Slashdot Mirror
Sony Shoots For 4-Filter CCD, 8 Megapixel Camera
Artifex writes "If you're looking to spend about $1200 on a new digital camera, check out this Digital Photography Review look at Sony's upcoming 8 MegaPixel Cyber-shot DSC-F828. The most interesting thing isn't the number of pixels in this prosumer-grade camera, but its 4-color filter CCD system. ['Instead of the traditional RGB color filter array, the new CFA is made up of Red, Green, Blue and Emerald (like Cyan) color filters.'] I've always been a strict Canon fan, but this is making me think twice."
The camera (as almost all other) feature a single CCD. It does however have a 4 color fileter in front of it. BTW, Nikon had that for about 4-5 years now.
Code poet, espresso fiend, starter upper.
It has only one CCD, but instead of having the regular RGB pattern on it, it has a four color pattern.
e ccd.asp
http://www.dpreview.com/news/0307/03071601sonyrgb
Because the whole point of photography is to print the pictures. What good is a high end digital camera if the pics can't be printed in a newspaper or magazine?
Support the First Amendment. Read at -1
The center frequencies to which our green and blue cones are sensitive are rather far apart, spectrally -- at least, compared to the G/R cones. Looks like the E sensor on this CCD is between the G/B cones.
(RGB sensors and emitters are generally calibrated to the center frequencies of our cones.)
This is a good idea that I'd never considered. More color information is always good, and we can always just define a transform to reduce it to human optics. If nothing else, this makes more data available for image correction and whatnot. I wonder if you can actually get the RGBE data out of the camera, or if it stores three-channel JPEGs like everyone else?
Well, in any case, tetrachromats rejoice.
Not necessarily. A photograph taken with a 35 MM can be blown up quite easily to 16x20 using a high end drum scanner (DS for example) It would not be as good as a 4x5 transaprency, but I have sen it done. At that size, for commercial uses, you would want at least a 4800x6000 resolution which is nearly 29 Mega-pixels.
What your printer at home produces and what commercial printing produce are 2 very different things. The screening technologies are very different and require different source data
It depends on a lot of circumstances, but the general concensus seems to be that a 35mm negative shot with a good lens contains about 11 megapixels of information.
Of course, pixels is only a small part of image quality. There's also s/n ratio, whitebalance, dynamic range, etc.
For holiday pictures and stuff like that, digital has reached equality with 35mm a few years ago. For professional work (especially medium size negatives) it still has a long way to go.
If you had bothered to look, you would have noticed that this particular sony camera uses standard compact flash cards, in addition to sony's normal memory stick.
It depends on how big you want to print the images. "Photo-quality" prints are typically 300 DPI. If you want to print a 4 x 6 image, that means you need and image that is 1200x1800 pixels to print at 300 DPI. If you want to print larger pictures, you need more mega-pixels. Also, the images in almost all digital cameras (except the Canon EOS1DS) is not 35mm. 35mm Film has an aspect ratio of 3:2. Digital cameras have the same aspect ratio as a normal computer monitor which is 4:3. When you are printing an image on 4x6 paper from a digitial camera, a small portion of the top and bottom is usually chopped off (unless you flattened the image). Digital cameras haven't yet passed 35mm film quality when it comes to how many pixels are captured. The closest is the Canon EOS-1DS(11 Megapixels) which rivals some consumer 400 speed films. But digital cameras still have a long way to go to rival 100 speed(or slower) film. But, the amount of detail captured in fine-grain 35mm film is almost never used by the typical consumer anyway. To make 4x6 prints, a 2 megapixel camera is about as good as a film camera.
The 10D is a great camera for the price, but readers should note that the $1499 MSRP is for the body only. Lenses must be purchased seperately. I read the article on the Sony a few days ago so I can't remember the focal length specs, but to get something similar on the Canon would require a serious cash outlay for lenses. In fact, I don't think there's a single Canon-mount lens that will cover that range. The EF 28-135mm f/3.5-5.6 IS lens might be closest, and is considered a bargain at around $500 - and that's inexpensive as far as lenses go. Most likely two lenses would be required to cover the Sony's focal length, and a good pair will run around $2500.
I should disclaim, I don't even own a 10D, although I plan to. I currently have a Powershot S400 because I need a small point-and-shoot, but I also want a full-featured camera for serious (albeit amateur) photography.
-- "Complacency is a far more dangerous attitude than outrage." -Naomi Littlebear
I did some Googling and found this guy has done some math:m .vs.dig ital.1.html
http://www.clarkvision.com/imagedetail/fil
Example:
Fujichrome Velvia has an lpm1.6 = 80 lpm. Equation 1 gives 10 megapixels for intensity detail, but
16 Megapixels is not that far away from the consumer market.
Me, I'll stick with my view camera. You'd need a few gigapixels to even approach an 8X10 contact print.
Most people won't need image resolution that high ... HOWEVER, it is good if you want to do anything with professional or production quality.
3264 x 2448 resolution will allow you to print a 21" x 16" print @ 150DPI. Thats enough for production quality posters, calendars, etc. It'll allow you to print an 8x10 at 300DPI, which is arguably better than 8x10 quality with regular 35mm film.
This camera is also good if you're a decent shot and want to sell your photos through stock shops like gettyimages, corbis, etc. You *NEED* to have resolution like this before production houses will even look at your stock images... because guess what they do with them?.?.?.?... that's right... posters, calendars, and other production which requires 150-300 DPI at reasonable visual sizes.
Skiers and Riders -- http://www.snowjournal.com
1) Blow-ups. 2.1Mpixel pictures look good on 6x4" prints, but if you enlarge them you'll notice the lack of resolution.
2) Detail. You'll notice the lower resolution in details like sharp and/or thin lines, like edges of a brightly-lit wall or power lines on poles.
3) Digital zoom benefits from a higher resolution as well.
It's a little like the difference between high-end and mid-range loudspeakers. Most people will hear the difference in quality when doing a direct comparison, and most people will see the difference in quality between a 2.1Mpixel print from a Photosmart printer, and a professional print from an SLR or a good digital camera, when both pictures are placed next to one another. 2.1Mpixel images are fine for the holiday happy snaps, but they'll look crummy in a glossy magazine. However for most people, the lower quality is enough precisely because they're just taking happy snaps, not professional photographs.
My camera is a 3.1Mpixel one, but I regularly use it at 2.1Mpixels so I can fit more pictures on the card. I'm happy enough with the quality I get, and as you said, it's close enough to 'real' photographs to serve my purposes. Photography is not my hobby... I just want to have some nice pictures.
If construction was anything like programming, an incorrectly fitted lock would bring down the entire building...
Color management would be handled the same as any other digital camera. Even though the CCD has a four color filter, the data is converted back to 3 color RGB (14 bit i might add... giving a slightly wider color gamut than current 12 bit digicams). If having a profile for your digicam is important I would recommend Monaco system's (now x-rite corporation) MonacoDColor application. This handy little app will allow you to perform easier color corrections simply by applying the cameras profile. It also intigrates well with Monaco EZColor (their prosumer monitor/printer calibrater).
so to sum it up... you do it the same way you always have... the new filter just gives more accurate color (epically since sony was well known for their clipped reds and yellows).
All the more important considering that you can get a 10 megasensor Sigma-based Digital SLR for around $1000 nowadays... See the DPReview site to compare this with other 6 Megapixel sensor.
There's an excellent source of information about what "three primary colors" actually means at of Dave Trapp's Sequim (WA) Schools science department site. There's a relatively simple explanation of how color vision works, then a facinating and highly detailed in-depth discussion of the issue.
Some interesting notes from Mr. Trapp:
* "All three [signals] are equally sensitive to blue light, two have expanded ranges that include green and yellow light, and the third signal includes sensitivity to red light."
* "While these paradigms of primary colors have worked well for human printing and light uses for over a century, it is likely that the three primary colors are not descriptive of the world, but rather an artifact of our eyes, the tools we use to perceive the world."
* "The real world does not have primary colors!"
He also discusses how the world would be perceived differently if we evolved a fourth cone, sensitive in the UV region. Very cool stuff!
Interestingly, though, he's no longer teaching science, and details the reasons on his site. Anyone who's ever been driven crazy in a class taught by a guy named simply "Coach" (and who on this site hasn't?) will sympathize with this good teacher's plight.
Stressed? Me? Of course not. Stress is what a rubber band feels before it breaks, silly.
Cyan is a blue-green mix, but I do agree that it seems more blue than green.
I've got a mind like a steel trap - it's got an animal's foot stuck in it.
Brahmastra. Thanks that was a good job comparing the resolution of a 35mm to Digital. My experience with a Nikon 5000 is that it's 5 megapixels is good enough to create great looking 8x10 images. But there is more to good quality photography than the resolution of the film, or of the digital chip.
IMHO the quality of the lens is a the single largest determinant in a camera's ability for producing creative photos, and is an area where many digitals fall way too short.
The first thing I do when shooting a photograph is decide what the subject is. The next thing I do is frame the photo in such a way that anything that is not relevant to the subject is excluded from the photo. A good lens on a 35mm slr gives you the ability to use selective focus so that your subject is sharp, and everything else is a blur. For this technique to work you need a longish lens, and a wide aperture. Most point and shoots and digitals fall short in both categories. (My $1500 CAD Nikon included) So the thing that interested me most about this Sony was it's Carl Zeiss lens 28 - 200 mm equiv. F2.0 - F2.8 aperture.
Zeiss is famous for quality, and 105 mm lens, at f 2.8 is a beautiful combination for selective focus portraits.
- and I'd rather they used the extra pixels for more resolution rather than an additional color channel. Foveon has a much better way.
The Sony DSC-S75 may allow you to take a picture with 3:2 aspect ratio, but in reality, it is just chopping off the top and the bottom of the image. This is because the CCD element in the DSC-S75 is still a 4:3 element. In case of the EOS-1DS, the CMOS sensor itself has an aspect ratio of 3:2, therefore giving a true 35mm image without any cropping.
Also remember that resolution of film depends on the film speed. So say, ASA100 film has smaller grains and more resolution than ASA400 film.
It depends how you define pro... no pro (in your definition) would be caught dead with a Sigma SD9 right now either, it's got far too many imaging issues (regardless of the theoretically preferable sensor configuration).
The Sony will appeal to folks straddling the line, who might have considered an E-20 in the past, or who are looking at the low-end SLRs, don't have an existing investment in compatible lenses, and need a good performer at a reasonable price. I use a variety of cameras all the time, and I am very seriously considering the F828 while I wait on the 6MP+ DSLR market to mature and come down to reality.
If I had my druthers, I'd buy an EOS-1Ds for myself right now - but short of that, this is the most thrilling camera for me when you balance the price, performance, and other factors. If you simply MUST have a standard lens mount, then you can't go too wrong with the EOS-10D right now.
Resolution is hardly everything... many folks will take a 3-5 megapixel DSLR over a 5-6 megapixel prosumer model any day.
Relevant factors:
Lens systems - most low-end digicams come with shit for lenses, little more than a transparent glob of plastic... chromatic abberations are common, spherical focus problems, color fringing, etc.
Chip - the size, filter configuration (3-color Bayer, 4-color Bayer, Foveon X3, etc), presence of microlenses or not, mfg process (CCD vs CMOS), and other concerns all impact imaging quality (subjective), depth of field, sample depth (8, 10, 12bit/component), noise (or lack thereof), pixel blooming, speed/sensitivity, etc
Camera - just like pro's don't say "Hell, that little point and click does 35mm too, why do I need my 10k worth of Canon gear to do the same", there's no comparing the qualities of a pro camera - the body, controls, post-processing (white balance controls, noise suppression, compression, etc), AF system, with that of a cheap consumer camera.
In other words, in many cases a 3MP image shot by a pro on a quality DSLR of a year or two ago will still surpass anything you can do with that 5-6MP Minolta you just got.
Yes, consumer magazines and newsprint typically have terrible resolution... but still, garbage in = garbage out (it's all relative). Besides, a pro shooting for a major magazine will use pro equipment - that's why it exists... it gives them the fine control they need to get a shot that will pass the photo editors muster. Snap shots on his wife's camera might suit in a pinch, but it won't go over well. I think you underestimate the challenges of professional photography...
Technically, you have three types of color sensitive cells in your retina (so-called S, M and L cones) and you need at least three variables to describe color as seen by humans. That said, nonlinearities in cones' reception make it beneficial for color reproduction to add more variables. Hence RGBE.
Also remember about CMY and CMYK problem in printing, where theoretically CMY would be enough to reproduce full gamut, yet CMYK is used due to ink impurities, which make it difficult to match exact black.
"Man in the Moon and other weird things" - wfmh.org.pl/thorgal/Moon/
The end result is each pixel on the Sony is only 1/8 the area of the pixel on one of the $1500 advanced amateur digital cameras (Canon EOS 10D, Nikon D100, Fuji S2, Pentax *ist D).
This means each pixel will receive very little light (and thus a low signal to noise ratio) and have images with a lot of electronic noise even at ISO 100. Noise manifests itself as colored dots that pepper smooth areas like skies.
If this camera had used one of the ICX413AQ 6MP APS-size sensors Sony sells Nikon and Pentax rather than the ICX456 used on this camera, it could have been a winner.
I love the Foveon idea and the much better resolution it provides. What keeps me from buying a camera with the Foveon sensor is the review I saw that showed it providing poor color reproduction: pure green is reproduced as olive drab.
Contribute to civilization: ari.aynrand.org/donate
There may be 8 million pixels in the camera, but that only equates to 2 million of each color!!! 3/4 of the data is interpolated in that 22.8mb picture!!!
Actually, there are 2 million red and blue sensors, and 4 million green ones.
Additionally, part of the data in that 22.8mb picture has probably been truncated (most sensors retrieve more than 8 bits of data; 12bit sensors are not uncommon).
So the raw acquired image on an 8mb camera is 12mb if the data is packed. 16mb if it's aligned on an int16 boundary. That means 1/3 to 1/2 of the interpolated image data is redundent.
the camera internally "guesses" at what the values in between would be.
It doesn't so much guess as "interpolate" using an algorythm derrived from some complicated math and physics knowledge.
The only current exception to this are cameras based on the Foveon chip
The foveon chips have their own set of problems, due to the size and arrangement of their sensors. Right now, the image quality they produce is not superior to some of the better CCD based cameras on the market.
So... until cameras get to ~30 MP, they won't be "equivalent" to the resolution of 35mm film. (that's not THAT far off). 10MP is tons for most uses, but it's not equivalent.
Actually this depends on the grain of the film you're using and the paper you're printing on. Is is also difficult to measure the absolute resolution of actual film, because each grain isn't the exact same size.
At the end of the day though, matching the resolution of 35mm film isn't as important as capturing enough data to create the size of print you want to produce.
That is correct, however unless you set your image to RAW format you will not be able to benefit from the extra resolution (JPG is 8 bit per channel). 9th bit might help as far as rounding go, but 10 or 14 does not make any difference as far as looking at pictures on your monitor. Even the top of line Matrox Paraphelia uses 10bit display at best. ATI and nVidia are nowhere near. Your LCD monitor (in case you use that) will have only about 5-6 bits per channel.
Always look at the bottleneck of your system, in the overall picture. Currently this stands at:
1. LCD 5 bit ($300) to 7 bit ($1000)
2. CRT 7 bit ($150) to 9 bit ($800)
3. Video Card 8bit($99) to 10bit ($500)
4. Camera 8bit ($200) to 14bit ($1500)
Code poet, espresso fiend, starter upper.
If you add up all the Red, Blue and Green elements in a digital camera's CCD, you end up with the number advertised... 5 or 6 megapixels or whatever. (In this example I will refer to a 5 megapixel camera.)
However this does not REALLY equate to that many pixels as we would normally think of pixels with other devices.
If your LCD monitor can support a maximum of 1280 x 1024 resolution, that multiplies out to be 1.31 megapixels.
But if we were to do the math the way digital camera manufacturers' marketing departments do their math, that same screen would be 3.93 megapixels... which is basically a lie.
What happens inside digital cameras is a certain bit of deception. They use the luminance factor from each of the 5+ million CCD sensors to achieve a semblance of the resolution advertised. However the color value for each of those so-called "pixels" is not independant, but rather is derived from the values of the surrounding pixels.
Therefore we have the baffling paradox of saving a RAW file at full resolution on a 5 MP camera and getting a 7.5 megabyte file; but strangely a TIFF file of the same resolution saves out at 15 megabytes in size. How can this be possible, you rightly ask? Just what is the camera adding to the raw sensor data to create a full resolution file which is somehow twice the size of the raw data? Here's what happens...
In the RAW file there may be 5 million 12-bit samples, half of which are green elements, with the other half evenly split between red and blue elements. Or, there may be 2.5 million 8-bit red, blue and green values each, with not all of them corrresponding to actual CCD elements.
In producing a 15 megabyte TIFF file from 7.5 megabytes of RAW sensor data, the camera's firmware defines a virtual 5 million simulated pixels, each of which has its 24-bit color values derived from the other adjacent physical "pixels." Then once 8 bits each of Red, Green and Blue data are derived for each virtual pixel om the memory array, the whole simulated thing saves out at 15 megabytes.
The ONLY cameras available which do not deceive you in this way are those new ones incorporating the Foveon CCD sensors, which are novel 3D arrays of elements, each element of which produces its own true RGB color values. With the Foveon CCD, each pixel is a true pixel, and the color definition is superior.
To be fair... the color interpolation firmware in standard CCD cameras has gotten so good, that it probably is worth putting up with the marketing deceptions and these artificially puffed up file sizes (200% of what they should be), at least for now.