8 MegaPixel Digital Sensor Unveiled

hdtv writes "Micron has unveiled an 8-megapixel digital sensor, that 'enables pocket-sized cameras and cell phones to capture bursts of 10 high-quality photos in a single second or even high-definition video.'" From the article: "'We're saying it can go in a point-and-shoot camera selling in the $200 to $300 range,' said Suresh Venkatrama, Micron's director of the digital camera segment. 'It brings high-quality digital video and photography down to the consumer space.' The new sensor is a type of chip known as a 'complementary metal-oxide semiconductor,' or CMOS. Analysts say the technology, which is also used in memory chips and microprocessors, will challenge the dominance of traditional light-sensing charge-coupled devices, or CCDs."

Comments

What's new exactly?

[TangoCharlie]

CMOS isn't new.
Digital camera's aren't new.
CMOS digital camera's aren't new.
So, what's new? So cameras can take 10 pictures in quick succession... Is that new? Erm.... no. My 3yrl old Minolta can store pictures in RAM before they get stored to the SD card so that you can take pictures quickly.

Nothing to see, please move along.

Where's the useful cut-off point?

[EmbeddedJanitor]

After a certain pixel count, there's not much point in going much further in consumer devices. You're just adding data, but not improving the viewable image. Why have an image that is higher res than a monitor or your eyeball's ability to process data? Where is that cut-off?

A bit like sound... once you get to CD quality, there's not much point in going any further because the speakers, amplifiers etc cause the most distortion and any improvement at theCD end will not make it to your ears.

Re:Where's the useful cut-off point?

[Intron]

"A high pixel count also means smaller physical sizes for each pixel on the sensor which means an increase in noise." Hunh? Externally generated noise hits a large pixel and changes it 25% or hits a small pixel and changes it all the way from black to white. Net result in image quality - looks the same because the three other small pixels around it didn't change. You can make the same arguments for internal random shot noise. The amount of noise hitting the camera doesn't change, so its net effect on the resulting image doesn't change.

Re:Where's the useful cut-off point?

[modecx]

Where is that cut-off?

There is no cut-off. That's the point. It may be that in our future, we have higer resolution displays, holograms, holo-deck, whatever. Who cares.

Personally, I won't say "That's Enough" until I can capture an entire landscape with such resolution that you can zoom in and clearly see the eyes of every damn ant that happened to be crawling around that day... Or better yet, print the whole thing at life-size, to cover a skyscraper or something. Okay, so that's probably not going to happen, and even if it did happen, it might not even be useful, but goddamn, it WOULD be cool.

Now, 8MP in a phone camera probably isn't going to be more useful than say, a 5MP sensor in a phone, as the resolution will certianly be limited by the cheap-ass plastic lens they put in front of it--and by any pocket-lint that it may have gained. The size of the aperture will limit the amount of light it can collect, and the noise generated by all those tiny sensors packed in there will probably not be a good thing for image quality--you'll just get even larger grainy, blury photos than you would with the 5MP sensor, which would be useful if you wanted to take big-ass grainy photos, I guess. I'm sure someone will find an art in it.

Re:Where's the useful cut-off point?

[DrDitto]

300dpi is considered a high-quality print.

Which means for an 8x10, you need an image that is 2400x3000 (or 7.2 megapixels).

Many claim that the human eye can indeed resolve the differences between 300dpi and 400dpi. At 400dpi, an 8-megapixel sensor falls well short of an 8x10 print.

Personally I shoot with a large-format 4x5" camera. This gives me 20" inches of film area, and when scanned at a modest 2400dpi, this gives me 115 megapixels. And my equipment (besides the scanner and film) is 30+ years old.

Re:Where's the useful cut-off point?

[Xyrus]

"If you're trying for 8x10" pictures, 8Mp is "enough"."

Don't know what camera you're using, but with a Nikon D70 6 MP camera I can blow up pictures a good deal larger than 8x10 and still have it look just fine. I'm not saying you could create a poster out of it but you can definately do larger than 8x10 with no significant loss of picture quality.

As was stated earlier, it all depends on the sensor size. A 6MP DSLR will give you much better pictures than a 6MP PAS.

~X~

Re:Where's the useful cut-off point?

[swillden]

However, I've also got an 8 mega pixel Canon digital SLR, and the picture quality is vastly superior

What lens do you have on that? I just bought a Rebel XT and although I was really impressed with the quality of the images, it was pointed out to me that the 18-55mm f/3.5-5.6 lens it came with really isn't a very good lens. It's not very sharp, especially in the corners, it's a slow lens, doesn't have very good depth of field, etc. At a friend's suggestion, I bought an inexpensive 50mm f/1.8 prime (non-zoom) lens, and I have been amazed by how much better the image quality is. I've shot the same scene with both lenses and the difference isn't subtle at all. Not only is the 50mm much sharper, but when you look at the pictures side by side, it's obvious that the 18-55mm gathers light unevenly. The picture is darker in all of the corners and in the right and left edges. Pictures with high contrast edges show pretty obvious color distortion with the 18-55 as well.

Oh, and if you're taking pictures indoors, that f/5.6 is just way too slow. You need a lot of light with that lens.

Note that I'm not actually knocking the 18-55; compared to my old camera (which wasn't junk, either) it takes *fantastic* pictures. But experimenting with the 50/1.8 has made me realize that the optics matter -- a lot. That seems like kind of a stupid thing to say, in retrospect, since it should be utterly obvious that optics are important to a camera, but I kind of assumed that the differences between lenses, other than zoom power, were subtle and maybe even subjective. They're not. Which makes it obvious that the tiny lenses that can be crammed into a compact camera are always going to be limited. No matter how many megapixels the sensor can capture, if the glass can't focus the light onto it properly, the pictures aren't going to be very good.

Consumer Level Camera Use

[fdiskne1]

My first thought was, "Cool! Higher resolution is dropping in price again." My second thought was, "Crap, now my users will be trying to email each other 3 megabyte and larger photos on a regular basis." I'm wishing there was some way new digital cameras could come with an education of what filesize means and how it relates to emailing and otherwise sharing with others. A large number of non-technical users have no idea of the concept. I've run across people wanting to email software CDs and copy DVD movies (inhouse, not MPAA) onto WAN-wide file shares.

Re:Consumer Level Camera Use

[Amouth]

i have given up on teaching users this.. when they ask why it is slow i tell them it is big and will take a while..... and some times they wait...

Sensors? Pfft...

[Locus Mote]

In terms of high-end photography, there are several requirements which rate MUCH higher than simple FPS:

Input Dynamic Range. This is the range of light values in a scene which the sensor "sees" and is able to record. In order to understand this, think of light at dusk reflecting off wet pavement in the distance. The super bright orange glare hitting your eyes is extremely high intensity light, while the shadowed sides of houses and trees and things are low intensity light. Both of these elements have detail that can be recorded. With a low dynamic range, one or the other can be exposed properly. With a high dynamic range it is possible to capture the detail in very dark shadows and very bright highlights without clipping. (Clipping is truncation to flat black or white pixels with no detail). Chemical film, especially positive film (slide film), has a dynamic range which obliterates the best digital sensors.

Falloff. This is the ability to clip gracefully. When using any type of transducer, whether it's a microphone, a square of film or a digital sensor, there is a response curve which maps values input values (light/sound) into recordable output values. In the age of analog (vacuum tubes, vinyl records and chemical film) the response curves were all based on Calculus. They literally rolled off (logarithmic) at the ends. This meant that as the microphone, vacuum tube or film overloaded, it did so gracefully with a smooth transition to clipping. In the digital world, our chips are "dumb". They can only do algebra, not calculus. Their falloff is linear. 8-bit = 256 values, 16-bit = 65536 values, etc. Anything above or below this is immediately clipped to white or black, on or off. The digital world is flat, if your input source is flat, you sail right off the edge into infinity.

Single Pixel Resolution. 99.99% of digital camera sensors use a single layer of matrixed sensors (Bayer array). These sensors are located in gangs of three, similarly to the pixels on an old CRT television. The problem is that each sensor can only see red, green or blue. There is a lot of jibber-jabber that I could go on with, but essentially, bayer sensors really only see 1/3 of the picture information their lenses dump on them. Chemical film is stacked in layers, thus each pixel location "sees" all three RGB. Currently only the Foveon X3 sensor in Sigma digital cameras is capable resolving all the information in each color channel at each individual pixel location.

Now, even if this new Matrix chip performs at even the sub-par level of today's CCD camera sensors, simply buying a camera with one in it does not by any means guarantee quality photography. Back when the sensor (film!) was interchangable from camera to camera, there was still intense competition between camera and lens manufacturers. This is because the sensor can only "see" the image that the lens and camera body deliver to it. The most important factor is the lens! Imagine rubbing vaseline on your glasses and walking around like that all day. This is life with a cheap camera lens. There's a reason why most professional lenses, without a camera body, cost betweed two and ten times as much as an entire point and shoot camera. If a lens is a valve for light, then a professional lens is like a firehose, a prosumer is like a garden hose, a point and shoot is a drinking straw and a cameraphone is a hypodermic needle.

--

That's my 2(6.022*10^23) cents worth.

over a dozen filters....

[Cryptnotic]

Actually, the Mars rover camera used a filter wheel with over a dozen filters (not just red, green, blue, but different shades and ultraviolet and infrared, if I recall correctly). They would take one image at each camera position with each filter, then those would be composited back on Earth. Then the camera was repositioned for the next position and the sequence was repeated. The resulting set of images were stiched together into large panoramas.