Luminous Landscape Forum
Equipment & Techniques => Medium Format / Film / Digital Backs – and Large Sensor Photography => Topic started by: hdomke on June 13, 2008, 09:04:18 am
-
I recently did a side-by-side comparison of the Hasselblad H3D2 with 39 MP back and my Canon 1DsMk3.
I took each file of the exact same shot and enlarged them to be 10-feet wide.
I then printed a 40-inch wide piece from the center of each images and compared them.
The Hasselblad was clearly sharper when viewed from 10-inches away. From 2 feet or further I could see no difference.
Still, I would like that kind of quality. Is it reasonable to expect that I could get that by stitching three of my Canon files into one? (see attached picture).
Shouldn't they be equivalent?
Any thoughts or suggestions are welcome.
[attachment=7030:attachment]
-
The Hasselblad was clearly sharper when viewed from 10-inches away. From 2 feet or further I could see no difference.
Still, I would like that kind of quality. ]
Putting 3 unsharp files side by side will make them sharper?
-
I recently did a side-by-side comparison of the Hasselblad H3D2 with 39 MP back and my Canon 1DsMk3.
I took each file of the exact same shot and enlarged them to be 10-feet wide.
I then printed a 40-inch wide piece from the center of each images and compared them.
The Hasselblad was clearly sharper when viewed from 10-inches away. From 2 feet or further I could see no difference.
Still, I would like that kind of quality. Is it reasonable to expect that I could get that by stitching three of my Canon files into one? (see attached picture).
Shouldn't they be equivalent?
Any thoughts or suggestions are welcome.
[attachment=7030:attachment]
[a href=\"index.php?act=findpost&pid=201341\"][{POST_SNAPBACK}][/a]
You raise an interesting, but not original question. I have both the 1DS MKIII and the Mamiya 645 AFDII/ZD and I frequently stitch (using PTGui) in exactly the way that you suggest. Both devices are capable of producing very high quality files in this way and for me, it is no longer really about resolution as I can get as much as I need these days using precisely these methods. Stitching is also often a good solution to optical issues, particularly for ultrawideangle type shots, though of course your subject does have to be largely static, so it is no surprise that the really high end MFDB are to be found in the studio or on fashion shoots. If you are having difficulty deciding which to invest in, you need to consider other factors as well. For me, the ZD back does give a more 3D type image, probably due to the slightly narrower focusing differential of the larger sensor (I do not want to encourage a technical debate on this, except to say that I see it empirically). Also, when properly processed, I slightly prefer the overall colour rendering from the ZD. Also, the Mamiya lenses show more uniform illumination and sharpness at the wide end. This is actually an advantage when stitching. If the subject matter is purely landscape, I would generally prefer the Mamiya, but the Canon is superbly versatile and much, much faster and less hassle to use in all situations.
-
You raise an interesting, but not original question. I have both the 1DS MKIII and the Mamiya 645 AFDII/ZD and I frequently stitch (using PTGui) in exactly the way that you suggest. Both devices are capable of producing very high quality files in this way and for me, it is no longer really about resolution as I can get as much as I need these days using precisely these methods. Stitching is also often a good solution to optical issues, particularly for ultrawideangle type shots, though of course your subject does have to be largely static, so it is no surprise that the really high end MFDB are to be found in the studio or on fashion shoots. If you are having difficulty deciding which to invest in, you need to consider other factors as well. For me, the ZD back does give a more 3D type image, probably due to the slightly narrower focusing differential of the larger sensor (I do not want to encourage a technical debate on this, except to say that I see it empirically). Also, when properly processed, I slightly prefer the overall colour rendering from the ZD. Also, the Mamiya lenses show more uniform illumination and sharpness at the wide end. This is actually an advantage when stitching. If the subject matter is purely landscape, I would generally prefer the Mamiya, but the Canon is superbly versatile and much, much faster and less hassle to use in all situations.
[a href=\"index.php?act=findpost&pid=201345\"][{POST_SNAPBACK}][/a]
Are you all forgetting dynamic range...?? Still won't be 16 bit
Snook
-
I also have a DB (P25) + 1Ds and have done stitching with both and have to agree with the above posts. Stitching canon files won't give you a DB file, just a huge 35mm file, which of course might be enough, just don't expect the same file quality.
-
Are you all forgetting dynamic range...?? Still won't be 16 bit
Snook
[a href=\"index.php?act=findpost&pid=201348\"][{POST_SNAPBACK}][/a]
I know that the Hasselblad is 16-bit and the Canon is 14-bit.
But what does that mean in the real world?
Better shadow detail?
When i look at the two images I created side-by-side I can see no differences other than fine detail. Tell me what to look for. If I can't see it, I'm not buying it.
Thanks!
-
I also have a DB (P25) + 1Ds and have done stitching with both and have to agree with the above posts. Stitching canon files won't give you a DB file, just a huge 35mm file, which of course might be enough, just don't expect the same file quality.
[a href=\"index.php?act=findpost&pid=201349\"][{POST_SNAPBACK}][/a]
In what way won't the file be a DB file?
Tell me what to look for.
I am from Missouri. We pride ourselves in being the "Show Me" state.
I need to see the difference.
Please educate me about what exactly I need to look for.
-
I know that the Hasselblad is 16-bit and the Canon is 14-bit.
But what does that mean in the real world?
Better shadow detail?
When i look at the two images I created side-by-side I can see no differences other than fine detail. Tell me what to look for. If I can't see it, I'm not buying it.
Thanks!
[a href=\"index.php?act=findpost&pid=201353\"][{POST_SNAPBACK}][/a]
Yeah, I agree with you. I think the ZD back is also a 14 bit device, so not all the backs are 16 bit, but so what? Show me a shot that could only have been done or is convincingly better quality with a 16 bit device.
-
Putting 3 unsharp files side by side will make them sharper?
[a href=\"index.php?act=findpost&pid=201344\"][{POST_SNAPBACK}][/a]
No, but using less enlargement will be sharper.
All files will appear soft if they are enlarged enough.
If I take three 35mm files and stitch them, I won't have to enlarge them as much to equal the MF equivalent. Hence they would appear sharper. Yes?
-
Are you all forgetting dynamic range...?? Still won't be 16 bit
Snook
[a href=\"index.php?act=findpost&pid=201348\"][{POST_SNAPBACK}][/a]
Exactly. You can stitch to get an increase in resolution (for stationary scenes), but the colour and DR will still be Canon.
-
12 bit: 4,096 shades of grey per channel.
14 bit: 16,384 shades of grey per channel.
16 bit: 65,536 shades of grey per channel.
16 bit will lend itself to greater shadow detail, more accurate color capture, smother tonal transitions, etc, etc.
-
In what way won't the file be a DB file?
Tell me what to look for.
I am from Missouri. We pride ourselves in being the "Show Me" state.
I need to see the difference.
Please educate me about what exactly I need to look for.
[a href=\"index.php?act=findpost&pid=201354\"][{POST_SNAPBACK}][/a]
Maybe you should move to the "Do it yourself and stop pestering others to convince you" state Seriously though, I'm just not very interested in convincing ppl to see it my way. I can tell you however that I see a huge difference in detail and dynamic range, which is particularly evident if you work the files a lot in post production. If you don't see it then just stick with 35mm.. it's as simple as that really.
-
Maybe you should move to the "Do it yourself and stop pestering others to convince you" state
[a href=\"index.php?act=findpost&pid=201362\"][{POST_SNAPBACK}][/a]
-
Maybe you should move to the "Do it yourself and stop pestering others to convince you" state Seriously though, I'm just not very interested in convincing ppl to see it my way. I can tell you however that I see a huge difference in detail and dynamic range, which is particularly evident if you work the files a lot in post production. If you don't see it then just stick with 35mm.. it's as simple as that really.
[a href=\"index.php?act=findpost&pid=201362\"][{POST_SNAPBACK}][/a]
Well, it is as simple as that of course, but your assertion that there is a huge difference is overblowing it and simply contributing to an impression (which may be unintended) of an elite/needing to justify my investment mentality, which is not helped by the 'just stick with 35mm' cheap shot.
-
Here we are again. I still don't understand, when some here have been kind enough to give an answer to the OP and some clues in what there "might be" a difference, where to look for these differences, which are the advantages, etc ..., why one cannot simply take a camera and do some shots to convince oneself or not, instead putting simply all and everything said in doubt. I guess nobody needs another thread with 200 + times somebody saying "prove it to me, up to now there is no clue".
Best regards,
Thierry
Well, it is as simple as that of course, but your assertion that there is a huge difference is overblowing it and simply contributing to an impression (which may be unintended) of an elite/needing to justify my investment mentality, which is not helped by the 'just stick with 35mm' cheap shot.
[a href=\"index.php?act=findpost&pid=201364\"][{POST_SNAPBACK}][/a]
-
Here we are again. I still don't understand, when some here have been kind enough to give an answer to the OP and some clues in what there "might be" a difference, where to look for these differences, which are the advantages, etc ..., why one cannot simply take a camera and do some shots to convince oneself or not, instead putting simply all and everything said in doubt. I guess nobody needs another thread with 200 + times somebody saying "prove it to me, up to now there is no clue".
Best regards,
Thierry
[a href=\"index.php?act=findpost&pid=201365\"][{POST_SNAPBACK}][/a]
Thierry,
I am sure I would not be alone in seeing some condescension in the post that I commented on. Also, I have offered that exact kindness at the start of the thread.
-
the ZD back does give a more 3D type image, probably due to the slightly narrower focusing differential of the larger sensor... I slightly prefer the overall colour rendering from the ZD. Also, the Mamiya lenses show more uniform illumination and sharpness at the wide end
[a href=\"index.php?act=findpost&pid=201345\"][{POST_SNAPBACK}][/a]
Those are three interesting specific ideas. Let me make sure I understand them:
1. "3D type image, probably due to the slightly narrower focusing differential"
Does that refer to the depth of focus issues of larger formats? When shooting MF at f/16 is really looks like f/8 on 35mm, hence more out-of-focus background and greater sense of depth?
2. "prefer the overall colour rendering"
Isn't this something easily adjusted in PS, or is it something else?
3. "lenses show more uniform illumination and sharpness at the wide end"
Of course, for landscape work (all I would be using it for) I would tend to shoot at f/11 or f/16. At those apertures, this becomes a moot point, yes?
Thanks for being specific!
-
Henry,
What you are describing is what I have done for the past 5/6 years, mainly with the Zoerk adapter with various 35mm digial bodies. Your point is well taken and I would agree with you that in regards to overal resolution, yes, you can get close. If you allow for the 20% overlap, you are looking at around a 53mp image from a MKIII. Zoerk will allow you to shift 20mm to each side and use various medium format glass, Mamiya, Hassy, Pentax. I mainly used the Pentax 35mm FA on my Canon bodies. If you just pan by setting nodal points you may even get a greater number of usable pixels. Three portrait across the sensor stitches to make up a Landscape orientation print and three landscape up and down stitches to make up a Portrait orientation print. The Zoerk will allow you to move the body and not the lens which makes for some very easy combinations.
It's all dependent on what your clients are looking for/what their predisposed ideas are. I wanted to be able to make large prints 30 x 40 or larger with less interpolation not as much create panorama type prints. For inkjet output this method can produce some very nice images.
With the newer software tools out there for stitching the overall workflow has gotten much better. Just using CS3's photo merge will many times handle the combination work.
Paul C
-
Drew,
I honestly do not see this as an "elitist" or "condescension" issue. The vast majority of the members here are ready to speak about the differences they see and have experienced. But they are certainly not ready to try to convince others and starting arguments: that is where all will say "stop" and "test it yourself".
If one is ready to spend such an amount of money to invest in such a MF system, then it is my opinion that one should also be ready to test this system seriously. And there is only one way to test, it is the do-it-yourself way. One will learn much more this way and able to take conclusions and understand (or not), then approve or (or disapprove) much more easily than asking others to show clues and proves. One should make one's own opinion, based on one's own experiences: some points may be irrelevant to some people.
Also, for some the differences suggested here and elsewhere so many times are obvious and do not need to be proven, for them. Working and looking at thousands of files during many years makes this understanding and conviction having some ground. But I have seen as well the contrary, which leads to make it obvious that there are different types of "seeing" and judging.
Best regards,
Thierry
Thierry,
I am sure I would not be alone in seeing some condescension in the post that I commented on. Also, I have offered that exact kindness at the start of the thread.
[a href=\"index.php?act=findpost&pid=201366\"][{POST_SNAPBACK}][/a]
-
Drew,
I honestly do not see this as an "elitist" or "condescension" issue. The vast majority of the members here are ready to speak about the differences they see and have experienced. But they are certainly not ready to try to convince others and starting arguments: that is where all will say "stop" and "test it yourself".
If one is ready to spend such an amount of money to invest in such a MF system, then it is my opinion that one should also be ready to test this system seriously. And there is only one way to test, it is the do-it-yourself way. One will learn much more this way and able to take conclusions and understand (or not), then approve or (or disapprove) much more easily than asking others to show clues and proves. One should make one's own opinion, based on one's own experiences: some points may be irrelevant to some people.
Also, for some the differences suggested here and elsewhere so many times are obvious and do not need to be proven, for them. Working and looking at thousands of files during many years makes this understanding and conviction having some ground. But I have seen as well the contrary, which leads to make it obvious that there are different types of "seeing" and judging.
Best regards,
Thierry
[a href=\"index.php?act=findpost&pid=201372\"][{POST_SNAPBACK}][/a]
Thierry,
I do not really agree with what you say on the need to do your own testing. Life is too short, not all of us have access to the equipment that needs to be tested in this way and so on. Also, there are the strong subjective elements to any such testing and all kinds of bias can be introduced. No, asking for the advice of other is a good way to short cut this, though that in itself is not infallible. In offering the advice we should be careful not to disrespect the person asking. After all 'Seriously though, I'm just not very interested in convincing ppl to see it my way' could be answered by 'well if it is so beneath you, don't bother to post a reply....it is as simple as that really
-
I wanted to be able to make large prints 30 x 40 or larger with less interpolation
[a href=\"index.php?act=findpost&pid=201371\"][{POST_SNAPBACK}][/a]
I should have mentioned the the output I am looking for will be often going to 5 x 8-feet or larger.
I am occasionally selling images now that are 30-feet wide (usually printed on vinyl).
My clients and I are VERY comfortable with the quality I can get in the 30 x 40-inch range using my 35mm DSLR. But now with my Epson 11880 printer, I am having more clients ask for 5 x 8 foot prints and bigger.
I just want to make sure that I am delivering the best file that is reasonable. I realize that most of my commercial clients won't do as much "pixel peeping" as we might do, but...
-
Those are three interesting specific ideas. Let me make sure I understand them:
1. "3D type image, probably due to the slightly narrower focusing differential"
Does that refer to the depth of focus issues of larger formats? When shooting MF at f/16 is really looks like f/8 on 35mm, hence more out-of-focus background and greater sense of depth?
2. "prefer the overall colour rendering"
Isn't this something easily adjusted in PS, or is it something else?
3. "lenses show more uniform illumination and sharpness at the wide end"
Of course, for landscape work (all I would be using it for) I would tend to shoot at f/11 or f/16. At those apertures, this becomes a moot point, yes?
Thanks for being specific!
[a href=\"index.php?act=findpost&pid=201368\"][{POST_SNAPBACK}][/a]
Henry,
Nothing very scientific in my response I am afraid. 1 and 2 are very subjective and cannot be quantified. All I can say is that for landscapes, I just slightly prefer the 'look' of the ZD files after a minimum of post-processing. As for 3, no I do not think that is a moot point at all. As a landscape shooter, I would prefer to shoot at the sweet point of f8-11 on both systems, unless I needed the DOF and generally the Mamiya lenses look better to me (perhaps not surprisingly as I tend to carry more Mamiya primes, though the 75-150 Mamiya zoom is excellent).
-
They will never be the same because the magnification ratio is different in the two formats which causes the look to be different. It's that simple unless you are just shooting brick walls - forget about pixels - these formats will just produce different looks.
The high pixel count DSLR's will not be able to shoot at small apertures due to diffraction. This means that you can't expect to have a sharp file with large DOF with the 1DsIII no matter what.
The high bit color rendering and greater dynamic range of the MFDB's contribute more over all than you realize. I absolutely don't buy it that if you printed these files and compared them that at 2 ft you could not see a difference. The MF files should jump out with better color and look even without considering detail.
Lastly its obvious to everyone that you are not going to be able to rely on stitching photos to provide a solution for most images. It's a very limited circumstance where you can stitch.
Hdomke,
If you only look at details you are missing almost all of what MFDB's offer. Seriously if you can't see the difference then something has gone wrong in your process. It should jump out.
Also might I make a suggestion to the moderators that threads like these be placed in the camera gear and techniques forum instead of the MF forum?
-
Drew,
I do respectfully disagree with you, as much as you do with me!
Although I am aware that it might be difficult to get access to the gear for some, it is not an "excuse" or valid reason to rely only on what others suggest, advice or show. Photography in general is a constant learning curve. Digital in particular needs such a serious learning path: others can help you, show you the way, give you hints and ways not to go, but eventually it has to be your choice and entire conviction based on experiences learned under your own conditions. The subjective elements and bias can be eliminated, with a little care, and that's where advices can help you, but not to the point to make you decide based on this advice.
Once you have jumped, understand convinced yourself, the learning curve will just start. It is by no way finished when you have decided to jump. And believe me, if this decision comes from the conclusion that what others have said is true, then it will be a much longer learning curve for you until your gear gives you satisfaction and you may actually be very disappointed.
I am absolutely on your side, when it comes to get advices and ways to go from others, and that's where everybody here is ready to jump in. Nobody says the contrary here.
And of course, disrespect has no place, agreed. If others tell you "do it yourself", it shouldn't be taken as such, IMO.
Best regards,
Thierry
Thierry,
I do not really agree with what you say on the need to do your own testing. Life is too short, not all of us have access to the equipment that needs to be tested in this way and so on. Also, there are the strong subjective elements to any such testing and all kinds of bias can be introduced. No, asking for the advice of other is a good way to short cut this, though that in itself is not infallible. In offering the advice we should be careful not to disrespect the person asking. After all 'Seriously though, I'm just not very interested in convincing ppl to see it my way' could be answered by 'well if it is so beneath you, don't bother to post a reply....it is as simple as that really
[a href=\"index.php?act=findpost&pid=201373\"][{POST_SNAPBACK}][/a]
-
Eric,
I do appreciate your opinion. I am sorry if I am going over old territory for you, but since I live on a farm in the middle of nowhere, I have only this forum to turn to...
If you could help me understand some of your comments:
"They will never be the same because the magnification ratio is different in the two formats which causes the look to be different."
What magnification ratio are you referring to? Surely not just focal length. Please explain.
"The high pixel count DSLR's will not be able to shoot at small apertures due to diffraction."
I don't believe diffraction is causing an trouble when I shoot at f/13 or f/16. Agreed? I am using Canon's "L" glass. The have MTF charts that look very good.
"The high bit color rendering and greater dynamic range of the MFDB's contribute more over all than you realize."
Can you describe what that might be? Is it more shadow detail, or something else? Also aren't bit-depth and dynamic range "joined at the hip"? One follows the other?
"I absolutely don't buy it that if you printed these files and compared them that at 2 ft you could not see a difference." I really did to this test. Perhaps you would have used different processing, but I was very careful. I used RAW files (DNG) and processed them in Lightroom. I output them as 16-bit files with ProPhoto color space. I enlarged each to be 10-feet wide in PS using Bicubic Smoother interpolation. I sharpened to taste (the Canon files needed much more sharpening) and then printed identical pieces from each file.
"its obvious to everyone that you are not going to be able to rely on stitching photos to provide a solution for most images."
I understand that is your opinion. You have more experience than I do. I'm trying to learn from your experience. Thank you for tolerating my naive questions.
"suggestion to the moderators that threads like these be placed in the camera gear and techniques forum instead of the MF forum"
But who else besides those who use MF every day can answer these questions??
Thank you for being tolerant of my lack of experience.
-
Yawn:)
-
Thierry,
No problem.
'Lastly its obvious to everyone that you are not going to be able to rely on stitching photos to provide a solution for most images' No, I disagree. For landscape work, you can rely on it for most images.
-
No, but using less enlargement will be sharper.
All files will appear soft if they are enlarged enough.
If I take three 35mm files and stitch them, I won't have to enlarge them as much to equal the MF equivalent. Hence they would appear sharper. Yes?
[a href=\"index.php?act=findpost&pid=201356\"][{POST_SNAPBACK}][/a]
Nope. Using less enlagement will simply give you more DPI. If you're at the point where an image falls apart because you've enlarged it too much you have a different issue alltogether. A soft image is a soft image...at any size.
-
You also forgot that you're moving the lens to shoot the 1Ds files, but you're not when shooting the single MF file.
-
"The high pixel count DSLR's will not be able to shoot at small apertures due to diffraction."
I don't believe diffraction is causing an trouble when I shoot at f/13 or f/16. Agreed? I am using Canon's "L" glass. The have MTF charts that look very good.
[a href=\"index.php?act=findpost&pid=201386\"][{POST_SNAPBACK}][/a]
The quality of glass will not overcome the basic physics that the Canon will be diffraction limited at f8. Scotty was right you can't change the laws of physics.
No MTF chart is going to change that.
-
I recently did a side-by-side comparison of the Hasselblad H3D2 with 39 MP back and my Canon 1DsMk3.
I took each file of the exact same shot and enlarged them to be 10-feet wide.
I then printed a 40-inch wide piece from the center of each images and compared them.
The Hasselblad was clearly sharper when viewed from 10-inches away. From 2 feet or further I could see no difference.
Still, I would like that kind of quality. Is it reasonable to expect that I could get that by stitching three of my Canon files into one? (see attached picture).
Shouldn't they be equivalent?
Any thoughts or suggestions are welcome.
[attachment=7030:attachment]
[a href=\"index.php?act=findpost&pid=201341\"][{POST_SNAPBACK}][/a]
if your object is static then I think you shouldn't have a problem getting high resolution shots with your stitched 1ds mk III images.
It's quite simple to gauge, just treat your 1Ds MKIII image as a crop of a H3D2 image and compared a similar crop on your H3D2 image.
As long as the process of stitching the files together doesn't introduce any loss of quality in the files, you can judge the quality of the shot based on the crop of the H3D2 image vs a single 1DS MKIII image with appropriate focal length.
The only problem with the 1DsMKIII is that the image has a different look to it due to the AA filter. How much of a problem it is for your images depends alot on your personal preference.
-
the Canon will be diffraction limited at f8
[a href=\"index.php?act=findpost&pid=201408\"][{POST_SNAPBACK}][/a]
Chris,
I had to revisit the term "diffraction" and I am not finding that the there is any limit of f/8. Can you send me a reference to that information?
Also, isn't diffraction measurable, and won't it show up in MTF charts?
In the book "EF Lens Work III" on p. 197 they say this about diffraction:
"A phenomenon in which light waves pass around the edges of an object and enter the shadowed area of that subject, caused because of the wavelike nature of light. Diffraction in a photographic lens is known for causing flare (diffraction flare) which occurs when light rays bend around the edges of the diaphragm. Although diffraction flare tends to appear when the diaphragm is smaller than a certain size, it actually depends not only on the diameter of the diaphragm but also on various factors such as the wavelength of the light, the lens focal length and the aperture ratio. Diffraction flare causes reduction in image contrast and resolution, resulting in a soft image. The laminated diffraction optical elements developed by Canon control the direction of the light by intentionally creating diffraction."
-
if your object is static then I think you shouldn't have a problem getting high resolution shots with your stitched 1ds mk III images.
[a href=\"index.php?act=findpost&pid=201409\"][{POST_SNAPBACK}][/a]
My subject will for this camera will be landscapes. They are mostly static.
It's quite simple to gauge, just treat your 1Ds MKIII image as a crop of a H3D2 image and compared a similar crop on your H3D2 image.
[a href=\"index.php?act=findpost&pid=201409\"][{POST_SNAPBACK}][/a]
Great idea. I will do that!
The only problem with the 1DsMKIII is that the image has a different look to it due to the AA filter.
[a href=\"index.php?act=findpost&pid=201409\"][{POST_SNAPBACK}][/a]
I wonder how the anti-alias (AA) filter affects the image. I would guess it would cause softness. Both cameras have IR filters, but only the Canon has an AA filter. I guess MF camera sensors are less prone to moire and therefore don't require an AA filter?
-
I saw the subject on the main forum list and thought I'd throw in my two cents:
I do mostly panoramic landscape photography and I don't think comparing crops of the two sensors is going to be very useful since you aren't gaining anything by making a couple 1ds frames equal a medium format frame. It's going to be much less work to just take a single MF shot rather than stitch even if the MF setup is worse than the 35mm setup.
On the other hand if you dramatically increase the amount of detail (the resolution) or increase the width of the image (like a pano format), and print large enough to use the resolution the stitching techniques can far surpass anything from a MF back. Provided the subject allows it, of course. If you don't print large enough to use the detail it's not going to make much difference, I'd expect, and stitching a large image can be a lot of work.
As another soapbox item: I find that viewing distance as a limiter of resolution is kind of a cop-out for many comparisons since if the detail is present in the photo people will get closer to inspect it. I find that even with 3x9 foot images viewers will inspect it at distances of a few inches if something catches their eye.
At wider angles of view stitching really starts to come into it's own and I think lenses are going to have much more of an impact on the image quality than the details of the sensors. Here a stitched image can dramatically improve overall image quality by allowing longer focal lengths and allowing you to use the best lenses for the subject rather than selecting a lens purely for adequate field of view. At extremes you could be talking about an MF fish-eye comparing to a lens like the Canon 135 F2 or even a Leica macro.
Of course you can always stitch with medium format and really make a statement!
Anyway, those are my thoughts, jumbled as they may be.
Doug
Jpg reductions of some of my humble efforts --> www.bcpanorama.com
-
Chris,
I had to revisit the term "diffraction" and I am not finding that the there is any limit of f/8. Can you send me a reference to that information?
[{POST_SNAPBACK}][/a]
This reference is a model of clarity and is interacive as well
[a href=\"http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm]http://www.cambridgeincolour.com/tutorials...photography.htm (http://index.php?act=findpost&pid=201413\")
-
1. "Magnification", "3D view", "better color" are BS, or more politely expressed, the imagination is playing a game with some.
2. Stitching can be completely adequate in some situations, like landscapes.
3. Stitching is not always working with architecture, and perhaps never in studio. Moving subjects are out of the scope in most caswes (when these moving objects are just the main subjects).
4. The dynamic range can be a big issue even there, where the stitching is the best, in landscapes. Although it is possible to capture an even larger dnamic range of the scenery in several frames than an MFDB can capture in one frame, the processing can be extremely laborous. Furthermore nothing help, if the large dynamic range of the scenery can not be separated in frames, except HDR - even more work, less fine result (and today's HDR results are not everyone's taste).
So, when time and effort are non-issue and it is not a problem, when a certain situation can not be solved adequatey, then yes, stitching. However, some can say justifiedly, that they can make several new projects in the time required to process a high-quality stitching.
There is another side of the issue. MFDBs too have limits; what if even a 39Mpix image does not suffice or the angle of view is not large enough? I hear already some: then stitching several MFDB images. Suddenly stitching becomes acceptable. LOL.
-
Both cameras have IR filters, but only the Canon has an AA filter
What do you mean with "IR filter"? An IR transmitting filter, replacing the hot mirror? If yes, then there is no AA filter there any more.
-
"There is another side of the issue. MFDBs too have limits; what if even a 39Mpix image does not suffice or the angle of view is not large enough? I hear already some: then stitching several MFDB images. Suddenly stitching becomes acceptable. LOL."
Exactly!
And now we have an even bigger file!
Nothing wrong with that!
Cheers,
Willem.
-
3. Stitching is not always working with architecture, and perhaps never in studio.
[a href=\"index.php?act=findpost&pid=201441\"][{POST_SNAPBACK}][/a]
Why can't you stitch in the studio? I did for years when I was shooting with a 2X3 sensor and didn't want to change the perspective by moving the camera back from the set to get the needed crop room.
-
Why can't you stitch in the studio? I did for years when I was shooting with a 2X3 sensor and didn't want to change the perspective by moving the camera back from the set to get the needed crop room.
I guess the MFDB owners of this forum wouyld be laughing their heads off comparing what you achieved with what they can achieve by one click.
-
I guess the MFDB owners of this forum wouyld be laughing their heads off comparing what you achieved with what they can achieve by one click.
[a href=\"index.php?act=findpost&pid=201449\"][{POST_SNAPBACK}][/a]
Well I was doing that with a "medium format" chip but I got so fed up with the limitations of the 2X3 chip proportion I bought an A75.
But you didn't answer my question.
-
There are several recent articles/tutorial/reviews that can be found on the main page of LL that go over why the newer high pixel count DSLRs are more diffraction limited due to smaller pixel wells.
As mentioned you can do all kinds of tricks with the DSLR's to get a better image - blending 2 exposures for more dynamic range, then stitching these HDR sets together but its a lot of work. Well you can go even further and stack f/5.6 shots at different focal distances to get more DOF. How much work do you want to do? And what is moving in your scene while you take all these shots?
With the MFDB you don't even necessarily need a tripod.... much much easier.
-
Perhaps,you should rephrase your initial question to "by stitching multiple frames
can I stretch the 'enlargeability',and get more bang for my buck,from my 1Ds3 ? '
If you are 'budget-limited' then this is a good way to enlarge a scene to a higher degree
before it starts to fall apart.
There's the inconvenience factor but there are no free rides
In my opinion,a properly executed 3-4 image stitch will get you somewhat closer to
a single frame 39Mp capture factoring out,of course, that now famous 3D effect and
enhanced shadow detail that the MFDB's bring to the table.
If budget isn't a factor and you have clients who are willing to pay for prints of this size
then,really,there should be no reason to consider the stitched 35mm option as the best
of what you can offer.
If you're prepared to stitch 35mm then stitching,even 2, 39Mp captures will set your bar
much higher.
Mark
-
Henry, you NEED to read this article. It will also be useful to the couple MFDB shooters with an open mind.
http://clarkvision.com/photoinfo/large_mosaics/index.html (http://clarkvision.com/photoinfo/large_mosaics/index.html)
-
Henry, you NEED to read this article. It will also be useful to the couple MFDB shooters with an open mind.
http://clarkvision.com/photoinfo/large_mosaics/index.html (http://clarkvision.com/photoinfo/large_mosaics/index.html)
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=201478\")
59 images! that guy has too much time on his hands.
now 4 images stiched with a 39mp back is easier way-
[a href=\"http://www.kapturegroup.com/quad/quad.html]http://www.kapturegroup.com/quad/quad.html[/url]
paul
-
Here is 32325x8829 image from 22 captures using ZD camera + 150mm AF lens + B&W polarizer. Shot with ZD mounted on a spherical bracket, mirrror lockup for each shot. Buffer wasn't an issue, shooting RAW + smallest jpg-s. Stitched using AutoPano Pro. Final file size 3.1 GB, PSB format.
(100% crops included - 1.9 MB)
In the first crop - I did not even see that very sad looking person during shooting!
The last crop - small pebbles from the mountain wall in the center of the image. This MF system is a microscope!
-
I think some MFDB users in this thread still haven't grasped the relationship between sensor size and resolution.
If you are stitching together two or three 1Ds3 images in order to get an image as detailed as a single shot from a P45 (for example), then you should use the same focal length of lens at the same f stop for both shots.
If you were to use the actual same lens on both cameras (using an adapter for the 35mm), there's no reason why the resulting images would be different in any respect, outside of differences in RAW converters and other peculiarities of design such as the differences between the CCD and the CMOS and AA filter issues.
The fundamentals with regard to resolution, DoF and even dynamic range are the same in the sense that a P45 is basically two 1Ds3 sensors joined together.
If the resulting image comes from two sensors stitched together, or two images from the separate sensors stitched together, the result is basically the same, provided the same lens is used
Whilst it's true that a P45 has 16 bit processing, considering all the puzzlement about the expected improvement of 14 bit over 12 bit in recent 35mm DSLRs, it's doubtful that the 16 bit of the P45 compared with the 14 bit processing of the 1Ds3 would be noticeable.
All of the above is based on the assumption that the objective in stitching is to get the same file size and aspect ratio as the single shot from the P45. If you change the aspect ratio of the stitched 1Ds3 image, as in a panorama, then clearly the 1Ds3 resulting stitch will be superior to the single P45 shot in every respect since you would have to use a wider angle lens with the P45 and crop to the aspect ratio of the 1Ds3 stitch.
-
Here is 32325x8829 image from 22 captures using ZD camera + 150mm AF lens + B&W polarizer. Shot with ZD mounted on a spherical bracket, mirrror lockup for each shot. Buffer wasn't an issue, shooting RAW + smallest jpg-s. Stitched using AutoPano Pro. Final file size 3.1 GB, PSB format.
(100% crops included - 1.9 MB)
In the first crop - I did not even see that very sad looking person during shooting!
The last crop - small pebbles from the mountain wall in the center of the image. This MF system is a microscope!
[a href=\"index.php?act=findpost&pid=201489\"][{POST_SNAPBACK}][/a]
Man! the crop of the lonely figure is just so forlorn. in comparison to the vast expanse of the landscape it only enhances the isolation of the subject.
What the heck is some one doing walking on that mountain with no pack? going to get some milk and sugar at the closest ezy-mart 5000 Kilometers away
-
My subject will for this camera will be landscapes. They are mostly static.
Great idea. I will do that!
I wonder how the anti-alias (AA) filter affects the image. I would guess it would cause softness. Both cameras have IR filters, but only the Canon has an AA filter. I guess MF camera sensors are less prone to moire and therefore don't require an AA filter?
[a href=\"index.php?act=findpost&pid=201426\"][{POST_SNAPBACK}][/a]
from my experience with a 1ds mk ii and a leaf 75s on a mamiya AFd, I find that the AA filter does cause a certain amount of softness and flatness that I haven't been able to overcome happily in photoshop...but that's only because I'm doing big prints for exhibiting so I get really close up.
I think part of the reason why people state this "3-d" look with MFDB is that with MFDB shots I get really defined edges on my subjects whereas with the Canon there's a feeling that the edges are abit smudged and blend in with the background. Very subtle but it's there.
I really wouldn't worry about moire. it's a very over-rated fear.
"At wider angles of view stitching really starts to come into it's own and I think lenses are going to have much more of an impact on the image quality than the details of the sensors. Here a stitched image can dramatically improve overall image quality by allowing longer focal lengths and allowing you to use the best lenses for the subject rather than selecting a lens purely for adequate field of view. "
I agree with the above statement. for the MF 35mm lenses, most are pretty poor.
-
Here is 32325x8829 image from 22 captures using ZD camera + 150mm AF lens + B&W polarizer. Shot with ZD mounted on a spherical bracket, mirrror lockup for each shot. Buffer wasn't an issue, shooting RAW + smallest jpg-s. Stitched using AutoPano Pro. Final file size 3.1 GB, PSB format.
(100% crops included - 1.9 MB)
In the first crop - I did not even see that very sad looking person during shooting!
The last crop - small pebbles from the mountain wall in the center of the image. This MF system is a microscope!
[a href=\"index.php?act=findpost&pid=201489\"][{POST_SNAPBACK}][/a]
This is all very impressing and boring too, always talking about resolution and MP, I am not sure if all of you "Fine Art Stitching Photographers" knows the work of Thomas Kellner. Sorry I forgot this is a MF forum, but I could not resist. T.
[attachment=7038:attachment]
[attachment=7039:attachment]
-
How many stops does a "blad" 39MP have?
And how many stops does the new canon 1DS3 have?
Dynamic range is the most important thing for me. Now that we are above 22Mps I dont care about them anymore. DR and real 16bit is what makes files sing.
I got the canon 1DS3 and I like it. The files are a bit stronger than the 1DS2 files.
Everytime a rent a Hassy I'm surprised how strong the files are. There is much more flesh on the bone for heavy color grading. I'm not talking about resolution here. Just the file quality.
I'm right on the fence on buying into a DMF system because of that.
So what's the DR difference between the canon and the hassy?
Is it realy that big?
Thanks
Tim
-
The basic answer is clearly yes.
The only question is whether you would need 3 or 6 1ds3 files.
I find my D3 stitched images to be totally superior to my ZD files whatever the print size.
I am at a loss as to why anybody who has done an actual comparison with real world images would answer something different.
Cheers,
Bernard
-
The german magazin profifoto did an extensiv comparison between hasselblad H3II 31/ 39 backs and a canon 1dsmk3 in their january issue.
Here are their findings:
dynamic range in f stops
iso 100
1Ds3: 11,4
h39: 12
h31: 11,4
iso 400
1Ds3: 11
h39: 10,5
h31: 10,6
resolution in lines / height + USM
1ds3 with 70-200/4 3467
h31 HC120 3344
h39 HC120 4145
Has anyone seen any other actual side-by-side test of D.R. between these cameras? (I am not talking about 14 vs. 16 bit conjecture, but Actual Tests...)
-
The basic answer is clearly yes.
The only question is whether you would need 3 or 6 1ds3 files.
I find my D3 stitched images to be totally superior to my ZD files whatever the print size.
I am at a loss as to why anybody who has done an actual comparison with real world images would answer something different.
Cheers,
Bernard
[a href=\"index.php?act=findpost&pid=201529\"][{POST_SNAPBACK}][/a]
Bernard,
I don't own a DB to do comparisons, but I see nothing in principle why a single DB shot would have any advantage over a stitched 35mm image of equal pixel count. unless the MF lens used were a better lens. If the resulting stitched 35mm image were of significantly greater pixel count than the single DB shot, then there would be no contest. The 35mm stitch would be better in all respects.
-
I invited a good friend of mine to enter in to this discussion. He happens to be an art school professor that teaches photography at a college in Texas. This was his answer:
it seems to me to be very much like the forums regarding speaker cable, "i can hear a difference, i don't care what the science is!"
I don't know. It seems to me that much of what is being discussed here is boils down to physics and optics. There should be a right answer. But it is interesting to compare high-end audio to high-end photography. I'm don't see photographers using the weird tweaks that audiophiles use (like thousand dollar speaker cables) but I wonder if he has a point.
Is this discussion of MF images being more "3D" really just an illusion? If it is real, can it be measured?
-
I invited a good friend of mine to enter in to this discussion. He happens to be an art school professor that teaches photography at a college in Texas. This was his answer:
it seems to me to be very much like the forums regarding speaker cable, "i can hear a difference, i don't care what the science is!"
I don't know. It seems to me that much of what is being discussed here is boils down to physics and optics. There should be a right answer. But it is interesting to compare high-end audio to high-end photography. I'm don't see photographers using the weird tweaks that audiophiles use (like thousand dollar speaker cables) but I wonder if he has a point.
Is this discussion of MF images being more "3D" really just an illusion? If it is real, can it be measured?
[a href=\"index.php?act=findpost&pid=201538\"][{POST_SNAPBACK}][/a]
The '3d effect' is the narrower DOF for a given aperture and angle of view caused by a large sensor that some perceive to'pop' the subject off the background (typically fashion and potraits)
It is not a myth but it is debatable whether the look can be replicated by choosing a wider ap on a smaller format - the answer - sometimes
However if you were stitching you have a larger sensor and the effect is the same !
IMO the blur filter on the canon makes it not as good pixel per pixel
so as BL says the question is "do I need 3 or 6 canon images to equal an MF capture"
Somewhere between I would imagine
Obviously 50 images from a canon are better than MF
Of course single shot capture is simple , allows for subject movment and minimises your time in front of the computer
S
-
Thanks to everyone on this forum for the comments!
It has been very educational. After reflecting on this I have decided to hold off on upgrading to a MF system. Here is my thinking:
- It is not much better (as I said, viewing big prints at a typical viewing distance showed no difference)
- Stitching three Canon images together should approach the 39 MP MF quality for landscapes.
- My clients will not see the difference anyhow since they don't do pixel peeping.
- Carrying two systems will be a pain
Also one experienced commercial photographer told me: "Buying that Hasselblad digital camera with three lenses was the worst business decision in my entire career" He does almost all his commercial work with his Canon.
And, as another photographer, one who has used a MF system for years said to me "I think you will have buyers remorse". He is thinking of getting rid of his 22 MP Phase One MF camera system and switching to Canon.
-
I don't own a DB to do comparisons, but I see nothing in principle why a single DB shot would have any advantage over a stitched 35mm image of equal pixel count. unless the MF lens used were a better lens. If the resulting stitched 35mm image were of significantly greater pixel count than the single DB shot, then there would be no contest. The 35mm stitch would be better in all respects.
[a href=\"index.php?act=findpost&pid=201532\"][{POST_SNAPBACK}][/a]
Ray,
That's exactly what I mean.
Cheers,
Bernard
-
Also one experienced commercial photographer told me: "Buying that Hasselblad digital camera with three lenses was the worst business decision in my entire career" He does almost all his commercial work with his Canon.
And, as another photographer, one who has used a MF system for years said to me "I think you will have buyers remorse". He is thinking of getting rid of his 22 MP Phase One MF camera system and switching to Canon.
[a href=\"index.php?act=findpost&pid=201547\"][{POST_SNAPBACK}][/a]
This is silly to say something so broad.
-
This is silly to say something so broad.
[a href=\"index.php?act=findpost&pid=201558\"][{POST_SNAPBACK}][/a]
I agree that those are broad statements, but those are not my words. Those are quotes from two friends of mine, both of whom are experienced professional photographers. The one thinking of getting rid of his MF system is a professional nature photographer.
-
Daniel Buck from FM did a series of stitched photos using the 1DS2 and 85mm f/1.2, wide open. It had an amazing 3D feel to it. The wider FOV and narrow DOF gave it a MF "look".
-
Nature as in shooting pictures of birdies in trees or flowers? Well I can see getting rid of the MF and going smaller.
I'm an experienced professional commercial photographer..26 years now shooting in SF and I would negate one your statements by saying "Buying my medium format digital system was the best thing I ever did".
-
Buying both a MFDB and a 1Ds Mark III was the best thing that I ever did.
-
Daniel Buck from FM did a series of stitched photos using the 1DS2 and 85mm f/1.2, wide open. It had an amazing 3D feel to it. The wider FOV and narrow DOF gave it a MF "look".
[a href=\"index.php?act=findpost&pid=201560\"][{POST_SNAPBACK}][/a]
You you by any chance have a link?
-
The so-called 3D look is not the product of DoF, but the wide angle view coupled with rectilinear projection. Most panoramic images are made with cylindrical projection, which is better for details, but it has a different effect. Plus, I guess thouse without a TS lens regard the perspective distortion as 3D effect as well.
-
I recently did a side-by-side comparison of the Hasselblad H3D2 with 39 MP back and my Canon 1DsMk3.
Hello Henry. Were you at the Frans Lanting seminar in St. Louis back in March? I think we may have exchanged pleasantries. Or maybe it was the Art Wolf seminar last year.
Anyway, I will probably get flamed by others on this forum for saying this, but there is some merit in what you are saying. I've done my own comparison between my D3 and my P45+ on my Mamiya 645AFD II. Using different focal lengths to ensure that I got the same number of pixels on the same field of view, I compared "pixel quality". I was fully expecting the P45+ to blow away my D3 at the pixel level, but it didn't. The P45+, at the pixel level, was better (crisper, mostly), but not dramatically so. In 14-bit mode, the D3 has tremendous ability to pull clean detail out of the shadows without suffering posterization and other artifacts. In that regard, I'd say it's on a par with my P45+. And the D3 actually has a little more highlight headroom than my P45+, so in that sense, the D3 is actually a bit better (tho again, the difference is slight). If the D3 sensor didn't have an AA filter, I'd say the pixel level quality would be a wash. I can't speak to the 1DsIII, but I would expect it to be similar.
Now, having said that, there are of course caveats. Obviously, if I were using both systems from the same place and shooting the same composition, the higher resolution of the MFDB is going to blow my D3 (and even your 1DsIII) away. Given that pixel quality is comparable, however, you could stitch and get almost the level of quality AND resolution of the MFDB. But stitching always makes life more complicated. Much more post processing. Difficulty with dealing with things moving in the frame between stitch shots. And there's the time factor - catching that fleeting light when you have to take a bunch of shots is obviously pretty difficult. All in all, stitching can get you a lot, but it comes at a high price, and many times just isn't practical. For all these reasons, I won't be giving up my P45+. And for landscape shooting, I would never use the D3 if I have the P45+ available, even if stiching were an option (remember, I can also stitch with the P45+ for even higher resolution).
The bottom line is that modern 35mm DSLRs with low noise sensors and true 14-bit low noise A/Ds deliver very very good pixel quality. Shadow detail retrieval is excellent. Color is very comparable. I feel that if it weren't for the AA filter, the latest high-end DSLRs would essentially match the pixel quality of MFDBs. The two big advantage MFDBs still have over high end DSLRs is resolution (due to the MPs) and crispness (due to the lack of an AA filter). For me at least, those two advantages, plus the issues with stitching, makes the P45+ a no-brainer for those high detail shots.
Regards,
Mort.
P.S. I should point out that I use my P45+ for landscapes. I am a resolution and crispness junkie, and if I had the patience, I'd be shooting 8 x 10 view cameras. So my D3 (and probably your 1DsIII) would always be lacking for me, even with stitching. If I were a studio shooter, with optimal lighting and high speed strobes, which I think might play to MFDBs strong points, or if I were shooting better glass than the Mamiya stuff I have (35 AF, 55-110, 80 AF, and 150 AF), I expect I would see a greater difference between my D3 and P45+ at the pixel level in favor of the P45+. This is one reason I am mulling switching my P45+ to Alpa or Cambo bodies and Rodenstock HR lenses.
-
Simply put, with regard to image quality comparison, stitching images only increases the number of pixels. All other image quality differences that are seen in unstitched images will remain the same (with the possible exception of noise).
-
I asked Kevin Sink, a successful nature photographer in Kansas City what he thought about this thread. Here is his email to me.
"Anyway, all of it is interesting but I think a bit silly. Shouldn't all you people be out shooting or working on images? Just kidding, sort of. For me, it comes down to just a ludicrous amount of money to spend on a camera. I mean really, how many trips to Patagonia could you take for the difference in price? All the arguments point to slightly better detail in the MF shadows, but shouldn't we all be griping to the manufacturers about how much these backs are?
Its just gotten so extreme I've lost my taste for the sharpness bleeding edge.... it all comes down to image capture anyway. Sometimes I feel like a manipulated gerbil - if you pay us this much more, you'll be a better photographer!!! Its about the craft of the artistic process, not chasing technology into bankruptcy!! Too harsh??"
Note: He has used Medium Format for years and is now using a Phase One Digital back.
-
I asked Kevin Sink, a successful nature photographer in Kansas City what he thought about this thread. Here is his email to me.
Note: He has used Medium Format for years and is now using a Phase One Digital back.
[a href=\"index.php?act=findpost&pid=201598\"][{POST_SNAPBACK}][/a]
Please modify your post above as the quote appears to be from me WHICH IT MOST CERTAINLY IS NOT! Thanks.
-
This is one reason I am mulling switching my P45+ to Alpa or Cambo bodies and Rodenstock HR lenses.
[a href=\"index.php?act=findpost&pid=201573\"][{POST_SNAPBACK}][/a]
[/quote]
Hi there,
I have tested side by side on landscapes/foliage/ etc
H2 Blad /35/80/120 lenses Phase P45
Canon 1Ds II and III with Canon WA's and Leica WA's
( the Leica 19 MM leaves the Canons in the dust here!)
Mamiya 645II with 28/35/80 lenses Phase P45+
Both single frame and Stitched files in Realviz
Against
Alpa SWA with P45+ and 24/35/47/90/180 Schneider Apo Digitars
Both single frame and Stitched files in Realviz
All files processed out in C1 or RAW Dev
Output to 200CM on our in house Lightjet Printer/ Fuji Archival Paper
The Schneiders blow the others away--period
Des W
-
Please modify your post above as the quote appears to be from me WHICH IT MOST CERTAINLY IS NOT! Thanks.
[a href=\"index.php?act=findpost&pid=201601\"][{POST_SNAPBACK}][/a]
TechTalk,
Sorry. I did not mean to suggest it was you. I don't know how to edit my post. Tell me how to do it and I will change it.
-
TechTalk,
Sorry. I did not mean to suggest it was you. I don't know how to edit my post. Tell me how to do it and I will change it.
[a href=\"index.php?act=findpost&pid=201611\"][{POST_SNAPBACK}][/a]
Click the "Edit" button on your post, remove my name and then click "Submit Modified Post" button below. Thanks.
-
Simply put, with regard to image quality comparison, stitching images only increases the number of pixels. All other image quality differences that are seen in unstitched images will remain the same (with the possible exception of noise).
[a href=\"index.php?act=findpost&pid=201576\"][{POST_SNAPBACK}][/a]
I don't believe that's quite right. Stitching not only increases the number of pixels but effectively increases the size of the sensor since each single image that is part of the resulting stitch is the product a sensor of a specific size.
Larger sensors usually have the advantage of greater dynamic range because they have a larger light-gathering area. That advantage disappears in comparison with a stitched image fom the smaller sensor.
Not only that, the reverse can be true when exposure is on automatic mode. Whilst conventional advice dictates that one use the same exposure for each shot to be stitched, one can increase dynamic range by using the camera in aperture priority mode so that dark parts of the scene receive a greater exposure and the brighter parts, such as a sky, receive less exposure.
This approach requires greater preparation of the images before stitching but is quite workable with modern stitching programs.
-
Larger sensors usually have the advantage of greater dynamic range because they have a larger light-gathering area. That advantage disappears in comparison with a stitched image fom the smaller sensor.
[a href=\"index.php?act=findpost&pid=201641\"][{POST_SNAPBACK}][/a]
Ray, it's not the sensor size that matters, but the size of the photosites within the sensor. MFDBs have bigger photosites and thus need a bigger area to contain whatever number of them. This is what gives them the DR advantage, not the overall dimensions of the sensor.
-
Those DR tests are done with the standard ISO definition of Dynamic range which while technically accurate are not useful for photographic comparisons. You'll see tons of tests all over the internet that show the 1D3 to have almost 12 stops too seemingly more DR that MFDB but in reality their useful DR is more like 9 stops.
If you want to believe a magazine article go ahead but I've tested this myself with Imatest software and a tranmissive step wedge. I have not tested the 1Ds3 personally but its reported to have less DR than the 1D3 which I did test - a trade off for more pixel density. Even the older phase backs from 2005 have more than 2 stops of useful DR and in some cases closer to 3 stops advantage over the 1Ds3. As is the case with the forums there will always be a Ray that wants to argue without testing - discarding other posters real information for their own speculation. I'm not going to go out of my way to do more tests or post endlessly - I'm just saying I've tested it myself with a controlled and rigorous test and there is a significant difference between the top DSLR's and even older MFDB in terms of DR. And actually the differences in color - smoothness of subtle color are just as big or bigger. On top of that you have the different lenses and magnification factors that also change the look.
There will still be the Ray's that will continue to argue - and there may be a few scenes/circumstances where you will get the same result but its only a few and i'm confident that in the majority of shots that a Stitched DSLR image is not going to equal a MFDB image. A low contrast scene without any movement that can be shot at f/8 or wider ... maybe something like that it will work. But anyone with a MFDB could also shoot the same scene and stitch their shots too and end up with a better file.
If you are still skeptical go rent a MFDB and shoot some different scenes and see for yourself.
-
Ray, it's not the sensor size that matters, but the size of the photosites within the sensor. MFDBs have bigger photosites and thus need a bigger area to contain whatever number of them. This is what gives them the DR advantage, not the overall dimensions of the sensor.
[a href=\"index.php?act=findpost&pid=201652\"][{POST_SNAPBACK}][/a]
Nick,
That might be partly true if we make a distinction between pixel size and pixel pitch. The pitch of the pixel relates to the number of pixels on the sensor. The pixels on Canon's first DSLR, the 3mp D30, were larger than the pixels on its 3rd DSLR, the 6mp D60, yet the dynamic range of the D60 was not less as a result.
The pixel pitch of the Phase One P45+ is very similar to the pixel pitch of the 1Ds3. Stitch two 1Ds3 sensors together and you have basically a P45+ in terms of both sensor size and pixel count.
I don't have any information on the DR differences between a single P45+ CCD pixel and a single 1Ds3 CMOS pixel. All else being equal, one might expect the P45 pixel to have a higher DR because the photoreceptor is a CCD with a greater fill factor than the smaller photodiode of the 1Ds3, which has to make room for other processing devices under the one microlens.
However, the very presence of other processors next to each photodiode is the strength of the CMOS sensor, and that is partly responsible for its low noise, which in turn affects dynamic range.
One could easily test this by taking two shots, one with a 1Ds3 and one with a P45+ using the same focal length lens from the same position. The P45 sensor is 48x36 and the 35mm sensor 24x36. One could turn the 1Ds3 vertically for the shot, then crop the P45 image in two. One could then compare the two images which effectively are from the same size sensor and consist of the same number of pixels.
It might be the case that the P45 image, cut in half, still has a slight DR advantage as a result of the greater exposure afforded by its lower base ISO of 50. It would be interesting to see the difference.
Nevertheless, setting aside such differences between the CCD and the CMOS design, the light gathering capacity of the sensor is largely dependent upon sensor size rather than pixel size, all else being equal.
-
Ray, it's not the sensor size that matters, but the size of the photosites within the sensor. MFDBs have bigger photosites and thus need a bigger area to contain whatever number of them. This is what gives them the DR advantage, not the overall dimensions of the sensor.
[a href=\"index.php?act=findpost&pid=201652\"][{POST_SNAPBACK}][/a]
Which is why I' rather stitch with a Nikon D3 than with a P45+...
Besides, PTgui Enfuse algo implementation makes DR limitations pretty much a thing of the past. Just shoot 3 frames at each location of your pano and you can reach results that even the best 22MP MFDB cannot dream of achieving in terms of highlight and shadow detail.
Cheers,
Bernard
-
12 bit: 4,096 shades of grey per channel.
14 bit: 16,384 shades of grey per channel.
16 bit: 65,536 shades of grey per channel.
16 bit will lend itself to greater shadow detail
Noise destroys much tonal detail in the shadows. I think a 16 bit 39MP camera cannot provide a diferent result than a 14bit 39MP camera. Maybe there is a difference in a MF 22MP camera at very low ISOs. I doubt it.
Read this superb document about noise:
http://theory.uchicago.edu/~ejm/pix/20d/te...oise/index.html (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html)
R.
-
Ok, that's the theory - now how about you check that it is really so in practice ?
I've tested the 1Ds3 and found that cropping *pixel to pixel* (same image features on same pixels) my Mamiya beat it. Not only on file quality (color and DR) but also on actual detail !
The Canon sensor is a compromise design - CMOS, microlenses, low-pass filter, small cells, NR etc all suck detail and DR away.
Edmund
I think some MFDB users in this thread still haven't grasped the relationship between sensor size and resolution.
The fundamentals with regard to resolution, DoF and even dynamic range are the same in the sense that a P45 is basically two 1Ds3 sensors joined together.
If the resulting image comes from two sensors stitched together, or two images from the separate sensors stitched together, the result is basically the same, provided the same lens is used
Whilst it's true that a P45 has 16 bit processing, considering all the puzzlement about the expected improvement of 14 bit over 12 bit in recent 35mm DSLRs, it's doubtful that the 16 bit of the P45 compared with the 14 bit processing of the 1Ds3 would be noticeable.
[a href=\"index.php?act=findpost&pid=201490\"][{POST_SNAPBACK}][/a]
-
Noise destroys much tonal detail in the shadows. I think a 16 bit 39MP camera cannot provide a diferent result than a 14bit 39MP camera. Maybe there is a difference in a MF 22MP camera at very low ISOs. I doubt it.
Read this superb document about noise:
http://theory.uchicago.edu/~ejm/pix/20d/te...oise/index.html (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html)
R.
[a href=\"index.php?act=findpost&pid=201680\"][{POST_SNAPBACK}][/a]
Nemo,
Thanks for the link to the article. It looks very impressive! However, it is a bit overwhelming for someone who is not an academic deeply versed in statistics. Perhaps you (or someone else) could summarize the take-home points from this article. After skimming the article, here are some specific questions:
1. In the real world, will prints from a 16-bit capture show more dynamic range than from a 14-bit capture?
2. What are the implications of this paper on "Expose to the Right"?
3. What are the implications of this paper on selecting the right ISO in low-light situations?
-
Hello Henry. Were you at the Frans Lanting seminar in St. Louis back in March?
...I feel that if it weren't for the AA filter, the latest high-end DSLRs would essentially match the pixel quality of MFDBs. The two big advantage MFDBs still have over high end DSLRs is resolution (due to the MPs) and crispness (due to the lack of an AA filter). For me at least, those two advantages, plus the issues with stitching, makes the P45+ a no-brainer for those high detail shots.
[a href=\"index.php?act=findpost&pid=201573\"][{POST_SNAPBACK}][/a]
Mort,
Thanks for your thoughts.
You are right, I was at the Frans Lanting and the Art Wolf seminars. Interestingly Frans Lanting creates huge prints from his Nikon 35mm DSLRs. They look great! I've seen prints of his 20-feet wide.
You mention the impact of the lens quality on the image. Since all wide-angle lenses are inferior to normal focal length lenses (chromatic aberration, contrast, softness at the edges) could it be argued that stitching together 3 images from a normal focal length lens (say a 50mm on my Canon) would have better lens quality than shooting a single image with a wide-angle lens with the Hasselblad (say the Wide Angle 50mm f/3.5 HC)?
-
Ok, that's the theory - now how about you check that it is really so in practice ?
I've tested the 1Ds3 and found that cropping *pixel to pixel* (same image features on same pixels) my Mamiya beat it. Not only on file quality (color and DR) but also on actual detail !
The Canon sensor is a compromise design - CMOS, microlenses, low-pass filter, small cells, NR etc all suck detail and DR away.
Edmund
[a href=\"index.php?act=findpost&pid=201685\"][{POST_SNAPBACK}][/a]
OK, Edmund, I stand corrected. But I'd still like to see the actual comparison you are referring to. Are we talking about same pixel pitch, pixel to pixel, or different pixel pitch, pixel to pixel?
I would not argue that a P21 or P25 pixel has greater dynamic range than a 1Ds3 pixel. The pixels are bigger than the 1Ds3 microlens. Bigger wells can hold more electrons. A single P25 image should have greater DR than a single 1Ds3 image.
But does half a P45+ image have greater DR than a full 1Ds3 image (same FoV), and if so, is it significantly greater?
We should also bear in mind, when DR is an issue with a stationary subject that lends itself to the stitching procedure, it's easy to increase DR either by varying exposure from frame to frame, or as Bernard suggests, bracketing the exposure of each frame. Not only does PTgui merge to HDR and stitch in the one procedure, but also Autopano Pro.
Also, when taking shots for stitching, the resulting image will usually exceed the size of any single shot from a DB. The additional file size will likely more than compensate for that slight extra crispness that is charcteristic of the sensor without AA filter.
-
it's not the sensor size that matters, but the size of the photosites within the sensor. MFDBs have bigger photosites and thus need a bigger area to contain whatever number of them. This is what gives them the DR advantage, not the overall dimensions of the sensor
This is common place; the D3 vs 1DsMkIII demonstrates this. However, now the MFDBs too have smaller sensels, so this is not the only, not even the main consideration.
The "secret" of the higher DR of MFDBs is in the larger bit depth and the different way the raw data is constructed.
Imagine following:
- make a shot with the 1DsMkIII
- preserve the state of the entire sensor
- create a raw file with ISO 200, 13bit
- create another raw file with ISO 400, 13bit
- create one more raw file with ISO 800, 13bit
- and one more, with ISO 800, 13bit
Now integrate the data in one image file with 16bit depth.
Got it?
Of course, the extra steps are not required: the MFDB creates a single raw file with ISO 800, 16bits - and it contains the entire dynamic range of the several shots with different ISO.
Note, that the P45+ is somewhat different from most other MFDBs in this relation.
-
Let me be clearer: If the 1Ds3 had the Kodak chip in it it would be a much better studio and landscape camera. The Canon's chip is an unsharp compromise design. As it stands, a center crop of 24x36mm out of the P45+ will beat the full frame of the 1Ds3 in file quality and detail. So anyone staring from a Canon for stitching is already at a disadvantage.
Edmund
But does half a P45+ image have greater DR than a full 1Ds3 image (same FoV), and if so, is it significantly greater?
[a href=\"index.php?act=findpost&pid=201697\"][{POST_SNAPBACK}][/a]
-
For 35mm stitching applications I would recommend Fuji S5 at base ISO. The dynamic range of this camera is on par wih MFDB. However, the crispness in S5 files is absent. Stitching many S5 RAF files will allow to create a print comparable to one from MFDB in terms of DR and smoothness of tones. Loss of sharpness due to interpolations in S5 RAW files reconstruction will need to be compensated with higher pixel count. "3D look" would be achieved to some degree in this case with proper aperture setting.
I have made many panoramas with S3 camera, capturing images in JPG format (as buffer limitations do not allow more than 3 raw files). Dynamic range captured in these JPG files rivals most DSLR's RAW formats. Processing images further in 32 and 16 bit allowed me to create panoramas in 200-600 megapixels. I carry all printing at 720 DPI, manually interpolating files to the output resolution. having greater pixel count, panoramas made from Fuji S3 jpgs do not require as much enlargement as single shot ZD. The print quality can be made extremely close.
Of course now that I stitch ZD images in the same procedure that applied to Fuji S3, ZD stitches win hands down!
-
What this often comes down to for me is, look in publication - any publication - billboards, magazines, point of purchase display, etc. - and tell me which camera was used for which image. For the most part, it's impossible to tell.
You can look in National Geographic and see images which look like they have amazing dynamic range, , colors, saturation and then you see that the story was shot by some dude who uses Canon 5Ds exclusively. Or it was shot on a Nikon D2x, which has pretty shitty dynamic range, noise issues, etc.
And then you walk into CVS and see some 2 foot beauty shots in the makeup aisle, and they look fuzzy and low resolution, even though they were shot with on a 39 mp back, or on 4x5 film. And you see Annie Leibowitz shooting ads for Disney, American Express, etc. with her Canons, *with zoom lenses attached.*
So I tell myself, while it can be fun to come on these forums and debate until we are all blue in the face, I see the biggest photographers in the world grabbing which ever camera they want to use that day and they shoot the job, take home huge pay checks, and spend no time on web sites like LL defending their decisions. Does anyone out in the real world know which cameras were used? No. Do they care? No. Could they have gotten the job done with 90% of the pro cameras available today? Yes. (I understand showing up with certain equipment in order to impress art directors & retouchers, but that's another story...and if you are shooting for galleries and making The World Largest Prints, then that might be another story too)
Of course these forum arguments are never going to go away. The only way to get beyond them is, like my heroes, you step away from the forums and focus on getting your work done, and stop to smell the flowers when you can.
-
Of course these forum arguments are never going to go away. The only way to get beyond them is, like my heroes, you step away from the forums and focus on getting your work done, and stop to smell the flowers when you can.
[a href=\"index.php?act=findpost&pid=201739\"][{POST_SNAPBACK}][/a]
I agree with some of what you say, especially in the pixel peeping type of arguments. They border on insanity and never seem to go away. Same with the brand war discussions, they seem to come from people with a vested interest in a specific brand, or in plain terms, they have an interest to sell you something.
Still, I do disagree that only good photographers stay away from forums or blogs like this.
Actually some photographers do share and they are on a lot of forums like aphotoeditor.com.
I've been a member of almost every photographers' organization on the planet and though I'm not much of a club joiner, I find the people that share usually get ahead because just the process of responding to someone's question will force you to analyze why and what you personally do.
I also find that all groups get splintered into different factions, some of it is everybody smacking each other on the back and saying well done, (and there is nothing wrong with positive reinforcement), but the other side is the small percentage that join just for personal gain.
The secret is learning how to filter out the good from the bad the selfish from the giving.
JR
-
Still, I do disagree that only good photographers stay away from forums or blogs like this.
Actually some photographers do share and they are on a lot of forums like aphotoeditor.com.
Hi James, I did not mean to imply that good photographers don’t blog. After all, I’m here ;-) When I said my heroes, I was referring to some specific photographers who are in fact, my heroes, and whom I have never seen on any forum to date (though they may be on some private ones…)
The secret is learning how to filter out the good from the bad the selfish from the giving.
JR
[a href=\"index.php?act=findpost&pid=201746\"][{POST_SNAPBACK}][/a]
Indeed.
-
What this often comes down to for me is, look in publication - any publication - billboards, magazines....
[a href=\"index.php?act=findpost&pid=201739\"][{POST_SNAPBACK}][/a]
203,
I appreciate your healthy skepticism and I agree with a lot of what you have said.
However, the reason I started this discussion was not because I am creating images for magazines. MF is vast overkill for that.
I am selling fine art prints that are very large. Now that I have an Epson 11880 (which can print 64-inches wide), I am offering 5 x 8-foot prints to my commercial clients.
When I sell something that size, I want to be sure that I get the best quality. I'm not talking about pixel peeping here, I'm talking about the general viewing public (my clients). I need to make sure that with these huge prints that I am offering the best that is practical.
That is why I am seriously considering going to Medium Format (or stitching 35mm).
-
203,
I appreciate your healthy skepticism and I agree with a lot of what you have said.
However, the reason I started this discussion was not because I am creating images for magazines...I am selling fine art prints that are very large. Now that I have an Epson 11880 (which can print 64-inches wide)...
[a href=\"index.php?act=findpost&pid=201751\"][{POST_SNAPBACK}][/a]
hdomke, I understand what you are asking about. (which is why I included the disclaimer: "and if you are shooting for galleries and making The World Largest Prints, then that might be another story too"
My dissertation above was in reference to the underlying discussion which has broken out here, which, it seems, (unfortunately) has little to do with your original question, and everything to do with fruitless pixel peeping.
Anyway I will stop distracting, and let you get back to the business of getting some answers ;-)
-
That is why I am seriously considering going to Medium Format (or stitching 35mm).
[a href=\"index.php?act=findpost&pid=201751\"][{POST_SNAPBACK}][/a]
Henry
I can quite categorically state that 5x8 foot images from a P45 or H3D2 will be visually better than from 3 stitched 1DsM3 images. It's subtle, but it's there in the shadows if you look for it and also in the overall accutance of the image.
But...
Will your clients see the difference? Even in side by side tests it's hard to see.
Does your business model enable you to spend the big dollars - i.e. will 1% extra quality earn you more money?
Better yet, shoot film. Seriously, unless you stitch MFDB images 6x12 roll film will produce a superiour print at that size. I know because I too make 4x8 foot prints and bigger, and over enlarged film always looks more satisfying to the eye than overenlarged digital files. You need a 23000px wide image to make an 8' print at 240dpi so even an MFDB will need a lot of interpolation. Film enlarges much better past native resolution limit. IMHO at least...
Also, why not stitch MFDB images for even greater quality? if your goal is the ultimate quality prints then this will get you there.
Ultimately only you can work out if the slight quality gain is worth it - but I assure you it is there. I did have some test images shot on a 1DsM3 and a H2D3 side by side but I can't lay my hands on them for some reason. Anyway, the Hassy did have better shadow detail - you'll just have to take my word for it.
-
Will your clients see the difference? Even in side by side tests it's hard to see.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=201803\")
Nick,
Thanks for the comment. You are right. I'm confident that my clients WON"T see the difference.
My business model would allow me to get the MF gear if there were any chance of a return on the investment. From what I can tell so far, that just won't happen.
Regarding your suggestion to shoot film; not in a million years. If I had to go back to film I would stop using a camera! I don't agree with you that "Film enlarges much better". I can always tell scanned film; it's the grain. Those great big chunks of grain. I love smooth, clean digital files.
Incidentally, I greatly enjoyed your recent review on this website: [a href=\"http://www.luminous-landscape.com/reviews/cameras/h3d-review.shtml]Hasselblad H3D On Location[/url].
-
I don't agree with you that "Film enlarges much better". I can always tell scanned film; it's the grain. Those great big chunks of grain. I love smooth, clean digital files.
[a href=\"index.php?act=findpost&pid=201807\"][{POST_SNAPBACK}][/a]
I reckon grain is more attractive ( or maybe less unattractive) than the plastic look of a digital file when enlarged too much. It's a value judgement I know, but overenlarged digi files just look awful. Given the choice of two less than optimal 'looks', I'd go with film every time.
However, print within the file's resolution limits and, yes, I'd never shoot film either.
-
Also, why not stitch MFDB images for even greater quality? if your goal is the ultimate quality prints then this will get you there.
[a href=\"index.php?act=findpost&pid=201803\"][{POST_SNAPBACK}][/a]
The main reasons are weight, bulk and above all shutter induced vibrations. The gap in terms of detail between the ZD and the D3 is much smaller once the ZD is mounted on a pano head...
I personnally feel that the Hassy H is the only credible platform for MFDB stitching thanks to their very stable leaf shutter lenses. I have not tried the latest Mamiya III body, but my ZD is not very satisfactory from that standpoint.
Besides, the overall slowness and shallow buffer of MFDB equipment is also often a problem when stitching dynamic subjects like a cloudy sky. The D3 is just much better at shooting many frames in a row.
I am not even going to discuss wide angle shooting since the quality of the wide Nikkors is clearly better than anything in MF world except the Schneiders and Roidenstock digital lenses. The only lens that might be close is the Hassy 28mm, but then again you end up being forced to buy a H3D.
My personnal view is that the Nikon D3x might become the best stitching camera, but we'll have to check the DR. If it is reasonnably close to the D3, then I'll give up my plans to go invest more in MF...
Cheers,
Bernard
-
I reckon grain is more attractive ( or maybe less unattractive) than the plastic look of a digital file when enlarged too much. It's a value judgement I know, but overenlarged digi files just look awful. Given the choice of two less than optimal 'looks', I'd go with film every time.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=201809\")
You are probably familiar with this, but I think that this is where dedicated plug ins like Genuine fractal do a better job since they manage to preserve a little better the acutance of edges after enlargement.
Another way to mitigate the softening of digital files after very large extrapolations is to add some form of film grain. By playing with the scale of the grain relative to the scale of the details I feel that one can convey a better impression of sharpeness. I am only starting to experience with this, but it seems promising.
I am sure that you know these techniques better than I do, but just in case this simple PS tutorial provides a good overview:
[a href=\"http://www.i-marco.nl/weblog/archive/2007/09/22/realistic_film_grain_for_digit]http://www.i-marco.nl/weblog/archive/2007/...grain_for_digit[/url]
Among the various plug-ins that are now showing up, this one seems to have good potential but more investigation is needed:
http://www.imagenomic.com/rg.aspx (http://www.imagenomic.com/rg.aspx)
DxO film pack is another candidate, but I have not tried it on very large files yet.
Regards,
Bernard
-
I reckon grain is more attractive ( or maybe less unattractive) than the plastic look of a digital file when enlarged too much. It's a value judgement I know, but overenlarged digi files just look awful. Given the choice of two less than optimal 'looks', I'd go with film every time.
However, print within the file's resolution limits and, yes, I'd never shoot film either.
[a href=\"index.php?act=findpost&pid=201809\"][{POST_SNAPBACK}][/a]
Nick,
I am running tests right now on a series of stitched MFDB files - landscape work.
We are finding that at 5m the MFDB quality is 'better' than the 8 x 10 at 1.5m. There is none of the 'plasticity' I had feared (and you mentioned) and all involved have been nothing short of awed at the quality. The transitions and detail rendition are beautiful.
The place where film takes over, however, is in long exposure land. As soon as long exposure noise from the chip appears - the interpolation turns the file to mush...
Murray
-
The main reasons are weight, bulk and above all shutter induced vibrations. The gap in terms of detail between the ZD and the D3 is much smaller once the ZD is mounted on a pano head...
I personnally feel that the Hassy H is the only credible platform for MFDB stitching thanks to their very stable leaf shutter lenses. I have not tried the latest Mamiya III body, but my ZD is not very satisfactory from that standpoint.
Besides, the overall slowness and shallow buffer of MFDB equipment is also often a problem when stitching dynamic subjects like a cloudy sky. The D3 is just much better at shooting many frames in a row.
I am not even going to discuss wide angle shooting since the quality of the wide Nikkors is clearly better than anything in MF world except the Schneiders and Roidenstock digital lenses. The only lens that might be close is the Hassy 28mm, but then again you end up being forced to buy a H3D.
[a href=\"index.php?act=findpost&pid=201811\"][{POST_SNAPBACK}][/a]
Bernard,
I am stitching with an Alpa, Digitar lenses and a Sinar back. There are NO shutter vibrations and the lack of distortion in the lenses really aids the stitching process when you re dealing with straight lines. I am shooting 24mm, 35mm (the 24mm is more problematic) and others - so very, very wide.
The results are incredible....
The problem is the slow buffer, and clouds do move, but 'smartbelnd' features in the programs seem to help...
Cheers
Murray
-
Bernard,
I am stitching with an Alpa, Digitar lenses and a Sinar back. There are NO shutter vibrations and the lack of distortion in the lenses really aids the stitching process when you re dealing with straight lines. I am shooting 24mm, 35mm (the 24mm is more problematic) and others - so very, very wide.
[a href=\"index.php?act=findpost&pid=201816\"][{POST_SNAPBACK}][/a]
Yes, I would expect this indeed. Many people try to do the same with medium format cameras though.
Regards,
Bernard
-
Let me be clearer: If the 1Ds3 had the Kodak chip in it it would be a much better studio and landscape camera. The Canon's chip is an unsharp compromise design. As it stands, a center crop of 24x36mm out of the P45+ will beat the full frame of the 1Ds3 in file quality and detail. So anyone staring from a Canon for stitching is already at a disadvantage.
Edmund
[a href=\"index.php?act=findpost&pid=201718\"][{POST_SNAPBACK}][/a]
That is interesting, Edmund. Pity most of us never get a chance to see such comparisons. You seem to be implying that a P21 would still provide a better quality image than the 1Dsd3 even if it were the same size sensor as the 1Ds3 (cropped to the P21 aspect ratio).
One wonders why some people who attempt 1Ds3/P21 or ZD comparisons find it necessary to put the 1Ds3 at a disadvantage by using the same F stop with both cameras.
-
Nemo,
Thanks for the link to the article. It looks very impressive! However, it is a bit overwhelming for someone who is not an academic deeply versed in statistics. Perhaps you (or someone else) could summarize the take-home points from this article. After skimming the article, here are some specific questions:
1. In the real world, will prints from a 16-bit capture show more dynamic range than from a 14-bit capture?
2. What are the implications of this paper on "Expose to the Right"?
3. What are the implications of this paper on selecting the right ISO in low-light situations?
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=201687\")
I contacted Dr. Emil Martinec about his fascinating paper: [a href=\"http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html]Noise, Dynamic Range and Bit Depth in Digital SLRs[/url] to see if he would answer the questions. He was kind enough to email this reply to me:
1. In the real world, will prints from a 16-bit capture show more dynamic range than from a 14-bit capture?
I know of no DSLR that would benefit from 16-bit capture. There are some MFDB's that use 16-bit capture, and I have read that they use the extra bits not because of increased dynamic range, but rather to do ISO settings in software after capture rather than via amplification in hardware during capture.
The problem is that current DSLR's have dynamic range that is limited by noise in the camera electronics. The best-in-class, the Nikon D3, is one the verge of meriting more than 12 bits, but certainly not more than 14.
2. What are the implications of this paper on "Expose to the Right"?
Expose to the right is a good mantra for better image quality (assuming it is done with proper attention paid to preserving those highlights one wants to preserve). It is shorthand for "higher signal with less noise is better", or even more succinctly, "the more light, the better". The point is that the more light the camera captures, the less apparent the effects of noise are going to be. For fixed ISO, this means that slower shutter speed/wider aperture allow more light to be captured, and this pushes the histogram to the right. If shutter speed/aperture are determined (say by DOF considerations, or freezing motion), then there is an advantage (mostly at low ISO, and especially on Canon DSLR's) to using higher ISO to push the histogram to the right, but the effect is less pronounced than using shutter speed/aperture to push the histogram to the right at fixed ISO.
3. What are the implications of this paper on selecting the right ISO in low-light situations?
I would say that for most DSLR's and raw capture, there is no point in going above ISO 1600; there is little difference between underexposed ISO 1600 (with exposure compensation during raw conversion) vs. proper exposure at higher ISO, assuming the raw converter being used properly implements exposure compensation (sadly, not all do). There is almost no penalty as far as noise is concerned. And a benefit is increased highlight headroom.
cheers, emil
Note: Dr. Martinec is a Physics Professor at the University of Chicago.
-
I contacted Dr. Emil Martinec about his fascinating paper: Noise, Dynamic Range and Bit Depth in Digital SLRs (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html) to see if he would answer the questions. He was kind enough to email this reply to me:
1. In the real world, will prints from a 16-bit capture show more dynamic range than from a 14-bit capture?
I know of no DSLR that would benefit from 16-bit capture. There are some MFDB's that use 16-bit capture, and I have read that they use the extra bits not because of increased dynamic range, but rather to do ISO settings in software after capture rather than via amplification in hardware during capture.
The problem is that current DSLR's have dynamic range that is limited by noise in the camera electronics. The best-in-class, the Nikon D3, is one the verge of meriting more than 12 bits, but certainly not more than 14.
2. What are the implications of this paper on "Expose to the Right"?
Expose to the right is a good mantra for better image quality (assuming it is done with proper attention paid to preserving those highlights one wants to preserve). It is shorthand for "higher signal with less noise is better", or even more succinctly, "the more light, the better". The point is that the more light the camera captures, the less apparent the effects of noise are going to be. For fixed ISO, this means that slower shutter speed/wider aperture allow more light to be captured, and this pushes the histogram to the right. If shutter speed/aperture are determined (say by DOF considerations, or freezing motion), then there is an advantage (mostly at low ISO, and especially on Canon DSLR's) to using higher ISO to push the histogram to the right, but the effect is less pronounced than using shutter speed/aperture to push the histogram to the right at fixed ISO.
3. What are the implications of this paper on selecting the right ISO in low-light situations?
I would say that for most DSLR's and raw capture, there is no point in going above ISO 1600; there is little difference between underexposed ISO 1600 (with exposure compensation during raw conversion) vs. proper exposure at higher ISO, assuming the raw converter being used properly implements exposure compensation (sadly, not all do). There is almost no penalty as far as noise is concerned. And a benefit is increased highlight headroom.
cheers, emil
Note: Dr. Martinec is a Physics Professor at the University of Chicago.
[a href=\"index.php?act=findpost&pid=201871\"][{POST_SNAPBACK}][/a]
Succinct answers indeed - all of this is in the paper of course, but it's good to have it summarised.
-
With straight and clear answers like these I guess Dr. M won't have a career in politics.
I'm going to tape over the ISO button on my cameras
Edmund
I contacted Dr. Emil Martinec about his fascinating paper:
Note: Dr. Martinec is a Physics Professor at the University of Chicago.
[a href=\"index.php?act=findpost&pid=201871\"][{POST_SNAPBACK}][/a]
-
I know of no DSLR that would benefit from 16-bit capture. There are some MFDB's that use 16-bit capture, and I have read that they use the extra bits not because of increased dynamic range, but rather to do ISO settings in software after capture rather than via amplification in hardware during capture
I strongly disagree with that. The difference between DSLRs and MFDBs is not primarily in the sensor technology but in the approach of presenting the data.
I described in #84 the relevant difference between these approaches. I am convinced, that Canon at al could achieve MFDB-like dynamic range any time with 16bit raw data.
-
I strongly disagree with that. The difference between DSLRs and MFDBs is not primarily in the sensor technology but in the approach of presenting the data.
I described in #84 the relevant difference between these approaches. I am convinced, that Canon at al could achieve MFDB-like dynamic range any time with 16bit raw data.
[a href=\"index.php?act=findpost&pid=201893\"][{POST_SNAPBACK}][/a]
Gabor,
Could you please explain on what grounds you disagree with Dr. Martinec regarding bit depth?
Dr. Martinec has presented a very careful analysis of his reasoning in the paper. Do you see a flaw with his analysis?
-
Gabor,
Could you please explain on what grounds you disagree with Dr. Martinec regarding bit depth?
Dr. Martinec has presented a very careful analysis of his reasoning in the paper. Do you see a flaw with his analysis?
[a href=\"index.php?act=findpost&pid=201898\"][{POST_SNAPBACK}][/a]
Mmmm... has he actually ever picked up a camera and taken a photograph though?
You can show me all the physics, maths, voodoo papers in the world.
hdomke, you started this thread asking for advice from Medium format owners, and anything they have offered you seem to reject? Therefore why ask the question.
A practicing photographer is more likely to give you the answers based on their experience as opposed to a science paper.
Good grief... lets go round again.
Jo S.x
-
Could you please explain on what grounds you disagree with Dr. Martinec regarding bit depth?
This is obviously somewhat "shocking". I tried to explain it in my #84; now I try a different way.
Let's take the Canon 40D as an example, based on DPReview's measurement (the 1DsMkII has not been reviewed yet). I don't agree with their result in general, they are JPEG based, which does not have much to do with the camera's capablity, but their result is suitable for the explanation.
1. The 40D loses exactly one stop from the highlights by one stop ISO increase from ISO 200 to ISO 1600. This does not appear that way in DPReview, because the middle gray moves as well. However, the fact remains, that ISO 400 cuts off the very top stop captured by the sensors.
2. DPR measured a loss of *total* dynamic range as 0.3EV from 200 to 400, 0.1EV from 400 to 800 and 0.1EV GAIN (this is ridiculous) from 800 to 1600. Together, the loss from 200 to 1600 is only 0.3EV. I am "generous", I assume 1 full stop loss.
3. The loss in the highlights zwischen 200 and 1600 is *three* stops, but the total loss is only *one* stop. This means, that the sensor can deliver all together *two* stops higher DR than it is doing at ISO 200.
4. Why does the 40D NOT deliver this higher DR? Because it delivers always a *crop* of the total dynamic range. With ISO200 it cuts off two stops from the shadows, with ISO1600 it cuts off three stops of the highlights.
This is, what most MFDBs are doing differently: they deliver everything in one - but that requires the 16bits.
If the 40D converted the ISO1600 data in 13 bits and added three bits to keep the highlights, its dynamic range would be two stops larger and I would not have to throw away some of my shots I just made in the Rocky Mountains, because the 40D could not capture the scenery even with the best possible exposure.
-
you started this thread asking for advice from Medium format owners, and anything they have offered you seem to reject? Therefore why ask the question.
[a href=\"index.php?act=findpost&pid=201902\"][{POST_SNAPBACK}][/a]
Jo S.x,
I am not rejecting the advice I am getting here! Actually, the advice I've been getting is quite diverse and thought provoking. There are a lot of conflicting opinions. I'm trying to sort it out.
The ideas on bit-depth by Dr. Martinec seem to be very carefully thought out. Even though I (like you) prefer real world experience to a theoretical explanation, I can't easily dismiss what Dr. Martinec says. It is very specific.
-
I use Nikon (D300) and MFDB (Hasselblad). In single shot there is no match between the DSLR and my MFDB. The MFDB generates much nicer images and is able to cope with certain situations much better than the DSLR (situations with a lot of contrast).
I can elaborate on how much they are better, sharper, etc.. but don't feel the real need because some of these things (maybe even most) are subjective in some way.
If I can I take the MFDB. If I cannot I use the Nikon (bad light, travelling light, fast moving stuff, plain lazy, etc..).
Both can be used to stitch. When stitching is possible in most cases you can also stitch with the MFDB in which case it generates a much nicer image again than a stitched DSLR file.
I don't theorize about whether a DSLR could have a DR as a MFDB if.... I just look at the files I get from my cameras.
Both machines have their uses & purpose. If you really want to know whether you should get a MFDB try a couple with various bodies/configurations to get some idea what you are facing.
-
Nick,
The place where film takes over, however, is in long exposure land. As soon as long exposure noise from the chip appears - the interpolation turns the file to mush...
Murray
[a href=\"index.php?act=findpost&pid=201813\"][{POST_SNAPBACK}][/a]
You're kidding about this right? Astrophotographers gave up film years before we did for digital sensors because of long exposure needs. There is no comparison to the detail in a digital file to a film file with long exposures. I shoot at ISO 1600 with my 1Ds and it looks better than anything shot on any 400 ASA film stock I ever used. Forget about any faster film. And the reciprocity and color shifts with film is absurd.
-
I recently did a side-by-side comparison of the Hasselblad H3D2 with 39 MP back and my Canon 1DsMk3.
I took each file of the exact same shot and enlarged them to be 10-feet wide.
I then printed a 40-inch wide piece from the center of each images and compared them.
The Hasselblad was clearly sharper when viewed from 10-inches away. From 2 feet or further I could see no difference.
Still, I would like that kind of quality. Is it reasonable to expect that I could get that by stitching three of my Canon files into one? (see attached picture).
Shouldn't they be equivalent?
Any thoughts or suggestions are welcome.
[attachment=7030:attachment]
[a href=\"index.php?act=findpost&pid=201341\"][{POST_SNAPBACK}][/a]
The weakest link here is the optics. Try using a Leica lens on the 1Ds for a better test and you may find much less difference. As good as Canon L lenses are, they are not good enough for the new sensors yet. The nicest feature of our Canon's is that we can use many different brand lenses for work that demands the finest detail.
-
This is obviously somewhat "shocking". I tried to explain it in my #84; now I try a different way.
Let's take the Canon 40D as an example, based on DPReview's measurement (the 1DsMkII has not been reviewed yet). I don't agree with their result in general, they are JPEG based, which does not have much to do with the camera's capablity, but their result is suitable for the explanation.
1. The 40D loses exactly one stop from the highlights by one stop ISO increase from ISO 200 to ISO 1600. This does not appear that way in DPReview, because the middle gray moves as well. However, the fact remains, that ISO 400 cuts off the very top stop captured by the sensors.
2. DPR measured a loss of *total* dynamic range as 0.3EV from 200 to 400, 0.1EV from 400 to 800 and 0.1EV GAIN (this is ridiculous) from 800 to 1600. Together, the loss from 200 to 1600 is only 0.3EV. I am "generous", I assume 1 full stop loss.
Point 1 is correct. To be more precise on point 2, engineering DR (raw saturation/read noise) on the 40D is 11.3 stops at ISO 100, 11.3 at ISO 200, 10.9 at ISO 400, 10.2 at ISO 800 and 9.3 at ISO 1600. Are the DPR numbers you are quoting based on measurements of raw or jpeg? They do both; the jpeg numbers are rather devoid of content. The raw numbers are the result of a somewhat flawed procedure. Some people prefer a notion of "useful" DR which is somewhat more subjective, but strongly correlated to, engineering DR.
3. The loss in the highlights zwischen 200 and 1600 is *three* stops, but the total loss is only *one* stop. This means, that the sensor can deliver all together *two* stops higher DR than it is doing at ISO 200.
Loss of engineering DR is two stops, compared to exposure change of three stops between ISO 200 and 1600.
4. Why does the 40D NOT deliver this higher DR? Because it delivers always a *crop* of the total dynamic range. With ISO200 it cuts off two stops from the shadows, with ISO1600 it cuts off three stops of the highlights.
This is a misunderstanding of the capture process. Canon is not "throwing away" DR by not recording it (subject to a caveat). ISO 200 doesn't have more DR in shadows because the sensor readout is limited by the noise of the variable gain amplifier that implements the ISO setting. At high ISO this is a rather minor component of read noise, but is the dominant effect at low ISO.
The caveat is that, because sensor readout is a non-destructive procedure in CMOS technology, one could preserve the state of the sensor while doing a read of the capture data at low ISO amplification, then do another read of the same data using high ISO amplification, and finally combine the two reads in an in-camera version of HDR blending to get at the full DR delivered by the sensels. This would combine the total DR in the way you are imagining, but it would cost a factor of two in buffer capacity and be a big drag on processing speed. Not out of the question for the near future though.
This is, what most MFDBs are doing differently: they deliver everything in one - but that requires the 16bits.
No. They simply record more bits, which does not in and of itself confer more DR. They do not garner more DR due to that; even the caveat above is not available, since a CCD sensor cannot be read non-destructively.
To the extent that the sensels are gathering more photons, there may be an improvement in quality over a larger portion of the DR for MFDB's over the 40D, since the S/N ratio climbs out of being dominated by read noise faster for larger photosites (this is at the pixel level; smaller pixels need not be under such a disadvantage when compared on a fixed spatial scale instead of the pixel scale).
If the 40D converted the ISO1600 data in 13 bits and added three bits to keep the highlights, its dynamic range would be two stops larger and I would not have to throw away some of my shots I just made in the Rocky Mountains, because the 40D could not capture the scenery even with the best possible exposure.
[a href=\"index.php?act=findpost&pid=201903\"][{POST_SNAPBACK}][/a]
Again, this is a misunderstanding of how the circuitry that processes the sensor data works. It's not simply a question of adding more bits to record overamplified highlights, camera company engineers are not so naive. The ISO gain amplifier and the ADC are matched to provide as much DR as they are capable of.
Emil Martinec
-
Mmmm... has he actually ever picked up a camera and taken a photograph though?
Yes, I have in fact. That doesn't prevent me from having an interest in the technical design aspects of the equipment I use. The two are not mutually exclusive. In fact, a quantitative understanding of the properties of one's camera leads to better informed decisions about exposure.
You can show me all the physics, maths, voodoo papers in the world.
Apparently it wouldn't help you as much as it does those who understand that math and physics are not voodoo.
A practicing photographer is more likely to give you the answers based on their experience as opposed to a science paper.
[a href=\"index.php?act=findpost&pid=201902\"][{POST_SNAPBACK}][/a]
Practical experience and quantitative testing both have useful things to say, especially in an area such as digital photography that combines both artful craft and sophisticated technology.
-
I am convinced, that Canon at al could achieve MFDB-like dynamic range any time with 16bit raw data.
[a href=\"index.php?act=findpost&pid=201893\"][{POST_SNAPBACK}][/a]
And they do not do this because ???
Serious question BTW.
-
And they do not do this because ???
Serious question BTW.
[a href=\"index.php?act=findpost&pid=201927\"][{POST_SNAPBACK}][/a]
I have read they do not do this because making the file 16-bit would make the RAW files even larger (and slower to process), with no real-world improvement in image quality.
-
I recently did a side-by-side comparison of the Hasselblad H3D2 with 39 MP back and my Canon 1DsMk3.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=201341\")
The following is from an email friend of mine, a professional nature photographer [a href=\"http://www.peterhaigh.com/]Peter Haigh[/url]. He is a self-described Luddite (http://en.wikipedia.org/wiki/Luddite), so he asked me to post his responses to the discussions on this forum. Most of his nature photographs are taken with medium format film. He had three areas that caught his eye:
1. Greater 3D effect
... some chap talked about the greater 3D effect he got from medium format digital capture. That really caught my eye because what has stood out to me and been one of the main factors in not going Canon is the images have always seemed so ‘flat.’ Now, obviously, a photo is 2D but I’ve always thought film gave the sense of greater depth than the Canon. Film and Cibachrome prints had a great sense of ‘depth’ if you will — a fullness. Although a good chunk of that goes away upon scanning and printing on the Epson, I still feel there is more left than from 35mm digital capture — and this is one of the reasons I have remained a holdout. (I might add I have experienced the same flatness with CDs and that is one reason I prefer vinyl which is ‘fuller’ in its presentation. So, the Hassie might bring back some of that depth.
2. Greater Bit-depth
A second factor that kept me out of 35mm digital capture was that previously the Canon (and all others I know of) were 12 bit capture — your new Canon is apparently 14 bit capture. My older Nikon scanner was 14 bit and I had, at times, some problems with banding in areas of very smooth gradation of tonality especially when doing some substantial adjustment in Photoshop. My new scanner is 16 bit and I haven’t run into that problem. It was only rarely that it happened but one hates to be limited in what they can do with the image.
3. Dynamic Range
The third factor brought up in the postings was dynamic range. I looked up the specs for both cameras and couldn’t find the dynamic range, but greater dynamic range is always a benefit — mainly in the shadows. The greater the dynamic range the less shadows tend to block up and, to the extent that they do, the less noisy they get when you apply Shadows/Highlights — mainly because you don’t have to apply it as strongly.
All in all, and based solely on paper as opposed to comparing output, I would think the Hassie would probably give you better images AND save some time and energy with stitching. (Although expensive, I’m confident it wouldn’t break your bank.) The above comments are based on my general experience (mostly subjective I might add) . So, at most you get what you pay for and I ain’t seen no check in the mail yet.
-
Loss of engineering DR is two stops, compared to exposure change of three stops between ISO 200 and 1600
I don't give a fig for the engineering definition of DR. I am using a camera for creating images.
This is a misunderstanding of the capture process. Canon is not "throwing away" DR by not recording it (subject to a caveat).
I'm afraid you misunderstood what I posted. "Canon" is throwing away DR, because they do not record all the possible data, which is in the sensor.
ISO 200 doesn't have more DR in shadows because the sensor readout is limited by the noise of the variable gain amplifier that implements the ISO setting. At high ISO this is a rather minor component of read noise, but is the dominant effect at low ISO
There is no law requiring, that only a 1/16th filled well can be read and converted with higher amplification. What about converting exposing to the right measured for base ISO and converting the now higher filled wells with high amplification?
Let's make it clear:
the only factual (i.e. not business) consideration against this is the amount of data and the processing time
Most DSLRs are designed for photographers, who want to shoot 3, 6, 10 frames per second. Landscapers are paying the prise for this slopsided approach.
No. They simply record more bits, which does not in and of itself confer more DR. They do not garner more DR due to that; even the caveat above is not available, since a CCD sensor cannot be read non-destructively
I base my opinion on actual analysis of actual MFDB images.
-
with no real-world improvement in image quality.
Well, my real word would improve a lot with higher dynamic range.
-
Well, my real word would improve a lot with higher dynamic range.
[a href=\"index.php?act=findpost&pid=201937\"][{POST_SNAPBACK}][/a]
It has been written many times that those last two bits (between 14 and 16) do not hold any image data which directly translates to an increase in dynamic range.
I believe that MF has better D.R., but it is said that this improvement in D.R. comes from data which is contained within the first 14 bits...
-
At some point you need to write the details down in equations what is being said here is being buried in the vagueness of the english language.In particular, I don't undrestand what Jpeg is doing in this discussion, the quantisation errors alone are spectacular and everything is crusged to 8 bits.
Edmund
This is obviously somewhat "shocking". I tried to explain it in my #84; now I try a different way.
Let's take the Canon 40D as an example, based on DPReview's measurement (the 1DsMkII has not been reviewed yet). I don't agree with their result in general, they are JPEG based, which does not have much to do with the camera's capablity, but their result is suitable for the explanation.
1. The 40D loses exactly one stop from the highlights by one stop ISO increase from ISO 200 to ISO 1600. This does not appear that way in DPReview, because the middle gray moves as well. However, the fact remains, that ISO 400 cuts off the very top stop captured by the sensors.
2. DPR measured a loss of *total* dynamic range as 0.3EV from 200 to 400, 0.1EV from 400 to 800 and 0.1EV GAIN (this is ridiculous) from 800 to 1600. Together, the loss from 200 to 1600 is only 0.3EV. I am "generous", I assume 1 full stop loss.
3. The loss in the highlights zwischen 200 and 1600 is *three* stops, but the total loss is only *one* stop. This means, that the sensor can deliver all together *two* stops higher DR than it is doing at ISO 200.
4. Why does the 40D NOT deliver this higher DR? Because it delivers always a *crop* of the total dynamic range. With ISO200 it cuts off two stops from the shadows, with ISO1600 it cuts off three stops of the highlights.
This is, what most MFDBs are doing differently: they deliver everything in one - but that requires the 16bits.
If the 40D converted the ISO1600 data in 13 bits and added three bits to keep the highlights, its dynamic range would be two stops larger and I would not have to throw away some of my shots I just made in the Rocky Mountains, because the 40D could not capture the scenery even with the best possible exposure.
[a href=\"index.php?act=findpost&pid=201903\"][{POST_SNAPBACK}][/a]
-
It has been written many times that those last two bits (between 14 and 16) do not hold any image data which directly translates to an increase in dynamic range
AFAIK only MFDBs create 16bit raw files currently, so "it" has not been written so times.
Anyway, the point is not, if 14 or 16 bits are enough, but how those bits are used. If one can start out with 13 bits and increases that with three bits for ISO 1600 (from ISO 200), then one ends up with 16bits. If one starts out with 10bits, then 13 bits are required to include the entire spectrum.
-
Mamiya ZD files are 12 bits and store the same dynamic range as Aputs 22, which alows 16 bit precision.
So let's stop counting bits in terms of dynamic range;)
Bits represent accuracy of data, thats all.
Captured dynamic range is limited by the sensor design.
-
Mamiya ZD files are 12 bits and store the same dynamic range as Aputs 22, which alows 16 bit precision.
So let's stop counting bits in terms of dynamic range;)
Bits represent accuracy of data, thats all.
Captured dynamic range is limited by the sensor design.
[a href=\"index.php?act=findpost&pid=201959\"][{POST_SNAPBACK}][/a]
So what exactly is being made more accurate/precise with the extra bits? As far as I know, it's just more levels/steps within the range of the chip, not more D.R.
Thanks.
-
So what exactly is being made more accurate/precise with the extra bits? As far as I know, it's just more levels/steps within the range of the chip, not more D.R.
Thanks.
[a href=\"index.php?act=findpost&pid=201965\"][{POST_SNAPBACK}][/a]
- all R, G, B intensity values are recorded with greater precision, reflecting minute variations. When converting to lower number of bits, some variations are obscured and ranges of measured values become represented by single numbers instead.
-
Mamiya ZD files are 12 bits and store the same dynamic range as Aputs 22, which alows 16 bit precision
The photographic dynamic range can not be separated from levels.
I don't have any Aptus raw files, but I do some Sinars, which show discernable details in the 11th stop.
I wonder if you have a ZD file showing details in the 11th stop (with wholly TWO levels to choose from?). If you think so, pls upload it.
-
You're kidding about this right? Astrophotographers gave up film years before we did for digital sensors because of long exposure needs. There is no comparison to the detail in a digital file to a film file with long exposures. I shoot at ISO 1600 with my 1Ds and it looks better than anything shot on any 400 ASA film stock I ever used. Forget about any faster film. And the reciprocity and color shifts with film is absurd.
[a href=\"index.php?act=findpost&pid=201918\"][{POST_SNAPBACK}][/a]
No, Not kidding at all!
I shoot with 8" x 10" film, 400iso for long exposure work. The work is exhibited at 1.5m prints, The exposures are often 7-9 hours. The 1Ds (which I also own) cannot do that with quality and no MFDB is able to offer that kind of exposure.
There is no comparison to an 8" x 10" drum scan with a 1Ds file at that time, enlarged and printed that big...
I have been shooting with MFDB and 8" x 10" side by side and comparing scenes, all with highest quality methods for performing huge enlargements for exhibition. As soon as my back exposures increase past 10sec, the noise inherent in the files does not interpolate well...it turns 'mushy' while the film grain holds together.
Under 1 sec, and particularly with stitching, the MFDB is enlarging beautifully with results beyond even low iso 8" x 10".
Murray
-
I don't give a fig for the engineering definition of DR. I am using a camera for creating images.
Well, then, perhaps it's time you paid attention. The engineering definition of DR is actually quite closely related to any other definition using shadow SNR of raw data to determine the noise floor, and determining DR as the ratio of raw saturation to said noise floor.
I'm afraid you misunderstood what I posted. "Canon" is throwing away DR, because they do not record all the possible data, which is in the sensor.
There is no law requiring, that only a 1/16th filled well can be read and converted with higher amplification. What about converting exposing to the right measured for base ISO and converting the now higher filled wells with high amplification?
[a href=\"index.php?act=findpost&pid=201936\"][{POST_SNAPBACK}][/a]
There is a defined voltage corresponding to the maximum digitized output of the ADC. The ISO gain determines what output voltage corresponds to a given number of photons in the sensel. If you increase the ISO, a smaller number of photons yields that saturating voltage input to the ADC. So if a full well yields this saturating voltage at ISO 100, then a 1/16 full well yields this voltage at ISO 1600. There is no process involving a single read of the sensor data that will recover the data in wells more than 1/16 full, if you have set the amplifier gain to ISO 1600. Simple as that. As I wrote previously, Canon engineers are not exceedingly stupid (at least on this point), they do know that ADC's saturate at defined input voltages.
So please explain more precisely how you think that Canon is supposed to extract this data that is in the sensor, in a way consistent with the operating principles of amplifiers and analog-to-digital converters.
-
Well, then, perhaps it's time you paid attention. The engineering definition of DR is actually quite closely related to any other definition using shadow SNR of raw data to determine the noise floor, and determining DR as the ratio of raw saturation to said noise floor.
There is a defined voltage corresponding to the maximum digitized output of the ADC. The ISO gain determines what output voltage corresponds to a given number of photons in the sensel. If you increase the ISO, a smaller number of photons yields that saturating voltage input to the ADC. So if a full well yields this saturating voltage at ISO 100, then a 1/16 full well yields this voltage at ISO 1600. There is no process involving a single read of the sensor data that will recover the data in wells more than 1/16 full, if you have set the amplifier gain to ISO 1600. Simple as that. As I wrote previously, Canon engineers are not exceedingly stupid (at least on this point), they do know that ADC's saturate at defined input voltages.
So please explain more precisely how you think that Canon is supposed to extract this data that is in the sensor, in a way consistent with the operating principles of amplifiers and analog-to-digital converters.
[a href=\"index.php?act=findpost&pid=201991\"][{POST_SNAPBACK}][/a]
You guys get it all worked out yet??
:+}
Snook
-
There is a defined voltage corresponding to the maximum digitized output of the ADC. The ISO gain determines what output voltage corresponds to a given number of photons in the sensel. If you increase the ISO, a smaller number of photons yields that saturating voltage input to the ADC. So if a full well yields this saturating voltage at ISO 100, then a 1/16 full well yields this voltage at ISO 1600. There is no process involving a single read of the sensor data that will recover the data in wells more than 1/16 full, if you have set the amplifier gain to ISO 1600. Simple as that. As I wrote previously, Canon engineers are not exceedingly stupid (at least on this point), they do know that ADC's saturate at defined input voltages.
[a href=\"index.php?act=findpost&pid=201991\"][{POST_SNAPBACK}][/a]
Emil,
One point I find interesting here is that not all cameras seem to employ this analog, pre-A/D amplification of the voltage.
Edmund has reported that his P45+ produces at least equally good images after underexposing at base ISO, compared to the same exposure used at a higher ISO.
John Sheehy also reported noticing this effect in certain P&S cameras. In other words, the false ISO settings, such as ISO 3200 in most Canon DSLRs, and ISO 12,800 and 25,600 in the Nikon D3, apply across the board in some models of digital cameras. Any setting above base ISO is essentially a false ISO that ultimately serves no purpose that can't be achieved in post processing.
One also wonders, if the analog signal can be boosted prior to A/D conversion to result in a higher ISO with less shadow noise than the unboosted, underexposed data at base ISO, why cannot a similar boost be applied at base ISO to lift the shadows? Is it simply a matter of the electronics not be sufficiently robust to handle the greater voltages?
-
The engineering definition of DR is actually quite closely related to any other definition using shadow SNR of raw data to determine the noise floor, and determining DR as the ratio of raw saturation to said noise floor
This is not of use for me. I am not shooting plain, unicolor surfaces all the day (well, I am sometimes, but only for test/demo purposes).
For me there is no photography without details, and that is the main point. I see the limit of the camera there, where it can not reproduce fine details any more. This includes the limit of noisiness as well, but it is more.
Therefor my camera needs to have enough levels to produce details. Of course it is a requirement as well, that those levels are useful, but that's a different, additional issue.
There is a defined voltage corresponding to the maximum digitized output of the ADC. The ISO gain determines what output voltage corresponds to a given number of photons in the sensel. If you increase the ISO, a smaller number of photons yields that saturating voltage input to the ADC. So if a full well yields this saturating voltage at ISO 100, then a 1/16 full well yields this voltage at ISO 1600. There is no process involving a single read of the sensor data that will recover the data in wells more than 1/16 full, if you have set the amplifier gain to ISO 1600.
Great. Now you have explained how it is working now in DSLRs, not why it is working that way.
How do you think most MFDBs are working? How do they realize the ISO gain?
-
You guys get it all worked out yet??
:+}
Snook
[a href=\"index.php?act=findpost&pid=201996\"][{POST_SNAPBACK}][/a]
Yep, must be where I am going wrong!
-
-
You guys get it all worked out yet??
[a href=\"index.php?act=findpost&pid=201996\"][{POST_SNAPBACK}][/a]
I have to admit I find the array of opinions a bit bewildering. When talking about the differences between medium format DSLRs and 35mm DSLR cameras can anyone provide an authoritative answer to any of these questions:
1. Is there anything to support the ideas that MF images are more 3D than those created with 35mm?
2. Could it be that the 3D effect that people refer to is simply that there is a narrower depth of focus at a given aperture for larger format cameras and that means that background is more likely to be out-of-focus. Is that what creates the sense of 3D?
3. What is the relationship between the dynamic range of the print and the bit-depth of the image that you get from the camera?
4. Do you know of any reason that the dynamic range of the print from a MF camera should be greater than that from a 35mm camera?
5. Is there any benefit that might be seen on print from cameras that claim to be 14-bit or 16-bit instead of 12-bit?
-
So what exactly is being made more accurate/precise with the extra bits? As far as I know, it's just more levels/steps within the range of the chip, not more D.R.
Thanks.
[a href=\"index.php?act=findpost&pid=201965\"][{POST_SNAPBACK}][/a]
An analogy may be of some help here:
Suppose a bunch of people are timing a race with stopwatches. The people timing the race have different degrees of hand/eye coordination, lag time in pressing the stopwatch, etc. The times they will record will thus fall over a range of values. This range of values is a measure of the level of ACCURACY of the recorded race time; suppose this range is a tenth of a second.
Now, what level of PRECISION of stopwatch is needed to time the race? Clearly, a stopwatch that records down to the nearest second is needfully imprecise and therefore inaccurate. A stopwatch that records in hundredths of a second is overkill, since the last digit doesn't carry any meaningful information -- the time is only accurate to the nearest tenth of a second. A stopwatch which records in tenths of a second is just about right -- neither overspec'd nor too imprecise for the job.
Notice I chose my words carefully, to distinguish accuracy from precision. A stopwatch measuring in the hundredths is more precise than one measuring in the tenths, but it is no more accurate since the whole timing accuracy is limited by that of the person using the watch (so long as the precision as as good as or slightly better than the person's accuracy).
This is the nub of the argument about bit depth too. Bit depth is precision of recording the illumination level of a pixel; noise is the limiting factor in the accuracy of the recorded data. Having bit depth that exceeds the noise is having more precision but not necessarily more accuracy. The substance of the debate is how many bits are enough precision given the current state of technology with regard to the noise limiting the accuracy.
On average, the error in precision in digitizing the signal from the sensor is 1/sqrt[12]~.29 of the quantization step size, a little less than a third of a raw level. The accuracy of the raw data is limited by the noise of the electronics, and is slightly over one 12-bit raw level at low ISO on the best DSLR's (Nikon D3, Canon 1-series); I haven't seen much data on MFDB's. Thus 12-bit data is sufficient to record to the level of the accuracy of the data from a single read of the sensor of DSLR's; if one wanted to be really conservative, 13-bit data is ample but that 13th bit is largely noise. Any bit depth beyond that, and the extra bits are essentially random noise.
-
OK, this is really getting more and more off the wall, but I am interested in answers to #4 and #5 in Henry's post just above...I am on my way to demo a Phase back as we speak, and I am bringing my 1Ds3 to shoot some side-by side.
-
Emil,
One point I find interesting here is that not all cameras seem to employ this analog, pre-A/D amplification of the voltage.
Edmund has reported that his P45+ produces at least equally good images after underexposing at base ISO, compared to the same exposure used at a higher ISO.
John Sheehy also reported noticing this effect in certain P&S cameras. In other words, the false ISO settings, such as ISO 3200 in most Canon DSLRs, and ISO 12,800 and 25,600 in the Nikon D3, apply across the board in some models of digital cameras. Any setting above base ISO is essentially a false ISO that ultimately serves no purpose that can't be achieved in post processing.
With a fixed exposure, the difference between using a higher ISO versus underexposing at a lower ISO and amplifying in software post-capture lies in the additional noise/error sources that get amplified. Basically, any amplification takes as input not just the signal but also any noise that was added upstream in the signal processing chain. The signal processing chain is
sensor readout > ISO amplification > quantization > post-capture manipulation
Thus, using the higher ISO, only the noise added by the electronics doing the sensor readout is amplified. Amplifying post-capture by software multiplication of the digitized raw data, one is amplifying in addition the noise of the ISO amplifier as well as the ADC, as well as the quantization error (the difference between the analog input to the ADC and the nearest quantized level that it can be digitized to).
To the extent that the ISO amplifier noise, ADC noise, and quantization error are small relative to the overall noise level, one can get away with amplifying post-capture. If they are large contributors to the noise level (as they are in deep shadows at low ISO in Canon DSLR's, for instance), one pays a substantial price for the underexposure. GLuijk had a nice demonstration of this in a recent thread for which I'm the OP.
One reason to use a higher bit depth is to eliminate quantization error as an issue; average quantization error is about .3 raw level, so if you are going to implement higher ISO as a post-capture software manipulation (as I'm told several MFDB's do) it makes some sense to use 16-bit capture instead of 12-bit -- quantization error is reduced by a factor of 16 and is completely negligible relative to other noise sources. Of course this doesn't change the fact that all those extra bits are merely giving a refined specification of those other noise sources, and not really helping to record the signal more accurately.
One also wonders, if the analog signal can be boosted prior to A/D conversion to result in a higher ISO with less shadow noise than the unboosted, underexposed data at base ISO, why cannot a similar boost be applied at base ISO to lift the shadows? Is it simply a matter of the electronics not be sufficiently robust to handle the greater voltages?
[a href=\"index.php?act=findpost&pid=202001\"][{POST_SNAPBACK}][/a]
Because in current designs the level of amplification is specified ahead of time, for all pixels being read; to do what you are suggesting, one would have to apply a different boost to pixels with few photons captured vs those with many photons captured. One might do this by splitting the signal and sending it to two separate amplifiers, one set to low amplification for the highlights and another to high amplification for the shadows. Quantize both, then combine the two using the high amplification output to replace the shadows in the low amplification output. The cost would be an extra amplifier/ADC combination (some cameras, like the D3, have parallel analog front end processors; the number would have to be doubled) and the processing overhead to combine the two data streams. If this could be done with no further introduction of noise, the Canon 1D3/1Ds3 and Nikon D3 would each gain about two stops of DR (and finally 14-bit capture would be justified, perhaps even 15-bit).
-
For me there is no photography without details, and that is the main point. I see the limit of the camera there, where it can not reproduce fine details any more. This includes the limit of noisiness as well, but it is more.
Therefore my camera needs to have enough levels to produce details. Of course it is a requirement as well, that those levels are useful, but that's a different, additional issue.
[a href=\"index.php?act=findpost&pid=202004\"][{POST_SNAPBACK}][/a]
The number of discernable levels IS the S/N ratio -- you can't tell that two levels which are different are telling you that they came from a different light level in the scene being photographed, unless the difference of levels is enough larger than the amount of noise present. This is why I said the lower end of DR is determined by one's criterion for an acceptable S/N ratio in shadows (which is in turn closely related to the noise in shadows, which is why I stated that engineering DR is closely tied to one's notion of "useful" DR).
Noise is not image detail, even if it's quantized in an exquisite fineness of levels.
-
I have just done a side-by-side test of the Phase P30+ and the 1Ds3. In my opinion the files look very similar to one anther processed in C1 4.1 in terms of look, sharpness, etc (though the Canon camera was further away than the Phase, so no real resolution test here. And the Phase does show some other very subtle advantages - but these are only seen at major pixel peeping levels.)
What struck me though is that the shadow detail, even when boosted to +2 exp. during RAW conversion, or within photoshop, shows no visible advantage in terms of shadow detail for the back.
(each at ISO 100, f9, no N.R. *and plus 2 exposure comp during RAW conversion*, just to see the difference one will encounter during extreme levels adjustments)
(I used to use Leaf backs, am now using 1Ds3 and renting MF backs, and may buy another back soon...)
Other notes on the Phase:
Lens feels much better than older mamiya 645 as far as I rememher, AF is good – way better than the Contax AF in a dimly lit room. The Contax hunted repeatedly…
Viewfinder seems about as large as the Canons, though the Canon is brighter. (50mm 1.4 on the Canon, 80mm on the Phase.)
Phase kit is a little heavier than the Canon with 50 1.4.
Operation of the Phase is very fast – seems like about a second between exposures
Zooming and scrolling on the Phase is awkward at best.
-
You guys get it all worked out yet??
:+}
Snook
[a href=\"index.php?act=findpost&pid=201996\"][{POST_SNAPBACK}][/a]
No! We need a sharper pencil. Bicubic sharpener perhaps
-
... plus 2 exposure comp during RAW conversion
[a href=\"index.php?act=findpost&pid=202129\"][{POST_SNAPBACK}][/a]
Are you saying that when you converted the RAW file that you adjusted the exposure level up 2 f/stops in the software?
If so, I guess that is a sure way to bring out noise in the shadows, but it sure seems extreme.
You said you are considering buying another MF back. Has this test changed your opinion? From what you said the difference between the Phase P30+ and the 1Ds3 were negligible, right?
If so, why would you buy a MF system?
(this is the issue that I am dealing with, so I would like your take on it).
-
Are you saying that when you converted the RAW file that you adjusted the exposure level up 2 f/stops in the software?
If so, I guess that is a sure way to bring out noise in the shadows, but it sure seems extreme.
You said you are considering buying another MF back. Has this test changed your opinion? From what you said the difference between the Phase P30+ and the 1Ds3 were negligible, right?
If so, why would you buy a MF system?
(this is the issue that I am dealing with, so I would like your take on it).
[a href=\"index.php?act=findpost&pid=202137\"][{POST_SNAPBACK}][/a]
Yes, +2 is extreme. I only went to +2 because +1 or +.5 did not show any bad issues in either camera in terms of shadow banding, etc. In short, I was looking for this 14-bit vs 16-bit medium format advantage which is being discussed in this thread and elsewhere on a daily basis. *People say that they have better shadow detail from their MF backs.* I saw no advantage there today.
And for the record, I usually get my exposure pretty correct during the shoot, and have only ever needed to boost shadows by around +2/3 when I had shot a cashmere sweater campaign, and later the A.D.s decided they wanted a lighter look...so yes, +2 is EXTREME in my opinion. It should therefore easily show differences in shadow detail, right?
I do have older Leaf back, which I no longer use. And I will do some more testing with the Phase p30+ and some others to see about the differences in detail, etc. (this test was conducted with strobes at a dealer, and during the demo the Phase had to be unplugged and reattached a couple times, as it had inexplicably frozen up.)
And yes, for the most part the differences were very negligible, IMO, especially since you are effectively looking at a 7' or 10' print from 12" away, as rendered on your 72dpi monitor.
-
Do you mind posting the raw files of your comparisons? (Yousendit is the way to go..)
-
Do you mind posting the raw files of your comparisons? (Yousendit is the way to go..)
[a href=\"index.php?act=findpost&pid=202143\"][{POST_SNAPBACK}][/a]
No I don't mind, I will post them later tonight or tomorrow. Any ideas how, aside from yousendit?
*and like I said, the Canon was further form the tabletop setup, so these files are no good for comparing resolution differences.*
And maybe when I do post them, I'll start another thread, as this one is already all over the map.
-
And maybe when I do post them, I'll start another thread, as this one is already all over the map.
This is a very good idea. And post please those, which you deem the best, most comparable. I suggest you to include a grey/white card as well.
I don't know any better service than yousendit. You don't need to register, as the individual files are under 100MB. You don't even need to use your own email address, use a fictional one both as sender and receiver, and post here the URL.
-
This is a very good idea. And post please those, which you deem the best, most comparable. I suggest you to include a grey/white card as well.
I don't know any better service than yousendit. You don't need to register, as the individual files are under 100MB. You don't even need to use your own email address, use a fictional one both as sender and receiver, and post here the URL.
[a href=\"index.php?act=findpost&pid=202145\"][{POST_SNAPBACK}][/a]
There was no grey card. I was at a dealer for about 45 min total. both cameras shot in "daylight" white balance. So you will have to white balance them yourself.
-
A point of information: I asked the product manager of Sinar - there is no analog ISO setting at all, the ISO setting in the file is just a datum used afterwards by the software.
Edmund
-
A point of information: I asked the product manager of Sinar - there is no analog ISO setting at all, the ISO setting in the file is just a datum used afterwards by the software.
I brought this surprize at the GetDPI forum. Funny is, that some owners appeared to have been offended by this insinuation, while I would be happy if my camera behaved that way (offering the entire dynamic range in every shot).
However, the software is not up to the proper processing of such images.
-
In short, I was looking for this 14-bit vs 16-bit medium format advantage which is being discussed in this thread and elsewhere on a daily basis. *People say that they have better shadow detail from their MF backs.* I saw no advantage there today.
[a href=\"index.php?act=findpost&pid=202142\"][{POST_SNAPBACK}][/a]
There is a similar debate about bit-depth in high-end audio.
Many audiophiles clamor for greater bit depth because it is theoretically better. They think it sounds better, but they have never tested it. Until a report came out last fall:
“In the September 2007 issue of the Journal of the Audio Engineering Society (Volume 55, Number 9), …a breakthrough paper that contradicts all previous inputs by the engineering community. They prove beyond a shadow of a doubt, with literally hundreds of double-blind listening tests at matched levels, conducted over a period of more than a year, that there’s no audible difference between the original CD standard (“Red Book”) and 24-bit/192-kHz PCM or 1-bit/2.8442-MHz DSD.”
In the world of high-end imaging there seems to be a similar acceptance that a higher bit depth file has to be better. It makes sense, so it has to be true right? Bit can you see it in the "real world"?
So far 203 is seeing no difference.
When I did tests with two different MF backs (Hasslblad and Phase One) I could see no difference in shadow detail when compared to the Canon 1Ds Mk3.
Has anyone else out there conducted careful side-by-side comparisons?
-
There is a similar debate about bit-depth in high-end audio.
Many audiophiles clamor for greater bit depth because it is theoretically better. They think it sounds better, but they have never tested it. Until a report came out last fall:
“In the September 2007 issue of the Journal of the Audio Engineering Society (Volume 55, Number 9), …a breakthrough paper that contradicts all previous inputs by the engineering community. They prove beyond a shadow of a doubt, with literally hundreds of double-blind listening tests at matched levels, conducted over a period of more than a year, that there’s no audible difference between the original CD standard (“Red Book”) and 24-bit/192-kHz PCM or 1-bit/2.8442-MHz DSD.”
In the world of high-end imaging there seems to be a similar acceptance that a higher bit depth file has to be better. It makes sense, so it has to be true right? Bit can you see it in the "real world"?
So far 203 is seeing no difference.
When I did tests with two different MF backs (Hasslblad and Phase One) I could see no difference in shadow detail when compared to the Canon 1Ds Mk3.
Has anyone else out there conducted careful side-by-side comparisons?
[a href=\"index.php?act=findpost&pid=202155\"][{POST_SNAPBACK}][/a]
OK, the guys has a point - but we'll never know because the CFA is different for the backs and for the Canons, and this also has an effect on the color precision.
Edmund
-
in the audio world happens exactly the same than in the world of imaging:
higher bit depth is better as long the data has to be edited or changed.
you easily can see that 8 or 12bit data can reach posterization, if smooth gradations are edited strongly. this happens much later if the images are 14bit depth.
the same happens with audio data. it has sense to record multitracks in 20 or 24bit depth, as well as the mixdown to stereo or multitracks, before the final mastering step is done.
finally mastered the difference between 16bit 44.1 or 48 khz and 24/96 bit/khz is not very important, the data will be reproduced in most cases nearly linear and the data is deep enough for that in 16bit.
same happens in the photo world. an image which is not more edited ( because its already finished ) wont look better in 16bit than in 8bit on todays printing and monitoring systems. this eventually might change in future with better reproduction systems,- but not today.
but during the production process the bit depth makes a huge difference.
but again: the mf system use only 14bits of image information, as the new 35mm hi-end cameras as well. two bits are "left "empty", although many dont want to believe this and claim to "see" the difference in 14 bit and ( imaginated ) 16 bit data.
-
higher bit depth is better as long the data has to be edited or changed.
[a href=\"index.php?act=findpost&pid=202164\"][{POST_SNAPBACK}][/a]
Certainly increasing the exposure by 2-f/stops during RAW conversion is some serious editing.
But you are saying he did not see any difference because the so-called "16-bit" Phase One back is really 14-bits, just like the Canon 1Ds Mk3?
-
yes and no.
but in any case the bit depth has not much to do with the dynamic range, and this is what you proof in raising up the level for 2 stops and looking in the shadows. you better can proof the bit depth and its influence if you raise up the contrast e.g. of a smooth colorband as a light blue sky ( or skin ) . deeper information will allow you by far more manipulation than lower information depth.
-
higher bit depth is better as long the data has to be edited or changed.
you easily can see that 8 or 12bit data can reach posterization, if smooth gradations are edited strongly. this happens much later if the images are 14bit depth.
during the production process the bit depth makes a huge difference.
but again: the mf system use only 14bits of image information, as the new 35mm hi-end cameras as well. two bits are "left "empty", although many dont want to believe this and claim to "see" the difference in 14 bit and ( imaginated ) 16 bit data.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=202164\")
The more I look into the matter, the more it seems to me that the only difference can come in the raw conversion process, as there is not much to see in added bit depth beyond DR at the level of the raw data. I took a 1D3 image, deliberately underexposed by about 6 stops so that the last six bits of its 14-bit raw data cover the range 0-255 and also encode almost all the image data (except for patch of sky on the middle right). I extracted one of the green channels of the Bayer array and lopped off those mostly empty highest six bits; I also rounded off the resulting 8-bit data to the nearest 6-bit level (what would have arisen had the 1D3 been a 12-bit camera). Because the noise is over four levels, it masks any tonal transitions one would think of as being posterized with those last two bits missing...
Here is this green channel raw data bit shifted and encoded as 8-bit grayscale tiffs, both the original and with the last two bits truncated by rounding to the nearest 6-bit level:
[a href=\"http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/dpr/ueIJ8bit.tif]http://theory.uchicago.edu/~ejm/pix/20d/te...pr/ueIJ8bit.tif[/url]
http://theory.uchicago.edu/~ejm/pix/20d/te...pr/ueIJ6bit.tif (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/dpr/ueIJ6bit.tif)
I find it VERY hard to tell the difference between these two files visually; the histograms do however show that one file has four times fewer levels than the other. And remember, what you see if you view the files has effectively six stops exposure compensation. Now, one can do silly things in post-processing to make the bit-truncated data posterize; median filtering does a pretty good job. I am beginning to think that any perceived issues with 12-bit data, when they are not in the mind of the viewer, are due to shortcomings in raw conversion software and not in the data per se. It should be possible to add back in random noise (or better yet, a proper dithering algorithm that only affects past the 12th bit) in the additional bits in such a way as to dither away issues in the conversion process so that even extreme edits have no effect.
-
yes and no.
[a href=\"index.php?act=findpost&pid=202169\"][{POST_SNAPBACK}][/a]
Rainer,
First you say "Yes" then you change your post to say "Yes and No"
Was gibt?
Do these MF backs that say they are 16-bit backs really have 16-bits of usable data, or is it really 14-bit?
If they really are only 14-(effective)bit backs, how can they call themselves 16-bit?
-
In is nonsensical to talk about MFDBs sensors in general. Even the P25+ and P45+ from PhaseOne are different.
Related to the P45+: it is by far not so, that there were only 14 bits there. I created a demo about this; everyone can make his/her own conclusion about the usefulness of these or those bits.
The file is in TIFF format, layered, over 8MB; the layers contain screenshots of raw data displays: 16bit; 14bit; only the low order bit; only the low order two bits.
The P45+ file is from Edmund. I hope he does not mind that I use a small crop of that for this demo. (I guess he still does not believe, that the shot has been underexposed by almost four stops.)
Demo (http://www.panopeeper.com/Demo/PhaseOne_14vs16bit_Analysis.TIF)
-
So far 203 is seeing no difference.
When I did tests with two different MF backs (Hasslblad and Phase One) I could see no difference in shadow detail when compared to the Canon 1Ds Mk3.
Has anyone else out there conducted careful side-by-side comparisons?
As I mentioned on page 4 of this thread, I've compared my P45+ against my D3. The only advantages I see with my P45+ is the total number of MPs, and the lack of an AA filter, which together yield crisper, higher resolution images. I see no real advantage to the P45+ on shadow detail retrieval and color fidelity. Specifically, here's what I said:
"In 14-bit mode, the D3 has tremendous ability to pull clean detail out of the shadows without suffering posterization and other artifacts. In that regard, I'd say it's on a par with my P45+. And the D3 actually has a little more highlight headroom than my P45+, so in that sense, the D3 is actually a bit better (tho again, the difference is slight). If the D3 sensor didn't have an AA filter, I'd say the pixel level quality would be a wash."
All that said, the P45+ is always going to be my first choice, because it does have the MP advantage and the lack of an AA filter. Those are still huge advantages, in my opinion, that can't be ignored, and can only be approached, with considerable limitations, using stiching.
This whole business of shadow detail and color has become a red herring, I think, at least when it comes to the latest generation of high end DSLRs (D3, 1DIII and 1DsIII). These new high end DSLRs have true 14-bit, low noise A/Ds, and for all intents and purposes are "just as good" as 16-bit sampling that you find in a MFDB, or at least so close in quality that it's virtually impossible to see the difference.
-
I have just done a side-by-side test of the Phase P30+ and the 1Ds3. In my opinion the files look very similar to one anther processed in C1 4.1 in terms of look, sharpness, etc (though the Canon camera was further away than the Phase, so no real resolution test here. And the Phase does show some other very subtle advantages - but these are only seen at major pixel peeping levels.)
[a href=\"index.php?act=findpost&pid=202129\"][{POST_SNAPBACK}][/a]
It's almost inconceivable that the P30 would not produce lower shadow noise than the 1Ds3. It not only has a slightly wider pixel pitch and a greater pixel count, but more significantly has a sensor which is double the size of the 1Ds3 sensor.
The question is, just how significant is the lower shadow noise? Unfortunately, because you haven't matched the FoVs of the two shots we can't tell how significant the lower shadow noise is, but it's clear to me that the P30 shot does have noticeably less noise in the comparison you've shown.
When comparing noise, it is essential to compare equal size files of matched FoVs.
Even if the sensors were the same size, it would be necessary to do this. An analogy would be the comparisons that have been made between the Nikon D3 and 1Ds3. Wild claims have been made by some that the D3 has significantly lower noise at high ISO than any other DSLR on the market, including the 1Ds3.
However, when the larger 1Ds3 files are reduced to the same size as the D3 files (a process which results in the discarding of some degree of noise as well as resolution), we find that noise levels are pretty much on a par.
The following image is a screen capture of a crop of your P30 shot reduced in size and superimposed on the 1Ds3 shot. You should be able to see clearly that the P30 image now exhibits less noise.
[attachment=7082:attachment]
-
When comparing noise, it is essential to compare equal size files of matched FoVs.
...
However, when the larger 1Ds3 files are reduced to the same size as the D3 files
No. A pixel is a pixel is a pixel and a digital image consists of pixels, of nothing else.
Have you purchased the 5D to downsize the image, so that it is even better? No-one I know would give many thousand dollars for a camera to regularly make a smaller image in order to make it appear a better camera.
The D3 has the advantage of large pixels vs the 1DsMkIII's small ones. The latter has the advantage of many pixels vs the D3's tiny image. These are trade-offs, which can not be balanced against each other in quality; the photographer's priority determines, what is more important.
Btw, it is not true, that downsizing an image yields the same result as if it had larger pixels.
-
The audio analogy is interesting because it highlights just how easy it is for people to believe things that are not necessarily true.
I think all of us have a tendency to believe what we want to hear and reject what makes us feel uncomfortable.
Having once attempted a comparison between the Nikon D3 and Canon 5D under less than ideal circumstances, I know how difficult it can be to get everything right. It takes time and painstaking attention to detail as well as an awareness of all the factors that might tip the balance one way or another in favour of one camera or system.
The recent article on LL, 'Do sensors outresolve lenses?' should make it clear why one should not be using the same F stop when comparing images from different size sensors even if they both have a similar pixel count.
I don't believe I've ever seen a proper comparison between 35mm and MFDB of similar pixel count. There's always been some major flaw in the methodology, at least in the comparisons shown on this forum.
Perhaps it's a promotional thing. There's a need to preserve the idea (perhaps myth, who knows?) that a DB pixel is inherently superior to any DSLR pixel of the same pitch or size.
-
No. A pixel is a pixel is a pixel and a digital image consists of pixels, of nothing else.
Have you purchased the 5D to downsize the image, so that it is even better? No-one I know would give many thousand dollars for a camera to regularly make a smaller image in order to make it appear a better camera.
The D3 has the advantage of large pixels vs the 1DsMkIII's small ones. The latter has the advantage of many pixels vs the D3's tiny image. These are trade-offs, which can not be balanced against each other in quality; the photographer's priority determines, what is more important.
Btw, it is not true, that downsizing an image yields the same result as if it had larger pixels.
[a href=\"index.php?act=findpost&pid=202180\"][{POST_SNAPBACK}][/a]
I have great difficulty in understanding your point, Gabor. Perhaps this is because you are so immersed in the minutiae of the digital process that you cannot see the woods for the trees.
Photographers are concerned with the final result. The final result is always, without exception, a print of a specific size or an image displayed on a screen of some sorts, at a specific size. You may change the size of the image at will, produce prints at various sizes, view images at different degrees of magnification on monitors of varying resolution, but it makes no sense at all from the point of view of the practising photographer, to compare camera systems by comparing different size images from those systems.
It makes little difference, (as regards noise) when comparing the D3 with the 1Ds3, whether the D3 image is upressed beyond its native resolution to make a large print, or whether the 1Ds3 file is downressed to make a small print, although I would recommend doing both to compare resolution. In both cases the noise will be affected. When the D3 image is interpolated, noise will also be interpolated (increased). When the 1Ds3 image is downsized, noise will also be downsized (reduced).
When comparing camera systems purely for image quality (as opposed to the many other features that might be significant such as weight, ergonomics, frame rate etc) it's absolutely essential to compare same size images of the same FoV.
-
The questions you are all debating about dynamic range and shadow detail have been definitively resolved. All those technical treatises we have been pondering were a waste of time, it's all totally irrelevant.
Real photographers know DR does not matter...
http://www.kenrockwell.com/tech/dynamic-range.htm (http://www.kenrockwell.com/tech/dynamic-range.htm)
Gotta love the guy!
-
You're welcome to publish these files - that's why I hand them out.
It's interesting that this demo clearly shows up the banding which made me declare that back as defective.
I think it would be worth applying the same technique to other images to see what one can learn.
As far as Sinar know, ISO is just a postprocessing setting in all the current backs including their own, which makes defining exactly what underexposed means a topic of debate.
Edmund
In is nonsensical to talk about MFDBs sensors in general. Even the P25+ and P45+ from PhaseOne are different.
Related to the P45+: it is by far not so, that there were only 14 bits there. I created a demo about this; everyone can make his/her own conclusion about the usefulness of these or those bits.
The file is in TIFF format, layered, over 8MB; the layers contain screenshots of raw data displays: 16bit; 14bit; only the low order bit; only the low order two bits.
The P45+ file is from Edmund. I hope he does not mind that I use a small crop of that for this demo. (I guess he still does not believe, that the shot has been underexposed by almost four stops.)
Demo (http://www.panopeeper.com/Demo/PhaseOne_14vs16bit_Analysis.TIF)
[a href=\"index.php?act=findpost&pid=202173\"][{POST_SNAPBACK}][/a]
-
The questions you are all debating about dynamic range and shadow detail have been definitively resolved. All those technical treatises we have been pondering were a waste of time, it's all totally irrelevant.
Real photographers know DR does not matter...
http://www.kenrockwell.com/tech/dynamic-range.htm (http://www.kenrockwell.com/tech/dynamic-range.htm)
Gotta love the guy!
[a href=\"index.php?act=findpost&pid=202200\"][{POST_SNAPBACK}][/a]
Ah yes, Ken Rockwell, doyen of technically minded photographers.
-
The questions you are all debating about dynamic range and shadow detail have been definitively resolved. All those technical treatises we have been pondering were a waste of time, it's all totally irrelevant.
Real photographers know DR does not matter...
http://www.kenrockwell.com/tech/dynamic-range.htm (http://www.kenrockwell.com/tech/dynamic-range.htm)
Gotta love the guy!
[a href=\"index.php?act=findpost&pid=202200\"][{POST_SNAPBACK}][/a]
I think this is one of the dumbest and least defensible opinion I've read from Ken Rockwell. But hey!, we all make mistakes. I've been known myself to occasionally make a mistake, although I know that's difficult to believe .
-
I put the RAW files in their own thread...
-
I think this is one of the dumbest and least defensible opinion I've read from Ken Rockwell. But hey!, we all make mistakes. I've been known myself to occasionally make a mistake, although I know that's difficult to believe .
[a href=\"index.php?act=findpost&pid=202213\"][{POST_SNAPBACK}][/a]
Actually, to be fair, I think what Rockwell is saying is not that DR doesn't matter, but rather that a photgrapher should learn how to use the DR available from the camera to cover the scene, rather than obsess over how much DR is available in the first place. It's not an entirely unreasonable statement.
The distinction between those two concepts might have bearing on the main thrust of this thread. Or, it might not.
-
Actually, to be fair, I think what Rockwell is saying is not that DR doesn't matter, but rather that a photgrapher should learn how to use the DR available from the camera to cover the scene, rather than obsess over how much DR is available in the first place. It's not an entirely unreasonable statement.
The distinction between those two concepts might have bearing on the main thrust of this thread. Or, it might not.
[a href=\"index.php?act=findpost&pid=202311\"][{POST_SNAPBACK}][/a]
Yes, his argument is not totally indefensible. If you are shooting in jpeg mode, reducing contrast can increase dynamic range somewhat; bringing in a truckload of lighting can improve the shadows; sticking a grey gel or plastic sheet over the windows can lower the contrast of the scene, and using a split ND filter for a landscape shot can allow one to use a single exposure which is suitable for both sky and land, provided the horizon is reasonably straight.
All this is true. However, there are much more convenient ways of improving DR, such as using a camera in the first instance which has a high DR, shooting in RAW mode instead of the jpeg format which Rockwell seems to always recommend, and bracketing exposure for digital blending or merging to HDR where the subject is stationary.
Also some statements in his article, such as the following, appear to be completely wrong.
It is trivial to design a digital camera with far more dynamic range than any other camera. All a designer does is lower the contrast.
Here, Ken could have written an article recommending the eschewing of those complicated and cumbersome procedures (sticking gel on the windows etc), and his article would still have been in character, like his article suggesting tripods are no longer needed because digital cameras have high ISO capability and image-stabilised lenses which enable shutter speeds which are generally adequate for most hand-held shots.
-
It's interesting that this demo clearly shows up the banding which made me declare that back as defective
I am not convinced, that it is really defective, and I am not convinced, that you should not use it with higher ISO (your dealer does not know the camera).
If you are interested on it, the post (yousendit) pls following shots (raw files, of course):
1. whatever, with the lowest ISO, metered for that ISO,
2. shots for all other ISOs with the very same exposure, i.e. successively overexposing,
3. one shot with the banding, the raw file and a crop showing the banding in the converted TIFF or high quality JPEG.
As far as Sinar know, ISO is just a postprocessing setting in all the current backs including their own, which makes defining exactly what underexposed means a topic of debate
Well, I could say "underexposure" means, that you have to adjust the intensity in raw processing so much, that the banding becomes visible.
However, I used the term differently. I admit it is not correct, and I don't want to redefine "underexposure", but there is no short term for this: the distance in EV measured from the very right.
In other words, my "underexposed by almost four stops" means, that the exposure could have been increased by almost four stops without causing relevant pixel clipping on raw level.
-
Gabor - I don't have that back anymore. It had centerfold issues and banding issues and got sent back. No idea what the software solution for those is. As for underexposed by 4 stops, I have no idea what that means these days, since I have found out that the back is always run at ISO 100...
Edmund
I am not convinced, that it is really defective, and I am not convinced, that you should not use it with higher ISO (your dealer does not know the camera).
In other words, my "underexposed by almost four stops" means, that the exposure could have been increased by almost four stops without causing relevant pixel clipping on raw level.
[a href=\"index.php?act=findpost&pid=202338\"][{POST_SNAPBACK}][/a]
-
As for underexposed by 4 stops, I have no idea what that means these days, since I have found out that the back is always run at ISO 100...
No, the P45+ is not. Axel's shots showed, that there is a gain between 400 and 800. However, the P25+, the P45 (not plus) and the P30 (not plus) appears to capture everything in a single ISO shot; though I don't have enough shots sufficiently proving this.
I became convinced, that theis is the way to go: capture everything possible (i.e. expose as high as you can under the given circumstances), convert everything at the highest reasonable ISO and write everything in the raw file. Let the raw processor think about what to do with it.
This requires huge storage and massive in-camera processing capacity, but I would happily sacrify the 6.5 frames per second of my camera if the DR became two stops higher in exchange.
What camera do you have now?
-
No, the P45+ is not. Axel's shots showed, that there is a gain between 400 and 800. However, the P25+, the P45 (not plus) and the P30 (not plus) appears to capture everything in a single ISO shot; though I don't have enough shots sufficiently proving this.
I became convinced, that theis is the way to go: capture everything possible (i.e. expose as high as you can under the given circumstances), convert everything at the highest reasonable ISO and write everything in the raw file. Let the raw processor think about what to do with it.
This requires huge storage and massive in-camera processing capacity, but I would happily sacrify the 6.5 frames per second of my camera if the DR became two stops higher in exchange.
What camera do you have now?
[a href=\"index.php?act=findpost&pid=202347\"][{POST_SNAPBACK}][/a]
I have another sample of the P45+
Edmund
-
It has been written many times that those last two bits (between 14 and 16) do not hold any image data which directly translates to an increase in dynamic range.
I believe that MF has better D.R., but it is said that this improvement in D.R. comes from data which is contained within the first 14 bits...
[a href=\"index.php?act=findpost&pid=201941\"][{POST_SNAPBACK}][/a]
Probably less than 14 bits are recorded. The advantage of a 16-bit analog to digital conversion is the additional number of levels that are available in all of the exposure zones recorded. Within each exposure zone, subtle and smooth tonal gradations rely on having an adequate number of levels to represent the source subject. This is especially important in the highlight and mid-tone regions where the eye is better able to differentiate small tonal variations.
The concept that noise floors should be the defining factor in how much bit-depth in a/d conversion is advantageous only makes sense if one disregards the tonality of the levels above the noise floor. In other words, yes--you only need a number of bits equal to the useful dynamic range that the sensor can output if all you want to do is record the minimum number of levels that represent that tonal range. However, you eye needs more than a handful of levels to perceive a smooth tonal gradation even in deep shadow areas. To record more levels than the minimum--you need more bit-depth. To record an adequate number of digital levels to properly represent the analog sensor output, you need some headroom in the analog to digital conversion.
Additional bit-depth (more levels in each zone) is also desirable when making post conversion tonal adjustments as has been noted. It is also useful to have more levels as the linear RAW files are converted into the nonlinear and higher gamma color space required for producing a useable image. As RAW levels are redistributed into the desired color space highlight levels will be compressed while shadow levels are expanded.
[attachment=7094:attachment]
-
Probably less than 14 bits are recorded. The advantage of a 16-bit analog to digital conversion is the additional number of levels that are available in all of the exposure zones recorded. Within each exposure zone, subtle and smooth tonal gradations rely on having an adequate number of levels to represent the source subject. This is especially important in the highlight and mid-tone regions where the eye is better able to differentiate small tonal variations.
The concept that noise floors should be the defining factor in how much bit-depth in a/d conversion is advantageous only makes sense if one disregards the tonality of the levels above the noise floor. In other words, yes--you only need a number of bits equal to the useful dynamic range that the sensor can output if all you want to do is record the minimum number of levels that represent that tonal range. However, you eye needs more than a handful of levels to perceive a smooth tonal gradation even in deep shadow areas. To record more levels than the minimum--you need more bit-depth. To record an adequate number of digital levels to properly represent the analog sensor output, you need some headroom in the analog to digital conversion.
Additional bit-depth (more levels in each zone) is also desirable when making post conversion tonal adjustments as has been noted. It is also useful to have more levels as the linear RAW files are converted into the nonlinear and higher gamma color space required for producing a useable image. As RAW levels are redistributed into the desired color space highlight levels will be compressed while shadow levels are expanded.
[attachment=7094:attachment]
[a href=\"index.php?act=findpost&pid=202368\"][{POST_SNAPBACK}][/a]
Techtalk,
It's the morning for me, my brain is not up to speed. But my BS detector has just been set off:
- How can you have more levels available in the zones if you are not recording them ?
- Why does having non-recorded but present low-end bits affect the Raw conversion ?
- To summarize, aren't you conflating the data captured by the camera with the data post conversion ?
Edmund
-
Techtalk,
It's the morning for me, my brain is not up to speed. But my BS detector has just been set off:
- How can you have more levels available in the zones if you are not recording them ?
- Why does having non-recorded but present low-end bits affect the Raw conversion ?
- To summarize, aren't you conflating the data captured by the camera with the data post conversion ?
Edmund
[a href=\"index.php?act=findpost&pid=202384\"][{POST_SNAPBACK}][/a]
It should have been set off. The number of effective levels in any exposure zone is the S/N ratio in that zone; one cannot distinguish two tonalities as being different unless their difference is greater than the ambient noise (otherwise the difference looks like the same tonality in the presence of noise fluctuations). Having more bits just divides the noise into a finer set of levels, with little or no effect on image quality.
The absolute number of levels in any zone is largely irrelevant; the true figure of merit is the S/N ratio. And it matters quite little what one does in terms of levels/curves/gamma corrections, since the noise spectrum gets pushed around in the same way that the signal does; for instance, stretching the levels makes gaps in the histogram, but the noise gets stretched too and dithers the tonal gaps.
Complaints about a lack of levels not providing enough detail or sufficiently smooth tonal transitions are really a complaint that there is not sufficiently high S/N in the image.