What would it take to make current cameras obsolete?

[willg]

Obviously with the introduction of faster processing times and noise levels, cameras like my 300d are not as desirable, but I can still create good pictures with it as long as I don't need ISO 1600+.

As far as the 5d goes, my only complaint right now is the build quality but I am usually thinking about that when I'm not actually shooting, just holding the camera so its probably all in my head:rolleyes:. Noise levels are great and I have very few complaints in that area especially when shooting in raw. And if somehow they got iso 6400 to look like my iso 100, then so be it, but I doubt that will happen this year.

I don't know what it would take for me to upgrade this fall...or even honestly want to/see real advantages to.

[peterdoomen]

Any camera is obsolete a few weeks after you bought it. But if it does what you need it to do, why upgrade? I'm still happy with my 20D, though the 30D has desirable features, I'm not upgrading or buying a new model since it's not the camera that hinders me in getting the pictures I want. Mostly, it's my own skill that is between fact and dream.

P.

[RLRiley]

I'm still happy with my 20D, though the 30D has desirable features, I'm not upgrading or buying a new model since it's not the camera that hinders me in getting the pictures I want. Mostly, it's my own skill that is between fact and dream.

I think thats often the case. I know it is for me and I've been shooting semi-seriously for about 15 years now. I've personally never felt the need to upgrade just to have the "latest and greatest" model. I do upgrade occasionally, but its generally only when I feel that my current collection of cameras is holding me back in some way. When my 20D can no longer achieve things that I want it to do and the problem is the camera and not the operator then I'll upgrade again, but for the forseeable future I'll be sticking with the 20D.

[willg]

I think thats often the case. I know it is for me and I've been shooting semi-seriously for about 15 years now. I've personally never felt the need to upgrade just to have the "latest and greatest" model. I do upgrade occasionally, but its generally only when I feel that my current collection of cameras is holding me back in some way. When my 20D can no longer achieve things that I want it to do and the problem is the camera and not the operator then I'll upgrade again, but for the forseeable future I'll be sticking with the 20D.


Thats exactly how I felt with the rebel...it was holding me back from doing some things (like shooting more than 4 raw frames every 30+ seconds :lol:). As for the 5d, I still have a lot of potential to pull from that machine before I am worthy of something superior

[John_Paul]

At this point, nothing... It's a splitting hairs contest among the cameras....the lenses & flash however, that's another thing all together..

JP

[René Damkot]

To me, a camera doesn't become 'obsolete' unless there is something I want it to do which it can't, and another camera can...

[DrPablo]

The current cameras will always do 99% of what the general public wants them to do.

But there is still a lot of engineering to do to get them to match the dynamic range, resolution, and price of corresponding film cameras.

I shoot large format (4x5), 35mm film, and a 300D. I get 460 megapixel images when I scan my 4x5 film. Yes, 460 -- and it would be higher if I used an 8000 dpi drum scanner. I'm about to print a 24x30 at 300 dpi, and I had to downsize to achieve that resolution. In fact I get 15-20 megapixel images from scanning 35mm. Digital cameras cannot yet match the resolution of larger film formats (larger MF and LF) -- and to approach the resolution of 35mm and 645 medium format is beyond what most people can pay (i.e. $7000). So resolution is something that still needs a lot of development (and may be difficult, because I doubt they'll ever make photosites that are the size of a film grain). This is not to say that most people need that resolution, but keep in mind that people who shoot 8x10 LF can make prints that are 16x20 feet.

Dynamic range and particularly highlight tolerance is also something lacking in digital cameras. You can overexpose black and white film by as much as 6 stops and still have highlight detail. Even slide film, with its narrow dynamic range, has better highlight tolerance than CCD and CMOS. Digital sensors may go the way of the Fuji S3 type system, with ND-filtered sensors to expand highlight tolerance.

Price is the other big area for improvement. The 300D rocked the world with its affordability; the D50 now holds that spot, but I'd imagine the leading DSLR-makers are going to keep trying to undersell one another until the price of a basic 6-8 megapixel DSLR approaches the $250 that one had to pay for a Rebel G 10 years ago.

[CyberDyneSystems]

Features and advances,. not MP

1. In addition to simply better Dynamic range from the CMOS, Auto-Iso adjustments (selectable) to individual pixels to give HUGE Dynamic range,
You'd be able to dial in how much ISO adjustment would be allowed to compensate for dynamic range, with 50-800 set, for example, bright areas would expose at ISO 50-100, and dark areas could bump ISO to whatever necessary up to 800 ISO to keep from blowing highlights. It would work with the metering system to figure out exactly what exposure is necessary to get the highlights correct, and then push only the mid range and darker areas to some ratio to maintain good exposure. Again,. you'd be able to dial in how much of this effect you'd allow for a given shot, or turn it off.

2. Low noise at higher ISO's Imagine 3200ISO looking like 100 ISO? Obviously then much higher ISO's would be usable as well. 128 or more.

3. Improved color CMOS, the obvious next step is 16 bit color native RAW.

4. GPS built into the camera or via an add in card.

5. Wireless, affordable and not clumsy

6. Built in flash units that will fire with an added flash, and trigger them wirelessly without having to buy an add on unit.

7. On Camera IS for 1.6x and 1.3X bodies (full frame would require new lenses using current tech, and EF-S would not work for the same reasons)

8. Square or octagonal CMOS sensors = reduced size of camera and no need for turning the camera to "portrait" mode, size and shape of images recorded could be selectable in camera and of course full frame images could be taken with the intention of cropping in post.

[Longwatcher]

I consider a camera (or any piece of equipment) obsolete only when it is broken and I can't get it fixed or it costs more to fix then getting a replacement. Thus my 1979 Hasselblad will pass into obsolesence sometime after Jan 07 when it breaks for the last time. All three of my Canon DSLR are still good at this time. Althought the D60 and 10D are approaching obsolesence due to cost of repair.

[willg]

The current cameras will always do 99% of what the general public wants them to do.

But there is still a lot of engineering to do to get them to match the dynamic range, resolution, and price of corresponding film cameras.

I shoot large format (4x5), 35mm film, and a 300D. I get 460 megapixel images when I scan my 4x5 film. Yes, 460 -- and it would be higher if I used an 8000 dpi drum scanner. I'm about to print a 24x30 at 300 dpi, and I had to downsize to achieve that resolution. In fact I get 15-20 megapixel images from scanning 35mm. Digital cameras cannot yet match the resolution of larger film formats (larger MF and LF) -- and to approach the resolution of 35mm and 645 medium format is beyond what most people can pay (i.e. $7000). So resolution is something that still needs a lot of development (and may be difficult, because I doubt they'll ever make photosites that are the size of a film grain). This is not to say that most people need that resolution, but keep in mind that people who shoot 8x10 LF can make prints that are 16x20 feet.

Dynamic range and particularly highlight tolerance is also something lacking in digital cameras. You can overexpose black and white film by as much as 6 stops and still have highlight detail. Even slide film, with its narrow dynamic range, has better highlight tolerance than CCD and CMOS. Digital sensors may go the way of the Fuji S3 type system, with ND-filtered sensors to expand highlight tolerance.

Price is the other big area for improvement. The 300D rocked the world with its affordability; the D50 now holds that spot, but I'd imagine the leading DSLR-makers are going to keep trying to undersell one another until the price of a basic 6-8 megapixel DSLR approaches the $250 that one had to pay for a Rebel G 10 years ago.

thats very interesting. do you think you could post a full frame resized and a 100% crop of one of your larger scans? I am interested to see how they look.

[Tim Larson]

Cyber,
Reading your post made me salivate! Great ideas. Being able to stop extreme movement and keep dynamic range would increase percentages drastically.
Full-frame, switchable to crops to increase frame rate would be sweet for me as well.

[dshootist]

Sounds like the idea of obsolescence is very subjective. A lot of the definition of obsolescence depends on the culture using the technology in question. People still use analog film and cameras to capture wonderful images right alongside digital users but some see film as obsolete. Same with records and CDs. Even though buggys haven't been the main mode of transportation in over a century, you'll still see them from time to time even if only in a very casual role. Just a thought... or three.:D

[DrPablo]

thats very interesting. do you think you could post a full frame resized and a 100% crop of one of your larger scans? I am interested to see how they look.

These crops are about 1/600th of the total film size, scanned at 4800 dpi. These scans are unsharpened.

http://www.pbase.com/drpablo74/image/64717752.jpg

http://www.pbase.com/drpablo74/image/64578027.jpg



http://www.pbase.com/drpablo74/image/63443543.jpg[/URL]

http://www.pbase.com/drpablo74/image/63443545.jpg[URL="http://www.pbase.com/drpablo74/image/63443545.jpg"] (http://www.pbase.com/drpablo74/image/63443543.jpg)

[willg]

very sharp indeed. the first resized picture has amazing detail, and the crop is impressive. so for the full size of the first one, what are the dimensions in pixels of the file. I want to upsize a sharp 5d file to that resolution and see how it compares because my 5d 100% crops are certainly sharper than the "prudential" (although obviously less magnified). It would be interesting to see how close photoshop can get

[OpticNein]

I'll upgrade to the 40D when it comes out... lol.

[farrukh]

I upsized 350D image to 34 megapixels in better quality then that PRUDENTIAL crop.
You can see that in this thread: http://photography-on-the.net/forum/showthread.php?t=105724

[René Damkot]

Cool. Maybe you could have a look at my thread over here (http://www.photography-on-the.net/forum/showthread.php?t=200008)? Some input would be nice...

[DrPablo]

To do the experiment from a 5D or any other image, you can do it two ways. You can either blow the entire image up to 400 megapixels, then crop, or you can crop 1/500 or so of the image area and use interpolation to blow it up to a web sized image.

Both images were made from 4x5 inch transparency film (Velvia 100F) and scanned at 4800 dpi. That yielded images of 461 megapixels and over 200 megabytes.

My scanning software is horrible, but I'm getting Silverfast, which supposedly will do a much better job, so I've found that all my transparency scanns are pretty unsatisfactory compared with looking at the transparencies on a light box (or even focusing on a big 4x5 inch ground glass focusing screen). The one with the Prudential Center is my best scan, but it's also my best 4x5 photo so far.

Regarding the sharpness of the Prudential crops, keep in mind that I've cropped a <1 megapixel image out of a 400+ megapixel image and blown it up to the size of the original. That's a 400x enlargement in terms of resolution, and in terms of linear dimensions it's a 20x enlargement. If I did that in real life the print would be more than 6x8 feet!

So it's a ton of resolution, but it's not meant to be examined at quite that degree of enlargement, especially at 72 dpi on a computer screen.

You'd probably get very good results with the 5D, but also remember that the 5D is a lower pixel density than the 1Ds Mk II and, for that matter, the D200. The D200 with 12 megapixels on an APS sensor has a 50% denser pixel array, which makes sense because you'd have to crop out 1/3 of the 5D's pixels just to make the sensor the same physical size of the D200. So the D200 would show more resolution than the 5D with a tiny crop by virtue of pixel density (I'm not making a case against the 5D -- it's this low pixel density that makes it so great at high ISO), but the 1Ds Mk II has a denser pixel array, so you'd have more successful tiny crops.

To get nearly this resolution you'd have to interpolate, though, both in RAW and subsequently. A point would come with extreme enlargements where interpolated detail is enlarged so much that it begins to soften the captured detail. Granted I could interpolate a 4x5 image. Someone just started a thread on largeformatphotography.info asking how to enlarge a 4x5 image to 16x20 feet.


By the way, I'm certainly not starting a film vs digital thing here. I love my 300D and I very nearly bought a 5D (but got my 4x5 instead). I'm just making the point that digital sensors, especially consumer-level ones, still have a long way to develop before the resolution of film is matched.

[DrPablo]

I upsized 350D image to 34 megapixels in better quality then that PRUDENTIAL crop.

LOL, let's compare apples to apples.

Your image is very well done, no doubt, but in terms of degree of enlargement it's not in the same league -- it's just not a valid comparison.

You've made a 4x enlargement in pixel resolution and a 2x enlargement in linear pixel content.

The Prudential crop was a nearly 400x enlargement in pixel resolution. and a 20x enlargement in linear dimensions.

Here is a re-scan (I just bought new scanning software yesterday). The letters in the sign were scanned at 9600 dpi. The letters are less than 1/1000th of the total film area.

http://www.pbase.com/drpablo74/image/64577780.jpg

http://www.pbase.com/drpablo74/image/64884197.jpg

http://www.pbase.com/drpablo74/image/64884201.jpg

[René Damkot]

For scanning software, IMO you could do worse then give Vuescan (http://www.hamrick.com/) a try...

[DrPablo]

Thanks, Rene,

I ended up upgrading the demo of Silverfast that came with the scanner. I have the Microtek i800, which comes with a Silverfast demo and some other bundled program. People speak very highly of both Vuescan and Silverfast.

The irony is I'm pretty adept at Photoshop, and I've quickly learned how to use the view camera (which is as labor-intensive as a camera can be), but the scanning is the limiting step for me. I'll be very happy once I get better at that.

Paul

[willg]

That image is just fantastic. It looks very 3d to me. Does that have something to do with the film, the larger format, the resolution, or what? The prudential crop at 460mp is way better than the 5d could manage (I don't even have to try). But, I have printed 12x18" prints from the 5d and the resolution is certainly not an issue at that level. 460mp is great in theory, but for me it just isn't necessary. It would be interesting if you downsized to 13mp and showed us a 100% crop though

[farrukh]

LOL, let's compare apples to apples.

Your image is very well done, no doubt, but in terms of degree of enlargement it's not in the same league -- it's just not a valid comparison.

You've made a 4x enlargement in pixel resolution and a 2x enlargement in linear pixel content.

The Prudential crop was a nearly 400x enlargement in pixel resolution. and a 20x enlargement in linear dimensions.

Here is a re-scan (I just bought new scanning software yesterday). The letters in the sign were scanned at 9600 dpi. The letters are less than 1/1000th of the total film area.



Ofcourse, Film is the resolution winner by a margin. But isnt that 400x enlargment overkill for the print requirments even for big billboards?
Also the colors are not so natural and not as accurate as digitals. Also i was wondering film does go through scanner's sensor to become a tiff or jpeg isnt it?

Also i was wondering if an 8mp 350d image can do 4x upscale with interpolation, then i guess 22mp 1dsMKIII would do 66mp or may be even more after interpolation with quite acceptable quality. which is is okay for most of the todays print requirment.

[DrPablo]

Thanks Will, it's actually my only really and truly successful shot I've taken so far. I'm still learning the medium!

I think the 3D effect has to do with 1) my choice of lighting (low lighting, 7:30 PM); 2) I used a very wide angle 90mm lens, comparable to 16mm on an APS-C (and wide angles accentuate depth); and 3) the sharpness of the image (LF cameras have a huge circle of confusion, meaning it's much easier to get them into razor sharp focus; plus the Schneider lenses are of truly legendary quality). I also used Velvia 100F slide film for this image, and Velvia is a very high contrast film -- it will accentuate the difference between bright and dark -- though of course the downside is that it is a pretty lousy film for high contrast settings.

The 460 MP is only useful if you're making a huge huge print, which with slide film is generally easier to do than getting a lab to make a contact negative from the slide and then an optical enlargement (i.e. making prints from slide film requires an extra step that's not necessary with negative film).

The other convenient thing about that degree of resolution, though, is that it helps you get rid of noise (from the CCD in the scanner). You have so much detail with a scan to that degree that merely downsampling will eliminate noise (because the noise is averaged with surrounding pixels as you downsize). So you have a ton of detail to spare, and can basically eliminate noise at super high resolution without losing image sharpness and detail.

As far as downsizing to 13 MP and showing a 100% crop, I'm not sure what that would reveal. First of all, 100% crop isn't a meaningful term, because it only defines the output size -- it doesn't describe how much of your image you've cropped out. A 100% crop of 50% of an image isn't the same thing as a 100% crop from 0.5% of your image, of course.

Now, I could scan it at a resolution that would produce a 13 MP image, but when you think about it all that means is I'm ignoring the vast majority of the fine detail in the image.

By analogy -- say you post a web sized image that you captured on your 5D, resized to 800x533 pixels. If I asked you to show a crop and zoom from 10% of the web re-sized image, it wouldn't make any kind of statement about the quality of your camera / sensor. Because if you take a 13 MP full frame image and resize it to 800x533 pixels, you've already discarded the vast majority of its information.

So I have an analog image, the transparency, that let's say for the sake of simplicty has 130 megapixels worth of true image data. If I scan it at 13 megapixels, then I've already deleted 90% of the image information for the experiment. So it wouldn't be a very valid way of comparing the two capture methods.

A better experiment would be to take 35mm slide film and scan it at 13 MP, because a scanned 35mm frame can probably produce a 15-20 MP image.

Or, since the ultimate question isn't how it looks at 72 dpi, how do the images look in print? It probably wouldn't be until the extremes of enlargement that one could differentiate the two, simply because the resolution of the print wouldn't be able to show the minute detail in the larger capture medium at small print size. In other words, you'd be hard pressed to see the difference between a 350D, a 5D, and a 1Ds Mk II if you're printing a 4x6. But if you're printing a 40x50 you certainly could.

[willg]

I guess my intention with the 100% crop is to show a more direct comparison with an image from the 5d. My 100% crops are sharp as far as 100% crops go, but it would be interesting to see how sharp the 460 to 13mp image is from corner to corner at full size. Does that make any sense? So instead of me upsizing the 5d, you downsize the 460mp file because most people here find a 13mp file to be more practical and every bit as useful as a 460mb file.

[DrPablo]

Ofcourse, Film is the resolution winner by a margin. But isnt that 400x enlargment overkill for the print requirments even for big billboards?

Be careful not to confuse area with linear dimensions. It's a 400x enlargement in terms of resolution, because resolution is a function of area -- pixels per square inch.

But the actual enlargement in terms of dimensions is the square root of that, i.e. 20x. So if I turned a 4x5 negative into an 80x100 inch print (not quite a billboard, but a large poster or mural), that would be a 20x enlargement linearly, but it would be a 400x increase in surface area.

Most people feel that you can enlarge by a factor of 10-12 without any loss of image quality. So 4x5 film can go to 40x50" or perhaps 48x60 (this is 4x5 feet, of course) without loss in quality. Full frame film, i.e. 35mm, can go to like 16x20 without loss in quality, but you're much more likely to see grain with 35mm film.


Also the colors are not so natural and not as accurate as digitals.

The issue of color accuracy is so complex just within the separate digital and film arenas that there's no way to make a general statement like that.

The accuracy of digital color capture is limited by Bayer interpolation, bit depth, and ICC color space. And this is assuming that you have a calibrated output device like a monitor or printer, and you haven't modified the white balance or saturation. By the time the light that enters your camera is translated to something visible it has been digested and processed many times.

Digital sensors just by their construction have an error-prone way of encoding color, and the only exception to this (at the consumer level) is the Fuji S3. With CCD and CMOS sensors you have separate red, green, and blue pixels. There are more green pixels than either of the others. Using Bayer interpolation the camera's processer (or RAW converter) derive colors based on the individual R, G, and B pixels. This is what our eyes do too, but our retinas have a lot more rods and cones than a digital camera has photosites. The Bayer interpolation process creates lots of color noise (just examine a RAW image in ACR with color noise reduction turned all the way down to 0).

Film is different because it produces R, G, and B images that are overlain on one another. There is no mathematical derivation of the color. It's just an optical combination of three overlain images.

That said, films have different color balances and different saturations. Velvia is a highly saturated film, Portra160 NC is a very unsaturated film.

Velvia, the film I used for the Prudential Center shot, is notoriously saturated, and is the favorite film of people who shoot landscapes. It produces vibrant colors, especially with that magic warm lighting at sunrise and sunset.

It's awful for portraits, though, because you don't want oversaturated skin / lips / eyes, etc. The less saturated films like NPS and 160NC are much better for portraiture.

Color accuracy is somewhat subject-specific, both in film and digital. It's analagous to dynamic range. B+W negative film can capture detail over 10-12 stops of dynamic range. But there's no print or monitor that can display 10-12 stops. What this means is that we're not necessarily striving for complete accuracy -- we're just trying to present a credible representation of what was in front of the camera.

So color film selection is very much analagous to digital color manipulation. Thankfully, with scanners, we can digitally manipulate color film images too.

As far as my picture goes, it to my eye looks completely like the scene did in real life (the Prudential one -- the other one with the clocktower was underexposed). But my other scans aren't so successful -- hopefully I'll find a way to calibrate my scanner.


Also i was wondering film does go through scanner's sensor to become a tiff or jpeg isnt it?

Yes, I scan it as a 48-bit HDR image and it's exported to Photoshop as a TIFF. Once the image is digitized it has the same properties as any other of the same bit-depth, resolution, etc.

Also i was wondering if an 8mp 350d image can do 4x upscale with interpolation, then i guess 22mp 1dsMKIII would do 66mp or may be even more after interpolation with quite acceptable quality. which is is okay for most of the todays print requirment.

Sure, I absolutely agree. I was never making an argument that DSLR sensors couldn't meet virtually everyone's needs. I have gorgeous enlargements from my 6.3 megapixel 300D, one of which I've sold. The biggest problem with a 6 or even 8 MP sensor is that the lenses greatly outresolve the sensor. In other words, if you have good lenses (say a nearly perfect lens like the 135 f/2L), your sensor just can't capture all the detail that lens can transmit. It also means that flawed lenses, slightly missed focus, noise reduction, etc, all reduce detail, and there's just only so much detail to spare. This is all academic 99.9% of the time; but it's real, and hopefully newer technologies will bring even better performance to the DSLRs we have.

[DrPablo]

So instead of me upsizing the 5d, you downsize the 460mp file because most people here find a 13mp file to be more practical and every bit as useful as a 460mb file.

I hear what you're asking, but again it doesn't make much sense. The original image is a transparency with a certain fixed amount of information in it. It doesn't make much sense to scan it to an end resolution of 13 MP to show a 100% crop, because it discards nearly all the information in the image.

So what your experiment asks is 'how much information is left in your image after you throw more than 90% of it out?'

I mean, if I scan my negative at 10 dpi the resulting image will be outresolved by my camera phone, but it doesn't make any kind of statement about the resolving power of the film or camera.

At any rate, here is I think what you're asking for. The first is the original image. I then scanned it to a resolution of 3200x4000, which happens to be 12.8 megapixels. The second image is a 100% crop from that scan.

http://www.pbase.com/drpablo74/image/64951465.jpg

http://www.pbase.com/drpablo74/image/64951581.jpg

[DrPablo]

And just because it's so fun, here's a 100% crop from a 9600 dpi scan (I think 9600 dpi is beyond the optical capabilities of a flatbed scanner, but at any rate that's as high as I can set it). I didn't bother to reduce noise at all in this one.

http://www.pbase.com/drpablo74/image/64952528.jpg

[willg]

Interesting post DrPablo, and another great shot. I understand what you are saying. Yes it does sound like a waste to scan it at a low resolution.

Would there be any difference if you scanned it at the highest resolution and then used software to resize to a lower resolution? Does that make sense? That might eliminate the quality of the scan as the limiting factor, but I am not really sure if it would be any different.

[DrPablo]

That's a great question, and in fact it's routinely done.

If your detail completely overwhelms your noise, then your noise will be absorbed into the surrounding detail as you downsize.

So most people will scan at a high resolution and then downsample (just using bicubic in PS or whatever).

I scan at usually 2x the resolution I need, run a 1-pixel median filter, and fade it to 25%. This gets rid of a lot of the noise, and in a high res scan there's no important detail that can't lose a pixel. I do all my noise reduction at high res, then downsample.

[Wilt]

Price is the other big area for improvement. The 300D rocked the world with its affordability; the D50 now holds that spot, but I'd imagine the leading DSLR-makers are going to keep trying to undersell one another until the price of a basic 6-8 megapixel DSLR approaches the $250 that one had to pay for a Rebel G 10 years ago.

I wonder why everyone fails to see the high cost of dSLR comparatively and is willing to accept that, without question? They all dismiss the cost differential with the commment 'but I don't have to pay for film and processing!'. Why not expect a much cheaper dSLR than we pay today *and also* no payment for film and processing?!?!?!

[Wilt]

I consider a camera (or any piece of equipment) obsolete only when it is broken and I can't get it fixed or it costs more to fix then getting a replacement. Thus my 1979 Hasselblad will pass into obsolesence sometime after Jan 07 when it breaks for the last time. All three of my Canon DSLR are still good at this time. Althought the D60 and 10D are approaching obsolesence due to cost of repair.

I also think of the fact that I have an Olympus OM-1 camera which I still can use today, 30 years after I bought it...then I think about the dSLR which I have today and wonder if anyone will be making BP-511a form factor substitute batteries, so I can still shoot with it 30 years from now!!!

[DrPablo]

I agree 100%, Wilt.

Honestly, I was extremely close to buying a 5D, largely so that my TS-E 24 would behave like a true wide angle tilt-shift lens. I used my TS-E on my Rebel G, and just loved its performace compared with on APS-C. But I was just afraid of paying $2800 or whatever on something that, for all its wonderful merits, is going to be upgraded every 2 years for the next decade.

And then someone I know from another forum offered me about $8000 of large format gear for $2000. So for less money I got something that really fits the bill for architecture and will last my lifetime.

This isn't at all to knock the 5D -- I'm sure I'll buy it or its replacement at some point when I've saved up enough and my 300D finally kicks it. Though I sure would love that Mamiya 7ii RF first :)

[Lightstream]

That's a great question, and in fact it's routinely done.

If your detail completely overwhelms your noise, then your noise will be absorbed into the surrounding detail as you downsize.

So most people will scan at a high resolution and then downsample (just using bicubic in PS or whatever).

I scan at usually 2x the resolution I need, run a 1-pixel median filter, and fade it to 25%. This gets rid of a lot of the noise, and in a high res scan there's no important detail that can't lose a pixel. I do all my noise reduction at high res, then downsample.

Great explanation :)

I used to do this (without knowing why) in the P&S days. If I shot at ISO 400 on a little P&S, which produced horrifying amounts of noise, I'd just resize the image to 50% or less since I was posting on the web, and the problems would "magically" go away.

So there is a good reason for more MP, even if you are throwing them away to some extent, their passing can still produce positive effects.

[DrPablo]

So there is a good reason for more MP, even if you are throwing them away to some extent, their passing can still produce positive effects.

Sure, as long as the closely packed photosites don't disproportionately generate noise. That's the risk, of course, with making a 22 or whatever MP sensor with current CMOS technology, (and that's why the 5D has better noise control than the 1Ds Mk II, lower pixel density). I'm sure you've noticed that if you take an ISO 1600 shot that is very noisy, the noise will in itself degrade the detail as you downsample, simply because it's recognized by the sampling engine and incorporated into the image.

But in my mind the main advantage of physically large sensors is that you're projecting your image onto a larger image circle. So the same fine details appear larger on the sensor and are picked up by more photosites. That way the image detail will greatly outweigh the interposed noise.

[MinisterStanley]

Any camera is obsolete a few weeks after you bought it. But if it does what you need it to do, why upgrade? I'm still happy with my 20D, though the 30D has desirable features, I'm not upgrading or buying a new model since it's not the camera that hinders me in getting the pictures I want. Mostly, it's my own skill that is between fact and dream.

P.

I agree. I looked at the 30D, and even contemplated upgrading to the 5D, but I realized that the limiting factor in my photographed (even after purchasing good glass) was my own lack of skill. I decided to stop tossing money in my photography and begin tossing more effort into learning the art (even though I have to admit to being a serious gearhead).

[Wilt]

I agree. I looked at the 30D, and even contemplated upgrading to the 5D, but I realized that the limiting factor in my photographed (even after purchasing good glass) was my own lack of skill. I decided to stop tossing money in my photography and begin tossing more effort into learning the art (even though I have to admit to being a serious gearhead).

Gee, one can only hope that all the wanna-be's slow down in throwing their money at L lenses and improve their shooting skills before demanding that the photos be razor sharp but horribly composed and lighted!!!

[DrPablo]

It's a great point, Wilt, and it's one that everyone overlooks.

If someone really and truly is learning this art, they will be much more limited by their skill than by the inherent weaknesses of their equipment. Just look at that 'post your best kit lens shot' thread. As maligned as the kit lens is, I'd rather see a slightly soft shot that is composed and executed with real vision and talent than see one with that 'L-quality' that is basically a snapshot.

This point completely drowns out 1) the oft debated differences between say the D70s and 350D, and 2) the changes introduced with each crop of camera. The lack of megapixels, Digic III, 45-point autofocus, ETTl-2, mag-alloy body, etc, was not nearly so limiting to me as my technique and my opportunities.

In fact when people over on Phototakers, where I spend a lot of time, ask what to spend their $2000 on, I usually suggest a week in Paris or the Grand Canyon or Denali -- go somewhere where you want to shoot pictures. Spend your money on that.

[Wilt]

This point completely drowns out 1) the oft debated differences between say the D70s and 350D, and 2) the changes introduced with each crop of camera. The lack of megapixels, Digic III, 45-point autofocus, ETTl-2, mag-alloy body, etc, was not nearly so limiting to me as my technique and my opportunities..

Within the past year, someone asked about 'what dSLR to buy' after attending a photo critique club and seeing the spectacular shots submitted by 'everyone with dSLR' vs. his point and shoot. I challenged him to figure out which shots were from my P&S and which shots from my dSLR, posting 4-6 shots. He guessed WRONG for EVERY shot, proving my point that an excellent photo can be made with any camera! Decades of photographers made award winning photos before L lenses were even a gleam in Papa Canon's eye!