Sensors and File Formats There are really two kinds of sensors that you can use for photography: electronic (digital cameras) and chemical (film cameras). Both depend on the fact that light has energy that can be released when the light is absorbed.
Film
Basically, when light hits certain molecules in the film, it changes them so that after you put the film through various chemical baths during development, light of certain colors can or can't pass through the film. So the number of the light-sensitive molecules that change in a certain area during exposure leaves a record of how much light was absorbed by that area. Then after developing the film, that information can be extracted by shining light through the film again. The standard size film for single-lens reflex cameras is 36mm x 24mm (35mm film).
Digital sensors
Digital sensors will usually be referenced to standard 35mm film by a crop factor, which is (aptly) the factor that they crop each dimension of the sensor relative to full frame. Full frame digital sensors are the same size, so their crop factor would be 1 (you get the same field of view as with a 35mm film SLR). Full frame sensors are still very expensive to make, so they're only on high-end SLRs, and a big part of the price difference is from that alone.
Most SLRs' sensors have a crop factor of about 1.5, which means that full frame sensors are half again as big on each side. This might not sound like much but it means that a full frame sensor has a little more than twice the area of one with a crop factor of 1.5. Changing the size of the sensor also changes the field of view—how much of the outside world your sensor can see at a given focal length. Imagine looking outside a window and the window shrinks. Now you see less of the outside world. Shrinking the sensor size is basically the same thing.
A bigger sensor gives you a bigger angle of view for a given lens focal length than a smaller one.
So what actually happens when light hits the sensor? The sensor is an array of tiny photodiodes (diodes only let electrical current go one way) with the cute name of pixels (short for picture elements, which sounds much too formal). When light energy gets absorbed in the sensitive area of a photodiode, the energy is converted into making mobile electrons (current), which then flow to one side of the diode. You count up the number of electrons that got freed during exposure and that tells you how much light got absorbed at that pixel.
If only it were that simple. You've now got three major issues:
(1) How do you count the electrons?
(2) How do you get color information if light of every color can create electrons at the pixel?
(3) How do you change the list of counted electrons at each pixel into something that makes sense to our eyes?
To solve (1)
You have to add little wires and electrical circuits to the photo diodes. That can be done, but all of it uses precious real estate that you could be using to collect light. Your sensor is less and less sensitive the more extra stuff you put on. To try to minimize that problem, they actually put microlenses over each pixel to concentrate incoming light onto the sensitive photo diode and not onto the dead zone with the wires and amplifiers.
This is one of the major advantages large sensors have over small ones: the bigger the area of each pixel, the more light-gathering each pixel can do. I would usually take a 6 megapixel picture from a large DSLR sensor over a 14 megapixel picture from a point and-shoot camera with a small sensor any day.
For (2) they do some tricky things. Most sensors filter the light coming into each pixel so that only light of one color gets to the photo diode for that pixel (the only counterexample I can think of is the Foveon sensor, which has layered photodiodes that get different amounts of light of each color based on their depth in the sensor. It has its own set of problems, though). The standard pattern of color filters over the pixels is called the Bayer pattern.
It turns out that you can make almost all the colors our eyes can see by mixing different combinations of red, green, and blue light. Each pixel in a monitor or TV has three tiny spots, and you get different colors by changing the relative intensity of those three spots. So if you know how much red, green, and blue light should be mixed in, you know which color the pixel should be in the photo. Sometimes you'll hear people talk about the red, green, and/or blue channels, and those values for each color are what they mean. To get the red and blue values for the green pixel in our example, a computer chip in the camera will interpolate: it just averages the closest blue pixels and the closest red pixels to get the complete RGB (red, green, blue) value for that pixel.
That's not the whole story, though, and we finally come to (3).
When our eyes look at something that gives off twice as much light as something else, we see it brighter, but not twice as bright. You'll see this if you sit by a window and check what your camera's automatic exposure does looking out the window and compare it to what it does looking in. (Since most people aren't interested in actually doing things like this, I'll do it for you.) I pointed my camera inside and then, without changing the aperture, I pointed it outside the door.
The shutter speed for proper exposure changed from 1/100 to 1/2500, so right now my camera is telling me there is 25 times more light outside than in, but it looks like maybe 3 times brighter to me. So who's right? We both are. To describe this, you would say that eyes have a nonlinear response to light—all that means is that the response doesn't follow a line: doubling light intensity does not double the perceived intensity.
Digital sensors are almost completely linear: twice the light means twice the electrons that get counted. So to make the pictures look right to us, the values the sensor records have to be stretched. This is also done by a computer in the camera. Or, you can save exactly what the sensor sees and have the stretching done by your own computer (it’s called a raw format and I'll get to it in a minute).
ISO Speed
So besides controlling the exposure of your sensor, there is another setting on your camera that lets you decide how sensitive the sensor is to light. They call it the ISO sensitivity, and it's a throwback to the days of film, when people would classify films based on how much exposure they needed to beproperly exposed. The higher the ISO speed of the film, the more sensitive it is to light, and the numbering is conveniently chosen so that ISO 200 is one stop more sensitive than ISO 100 (it needs half the light ISO 100 needs), ISO 400 is two stops more sensitive than ISO 100 (it needs 1/4 the light ISO 100 needs), and so on. In digital cameras, what you actually change is how much you amplify the number of charges made during exposure before reading it out.
So set your camera to the highest ISO and then you don't need fast lenses or long exposure times, right? Well, it's not a free lunch. The more you amplify the signal, the more you amplify the noise from each pixel. It's about the law of large numbers: each photo diode doesn't even come close to absorbing every photon (light particle) that goes through it (it's a 25% kind of thing), and if you give it enough photons to get a good feel for how much light there is, you get about the same number of electrons for a given amount of light every time. But if there are only a few photons to catch, the value from one photo to the next (or from one pixel to the next) will probably be wildly different. Amplify those differences and...well...yuck. The general rule is to use higher ISO only when you have to, and use the smallest possible ISO in every situation.
File formats
I mentioned before that you don't always need to have your camera convert the raw data from the sensor's electron counting into something reasonable. Some cameras support a raw format where you just store the red, green, or blue value (but not all three) from 6 each pixel, as captured. I'll talk more about raw files in a minute, but for now I'll mention the other types of files so we can compare them. JPEG or JPG Jpeg files are by far the most common these days.
They are a compressed format, which means they don't actually store the RGB values for each pixel, but through some complex math, it stores the way to get pretty close to each pixel's RGB value when you reverse the process. Jpegs are 8-bit, which actually means that the red, green, and blue channels are each broken down into 256 levels (or 28 , hence 8-bit) before the compression. Somewhere black on your image might have RGB values of [12, 7, 11] and a pixel in the sky might be [50, 85, 220]. The major advantage with jpegs is that they make files so much smaller.
TIFF or TIF
Tiff files can be either 8-bit or 16-bit (256 levels per channel or 65536 levels per channel). There are compression schemes for tiff files, but they don't compress files nearly as much as jpeg compression does. Tiffs store an RGB value for every pixel, so tiff files get huge: 16 bits is two bytes, and with three channels, that's 6 bytes per pixel. A 10- megapixel 16-bit tiff with no compression, then, will set you back about 60 megabytes, and an 8-bit version will be half that. Yikes. You can see why tiffs are usually used in commercial settings where even the slight loss of quality you get with jpegs is unacceptable.
RAW (CR2, NEF, DNG, etc.)
Finally, back to my favorite file type and the one that makes the most sense to me for backups: raw. Even if your camera's sensor reads 12 bits (4096 levels) for each pixel, you can have a raw file that is smaller than an 8- bit tiff because you only keep one channel per pixel—the file is stored before the interpolation of the reds, greens, and blues. For example, that same 10-megapixel picture in uncompressed raw format will have 1.5 bytes per pixel, and you get a 15 megabyte raw file.
That seems like a lot, but you don't throw away any of the information the sensor records, which can come in handy if you need to change your pictures later (and let's face it, we often do). My personal opinion is that if you have a camera that supports raw capture, you should always use it so information that might be useful someday isn't thrown away. It can't work miracles on really bad pictures, but it can help and it's only slightly more work to process raw, anyway.
For the still-curious
A white paper on how you go from linear capture from a digital sensor to something logarithmic like your eye would see. [http://www.adobe.com/digitalimag/pdfs/linear_gamma.pdf] A really good article on shooting raw. [http:// www.bythom.com/qadraw.htm]
No comments:
Post a Comment