposure HDR image, or give a standard exposure using the full resolution the sensor is capable of. Even then, RAW support for the altered array was challenging for third party software, and created a lag in support for the cameras. This seems to now be resolved, but note also that with in-camera HDR generation, the need for external RAW processing is somewhat reduced. It may be part of Fuji’s story that due to the diagonal arrangement of square photosites, the EXR sensor would tend to show less moire on horizontal and vertically repeating patterns, which may occur more commonly in typical subjects and compositions than do diagonal ones.
  • Light field cameras: The Lytro has been the first consumer-oriented camera using a light field sensor that can simultaneously record pixels at different focal depths. This means that different depths of the image can be independently explored in a 2D viewer. However, the camera has low 2D resolution, and is strongly tied to Lytro’s desktop and web offerings. German light field specialist Raytrix is growing its portfolio of cameras, with models ranging from 1 to 7 megapixels. Internally, Lytro has a standard flat CMOS sensor situated behind micro lenses that split incoming beams depending on the direction of incoming light. The different planes of focus are then “triangulated” using software.
  • Colour splitting: While Panasonic’s patent application is not the first use of beam splitting with a prism to harvest the full amount of light, it does reduce such splitting to a micro-level that some are anticipating will lead to revolutions in compact and low-light photography. This is in contrast to earlier ideas from Foveon, wherein three different sensors were used, each dedicated to one colour, and positioned relative to one central beam splitting prism.

Given the fact that especially the smaller camera manufacturers such as Panasonic, Pentax Ricoh, Fujifilm, Olympus and Sony*, are struggling to be profitable, it would be interesting to see new consortia for technology-sharing spring up – perhaps between Fujifilm and Panasonic.

* Sony may be a big corporation and supplying nearly everyone else with sensors, but they don’t ship a lot of cameras compared to the “big two”.