Slashdot Mirror
Seitz's 160 Megapixel Digital Camera
An anonymous reader writes "Digital cameras had been lagging behind Moores law for a while, but Seitz has taken a massive step forward with their announcement of a 160 Megapixel digital camera! At almost 20" long, with a price tag of around $36,000, and with on-board gigabit ethernet to copy off the image it's not exactly going to take on the consumer market, but how long before we see this resolution in a mobile phone?
Even with todays current range of digital cameras massive images are possible — such as the amazing 720 Megapixel image of Sydney Harbour"
Even with todays current range of digital cameras massive images are possible — such as the amazing 720 Megapixel image of Sydney Harbour"
From the submission: but how long before we see this resolution in a mobile phone?
Enough with stupid tag questions already! Would submitters and editors please stop with this insanity - we don't need to be *led* into a discussion, we're good enough already.
Global warming is a cube.
The future of high-rez pr0n!
megapixels without good non-fixed lens == pissing away bits.
Makes for great marketing though. Let them megahur^H^H^Hpixels fly! See, the megahurtz race didn't come back to bite the industry too hard, so no reason to learn.
Ahem. Gigapxl Project.
The sysadmins that host the 720 megapixel image of Sydney are probably not going to be sending you thank-you cards, I'm guessing.
That's it. Link a 720 megapixel image, on the front page of Slashdot, from an Aussie server, just as North America is getting into the office and commencing "working." ;)
Maybe they need to replace the camera on Hubble Space Telescope with one of these cameras? Seems like only the government can afford one of these things, and can blame the manufacturer if the technician drops the camera when taking it out of the box.
to say: wow. I think my jaw just dropped.
Religion is what happens when nature strikes and groupthink goes wrong.
It's a lens with a scanner !
Fast scanner, big resolution scanner!
But a lens with a scanner !
Ceci n'est pas une Signature !
I have no idea how the photosensitive surface of a digicam works but I'd guess they are built using semiconductors. Obviously a higher element density on semiconductors means more pixel receptors on the surface then.
Justice is the sheep getting arrested while an impartial judge declares the vote void.
160MP in a pinhole camera. That'll be sure to produce some great results...
This guy's the limit!
What is the big point in churning up the pixel count, if the dynamic range is the same old 1.0e03? Human retina has a dynamic range 1.0e06, three orders of magnitude better. And it has about 2.7 million rod cells and cone cells. One can create amazing speakers with absolutely perfect sound fidelity at 150 KHz, but human ear cant hear it. There could be some applications not involving human hearing/cdplayers/boom boxes. But at that point it is not really a "speaker". Same way at 160e06 pixels or 720e06 pixels it is not a "digital camera". It is some exotic machine with really pathetic dynamic range and huge number of pixels.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
I find it interesting that Linux is mentionned as a control platform from the start. I doubt that many professional photographers actually use it since the tools aren't quite up to par with the commercial stuff for a number of uses (no "but Gimp can do it too whining please"). Were there that many requests for Linux support ? Or are the makers just Unix hackers ?
Even the portable control device is apparently by default a Sharp Zaurus.
May contain traces of nut.
Made from the freshest electrons.
20" long eh? That's almost big enough to *SANTIZED BY FCC* in one shot!
stuff |
Lets do some back of the envelope math here, shall we? While it will depend heavily on how well the scene compresses, my digital camera averages about 400kb for a 2 megapixel shot at "pretty decent for mailing home quality" JPG. So that Aussie harbor is probably weighing in at above 300 MBs, and its a static file. This sounds like a job for bittorrent, not for "Hey, I've got a bright idea, lets Slashdot their webpage as the morning rush comes in".
Help poke pirates in the eyepatch, arr.
The reason people use DSLRs is because even at todays 6-8 Megapixels the lens is the weak element.
Add all the pixels you want, without a bigger and better lens it doesn't matter.
Sure we can improve on the dynamic range and noise of the sensor, but the megapixel days are over.
Since pixels need to collect photons in order to generate the electrons that form an image, the smaller you make them, the less responsive they are. With smaller and smaller pixels, you either need longer exposure times (opening yourself up to blur if the subject is moving), or larger lenses (which cost mucho mucho dinero). People are already making pixels at 2.5um pitch. You are unlikely to see any further major reduction in that size, given the constraints of responsivity.
It may look like I'm doing nothing, but I'm actively waiting for my problems to go away.
--Scott Adams
he's in the second tall building from the left, 12th floor, 6th window. he's the one screwing his secretary.
I am not sure that scan back digital cameras will ever be very useful for much more than studio work, as illustrated by the awful ergonomics of this particular one.
Molding finger grooves into anything is a silly idea. Molding them into a camera that couldn't possibly be handheld has to rank up there in the sillydom world.
"Sacrifice for the good of The State" - The State
I highly recommend giving Dans "Enough already with the megapixels" article a read. He explains the situation more clearly than I ever could.
Censorship is obscene. Patriotism is bigotry. Faith is a vice. Slashdot 2.0 sucks.
Someone explain to me how that applies to Digital Cameras pixal density.
http://www.howstuffworks.com/digital-camera.htm
KFG
This promo image of the camera makes it look like goatsee...
---- You are fully entitled to my opinion.
The article is quite skimpy on the technology. By the size of the camera I am guessing all they have done is to split the image optically into some 16 or 20 pieces and are using some 18 or 21 CCD image sensors to capture the image. They read these chips in parallel and load it into an internal buffer. The dynamic range if each CCD sensor is exactly same as what you could get in 5 Megapixel camera.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
This is why even a pretty good 6MP sensor - like, say, a Nikon D50 today - can produce much better pictures than a crappy 8MP point-n-shoot camera sensor does. By better I mean cleaner, less noisy, with more realistic color, etc, etc. Even with good lenses, the tiny sensors just aren't getting the light information they need to do well.
You're special forces then? That's great! I just love your olympics!
Anyone notice that the data storage device itself is a mac mini in a bag?
The CB App. What's your 20?
"With smaller and smaller pixels, you either need longer exposure times"
If the gaps between the pixels are getting larger, yeah... but if you have four pixels that are quarter the size, they receive quarter of the amount of light in the same amount of time... put the four together, and you end up with the same amount of light, no?
The revolution will not be televised... but it will have a page on Wikipedia
With other technological advances there are reasons for the extra speed, larger storage capacity, etc. Photography however suffers from other limitations that make anything above 8-16Mpixels virtually useless.
At best computer monitors have a resolution of 1600x1200, so without significantly zooming out, you can never display the entire picture on the screen. Printing is the only area where more Mpixels are needed, but even there, at 8.5x11 8-16 Mpixel images are crisp enough. There ARE areas where extremely high resolutions are needed, but they're definitely not consumer level.
I have an 8Mpixel camera, and I am not likely to want more any time soon.
The camera is fairly specialized - it's a panoramic camera for commercial photographers. 160MP makes sense if you're going to do very large mural size prints. Think, giant travel scenery or product posters at trade shows or other commercial venues.
[Insert pithy quote here]
I think that I will *not* be taking pictures of my coworkers with this. I don't want to see anyone I know in that kind of detail. My most of my co-workers look like this anyway. Why would I want a closer-in shot to see the pores, etc.
2 cents,
QueenB
HDGary secures my bank
We take up a collection to purchase one of these for the playboy photographers.
[..]but how long before we see this resolution in a mobile phone?
I do realize that the future will bring us things that we simply cannot understand the use of today, such as computers exceeding today's super computers. But I doubt that just because the tech is going to be there, that we will see 160 MP consumer cams. Eventually, people will stop hearing megapixel and instead listen to other intuitive features. Maybe built-in software with 3D depth readability and such?
Scanners are about the same. Back when we had 300 DPI scanners, it was all about DPI. Now that scanners can make the balls of a fly look hairy with perception, there's no need for more.
OT: If you're thinking that 160 MP is a lot, how about 4 GP? You can check the proof right here.
Full Tilt
"Even with todays current range of digital cameras massive images are possible -- such as the amazing 720 Megapixel image of Sydney Harbour"
To be fair ~ That image was made with 169 images from a Canon EOS 10D that has 6.3 Megapixels and then the multiple images were stitched together using AutPan Pro.
Was the Goatse man inspiration for this design?
I would love to own one, though.
"how long before we see this resolution in a mobile phone?"
long.
-- http://frobnosticate.com
The Gigapan creates multi-gigapixel panoramas using an off-the-shelf digital camera. The downside is that many pictures are taken and then combined into a single panorama. The upside is that it's much higher-resolution and a lot cheaper than $36K. Check out the site -- if you zoom in on the Golden Gate bridge you can see the speed limit sign :-).
I think it came from the part that says "but how long before we see this resolution in a mobile phone?"
HTH.
Granted it is difficult to get great pictures using a small fixed lens (ie cameraphones), and additionally just increasing pixel count doesn't directly increase image quailty, so I would think that part of the reason that we are holding at roughly VGA resolution cameraphones has something to do with storage requirements. *Most* camera phones are now the standard/base model phones. They don't spend the extra money on hardware for external storage, and also most of their users don't want to spend the additionaly money on storage. All of that to say, even if this was "available" for cameraphones, until we can easily have/add several gigs of storage, the ability to have several photos will probably be more useful that having *better* photos.
When I have a kid, I want to put him in one of those strollers for twins and then run around the mall looking frantic.
Insert comment here about how this technology could be used to render highly detailed and accurate images of the undraped human form.
"How to Do Nothing," kids activities, back in print!
It doesn't matter if you have a bijillion megapixels in your cell phone camera. Tiny crappy lense = crappy pictures.
The masses are the crack whores of religion.
I've never seen Sydney in person, but now I know that just about every building in the city is tagged with "Scott Howard" graffiti. The guy is just out of control.
-USR1
My phone has a 1 MP camera on it, my camera has 5MP, but an option to shoot at 1MP. Now, if i take a 1 MP picture with both of the same scene, the one i took with my camera looks exactly like the one i took with my phone, but without the snowstorm that seems to be raging on in the phone version. 700+ MP is not going to be any better in terms of actual image quality then current professional >50 MP cameras.
my capcha was condom
Even if you are surrounded by walls.
You just have to put some big panoramas on the walls.
We had that on one wall of our living room. Imagine that on the 3 walls of your garden. You add a powefull light for the sun and voilà!
We can now all go build condo on the grand canyon without destroying the view as it has already been digitalized at 160 Mpixels.
And If you are bored with the Grand Canyon after a while, you can replace the view by a view of Antartica. It will be refreshing.
Strange. The gear of those wind surfers is definitely pre-2000...
Seitz: 160 megapixel in a 60x170mm sensor = 15,686 pixels per mm^2
1Ds: 11.4 MP in a 35.8x23.8mm sensor = 13,379 pixels per mm^2
Rebel: 6.3 MP in a 22.7x15.1mm sensor = 18,379 pixels per mm^2
The digital rebel has a higher pixel density than the Seitz. According to your quote, that makes the Seitz more responsive than the rebel but less than the 1Ds.
Like usual around here, the invocation of Moore is just to get
(I prefer Canon so substitute in your preferred cameras where you see fit.)
:wq
Cell phone cameras will never be an acceptable substitute for the hobbyist or pro-sumer. Megapixels are only a small part of what makes a quality photograph. Even if a cell phone were to incorporate manual settings for shutter speed, aperture and focus, there will always be one area in which they can not match a decent camera, and that is the optics. There's just not enough room on a cell phone body to incorporate a quality lens.
Due to circumstances beyond my control, I am master of my fate and captain of my soul.
These guys:
http://www.betterlight.com/products4X5.asp
Have been making high resolution scanning backs for large format cameras for years now.
Yes, but you start to get more noise in the photo. By dividing the total into smaller and smaller quantities, the sampling error will grow. The only solution is to come up with better and better measuring mechanisms, which they do with each generation of sensors, but the improvements are gradual.
Longer exposures overcome this limitation. It's kind of like using a stopwatch. How precicely can you measure a millisecond? How about a second? A minute? Your measuring error is pretty consistant no matter what length of time you are measuring. The difference is, the longer the span of time you are trying to measure, the less significant your error is.
Well such scanner cameras aren't new. Many years ago, back when consumer digital cameras only stored 320x240 16 greyscale monocrome pictures, I have heared of one of those taking about a minute to take a high resolution picture.
What is rather amazing, however, is the speed of that camera. It can actually scan the whole picture in a single second. That's almoust like a real camera.
My dream is to have a fisheye-lens and a wicked amount of detail. That way I can take a picture without knowing exactly what I'm photographing. When I get home I can find many interesting high resolution photos of stuff I didn't even see when I was there.
That would open up for a completely different kind of photography. Put this in a mobile phone, and take one of those boring pictures of your friend looking very uninteresting on the bus, but now in the same picture you may find an interesting scene happening on the side walk.
Yeah yeah, it might not be worth the time once you get used to it, but I'd sure like to try.
Well, if you really want to get into it, the Seitz sensor (actually made by Dalsa) is a TDI sensor. It is referred to as a "high-sensitivity linescan". In linescan sensors, there is a single row of pixels. You capture the image one row at a time. A TDI sensor works the same way as a linescan, capturing the image one row at a time. However, the difference is that there are more rows of pixels on the sensor. The electrons are moved through the sensor at the same rate that the image moves across it. This means you can get longer integration times per row, without decreasing the speed of the sensor readout.
Dalsa has a website that describes the different types of sensors and has diagrams that explain the functioning of TDI sensors.
The Dalsa sensor itself is not 60mm x 170mm. It is 60mm tall, and scans across an area 170mm long. The sensor itself actually pans across the back of the camera, to capture the entire image.
It may look like I'm doing nothing, but I'm actively waiting for my problems to go away.
--Scott Adams
You just slashdotted my country.
It's a Bagel.
We're not talking science fiction. The concept has been tested in practical application and yielded orders of clarity beyond the diffraction limits of the wavelengths of light being captured.
Yes . . . and no. Each photon that hits a pixel will create one electron (gross simplification). The electrons are shuffled through the CCDs in packets until they get to an amplifier at the end, which converts the charge into a voltage that can then be fed through an ADC to give you a digital number. Each pixel gets read out separately.
With a 10um x 10um pixel, you'll collect (say) 100,000 electrons at a certain exposure time. This translates to a digital number of 1024 (or whatever). If each pixel is only 5um x 5um, each pixel will collect 25,000 electrons, and the resulting digital number is 256. To get the same output with pixels 1/4 the size, you need 4x the light - either a larger aperture (or lens) or a longer exposure time.
The numbers can't just be gained up (digitally or in the analog world) because of noise. The amplifier on the CCD will give you a certain number of electrons in noise. If you get a larger signal through gaining it up, you also gain up the noise. You'll end up with a smaller dynamic range in the camera, where dynamic range is defined as the signal level at saturation divided by the noise level.
Now, there are some people who use smaller pixels, but use something called "binning" in low-light situations. In CCDs, it is possible to add the electrons from several different pixels together before doing the charge-to-voltage conversion. However, this results in a loss of resolution. With 2x2 binning, your 2k x 2k sensor effectively becomes a 1k x 1k sensor with four times the sensitivity.
It may look like I'm doing nothing, but I'm actively waiting for my problems to go away.
--Scott Adams
I read an article a few years back rating film resolution. They used "Pro" 35mm cameras with the best available lenses at the time, a good tripod, and test-pattern images. The best films rated in at a bit over 100 line-pairs per millimeter. That's 100 black lines with 100 equally-sized white lines between them, or 200 dots per millimeter. When you digitize, you play it safe and double that number to 400 dots/mm.
400 dots/mm on 24mm X 36mm film is 9600x14400 dots, or 138.24 megapixels.
When we can squeeze 138.24 megapixels down to a 24mm X 36mm area, "we have arrived." I'm putting my money on this being available in high-end-yet-still-under-$2000 cameras by 2012.
By the way, for some applications, such as portraiture, 8 megapixels produces beautiful 20"x30" prints. However, some applications demand better, particularly those involving severe cropping and expanding.
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
in the cellar.
I guess if you're printing the photos billboard-sized, and stand 3 feet away to view them, you need this resolution. Otherwise it just eats up disk space. I'm a pro with 8 megapixel Canon 1D2N and 20D's, and with a top-end lens, 13x19 inch prints look fantastic. Heck, a billboard from my camera would look OK - if your viewing distance is a couple hundred feet.
Hmm. What a dilemma. I care about the environment, but should I risk using rechargeable batteries in this thing?
http://outcampaign.org/
. . . making a digital camera that doesn't have a smear or blur issue during low light situations. Or one that will allow the user to take more pictures in quicker succession. I can get about 2 a second, which is okay, but lets advance some of the other features before we make more megapixels. What will that do for me anyways unless I want to make a Citizen Kane sized print of myself.
Can I bum a sig?
http://www.roundshot.ch/pictures/Seitz-6x17-handhe ld.jpg
When I want to do a proper landscape, I put my DSLR in the bag and pull out the $200 Graflex Crown Graphic from 1947. The 4x5 transparency film will be scanned at 2400 dpi on my $300 Epson 4990 flatbed scanner to give me just about 100 megapixels... Largely enough to get my 2GB desktop computer to it's knees when opened in photoshop :-)
Now, thats a deal.. Scanning medium and large format transparency gives fantastic images when scanned; you can the advantages of film (tones, size, and free 70 years backup of the image...) and the advantages of digital (photoshop!)
So, how much longer do I have to wait for the Hi-Res porn to arrive on the market?
Knowing Google's lust for data collection, the Soviet Union is still alive and well inside the psyche of Sergey Brin....
BetterLight makes many scan backs for large format, essentially what the Seitz back is, but the BetterLight models have a less panoramic ratio, and are cheaper. See http://www.betterlight.com/superModels.asp for their list of the high end models, but the highest end is the Super10K-2, 10200 x 13800 pixels - Native CCD resolution, 402 MB max. file 24-bit RGB (804 MB in 48-bit RGB). They don't call it a 140 megapixel back, because that would imply that it takes the entire 10200x13800 image at once, instead of by scanning, and they're a bit more honest than Seitz. Doing the scan in 1 second is impressive, but it's still a scan back.
Sounds just like the aperture size on a mechanical camera.
Justice is the sheep getting arrested while an impartial judge declares the vote void.
Maybe we can use the 160 megapixel camera to take pictures of the servers hosting the Sidney image exploding, just like the Cox laptop.
If I knew the wedgies I gave you back in 6th grade would have resulted in this . . . I might have taken a moments pause.
how long before we see this resolution in a mobile phone?
Never. The basic limit of resolution you can get is set by the Rayleigh criterion:
sin theta = 1.22 * wavelength / lens diameter
where theta is the angular diameter of the smallest detail that can be resolved.
Using a 5*10^-7 m (green light, more or less in the middle of the visible spectrum) and a 0.01 m diameter lens (which is generous for a mobile phone), this gives us a 3.5*10^-3 degree angle as the minimum amount of viewfield that can be covered by one pixel. Thus, a picture with a 20 degree viewfield* would be, at most, 5700 pixels in each dimension, or 32.5 megapixels.
*Of course, a viewfield could be wider, but getting a wider-angle picture without distortion raises a whole other batch of problems if you have to do it in such a small package.
/. If the government wants us to respect the law, it should set a better example.
As someone who derives income from photography and photo retouching, I'm calling you out on knowing nothing about actual "professional photographer" imaging workflow.
"there are a few things that GIMP doesn't do or doesn't do well (like Photoshop's Layer Effects)" which is extremely useful and necessary to mine and others' photo retouching processes. Yes, you might be able to duplicate the effect in GIMP by using a workaround, but not as quickly as is possible with the variety of layer effects and adjustment layers available in Photoshop. As far as I and many of my peers are concerned, photo retouching just isn't done without layers.
looking at the website, it appears these features are also missing or incomplete:
-RAW image file editing
-colorspace management
-good documentation
-commercial support
Gimp would be well served to just duplicate Photoshop instead of trying to reinvent the wheel. Photoshop works. Well. It has years of development and millions of hours of testing by people that are serious about image editing. Gimp works. Not as well as photoshop. It's built by computer geeks for computer geeks. It's used by people that don't take image editing seriously enough to shell out a couple hundred dollars for Photoshop. Why would a professional photographer with thousands of dollars in camera equipment wrestle with a non-standard hackjob of Gimp and associated utilities when a couple hundred dollars buys them a fully functional all needs met solution known as photoshop? Hell, most digital SLR cameras these days come with a copy of Photoshop Elements and cheap options to upgrade to a license of the full Photoshop.
Any photographer that's actually making money isn't going to waste time cutting at best 1% of their overhead cost by using an inferior tool. The time wasted using Gimp instead of Photoshop is better used marketing, working for clients, or learning to take photos that need less retouching in the first place. Gimp should just give up on the whole "it can work for professional photographers too" thing. There's no incentive for us to switch to Gimp.
More like grain size in conventional film. Faster speed film (higher ASA) have larger grains, with faster response. The fine-grained films (low ASA numbers) require longer exposure times because the light-sensitive grains are smaller.
It may look like I'm doing nothing, but I'm actively waiting for my problems to go away.
--Scott Adams
"That being said, most professional photographers use Macs because they know the interface well."
Er... the reason they know the interface well is because they all use it, not the other way around. The Mac is the workhorse of the graphics industry because Apple has accommodated the needs of that industry. If you are just taking snapshots for your mum's webpage the yes, any old PC with a halfway decent image editor will do. But if you are doing graphics layouts for magazine spreads where the final result is high quality print, then you absolutely need CMYK color management. That is built into the Mac. Not Windows and not Linux. Mac is king of the graphics niche for precisely that reason.
Printers speak CMYK, they do not understand RGB. Most software allows you to tweak away until you get something acceptable, but a professional printshop is not going to bother. It has to be right when they get it. So the Mac can speak directly to the printer and get it right. Yes, you can buy expensive software and spend some time calibrating Windows for that, and with a little more effort you can get Linux there too, but why bother when the Mac does it out of the box? Until MS & Linux developers understand that, Mac will remain king.
Some clarification - in a standard CMOS process you'd be getting about 0.3 electrons per photon, in an optimised CMOS process you might get 0.4 electrons per photon or maybe 0.5 electrons if you're lucky.
Cheers,
Roger
Do you have any better hostages?
It's because the Mac was designed by artists, for artists. They all speak the same language. Duh.
And now, a PSA from David Lynch.
The tag says that there will never be a 160 megapixel camera phone but I say "never say never". I mean look how dumb Bill Gates looks now with his "You'll never need more memory then this" quote.
(eg: if you've actual light intensities of 2 4 8 4 2, and your sensor's pixel size is 3 times that - so it overlaps 3 of these light intensities at a simgle shot, then you will get values of 2 6 14 16 14 6 2, assuming you can move the sensor one third of the distance covered by a single pixel and the extreme limits of the sensor allow the last pixel to be 2/3rds covered. Simply subtracting out what you've already collected within the window allows you to infer what the remaining 1/3 of the window MUST be for the values to be correct.)
This use of filtering is by no means your only option. I would also totally disregard all colour sensors and use a colour pinwheel. Scan with a red filter, then scan with a green filter, then scan with a blue filter. The technique was used by a Russian photographer to make colour photographs of very high quality in about 1910 (although by having microsecond filter switches, rather than several minutes of ripping the camera open and swapping plates, you could avoid the distortion caused by motion). It was also used by John Logi Baird, when he invented this strange thing called television, but his colour displays were absurdly primitive even for the time and were dumped in preference to black-and-white rasterised displays. Regardless, the use of a high-resolution monochrome system with a colour wheel to create the illusion of a colour display has a long and illustrious history.
Greyscale CCDs and other light-sensitive devices are often at vastly higher resolutions than colour systems, so by using one such device and a rotating wheel to filter correctly, you would be able to produce a much higher resolution device than you could from a purely photosensitive device alone.
A third method is instead of using just a converging lens, if you added a diverging lens just after the focal point, you could spread the image out. Instead of trying to mimic pixels closer together, simply move the light rays further apart. The effect is logically the same. You could do the same thing by moving the sensor further away, as the density of the light falls off with the square of the distance, but it's hard to get a good photograph with a thirty-foot camera.
Of course, there's no reason you couldn't combine the techniques to produce even greater resolutions yet, or special-purpose photographs that cannot be produced using a conventional system. For example, if you had a four-colour pinwheel, you could produce photographs that people with tetrachromatic vision could correctly see. By using subtraction/bitmasking for pixel interpolation, it may be possible to devise a cheap way to handly high contrast/high dynamic range images that conventional digital cameras are useless at. Mind you, you'd need a camera that supported JPEG2000 or OpenEXR, which most cameras don't, but that's a relatively minor software issue.
All in all, camera technology is advancing for those people who have a particular need for it to, but vendors have no interest in supplying Joe Average with high-tech gizmos. Aside from the fact that Joe Average can barely take photos, the professional guys have large bags of money and are willing to pay. Who, in their right minds, would sell a dirt-cheap gigapixel high-dynamic-range eight-colour motion-cancelling tea-making camera for a hundred bucks to ten thousand wannabes, when they can sell the exact same camera for a hundred thousand to the ten photographers in the world who could afford them? For every dollar of overhead, the second option is nine thousand, nine hundred and ninety times better for the vendor.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Apart from the fact that the controlling device is a Zaurus (or other PDA), did you guys notice that the storage device is a Mac mini ???
Storage device: Portable Mac Mini 1.66Hz Intel Core Duo (2 MB Cache, 2 GB RAM, Mac OS X, Windows XP)
G.
To be fair, there's less data present in a 2MP digicam shot than you might think. There aren't actually two million color pixels-- there are 1 million green pixels, and 500,000 red and blue pixels arranged in a grid. So, you've got 2MP of decent luminance data and significantly less color information.
On a nice 1600x1200 monitor, you should be able to see an increase in sharpness by viewing a scaled-down 8MP image over a 2MP image, all other things being equal. The scaled-down image will have complete color and luminance information at each pixel on the monitor, not a dodgy interpolated version of it.
And, of course, you're acting like there aren't any monitors above 1600x1200-- when Dell will happily sell you an LCD with a native 2560x1600 resolution. And let's not forget actually printing your pictures out at a size above 5x7.
It's effectively a scanner with a fast data transfer; not like this is some wonderful new development in CCDs. The price seems pretty steep when you compare it with the PhaseOne P45 back that incorporates a 39Mpixel non-scanning back, too. I'm sure it's very nice, and has its niche, but I'm not convinced it's News That Matters...
"'I pass the test,' she said. 'I will diminish, and go into the West, and remain Galadriel.'"
- JRR Tolkien.
Can we vote this up for weirdest post of the day? The guy's claiming that when he uses his 5MP camera in 1MP mode it produces images that look like images from a 1MP camera! To make matters worse, considers two images to be 'exactly' alike if one appears to have a 'snowstorm' of noise raging on it! That's the weirdest thing I've seen since this morning.
Doesn't it make you feel good to know that our freedoms are protected by politicans, lawyers and journalists.
Moore's law applied to transistors. Ok, so some cameras (those with CMOS sensors, i.e. some of Canon's) have transistors. But how does Moore's law apply to cameras specifically? It's not like we need to figure out how to make to small transistors all over again from scratch.
Heh - some more clarification: In a standard CMOS process, there is a 30% chance that a photon hitting the sensor surface will:
:)
1. Strike a light-sensitive portion of the surface
2. Penetrate deep enough to generate an electron-hole pair inside the silicon
3. Not be so deep as to prevent the capture of the electron in the pixel's depletion region
4. Generate an electron-hole pair that does not recombine before the electron is captured in the pixel's potential well.
The number (30%) is referred to as the quantum efficiency. CCD sensors have higher QE than CMOS because they have a larger percentage of the pixel that is light sensitive (higher fill factor).
Even more clarification: the quantum efficiency will be different for different wavelengths of light. Shorter wavelengths will have trouble penetrating into the silicon, and longer wavelengths will generate electron-hole pairs too deep in the silicon to be captured. Also, the colour filters applied to the sensors will absorb different wavelengths to different degrees.
Huzzah clarifications!
It may look like I'm doing nothing, but I'm actively waiting for my problems to go away.
--Scott Adams
Moore's law applied to transistors.
No, Moore's law applies to dicrete componants. Semiconductors. Transistors are one type of semiconductor, so are the discrete sensor componants of a CCD. It's like a tiny LCD screen working in reverse.
It's not like we need to figure out how to make to small transistors all over again from scratch.
No, but you do need to learn how to make 'em smaller than you did the last time, and that's what Moore's Law is all about. The rate at which we learn to do that. Make 'em smaller, get more of 'em on a chip of the same area.
"Consumer grade" cameras use sensor chips of a fixed size, so more discrete sensor componants must be crammed onto the chip to get greater resolution.
The camera the article is about doesn't do this. It doesn't have greater pixel density. It uses a bigger chip to store more pixels, producing a larger, panoramic image. Thus it has nothing to do with the editorial comment about phone cameras.
KFG
or down. or not at all?
A12A.713 is the root of ASC('evil')
> Dynamic range 1 : 2,600 (11 f-stops)"
Which is quite a bit better than most dSLRs, which are usually around 8 stops or so.
Yeah, next thing you know, they'll be cramming view cameras in there, too.
My esteemed article author, do you know what a scanning back is? It's like a flatbed scanner, only it scans the projected image from a lens one line at a time. But like a flatbed, there is a moving sensor which captures one row of pixels at a time. Only someone ignorant of what this camera is would suggest putting one in a mobile phone. Yes, my irony meter works perfectly, so does my ignoramous meter.
And this is not the largest one you can buy. That honor goes to the BetterLight SUPER 8K-HS, which has 384 Mpixels at 12,000 x 15,990.
http://www.betterlight.com/
Edith Keeler Must Die
Bah! You spoiled my explanation by explaining it properly :)
I'm not sure I agree with you here. My understanding is that QE is the chance that a photon hitting the light sensitive area will generate an electron-hole pair. The fill factor is nothing to do with it, but still important of course. As it is, 30% is widely used as "the" QE for a standard CMOS process - it wouldn't be possible to quote that without also specifying the fill factor if it depended on fill factor. A better reasoning - the unit for QE is A/W - no mention of area there. It's also true that there are camera optimised CMOS processes that offer improved QE - for instance the AMS C35B4 OPTO process which is a drop in replacement for their C35B4C3 process - change from "small" QE to "not quite as small QE" without modifying the design. Alas, I have no data on the opto process but I think that it offered a QE of 40%. CCDs work on the same principle - they are for cameras so the process is optimised to give higher QE.
The corollary of Moore's law is perhaps that the feature size will halve every however often it was that he said - so it does have some effect. What some people don't know is that as the x and y feature sizes shrink, so does the z - although not necessarily as rapidly. As you say the z dimension affects the wavelength sensitivity, so by shrinking it we'll be reducing the ability to detect longer wavelengths. I'm sure that this is one reason why processes like the 0.35um one I mention above is still in wide use in camera research whilst most people here would scoff at the very idea that anybody was using such an "archaic" technology. Going from that 0.35um process to the UMC 0.18um process and the depth of the N-well junction just about halves. Price and ooh, some headroom in analogue design are other reasons for using such a process of course.
Roger
Do you have any better hostages?
I didn't mean to assault the GIMP community in general, it is a good solution for a lot of image editing needs.
I did mean to assualt all the GIMP zealots out there that have adjusted contrast on a few photos for their website using GIMP, and then insist based on their limited experience that GIMP is all a professional photographer would ever need in an image editing program. These zealots have no idea what a professional photographer needs out of a image editing program, yet they tout GIMP as a "professional level" tool so they can get off on being smarter than me for using tools that are supposedly "as good as" mine without paying for it.
My original Parent poster is about the 10th geek I've heard this month proclaiming GIMP and a magical combination of other utilities as a professional solution and how nobody needs Photoshop. Quite frankly, spewing that kind of delusional misinformation is a great disservice to anyone trying to make their way in the world of photography. Anyone looking to develop professional level skills should spend the $100 to buy into a real professional level system like Photoshop instead of spending time learning a tool like GIMP that they will eventually discard when they inevitably buy Photoshop so they can take their work to the next level with a better set of tools.
Because Photoshop is an industry standard tool with tons of documentation, plugins, actions, etc., made by the user community, it is the tool of choice for pros. I don't mean to say Photoshop is better in a sense of "well our interface is more intuitive etc." but for goodness' sake, no image program can be used at a professional level without spending significant time learning how the program works. And there comes the crux-- 99% of the people a budding professional photographer will try to learn from will be using Photoshop. For that reason alone I don't feel anyone serious about photography, regardless of their ambitions to derive income from it, should spend time learning to edit in GIMP, unless they're a glutton for pain and want to reinvent the wheel for almost every complex image manipulation technique they'll use. As mentioned also before, there are issues like colorspace management that GIMP just doesn't handle well/at all, so even if one masters GIMP, they're still limited by their tool (GIMP) in dealing with the rest of the world. Color profiling and colorspace management is a HUGE deal in the pro world, because reprinting and tweaking your files untill they look right is damn expensive.
Some version of Photoshop comes bundled with many of the Pro and Semi-Pro digital cameras on the market, so GIMP can't compete with Photoshop on price for people that buy one of these bundles. If one somehow manages to buy a digital SLR without a free copy of Photoshop, a copy of Photoshop is pretty cheap. I net enough profit from a typical single print sale of one of my photos to pay for a copy of Photoshop elements. I'm not even that good a photographer.
The cost of Photoshop is so easily recouped over the long term, I can't fathom why anyone running a business that relies on image editing to make their product would waste so much time trying to save a few dollars by not using the best tool for the job. Even the ~$600 cost for a full version of Photoshop CS 2 pales in comparison to the amount of money a typical photographer has tied up in camera equipment, and value of the time spent trying to make GIMP do what Photoshop does better. Of all the places a photographer in business could cut costs, Photoshop is about the least worthwhile item to cheap out on.
Not so much.
First let me say that it's pretty unquestionably true that increasing clock speeds was an important part of making the CPU actually faster to some extent. No one's trying to suggest all computers should be running at 15Mhz or anything, nor trying to complain about an increase in Mhz when that was the most efficient improvement (which I agree should generally have preceded dual-core designs)
HOWEVER, to a significant extent the increase in Mhz was done at the _expense_ of improved performance. That is, development of chips that could have a higher advertised Mhz was prioritized over chips that actually performed faster (regardless of why they performed faster)
Late Pentium IIIs are faster clock-for-clock than early PIVs. The Pentium Pro was faster clock-for-clock than PIIs. The Pentium M kicked everything's butt in the Intel lineup when it came out, and Core Solo was still clock-for-clock way faster than most everything else single-core in the roadmap.
AMD tried to make their chips WORK faster and then simply made up a rating number to compete in the fake Mhz war, which was why you got models like "3200+" which basically meant "like a 3.2 Ghz Pentium, even though it's slower"
So the real thing people are complaining about is mostly that Intel, especially with a big part of the Pentium IV line, made a choice to develop for "Mhz" so they could market with it instead of developing for performance.
Looking for freelance Actionscript (Flash/Flex) or ColdFusion work and/or freelance developers. Email me, put Slashdot
I'm not sure I agree with you here. My understanding is that QE is the chance that a photon hitting the light sensitive area will generate an electron-hole pair. The fill factor is nothing to do with it, but still important of course.
:)
Bah, you spoil my explanation by pointing out the flaws in it
Okay, fine, I'll concede to you that QE which takes fill factor into account is normally referred to as effective QE. It's a way to compare different sensors without having to use two different numbers, and works because fill factor and QE can directly be multiplied together to get the behaviour of a sensor.
I think the 30% QE that you've quoted for CMOS processes does take into account fill factor. With a huge pixel that is mostly photodiode, there's no reason why the effective QE couldn't be better than a CCD sensor, which uses photogates.
It may look like I'm doing nothing, but I'm actively waiting for my problems to go away.
--Scott Adams
We're talking about dynamic range, which isn't related to the aperture range of the camera/lens combo except that both are usually measured in f-stops.
Dynamic range in this context is the range of intensities that the sensor can capture accurately in one image.
For example, say you're shooting into the sun, and have something silhouetted in front of it. If you have enough dynamic range you can pull detail from all parts of the silhouetted area, and the sky at the same time. If you have insufficient dynamic range you'll clip either the shadows/highlights or both, resulting in areas with no real image information - flat white or solid black in other words.
Generally speaking larger photosites should result in better dynamic range, as the signal to noise ratio should be better with larger photosites - all else being equal.
More on dynamic range here
Yeah, go on then. It's obviously useful being able to compare sensors...
A colleague of mine focussed a laser so all of the light power was entirely within a single CMOS photodiode and came up with values around the 30% mark, depending on wavelength and the type of diode of course. The fill factor has no affect there to be certain. That doesn't mean that other sensors have an effective QE of 30% of course. I could be misremembering his results as well.
Shame nobody else will read these comments :)
Except that CMOS probably uses more process layers, leading to more boundaries for the light to pass through and so less light at the diode.
Do you have any better hostages?