Poll

Which of these features would add to your creative use of digital photography?

My current camera meets my needs for the foreseeable future
- 14 (17.9%)
A camera exists that does everything I want, but I don't have it.
- 5 (6.4%)
Give me more resolution/detail (above 24 mp) - I really do print that big!
- 3 (3.8%)
I'm not satisfied with ~10 stops of dynamic range, or with present tonal scales
- 20 (25.6%)
The present top-end (D3x, etc...) image quality is fine, but I want it in a smaller/lighter/cheaper camera
- 10 (12.8%)
Get rid of Bayer sensors already - I've got plenty of detail, but want better color
- 11 (14.1%)
I want a still/video hybrid camera, and HD video is fine (improved version of 5D mkII video capability)
- 2 (2.6%)
Please, Mr. Jannard (RED), can I buy a Scarlet - I want to shoot video at 24 mp along with stills
- 3 (3.8%)
My most important feature is something exotic (tilt sensor or the like)
- 4 (5.1%)
I'm waiting for something that exists to become smaller/lighter/cheaper (D700x, EOS-3D)
- 6 (7.7%)
I feel the need for speed - the slow frame rates of today's highest image quality cameras aren't fast enough.
- 0 (0%)

Total Members Voted: 43


Pages: [1]   Go Down

Author Topic: Where to from here (what improvements in image quality do we need)?  (Read 5771 times)

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044

We're getting close to a number of limits, both on possible DSLR image quality and on what we actually need. First of all, how big do you intend to print? This is different for each photographer, but anything above 16x24 inches is probably uncommon with prints above 24x36 being very rare. Most modern high-pixel count DSLRs will produce a gallery-quality 16x24 inch print, and there's at least one, and maybe more, that prints 24x36 inch prints with superb detail (stands up to inspection from 8 inches away). I shoot the D3x, and can say from experience that it prints 24x36 inches very well. The Alpha 900 or one of the modern Canons might as well, but I've never tried it. Any more detail extraction (a combination of resolution, AA filtration and lens quality) than the D3x is capable of will only be visible on VERY large prints (probably beyond what a 24 inch printer can do). 44 inch printers are a major commitment, in terms of price, space for the equipment and display space for the prints. A 44 inch printer is roughly the size and weight of an upright piano, and the images it produces cannot be matted on standard boards or displayed on most walls (framed, a 40x60 inch print is close to 6 feet wide). Even a 24 inch printer sits on its own stand, and is an ugly piece of furniture! I've just figured out where one is going to go in my workroom. For what I do (landscapes and macros), a DSLR with detail sufficient for 24x36 inch printing, plus wide dynamic range and a beautiful tonal scale is what I need, and I bought it a few weeks ago. I was only willing to spend the money because I feel that it is a 5-10 year camera (I can't see what I'd really use that it can't do).
     The newest DSLRs are also giving us detail over a very wide dynamic range - wider, in fact than film ever did. The Zone System accounts for 11 stops of range (Zones 0 through X), but only 9 of them hold detail (X is paper white and 0 is maximum black without detail). Some modern DSLRs are holding good detail over 10 stops, maybe approaching 11 with careful post-processing (we've added 1 to 2 stops beyond where the Zone System was meant to go). There's nothing that says that a photographer couldn't USE, say 14 stops and bring them into the smaller range of the printer in post-processing , but it's a tool we've never had before. The quality of those tones is rapidly increasing - the banding and harsh transitions that used to scream "digital" are pretty much gone.
    As we reach these high image qualities (the D3x is easily equivalent to a 6x9 cm film camera, better in some ways, and the higher resolution medium-format backs equal 4x5 film), the laws of physics are beginning to rear their ugly heads and say "Thus far shall you come, but no farther". Between diffraction, lens resolution limits and unavoidable quantum noise, the maximum EFFECTIVE resolution of a FF35 sensor is somewhere between 25 and 40 megapixels. You can certainly build a higher resolution sensor (the pixel density of an average digital compact would be around 300 MP if spread over a FF35 sensor), but you won't get a commensurate improvement in image quality. Even at the low end of the 25 to 40 mp range, lenses are affected by diffraction above about f8 (so absolute maximum image quality means staying between f4 and f8). By the upper end of that range, the effect of diffraction appears by f5.6, so a 40 mp DSLR would need superb lenses used between f4 and f5.6 to extract maximum detail.
         Quite a few cameras are in the lower end of this range (not only the 20+MP FF35 DSLRs, but 10-12 MP APS DSLRs and 50ish MP MF backs), and the best of them are capable of very good quality per pixel with careful handling. Leaving aside digital compacts, there have only been a few penetrations into the upper reaches of this range of pixel densities (all are APS DSLRs, and the most notable is the Canon EOS-50D (which would be 38 mp if full-frame)). The 50D has received mixed reviews, but one review it has NOT gotten is "the full effect of the extra resolution above the 40D is noticeable". Some reviewers say that the overall image quality of the 50D is actually not as good as the older 40D, while others say that it has a noticeable positive difference, but less than one would expect from the raw numbers. The 50D is actually easier to get good quality out of than a full-frame DSLR of the same pixel density would be, because it doesn't need lenses with that level of resolving power over as large an area. A full-frame 38 mp DSLR would display its maximum image quality in vanishingly few situations!
    There are several possible directions to go, given that "more resolution" will soon be ineffective. One possibility is non-Bayer sensors, which will affect color performance more than anything else. Another is further exploration of still-video hybrid cameras, where still performance is more or less what we have now, but video becomes an increasingly important part of the package. RED is beginning to take this route. A third option is smaller, lighter or cheaper cameras with the same image quality as the big models (although perhaps compromised in some other area of performance). There are other, more exotic possibilities, like a tiltable sensor (could add creative control over depth of field, but could only be focused by using Live View). Of course, the speed and performance of existing cameras could be increased - no law of physics prevents making a camera with the resolution of a D3x but the speed of a D3!
     Which of these would you like to see? Does the camera you'd like to use for the next 5-10 years already exist? Do you own it, and if not, why not (too expensive? too heavy? wrong lens mount?)? If it doesn't exist, what could your favorite camera manufacturer give you that would cause you to use digital cameras like we all used to use film cameras - buy one to use until it breaks or wears out.

                                                                  -Dan
« Last Edit: January 11, 2009, 10:36:06 pm by Dan Wells »
Logged

brianc1959

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 53
    • http://
Where to from here (what improvements in image quality do we need)?
« Reply #1 on: January 11, 2009, 11:57:25 pm »

I think that there will be at least one more resolution doubling in the 24x36mm format to ~50 megapixels.  Its true there are very few lenses that will be completely satisfactory all the way to the corners at 50 mp, but this will improve in due course.  I also think the days of the DSLR are numbered, and that short BFL EVF cameras will become dominant.  One reason is that elimiinating the BFL constraint will allow for significant lens improvements.  Its entirely feasible to build ~f/2 prime lenses that will outresolve a 50MP sensor at f/2 over the whole image area.

Likewise, I think the resolution of MFDBs will double at least once more to ~100 megapixels, which will allow for 8x10 film quality in a compact digital package.

In practice, going beyond a resolution doubling would be very desirable, as that will help eliminate the AA filter once and for all.
Logged

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044
Where to from here (what improvements in image quality do we need)?
« Reply #2 on: January 12, 2009, 12:42:52 am »

Quote from: brianc1959
I think that there will be at least one more resolution doubling in the 24x36mm format to ~50 megapixels.  Its true there are very few lenses that will be completely satisfactory all the way to the corners at 50 mp, but this will improve in due course.  I also think the days of the DSLR are numbered, and that short BFL EVF cameras will become dominant.  One reason is that elimiinating the BFL constraint will allow for significant lens improvements.  Its entirely feasible to build ~f/2 prime lenses that will outresolve a 50MP sensor at f/2 over the whole image area.

Likewise, I think the resolution of MFDBs will double at least once more to ~100 megapixels, which will allow for 8x10 film quality in a compact digital package.

In practice, going beyond a resolution doubling would be very desirable, as that will help eliminate the AA filter once and for all.

What about diffraction? The two cameras you postulate are diffraction limited at any aperture smaller than f4. This doesn't mean that they are worse than lower resolution cameras - just that they aren't better from a resolution perspective. Your point about the AA filter is interesting, and could be an advantage of a 50 mp camera diffraction limited to 25 over a 25 mp camera. I also like your EVF idea (if the EVF is good enough) both because of the lens design advantages you mention (especially in medium format, where backfocus to clear the mirror is huge) and because of increased versatility - you can overlay anything you want, from a histogram to a level to a grid, anywhere you want it on an EVF, and modular viewfinders become easy.

                                                             -Dan


Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Where to from here (what improvements in image quality do we need)?
« Reply #3 on: January 12, 2009, 12:59:18 am »

Dan,

A lot of good points. I made a couple of comparison prints in A2  between my Alpha 700 which is 12.5 MPixel APS-C and my Alpha 900 with 24.4 MPixels and full frame. I could see very little difference in prints but a lot of difference on screen.

http://luminous-landscape.com/forum/index....st&p=250746

On the other hand experienced MFDB owners clearly state that MFDBs like P45s give better prints than the A900, I cannot explain and I haven't checked the math but I have no reason to deny that MFDBs print better than 24 MPix DSLRs.

Pixels are probably not the only thing that matters. At this stage I guess that most pictures we take are not limited by resolution but:

- Focusing errors
- Camera shake
- Depth of field
- Diffraction
- Camera lenses

If we don't shoot test charts under lab conditions, that is.

On the other hand, we may have reached "diminishing returns". Improvements are still possible but their effect may be small, some improvements will only be visible on very large prints.

A couple of areas where improvements may be possible:

1) More pixels used to increase dynamic range like in Fuji C5.
2) Quadrupling resolution it would be possible to avoid either interpolation or the antialiasing filter. This would give us the same number of effective pixels than today but with marginally less artifacts.
3) Modifications to CFA (Color Filter Array) possible include luminosity pixels.

Best regards
Erik







Quote from: Dan Wells
We're getting close to a number of limits, both on possible DSLR image quality and on what we actually need. First of all, how big do you intend to print? This is different for each photographer, but anything above 16x24 inches is probably uncommon with prints above 24x36 being very rare. Most modern high-pixel count DSLRs will produce a gallery-quality 16x24 inch print, and there's at least one, and maybe more, that prints 24x36 inch prints with superb detail (stands up to inspection from 8 inches away). I shoot the D3x, and can say from experience that it prints 24x36 inches very well. The Alpha 900 or one of the modern Canons might as well, but I've never tried it. Any more detail extraction (a combination of resolution, AA filtration and lens quality) than the D3x is capable of will only be visible on VERY large prints (probably beyond what a 24 inch printer can do). 44 inch printers are a major commitment, in terms of price, space for the equipment and display space for the prints. A 44 inch printer is roughly the size and weight of an upright piano, and the images it produces cannot be matted on standard boards or displayed on most walls (framed, a 40x60 inch print is close to 6 feet wide). Even a 24 inch printer sits on its own stand, and is an ugly piece of furniture! I've just figured out where one is going to go in my workroom. For what I do (landscapes and macros), a DSLR with detail sufficient for 24x36 inch printing, plus wide dynamic range and a beautiful tonal scale is what I need, and I bought it a few weeks ago. I was only willing to spend the money because I feel that it is a 5-10 year camera (I can't see what I'd really use that it can't do).
     The newest DSLRs are also giving us detail over a very wide dynamic range - wider, in fact than film ever did. The Zone System accounts for 11 stops of range (Zones 0 through X), but only 9 of them hold detail (X is paper white and 0 is maximum black without detail). Some modern DSLRs are holding good detail over 10 stops, maybe approaching 11 with careful post-processing (we've added 1 to 2 stops beyond where the Zone System was meant to go). There's nothing that says that a photographer couldn't USE, say 14 stops and bring them into the smaller range of the printer in post-processing , but it's a tool we've never had before. The quality of those tones is rapidly increasing - the banding and harsh transitions that used to scream "digital" are pretty much gone.
    As we reach these high image qualities (the D3x is easily equivalent to a 6x9 cm film camera, better in some ways, and the higher resolution medium-format backs equal 4x5 film), the laws of physics are beginning to rear their ugly heads and say "Thus far shall you come, but no farther". Between diffraction, lens resolution limits and unavoidable quantum noise, the maximum EFFECTIVE resolution of a FF35 sensor is somewhere between 25 and 40 megapixels. You can certainly build a higher resolution sensor (the pixel density of an average digital compact would be around 300 MP if spread over a FF35 sensor), but you won't get a commensurate improvement in image quality. Even at the low end of the 25 to 40 mp range, lenses are affected by diffraction above about f8 (so absolute maximum image quality means staying between f4 and f8). By the upper end of that range, the effect of diffraction appears by f5.6, so a 40 mp DSLR would need superb lenses used between f4 and f5.6 to extract maximum detail.
         Quite a few cameras are in the lower end of this range (not only the 20+MP FF35 DSLRs, but 10-12 MP APS DSLRs and 50ish MP MF backs), and the best of them are capable of very good quality per pixel with careful handling. Leaving aside digital compacts, there have only been a few penetrations into the upper reaches of this range of pixel densities (all are APS DSLRs, and the most notable is the Canon EOS-50D (which would be 38 mp if full-frame)). The 50D has received mixed reviews, but one review it has NOT gotten is "the full effect of the extra resolution above the 40D is noticeable". Some reviewers say that the overall image quality of the 50D is actually not as good as the older 40D, while others say that it has a noticeable positive difference, but less than one would expect from the raw numbers. The 50D is actually easier to get good quality out of than a full-frame DSLR of the same pixel density would be, because it doesn't need lenses with that level of resolving power over as large an area. A full-frame 38 mp DSLR would display its maximum image quality in vanishingly few situations!
    There are several possible directions to go, given that "more resolution" will soon be ineffective. One possibility is non-Bayer sensors, which will affect color performance more than anything else. Another is further exploration of still-video hybrid cameras, where still performance is more or less what we have now, but video becomes an increasingly important part of the package. RED is beginning to take this route. A third option is smaller, lighter or cheaper cameras with the same image quality as the big models (although perhaps compromised in some other area of performance). There are other, more exotic possibilities, like a tiltable sensor (could add creative control over depth of field, but could only be focused by using Live View). Of course, the speed and performance of existing cameras could be increased - no law of physics prevents making a camera with the resolution of a D3x but the speed of a D3!
     Which of these would you like to see? Does the camera you'd like to use for the next 5-10 years already exist? Do you own it, and if not, why not (too expensive? too heavy? wrong lens mount?)? If it doesn't exist, what could your favorite camera manufacturer give you that would cause you to use digital cameras like we all used to use film cameras - buy one to use until it breaks or wears out.

                                                                  -Dan
Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Where to from here (what improvements in image quality do we need)?
« Reply #4 on: January 12, 2009, 01:17:14 am »

Hi,

I'd suggest that many of the improvements are possible but may be more or less marginal. Removing the AA filter may give a 15-30% increase in effective resolution and less artifacts. Anybody seen artifacts recently? Most issues with aliasing should be in the studio/catalogue area and MFDBs without AA filter are very often used in those settings, yes, they have problems with moiré.

Although I like EVFs I'd suggest that they will not work very well in darkness. Another advantage with SLRs is that the sensor is exposed just under exposure. With EVFs the sensor needs some protection. Pointing the camera into the sun could easily destroy the sensor on an EVF camera.

In my view EVF is the natural evolution with digital but it may be that the technology is not here yet.

Best regards
Erik

Quote from: Dan Wells
What about diffraction? The two cameras you postulate are diffraction limited at any aperture smaller than f4. This doesn't mean that they are worse than lower resolution cameras - just that they aren't better from a resolution perspective. Your point about the AA filter is interesting, and could be an advantage of a 50 mp camera diffraction limited to 25 over a 25 mp camera. I also like your EVF idea (if the EVF is good enough) both because of the lens design advantages you mention (especially in medium format, where backfocus to clear the mirror is huge) and because of increased versatility - you can overlay anything you want, from a histogram to a level to a grid, anywhere you want it on an EVF, and modular viewfinders become easy.

                                                             -Dan
Logged
Erik Kaffehr
 

dwdallam

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2044
    • http://www.dwdallam.com
Where to from here (what improvements in image quality do we need)?
« Reply #5 on: January 12, 2009, 02:47:33 am »

One thing that was missed here and it always seems to get missed is that you are not always printing with the full resolution, unless you nail your composition perfectly each time. More resolution does not create a better 20x30 print after 24MP, or 21, or even 13 for that matter, with proper post processing, but the reason we want more resolution is when we need to crop. Even mild cropping EATS resolution. Give me more until the physics of more becomes impossible. Headroom is always nice to have. Always.
Logged

wollom

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 61
Where to from here (what improvements in image quality do we need)?
« Reply #6 on: January 14, 2009, 12:03:03 am »

Quote from: Dan Wells
Of course, the speed and performance of existing cameras could be increased - no law of physics prevents making a camera with the resolution of a D3x but the speed of a D3!

Maybe adding increased buffer depth to the survey would be useful.

Wollom
Logged

danlo

  • Newbie
  • *
  • Offline Offline
  • Posts: 15
Where to from here (what improvements in image quality do we need)?
« Reply #7 on: January 14, 2009, 02:24:36 am »

I want better colour rendition.. .. In D3s Nef-files it´s not good at all, and that says alot about the rest of the cameras in the dslr-market.
Logged

Geoff Wittig

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1023
Where to from here (what improvements in image quality do we need)?
« Reply #8 on: January 14, 2009, 10:41:00 am »

1) A dedicated B&W camera without the Bayer matrix or AA filter. In principle this should provide a huge increase in real resolution and image quality for those of us who love monochrome.

2) Current higher end D-SLR's already provide resolution that is really pushing the limits of current lenses. I'd rather see Canon spend the money to develop a really good circa 20 mm lens that is truly sharp right to the corners, before they cram more pixels we can't effectively use into current sensor sizes.

3) Further improvements in high ISO performance; the potential for revolutionary new photographic work in very low light levels is very exciting.

4) In-camera HDR technology to greatly expand dynamic range in an automated way. Finally, back-lit landscapes that look great without hours in front of the computer using clumsy software tools.
Logged

marc

  • Newbie
  • *
  • Offline Offline
  • Posts: 16
Where to from here (what improvements in image quality do we need)?
« Reply #9 on: January 14, 2009, 02:24:13 pm »

Image stabilization in camera. I know some DSLRs have it, but not the ones I want.. I don't understand why Canon and Nikon don't offer this. Even if it works better in the lens it could just be turned off in camera when present in the lens.

What's even more confounding is that they don't even bring out lenses in the wide to mid focal range with IS, save for a very few (and they typically extend into the telephoto end). Where are the non-tele primes with IS? Where are the wide zooms with IS? This technology is standard in compacts these days, yet the two biggest players in the DSLR market are treating it like a luxury item only needed for specialty telephoto lenses.
« Last Edit: January 14, 2009, 02:25:33 pm by marc »
Logged

Plekto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 551
Where to from here (what improvements in image quality do we need)?
« Reply #10 on: January 14, 2009, 09:18:30 pm »

For me, it's the Bayer pattern.

See, we can go to 50 MP(technically sub-pixels) or whatever, but all that does is give us more data/larger files, and tons of loss when the printer has to compress it down to something printable.

What we need is a 1:1 pixel(full color range) to location ratio.  This would result in smaller sizes, easier manipulation, and less overhead and processing time/memory/etc.  6x7 done this way would be a meager 35MP.(equal to 2400dpi scanned film)  Foveon looked promising, but it's been several years now and they've basically dropped the ball.

But Bayer patterns are still in need of retirement.  They cause tons of problems that aren't really solvable by just upping the sensor resolution.
Logged

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044
Where to from here (what improvements in image quality do we need)?
« Reply #11 on: January 15, 2009, 02:18:27 pm »

Very interesting results so far... The largest group of us wants more dynamic range. Have those who voted for more DR tried the D3x? People didn't quite believe me when I posted, back in December, that I was subjectively seeing almost two extra stops in my landscapes above the older Canon EOS 1Ds mkII. Not only do I still think those stops are there, hundreds of images later, but DxOmark just found them... The D3x scored 2 stops higher on DxOmark's dynamic range test than the older Canon, and a full stop more than ANY (non-Fuji) DSLR. The reason I mention this is NOT as a "Nikon is better" post, but to say that the D3x's dynamic range is a preview of what other manufacturers (and Nikon in other bodies) will soon be bringing us. There's an extra stop in the pipeline, which Nikon has now (and I'm shooting with all the time, and have come to depend on), and everybody else will have soon. Would the 11 people who wanted more DR be satisfied with the extra stop (more than that, if your camera is anything except a D3, a D700, an Alpha 900, a 1Ds mkIII, or a 5D mkII - the non D3x DR champs) they'll soon have, or do you want even more than that?
      The next largest group are satisfied with their cameras - the present cameras really are getting good! I spent $8000 on a camera that I hope to shoot for 5 years or more (otherwise, WHY spend that on a camera)!, and if we add them to the group who know what they want, and it exists, but they don't yet own it, they actually outnumber the DR desirers.  These folks (close to half of us) are satisfied with today's DSLR choices (if not necessarily with their prices). The "satisfied plus price sensitive" group actually exceeds half of the sample, and approaches 2/3, if we add in the "easy requests" like a D700x - an existing camera in a different body style.
     The biggest group that want real change are looking forward to Foveon or other non-Bayer image sensors. This may merge with the DR request if Fuji can ever get the pixel density of the Super CCD (which is sort of a non-Bayer sensor) up. If we add this group to the DR group, close to half of the sample want a fundamentally different image sensor.
    Nothing else got more than a few votes - a couple people want resolution, a few want video (about equally divided between HD mode only (Canon) and high-pixel RAW video (RED)), and a few want some exotic sensor feature
     This is quite different from the manufacturers' priorities. The photographers who answered my unscientific survey want our D3x-level image quality in a range of bodies, lens mounts and price points, then we want better (but not higher resolution) sensors. I suspect that some percentage of the DR group would be satisfied with D3x DR, if they had a chance to try one in their preferred body style and lens mount, and that many of the others in this group will be most easily satisfied by an unconventional sensor.
     The manufacturers, on the other hand, seem to be ignoring what photographers (at least those active here) want - we haven't seen any development on unusual sensors in quite a while. One of the "most wanted" smaller/lighter/cheaper cameras is the Canon 3D (a non EOS1 size body with seals and pro AF), which has been on that list for many years - Canon doesn't seem in any hurry to release it (hint to any Canon execs reading this - it will sell) . We'll see how Nikon does with the D700x, which would probably take the other top spot on the most wanted list...
    A lot of development seems to go into more resolution (50D), video modes and more speed - all low ranking in this poll (more speed didn't even get a mention)


                                                                            -Dan
Logged

AJSJones

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 357
Where to from here (what improvements in image quality do we need)?
« Reply #12 on: January 15, 2009, 08:22:19 pm »

Quote from: ErikKaffehr
Although I like EVFs I'd suggest that they will not work very well in darkness. Another advantage with SLRs is that the sensor is exposed just under exposure. With EVFs the sensor needs some protection. Pointing the camera into the sun could easily destroy the sensor on an EVF camera.

In my view EVF is the natural evolution with digital but it may be that the technology is not here yet.

Best regards
Erik
It may be close than you think!  
If you consider the view through an optical VF on a FF looks like a ~ 3x5 at ~12" and all it needs is 300-400ppi's worth of content at that apparent distance, you only need 1200x2000 pixel dimensions and a viewing lens to make it look a similar size.  Here is an announcement of a 0.44" "OLED on silicon" microdisplay that has 11 micron pixels  for 600x800,  so a 1.25" display like that would more than meet our needs of 1800x2400.  It could simulate exposure well, have adjustable brightness, allow 5x and 10x focusing just like LiveView, turn on only when an eye is sensed (to save battery life).  All that would be needed would be better AF than current LiveView options...
Logged
Pages: [1]   Go Up