Naked Science Forum
Non Life Sciences => Technology => Topic started by: Don_1 on 21/09/2009 14:24:09
-
The digital camera concept dates back to the very early 1960's, but it was not until 1990 that the first digital camera came on the market. In 1991 Kodak released the DCS-100, the first professional digital camera, with a staggering 1.3 megapixel sensor. WOW!!!
Nikon soon beat Kodak at its own game with their D1 at a mind blowing 2.74 megapixels, with Canon bringing up the rear with their 1Ds at a galactic 4.1 mp.
Today we have point & shoot cameras of 12+ megapixels. While the professional cameras leap ahead with DSLR's boasting 20+ mp and the rumoured 30+ mp Nikon D4x and medium format cameras with 50 mp (such as the Hasselblad) and even 60+ mp for the Phase One.
So how many pixels will digicams have in the future? Taking the sensor of a DSLR, such as the Nikon D3, Canon 1DsIII or Pentax K7, what will be the mp count in the future?
-
The "wires" on a chip can currently* be made ~1/10th the wavelength of visible light ...
http://en.wikipedia.org/wiki/Electron_beam_lithography
However I think lower limit on photosensor size would be the wavelength of light it is designed to detect.
[* pun unintended]
Diffraction by the aperture in the lens would also be a factor limiting the maximum possible resolution ...
http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm
-
There's nothing to stop you sticking sensors together. How many pixels is google earth?
-
What a remarkable coincidence. I have just finished making the World's first 200 mm diameter, 100 Terapixel sensor.
[ Invalid Attachment ]
-
... I have just finished making the World's first 200 mm diameter, 100 Terapixel sensor.
If it is then all your pictures will look like they've been taken through a mesh fence...
[ Invalid Attachment ]
-
How many pixels can I have?
How deep is your wallet?
Basically it boils down to how small you make your pixels, and the overall size of your sensor, i.e. how much area of silicon.
You can make pixels smaller, but as you do so the ratio of "active" pixel to dead-space inbetween them (known as "aperture ratio" of the sensor) gets worse so you lose sensitivity. Very small pixels cannot store much charge, and have tiny tiny signals so your dynamic range (difference between 'dark' noise and saturated max brightness you can record) decreases. This leads to clipped highlights and noisy images - or on excessive reliance on 'noise reduction' processing which generally makes the images look lousy/weird.
The manufacturing cost of a sensor is substantially proportional to its area, all else being equal. Double the size to double the number of pixels, and you double the cost.
So professional high megapixel cameras will have "full frame" 35mm sensors, or "large-format" sensors. Large sensors inevitably means larger cameras (look at the geometry of the optics), so you won't be seeing large sensors in pocket-sized point-and-shoot cameras.
In reality though, the lens performance is often a limiting factor - you might have a 12mega-pixel camera, but how often does the image appear pin-sharp when you view it at 1:1 pixels?
So to actually benefit from more pixels on the sensor, you will require a bigger, heavier, and much more expensive lens. Unless you plan on taking every image on a tripod, or in very bright sunlight, you might need a sophisticated anti-shake system too.
Unless you need absolutely razor-sharp images when printed at A4 or A3 size, you shouldn't need more than around 12Mpixels - but you might need a better lens.
-
If it is then all your pictures will look like they've been taken through a mesh fence...
Rats! I hope you realize I must have spent almost four quid developing this.
-
Techmind, all you say is quite true. My wife has a 10mp Pentax compact, which takes a good pic, but my 12mp Nikon DSLR is far and away superior, mainly because of the lens.
I was just curious about the sensor pixel capacity, since in the past 10 years the DSLR (35mm equiv) has gone from a mere 4 mp to 20+ mp and could well top 30 mp next year.
As you say, a 12 mp sensor will give poster prints of very high quality, but go much beyond and you will begin to see pixelation. One big problem is the alignment of any lines which are not horizontal or vertical. As individual pixels are expanded, any diagonal or curved lines become misaligned. The more pixels, the greater the expansion of an image before this becomes visible.
So what would be the maximum number of pixels that can be crammed into a defined area. Take, for example, the Canon EOS 7D. This has 19 mp on a sensor measuring 22.3 x 14.9 mm, while the Sony A550 has 14.6 mp on a sensor measuring 23.4 x 15.6 mm. What would be the maximum number of pixels that the Canon's 22.3 x 14.9 mm sensor could have?
-
Techmind, all you say is quite true. My wife has a 10mp Pentax compact, which takes a good pic, but my 12mp Nikon DSLR is far and away superior, mainly because of the lens.
I was just curious about the sensor pixel capacity, since in the past 10 years the DSLR (35mm equiv) has gone from a mere 4 mp to 20+ mp and could well top 30 mp next year.
As you say, a 12 mp sensor will give poster prints of very high quality, but go much beyond and you will begin to see pixelation. One big problem is the alignment of any lines which are not horizontal or vertical. As individual pixels are expanded, any diagonal or curved lines become misaligned. The more pixels, the greater the expansion of an image before this becomes visible.
So what would be the maximum number of pixels that can be crammed into a defined area. Take, for example, the Canon EOS 7D. This has 19 mp on a sensor measuring 22.3 x 14.9 mm, while the Sony A550 has 14.6 mp on a sensor measuring 23.4 x 15.6 mm. What would be the maximum number of pixels that the Canon's 22.3 x 14.9 mm sensor could have?
It might be possible to ultimately improve resolution further using optical antennas. But I'm not counting on it! [:D]
Here's a link http://www.optics.rochester.edu/workgroups/novotny/antenna.html
-
As you say, a 12 mp sensor will give poster prints of very high quality, but go much beyond and you will begin to see pixelation. One big problem is the alignment of any lines which are not horizontal or vertical. As individual pixels are expanded, any diagonal or curved lines become misaligned. The more pixels, the greater the expansion of an image before this becomes visible.
So what would be the maximum number of pixels that can be crammed into a defined area. Take, for example, the Canon EOS 7D. This has 19 mp on a sensor measuring 22.3 x 14.9 mm, while the Sony A550 has 14.6 mp on a sensor measuring 23.4 x 15.6 mm. What would be the maximum number of pixels that the Canon's 22.3 x 14.9 mm sensor could have?
I understand that the sweet spot is somewhere around 10µm pixel pitch, but as explained earlier it is a compromise and may shift as technologies develop.
Another thing to bear in mind is that digital image sensors usually use a Bayer colour filter, with half the pixels (in a pettern like all the white squares on a chessboard) sensing green, a quarter sensing red, and a quarter sensing blue. The quoted "megapixel" count counts all these colour subpixels, which is a bit of a con. Since most of the perceived sharpness comes from the green pixels, the "real" image detail would be more realistically described with a figure approximately half the claimed megapixel count.
See also: http://en.wikipedia.org/wiki/Bayer_filter
With regard to the jaggies on non-aligned straight edges: you shouldn't really see these in a well-designed optical system. Jaggies either imply you've overdone the digital sharpness-enhancement, or that your original image has been undersampled (aka aliassed) - which could mean that your lens is too sharp for your sensor. If properly sampled, the jaggies should be filled in with mid-colour (in between the two colour areas) pixels making the boundary appear smooth.
Other symptoms of aliassing include moire interference patterns on regular textures such as bricks, roof tiles, or close-ups of fabrics and hair. Better cameras will use an optical anti-aliassing filter in front of the sensor; this blurs the image very slightly so as to ensure that the light detected by each pixel in the sensor is representative of all the light falling on the surrounding area between the next adjacent pixel (ideally of the same colour-type).
See also: http://en.wikipedia.org/wiki/Anti-aliasing
(Be aware that while sampling-theory and aliassing issues are well understood mathematically, because the symptoms are not uncommon, there is a lot of material written on the subject by non-experts which can be misleading.)
Sensors with smaller pixels (more dead-space between them) will tend to be more prone to moire problems, although if the sensor has more 'resolution' than the lens (which will also tend to happen with small pixel-pitch), then you can save cost and omit the filter.
Professional video cameras use 3 CCDs, one each for red, green, and blue, and a colour-splitting prism. This arrangement removes the need for a Bayer filter (which by definition wastes around 2/3rd of the light) and allows you to have bigger pixels. This gives you better sensitivity and lower noise - at the expense of using three sensors very precisely aligned - and the extra physical space that takes.
See also: http://en.wikipedia.org/wiki/Three-CCD
Note that as well as being present in the original image, aliassing artifacts often arise in the resampling/resizing of images for display if a faster, rather than mathematically pure, algorithm is used.
I have a Canon G9, to-of-the-range compact camera with 12Mpix, which I am very pleased with. In use, some of my pictures really are stunningly sharp to the pixel-level (perhaps when at some sweet spot of the zoom and aperture settings), while most are good, but not razor-sharp. That said, the default level of sharpening is slightly excessive on a pin-sharp optical image, and I prefer the more natural less-sharpened image (which can be set on the custom-colour menu).