 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
| PhotoKB HomeDiscussion Groups Digital Photography Digital PhotoDSLR CamerasZLR CamerasPoint & Shoot CamerasFilm Photography 35 mmLarge FormatMedium formatDarkroomFilm and LabsOther EquipmentPhoto Technique Nature PhotographyPeople PhotographyTechnique GeneralGeneral Photo Topics General TopicsAustralian PhotographyUK PhotographyDirectoryPhoto Clubs |
Photo Forum / Film Photography / Medium format / November 2007Medium format versus digital sharpness
How many megapixels would a digital camera image file need to equal the sharpness of a six-by-seven centimeter or a four-by-five-inch transparency? Robert > How many megapixels would a digital camera image file need to equal the > sharpness of a six-by-seven centimeter or a four-by-five-inch > transparency? The short answer is that there are no affordable digital systems that compete with 6x7 and 4x5. Your mileage will vary, but my experience is that for practical purposes, 12.7MP (the Canon 5D) acts very much like 645. But 6x7 scanned on a Nikon 8000 is noticeably better than the 5D. So the new 21MP Canon 1DsIII should give 6x7 a run for its money. (Although the only Canon wide angle lens up to that is probably the new 14/2.8 II L. And maybe the new 16-35/2.8 when stopped way down. Maybe.) I'd guess the 39MP digital backs would be encroaching on 4x5 territory. David J. Littleboy Tokyo, Japan > So the new 21MP Canon 1DsIII should give 6x7 a run for its money. (Although > the only Canon wide angle lens up to that is probably the new 14/2.8 II L. > And maybe the new 16-35/2.8 when stopped way down. Maybe.) the keyword here is of course: "should" > I'd guess the 39MP digital backs would be encroaching on 4x5 territory. doubt it. my experience here, with a very good scanner and as sharp a lens and film combo as one can find: 35mm = 14MP 6X4.5 > 20MP 6X7 = I never seen a digital that comes even close, but as a wild guess, I'd say at the very least 30MP Now, a guess: 4X5 = my guess: around 100MP Caveat: I don't subscribe to the notion that one should compare a highly processed digital image to a raw film scan. My scanned images show the results of not only the film/lens/scan combination used but also the post-processing that I apply which is similar to what a digital camera applies to its raw file: de-noise+sharpen. Under those conditions, I have no doubts whatsoever in putting forward the above numbers. Mind you, my results are based on modern scanners with modern film and modern software. Not 7 year old samples. I do use a dslr as well, so I know first hand when it comes to digital versus film comparisons. >> So the new 21MP Canon 1DsIII should give 6x7 a run for its money. >> (Although [quoted text clipped - 11 lines] > 35mm = 14MP > 6X4.5 > 20MP That's simply not my experience. 645 is very close to 12.7MP digital. It's simply not possible to justify using 645 if there's a 5D at hand. For print sizes at which 645 is adequate, the 5D is also. At print sizes where 645 is losing it, the 5D is too. > 6X7 = I never seen a digital that comes even close, 6x7 certainly came out ahead of 12.7MP digital in the test that I did. I just compared a 35mm lens on the 5D with the 65mm lens on the Mamiya 7. I should go back and shoot the center of the scene with a 35 x 1.4 = 50mm lens to compare the M7 with a hypothetical 25MP digital. The problem, though, is that there has to be really high contrast detail in the subject for it to appear on the film. Urban landscapes with signs and auto license plates and contrasty detail. Foliage that the 5D turns to mush also ends up as mush in 6x7. And then there's shadow detail... In actual use, I find 6x7 to be enormous compared to 645 (I think that's because I target 12x18 prints, and that's just begining to be pushing it for 645, but 16x20 from 6x7 is like falling off a log). But it's only 50% more film area. > but as a wild guess, I'd say at the very least 30MP It _ought to be_ somewhere in the 24 to 30 range. At 20x24, 6x7 is a 9x enlargement, which is just over what I consider the reasonable range for film. At 20x24, 28MP is a 240 ppi print, and sharp 5D images look very nice at 240 ppi (you need to get very close to see it as anything but razor sharp). > Now, a guess: > 4X5 = my guess: around 100MP Only if you get drum scans. But real 4x5 landscape work happens at f/32 and f/45. At f/32, MTF50 is 25 lp/mm and you really won't see much _photographically significant_ detail above 1500 ppi in a scan. (Grumph: how big is a real 4x5 frame? It's not 4x5...) That's under 50MP. Also, I doubt most 4x5 cameras hold the film particularly flat... > Caveat: I don't subscribe to the notion that one should > compare a highly processed digital image to a raw film scan. [quoted text clipped - 4 lines] > Under those conditions, I have no doubts whatsoever > in putting forward the above numbers. Yep. We're on the same page as far processing is concerned. It's just that I don't find 645 to be worth more than 12.7MP in real life for real prints. If I were making 16x20 prints, I'd be shooting 6x7. But both 645 and 12.7MP digital make very nice 12x18 prints. David J. Littleboy Tokyo, Japan > > 35mm = 14MP > > 6X4.5 > 20MP [quoted text clipped - 3 lines] > sizes at which 645 is adequate, the 5D is also. At print sizes where 645 is > losing it, the 5D is too. not IME. don't get me wrong: the 5d is very good, but 645 will beat it. of course, like I said: I'm assuming top of the line film/lens/scanner combo, just like the 5d is top of the line dig. > 6x7 certainly came out ahead of 12.7MP digital in the test that I did. I > just compared a 35mm lens on the 5D with the 65mm lens on the Mamiya 7. I > should go back and shoot the center of the scene with a 35 x 1.4 = 50mm lens > to compare the M7 with a hypothetical 25MP digital. I've tried a 127 K/L Mamiya and it is simply astonishing. And you've seen some of my prior work with a 180 "C" and the rudimentary scanning I was doing back then. > auto license plates and contrasty detail. Foliage that the 5D turns to mush > also ends up as mush in 6x7. And then there's shadow detail... disagree completely here. that is precisely where I see the difference: foliage does NOT turn to mush with 6x7, quite the opposite. it does on the 5d because of the bayer mask on greens: no such animal on film/scanner. and in 35mm, but that is a limit of the negative size itself: that's why we use larger. as for shadow detail, it's all a question of exposure. > In actual use, I find 6x7 to be enormous compared to 645 (I think that's > because I target 12x18 prints, and that's just begining to be pushing it for > 645, but 16x20 from 6x7 is like falling off a log). But it's only 50% more > film area. biggest problem I have with 6x7 is keping it flat on the scanner... getting there, but it's a hard slog. > film. At 20x24, 28MP is a 240 ppi print, and sharp 5D images look very nice > at 240 ppi (you need to get very close to see it as anything but razor > sharp). the problem I have with all this digital stuff is precisely the definition of "sharp". A vertical or horizontal edge that is straight is rendered as sharp by just about any digital at anything over 6MP. but when it comes to anything outside those two special cases, forget it: jaggies come in. they might be sharp jaggies, but a sawtooth is not my definition of a straight edge! and that is what I get when I compare a diagonal in film against the same in digital. nature being what it is, diagonals are everywhere. maybe it's just me, but I find a jagged edge totally unnatural and objectionable. it totally spoils an image. > Only if you get drum scans. But real 4x5 landscape work happens at f/32 and > f/45. At f/32, MTF50 is 25 lp/mm and you really won't see much > _photographically significant_ detail above 1500 ppi in a scan. (Grumph: how > big is a real 4x5 frame? It's not 4x5...) That's under 50MP. hmmmm, I doubt anyone would be able to keep 4x5 flat other than in a drum scanner. that would be the problem, I reckon. assuming of course high rez. flatbed scanners are getting good, but they are no match for proper film scanners. yet. as for the lpm thing: that is also proportional to the size of the negative/lens combo. a 4x5 coverage lens closed down to f32 has a much wider aperture than a 35mm lens at f32.the difference is marked. you can even see it with 6x7 mamiya lenses: try it. the usual formula for mtf calc from the aperture must be corrected for the negative/lens coverage size. I don't know the exact formula, what I can tell you from real observation is that f32 in a 6x7 lens is about the same as f8 in a 35mm lens. and that is a lot more than 25lpm. > Also, I doubt most 4x5 cameras hold the film particularly flat... the scanner side would be the problem. the camera film holders, from what I've seen in a recent 4x5 seminar, are very good at keeping things flat. > Yep. We're on the same page as far processing is concerned. It's just that I > don't find 645 to be worth more than 12.7MP in real life for real prints. If > I were making 16x20 prints, I'd be shooting 6x7. But both 645 and 12.7MP > digital make very nice 12x18 prints. I'm not arguing that 12.7MP dig won't make very nice prints. all I'm saying is that I can get better from 645. like I get a lot better than 8MP from 35mm: http://members.iinet.net.au/~nsouto/photos/pink%20wave%20small.jpg and http://members.iinet.net.au/~nsouto/photos/by%20the%20lake%20small.jpg are a few recent examples. that's the fuji pro 160s you and I like so much, on a zeiss ZM, with a voigtlander Ultron 28/1.9 lens, hand-held. aperture was f5.6 for the couple by the lake, f11 for the flower bed. astia 100 is better than this, but not by much. both images have been *downressed* to the size a 10MP camera will provide, in this case the 40D. I think you'll agree that there is a lot more detail in there than 8MP. I could post the 5600X3700 fullsize 4K scans but I don't want to shame the 20MP digital backs too much! :-) still: for handheld photography with a rangefinder, I'll take these anytime over a 8MP dig. BTW: the 645 photos on that same day with a Fuji G645S came out absolutely *stunning*: Astia 100 at 1/3 under is absolutely out of this world! when I get some time to scan them at 4000, I'll put them up for your critique. > biggest problem I have with 6x7 is keping it flat > on the scanner... getting there, but it's a hard slog. On the 8000/9000, get the glass carrier. Then use the provided masks (or cut your own) to hold the film slightly above the lower (non-AN) glass. Put the film on the mask, and use the upper (anti-Newton) glass to press it flat. (I think placing the film so the center, rather than the edges, is higher is better.) Use the AF function to measure the position of the film at several places around the frame and make sure it hasn't buckled. It's a pain, but you can get corner to corner sharp scans. >> film. At 20x24, 28MP is a 240 ppi print, and sharp 5D images look very >> nice [quoted text clipped - 6 lines] > anything over 6MP. but when it comes to anything > outside those two special cases, forget it: jaggies come in. The demosaicing algorithms are getting better. Lightroom 1.2 and ACR 4.2 are very good; smoother edges and less Moire than, for example, Raw Shooter Premium. > they might be sharp jaggies, but a sawtooth is not my > definition of a straight edge! and that is what I get when I > compare a diagonal in film against the same in digital. I don't get the people who overenlarge digital. But then I don't get the people who overenlarge film, either... > nature being what it is, diagonals are everywhere. > maybe it's just me, but I find a jagged edge totally > unnatural and objectionable. it totally spoils an image. No argument here. Oversharpening can aggravate the problem, too. >> Only if you get drum scans. But real 4x5 landscape work happens at f/32 >> and [quoted text clipped - 13 lines] > down to f32 has a much wider aperture than a 35mm lens > at f32.the difference is marked. No. The diffraction effect gets normalized by the focal length in the f number calculation, so the MTF due to diffraction at the film plane in lp/mm is dependent on the f number, not the physical size of the aperture. (Essentially the smear due to diffraction gets magnified by the focal length.) > you can even see it with 6x7 mamiya lenses: try it. the > usual formula for mtf calc from the aperture must be [quoted text clipped - 3 lines] > observation is that f32 in a 6x7 lens is about the same as f8 in > a 35mm lens. and that is a lot more than 25lpm. The formula is 1600/(f number) for zero MTF and 800/(f number) for 50% MTF in lp/mm regardless of the focal length. > I'm not arguing that 12.7MP dig won't make very nice prints. > all I'm saying is that I can get better from 645. [quoted text clipped - 13 lines] > camera will provide, in this case the 40D. I think you'll agree > that there is a lot more detail in there than 8MP. Uh, no. I won't. The flower bed photo does not look as good per pixel as I'm getting from the 5D + Tamron 28-75/2.8. Here's a hacked version of the flowers image. (I'll take it down shortly.) http://www.pbase.com/davidjl/image/87393300/original Tell me how much better or worse you think I've made it than your original image. (This is, of course, a trick question<g>, so look _very_ closely.) > still: for handheld photography with a rangefinder, > I'll take these anytime over a 8MP dig. > > BTW: the 645 photos on that same day with a Fuji > G645S came out absolutely *stunning*: Astia 100 at > 1/3 under is absolutely out of this world! Interesting that you like Astia. I find it soft and lacking in pop and much prefer Provia 100F. Go figure. But the GS645S lens is painfully sharp. David J. Littleboy Tokyo, Japan > I don't get the people who overenlarge digital. But then I don't get the > people who overenlarge film, either... well, that's what I gotta do for printing at 300 or 600. I wouldn't look at the images per se, but to get the best from my 600dpi printer I've got to start with very large images. tried with 240 and it's just not the same. upressing for printing won't get the best either: looks much better if I start with an image that is already the right size. hence the very large scans. HP8750 here. > No. The diffraction effect gets normalized by the focal length in the f > number calculation, so the MTF due to diffraction at the film plane in lp/mm > is dependent on the f number, not the physical size of the aperture. > (Essentially the smear due to diffraction gets magnified by the focal > length.) Interesting. thanks for that. I'll do some more reading on this. the effect I've noticed is that the actual diameter of f32 iris in a 6x7 lens is actually as big as f8 in a 35mm lens. hence why I feel it could indeed do as good as 35mm at that diameter: after all, diffraction is caused by size of iris, no? but like I said: I'll do some more reading on this, now you've made me curious as to why! > The formula is 1600/(f number) for zero MTF and 800/(f number) for 50% MTF > in lp/mm regardless of the focal length. yeah, but the focal length is not the issue I see here. it's the size of the lens itself. try it: put a 50mm 6x7 mamiya on f22 side by side to a 35mm 50mm at f22. the actual diameter of the iris on the 6x7 will be noticeably larger than the one on the 35mm lens. > > both images have been *downressed* to the size a 10MP > > camera will provide, in this case the 40D. I think you'll agree > > that there is a lot more detail in there than 8MP. > > Uh, no. I won't. The flower bed photo does not look as good per pixel as I'm > getting from the 5D + Tamron 28-75/2.8. David: I said "than 8MP". you're telling me that it's not as good as what you get at 12MP? of course it isn't! this is 35mm with a not-leica lens, handheld: don't expect miracles! > Here's a hacked version of the flowers image. (I'll take it down shortly.) > > http://www.pbase.com/davidjl/image/87393300/original > Tell me how much better or worse you think I've made it than your original > image. (This is, of course, a trick question<g>, so look _very_ closely.) thanks. NP with keeping it there, I'm not as anal as others about copies of my work. mostly because I don't sell it, I guess! ;-) you've applied subtle sharpening. very well done too, thanks. this was a lanczos irfanview downress of a fullsize scan: I've noticed that it slightly "hazes" images when I downress, unless I use the "sharpen" checkbox of resize. which ends up over-sharpened! can't win... :-) this is a crop of the full size scan: http://members.iinet.net.au/~nsouto/photos/pink%20wave%20crop1.jpg crop of the 10MP downressed image I posted before: http://members.iinet.net.au/~nsouto/photos/pink%20wave%20small%20crop1.jpg both can be further sharpened, of course! but definitely either of them is better than anything I've ever seen at 8MP. the full size one, fully sharpened, I reckon is right up there with the best I've seen at 12MP. of course I need to work more on the colour balance of shadows! next scan. note that I picked a section away from the centre: any lens is good in the middle, it's away from it that shows the diffs. and the lens was focused to about 8 feet, so far areas are relying on f11 handling it: not always perfect, but certainly usable... > Interesting that you like Astia. I find it soft and lacking in pop and much > prefer Provia 100F. Go figure. But the GS645S lens is painfully sharp. astia 100 at -1/3 gets a LOT more pop, try it! standard, it's nowhere near as good IMHO. but what really strikes me with it is the near absence of grain, even at a 4k scan. pro160s is very, very good but astia beats it fair and square. and now I'm off to bed: it's late here and I've got to work tomorrow. then I got a roll of portra to scan and some more testing to do with the new Sigma 12-24: that is an awfully good FF lens, it works wonders with the F100. got tired of waiting for Nikon to do something in this area, tried the Sigma and really liked it! works incredibly well with film, it's not bad at all either with the D80: by the time I've got it sorted out, the d3 will have come down in price and I'll definitely give it a go with my coll^H^H^H^Hstable of FF Nikkors! ;-) thanks a lot for the heads-up, much appreciated. >> I don't get the people who overenlarge digital. But then I don't get the >> people who overenlarge film, either... [quoted text clipped - 20 lines] > but like I said: I'll do some more reading on this, now you've > made me curious as to why! f/32 is 4 times smaller then f/8 so to get the same resolution you would need film that was 4 times larger, or 16 times the area. And by resolution I am not talking about line pairs/mm but rather total number of line pairs across the image. So take 24mm x 4 and you get that you would need film that is almost 100mm, almost twice as wide as MF film. But then there is no film that I know of that will make full use from the resolution of a good lens at f/8, f/16 is closer to the mark. So the total number of line pairs you can get with MF at f/32 would be about the same as you can get with film at f/16, which will no be much different then 35mm at f/8. >> The formula is 1600/(f number) for zero MTF and 800/(f number) for 50% MTF >> in lp/mm regardless of the focal length. [quoted text clipped - 4 lines] > the actual diameter of the iris on the 6x7 will be > noticeably larger than the one on the 35mm lens. When talking about the resolution at the film plane the f/number is what matters, the FL of the lens only matter when talking about resolution in the object space. It is my experince that my film scans (4000 dpi) viewed at 50-75% "looks" like direct digital output from a good dslr view at 100%. How to convert the "film MP" i don't know......but a division by 3 may be a good factor to use....or maybe not. A scanned 24x36 is about 4000 x 6000 dots.......24 MP / 3 = 8 MP. But once I made at test against a 8MP dslr and a 24x36 Velvia of same scene. Here the film scan showed much more details.....but of course you had grains which you did not have in the digital image. Can see I still have my test images at pbase. This is at least some kind of documentation. But of course there are a lot of parameters that can have an influence of the results. http://www.pbase.com/mxp/velvia_and_dslr Max >> I don't get the people who overenlarge digital. But then I don't get the >> people who overenlarge film, either... [quoted text clipped - 103 lines] > > thanks a lot for the heads-up, much appreciated. > It is my experince that my film scans (4000 dpi) viewed at 50-75% "looks" > like direct digital output from a good dslr view at 100%. Agreed. > How to convert the "film MP" i don't know......but a division by 3 may be > a good factor to use....or maybe not. [quoted text clipped - 8 lines] > > http://www.pbase.com/mxp/velvia_and_dslr The detail in the Velvia is lovely (it nails the grating on the square window above the black car, which the digital misses completely), but the grain noise is seriously objectionable. And it's not doing as well as the digital on the foliage in the background (were they taken with the same lighting/at the same time?). I only shot a small amount of Velvia 50, and gave up on it quite quickly because the noise became objectionable in prints before the extra detail could be used. Provia 100F and Velvia 100F are quite a bit better in that department. David J. Littleboy Tokyo, Japan >> It is my experince that my film scans (4000 dpi) viewed at 50-75% "looks" >> like direct digital output from a good dslr view at 100%. [quoted text clipped - 19 lines] > digital on the foliage in the background (were they taken with the same > lighting/at the same time?). Taken at the same time (+- 5 min.) but not same lightning as it shift all the time in Scotland :-) .....a bit more sun at the digital capture. > I only shot a small amount of Velvia 50, and gave up on it quite quickly > because the noise became objectionable in prints before the extra detail [quoted text clipped - 3 lines] > David J. Littleboy > Tokyo, Japan > Here the film scan showed much more details.....but of course you had grains > which you did not have in the digital image. correction: you have "grains:"also on the digital image. in fact, an unprocessed raw file is a complete mess in terms of IQ. what you have to do is apply to the scanned velvia the same degree of correction that your digital camera is applying to the raw file. namely: de-noising and sharpening. in the case of film, this takes the form of de-graining and sharpening. same principle. this is why you look at my large film scans and you see virtually no grain: I'm comparing apples with apples. not to say that the digital camera onboard processing isn't more convenient: it of course is, that is why folks use them more and more. >> Here the film scan showed much more details.....but of course you had >> grains [quoted text clipped - 12 lines] > onboard processing isn't more convenient: it of course > is, that is why folks use them more and more. It seems I can remove color noise in the film scan quite easy without loss of details. But if I try to remove the grain noise then it will harm some of the very fine details? ....but if these details are unusable for printing......then. But in fact I like better the way details in film disappear slowly into the grains. In digital it disappears more "ugly".....and after all......I don't find the grains look to bad in the prints. > It seems I can remove color noise in the film scan quite easy > without loss of details. But if I try to remove the grain noise [quoted text clipped - 3 lines] > digital it disappears more "ugly".....and after all......I don't find the > grains look to bad in the prints. quite right, IMHO. go here: http://wizofoz2k.deviantart.com/gallery/ to see many examples of film scans with very little scanner "grain". and some dslr shots as well. click on thumbs to see details, then on image to see bigger >> I don't get the people who overenlarge digital. But then I don't get the >> people who overenlarge film, either... > > well, that's what I gotta do for printing at 300 or 600. No. I don't get people who think that dSLR images look good at 20x30. (Upsampling to 300 ppi at the target size and then sharpening carefully helps, of course, but the information just isn't there.) Still, ScottW's point that if you have a decent image, the larger you print the better it looks, is spot on. >> > both images have been *downressed* to the size a 10MP >> > camera will provide, in this case the 40D. I think you'll agree [quoted text clipped - 8 lines] > isn't! this is 35mm with a not-leica lens, handheld: > don't expect miracles! No. I'm claiming that a 10MP crop from a 12.7MP 5D image would look better. >> Here's a hacked version of the flowers image. (I'll take it down >> shortly.) [quoted text clipped - 10 lines] > you've applied subtle sharpening. very well done too, > thanks. For the record: Photoshop CS3, Unsharp Mask, Amount = 100%, R = 0.9, radius = 0. But it's what I did _BEFORE_ the sharpening that was sneaky. I _downsampled_ it to 4MP (2500 in the long direction) and then _upsampled_ it to the original dimensions. Which means that the resampled image has, at an absolute maximum, 4MP of real information. And since it looks just as good as the original, it means the original image has, at an absolute maximum, 4MP of real information. (Had I been more energetic, I would have tried several different levels of downsampling (to find an amount of downsampling that clearly loses information and thus get a lower bound on the amount of actual information in the image), but the first guess was enough to make the point.) Thanks to ScottW for this reality check technique. > this was a lanczos irfanview downress of a > fullsize scan: I've noticed that it slightly "hazes" images > when I downress, unless I use the "sharpen" checkbox > of resize. which ends up over-sharpened! can't win... :-) Yes. You have to sharpen by hand and watch very carefully what it's doing to the image. David J. Littleboy Tokyo, Japan > But it's what I did _BEFORE_ the sharpening that was sneaky. > [quoted text clipped - 3 lines] > Which means that the resampled image has, at an absolute maximum, 4MP of > real information. Nope, sorry. Totally wrong, as I have explained many times before: your process for this is completely wrong, even though the principle is sound. You have downsampled STARTING from a jpg image, which is lossy to start with. A jpg image does NOT have all information, so when you downsample and upsample, you end up with the same amount of information. Had you downsampled STARTING from a lossless image file, like a TIF, you'd have lost size when you upsampled, every single time. > And since it looks just as good as the original, it means the original image > has, at an absolute maximum, 4MP of real information. Looks are completely deceiving, what counts is the size of the file that defines the amount of information in it. Do it with an original lossless tif file BEFORE any jpg conversion and you WILL lose size. Which means you lost information. > Thanks to ScottW for this reality check technique. Yeah, Scott keeps insisting on that caper, even though I have demonstrated with facts to him that his process is completely wrong. > Yes. You have to sharpen by hand and watch very carefully what it's doing to > the image. Tell me about it! ;-) >> But it's what I did _BEFORE_ the sharpening that was sneaky. >> [quoted text clipped - 8 lines] > before: your process for this is completely wrong, even > though the principle is sound. It's completely correct. Mathematically, even. > You have downsampled STARTING from a jpg image, > which is lossy to start with. A jpg image does NOT have all > information, so when you downsample and upsample, you > end up with the same amount of information. Had you > downsampled STARTING from a lossless image file, like a > TIF, you'd have lost size when you upsampled, every single time. You are into what is getting painfully close to what is essentially religion: a belief in something that's simply not there. Remember, the object of the game is photography. _Appearance_ is everything. It's what the game is. If two images look essentially identical, then they look identical. And that's what photography is about: how images look. >> And since it looks just as good as the original, it means the original >> image >> has, at an absolute maximum, 4MP of real information. > > Looks are completely deceiving, what counts is the size > of the file that defines the amount of information in it. No. Looks are the one thing that are not decieving in this game. It's what the game is about. If you take a 5D image with 12.7 MP in it, interpolate it up to be a 50MP tiff to compare it with a 4000 ppi scan of a 645 frame, it still has at most 12.7MP of information. Even though it looks very much like the 645 scan. If that 50MP tiff looks just as good as the 645 scan, then the 645 scan has at most 12.7MP of information. Information is real, it's not magic. > Do it with an original lossless tif file BEFORE any jpg > conversion and you WILL lose size. Which means you > lost information. This game can be played with tiffs as well as jpegs. Try it. Take a 4000 ppi scan, downsample to 2800 ppi, and upsample again to the original dimensions. You won't see any difference. But the intermediate file has 1/2 the bits. So the big file has at most 1/2 the information the file size implies. David J. Littleboy Tokyo, Japan > It's completely correct. Mathematically, even. No. You are starting from a file that already has lost information, and using that as "proof" that information is not there! There is no maths there, just fallacy. The only way you can convince me with numbers and your process is if you start from a non-lossy comnpression file. Like a TIFF. And we have done that before and it has been proven that there is indeed information loss and the images HAVE the detail you insist is not there. And PLEASE!: doi not even attempt to convert my jpg intop a tiff! That would be indeed proof you don't have the foggiest what we talking about! > You are into what is getting painfully close to what is essentially > religion: a belief in something that's simply not there. No. I'm simply using facts, rather than wrong beliefs. Try this: pick up a jpg of the SAME compression ratio, but now of a dslr image. Apply the same process. You will end up with exactly the same results you are seeing in film scans. Please: it is OBVIOUS where the problem is with your process. Just repeat the tests in an impartial and objective way, with BOTH tiff and jpg of BOTH scans AND digital. Rather than trying to "prove" what is not there. > Remember, the object of the game is photography. _Appearance_ is everything. > It's what the game is. No. You and Scott are the ones that came up with the "proof" and "objective" tests. Which are anything but, as I have proven many times. As for appearance, I'll put up the one from my film scans against any 10MP digital, anytime. If that is indeed now the criteria. But don't go and change it at will, because I won't let you. > If two images look essentially identical, then they look identical. And > that's what photography is about: how images look. Well then, why use a fallacious "objective test" to "prove" that digital is "better"? > >> image > >> has, at an absolute maximum, 4MP of real information. [quoted text clipped - 4 lines] > No. Looks are the one thing that are not decieving in this game. It's what > the game is about. No. THIS particular game is about your assertion that an image that is full of detail only has 4MP of useful information. It's right there, for anyone to read. Don't try to chage the subject now. Like I said: pick up a 10MP digital jpg, apply the same process and you'll end up with the same result. Which proves what I've been saying all along: this process you and Scott apply as "proof" of lack of detail is completely wrong and doesn't have a leg to stand on. Completely independent of what an image looks like or the importance of "looks". > If you take a 5D image with 12.7 MP in it, interpolate it up to be a 50MP > tiff to compare it with a 4000 ppi scan of a 645 frame, it still has at most > 12.7MP of information. Even though it looks very much like the 645 scan. But if you take that same image, save it as a jpg, then downress it to 10MP and back up to 12MP, then save it again as a jpg, you end up with the same size file. And THAT is what neither you nor Scott have EVER tried. And why your process is completely wrong. Please, don't change the subject now to wild interpolations: you are applying a process that REDUCES the size of the file, then expands it again. That has NOTHING to do with what 50MP looks like, or if it is similar to 645 scans. > If that 50MP tiff looks just as good as the 645 scan, then the 645 scan has > at most 12.7MP of information. > > Information is real, it's not magic. Then why apply completely subjective criteria to it, like "looks just as good"? Surely with all the technology we have nowadays there is a safer method than just "looks just as good"? Hint: there is, but it doesn't consist in downressing files compressed with a lossy algorithm to start with. > This game can be played with tiffs as well as jpegs. Try it. Take a 4000 ppi > scan, downsample to 2800 ppi, and upsample again to the original dimensions. > You won't see any difference. But the intermediate file has 1/2 the bits. So > the big file has at most 1/2 the information the file size implies. I have! WITH tiff files. And have done it here publicly! And it proved what I said: if you do it with a tiff file, you end up with a COMPLETELY different result from yours with a jpg. On the SAME image! And what was the reaction from you and Scott? To fall back into the "looks good enough" argument! Like I said: if you want an objective test based on facts, then STAY that way all the way through. Falling back on subjective arguments is indeed bringing in "belief" into what doesn't need it. >> It's completely correct. Mathematically, even. > [quoted text clipped - 112 lines] > Falling back on subjective arguments is indeed > bringing in "belief" into what doesn't need it. Your tiff images are still up on my website and can be gotten here. http://www.sewcon.com/noons/ I did take one of them 06-08-04try-nis4-du.tif and down sampled it to a 8MP image and then back up to your original size, no difference that I can see. Really a jpeg has the same detail as a tiff, just more artifacts. Scott > Your tiff images are still up on my website and can be gotten here.http://www.sewcon.com/noons/ > > I did take one of them 06-08-04try-nis4-du.tif and down sampled it to a > 8MP image and then back up to your original size, no difference that I > can see. sorry, thoise are old scans, done when I didn't know as much about the 9000 as I know now. You'd have to get new files. And not use lanczos: it introduces its own "detail"... > Really a jpeg has the same detail as a tiff, just more artifacts. Disagree. The whole purpose of jpg compression is to save space by losing detail. And that is what happens when you start from a jpg file: you have lost detail right there. You might not see it, or it might not be objectionabled: that's subjective. But you have lost it. Simple as that. >> Your tiff images are still up on my website and can be gotten here.http://www.sewcon.com/noons/ >> [quoted text clipped - 6 lines] > You'd have to get new files. And not use > lanczos: it introduces its own "detail"... Your FTP account is still there, so feel free to up load your current best stuff. >> Really a jpeg has the same detail as a tiff, just more artifacts. > [quoted text clipped - 6 lines] > subjective. But you have lost it. > Simple as that. So show me one tiff photo that can't be turned into a jpeg without loss of detail. Scott > Your FTP account is still there, so feel free to up load your current > best stuff. No need. Go to http://members.iinet.net.au/~nsouto/photos/ adn check the sizes of "pink wave org.jpg" and "pink wave tst.jpg". Look at them if you want. Here is how I created them: 1- load the original tiff scan file in GIMP. 2- downress with "Cubic" to 4000X2649, or approximately 10MP. 3- make that into a jpg labeled "org", with 100% "quality" and sub-sampling 4:2:2. that's the 1st file above. 4- open that file, then downress by 50% using Cubic, then upress back to 4000 width. 5- Save with same settings, labeled "tst". that's the 2nd file above. Now: David tells me he got no difference in size, with jpgs, when he did a similar process. As you can see, there is indeed quite a significant difference in size, even though we are using jpgs with their inbuilt "loss". So, is this or is this not a 10MP detail image? And this begs the question: why is David getting a "no change in size"? Is he using a different handler for jpgs? This is why I feel it's of capital importance to disclose ALL details about the process used: had I used Irfanview with lanczos for example, the results would have been incredibly different! And note that I downressed from what is for all intents and purposes a > 16MP original: had I used that one, there would have been even more difference. I know, I tried. > So show me one tiff photo that can't be turned into a jpeg without loss > of detail. er sorry, I don't get this one. I reckon ANY tiff image will be turned into a jpg with loss of detail. Thats is my whole point. You get more or less loss depending on how much you tweak the jpg settings, but you do get SOME loss. > Now: > David tells me he got no difference > in size, with jpgs, when he did a similar > process. I don't recall ever saying or intending to say anything about the size of jpegs in this discussion. In this thread, I've only been concerned with the visual appearance of the files, since that's what determines how good the prints will look. To the best I can tell, jpeg or not is irrelevant, since (if it's an effect at all) it's not large enough to be significant. David J. Littleboy Tokyo, Japan >> Your FTP account is still there, so feel free to up load your current >> best stuff. [quoted text clipped - 28 lines] > "loss". So, is this or is this not a 10MP > detail image? I did not see where David said there was no difference in the size of the file, what matters is not if there is a difference in the size of the image after down sampling and back up, what you are looking for is a visible loss of detail. I took your original image and did the down sample and back up and could not see a change in detail. In other words I can produce a file that has about 2.5MP that has all the visible detail that is in your image. I have resized one of my images from the 350D to 4000 pixels wide, you can see it here. http://www.pbase.com/konascott/image/87798130/original This is a far sharper image then your 4000 pixel wide image, here is a crop from both side by side. http://www.pbase.com/konascott/image/87798324/original The resolution is set at 300ppi, try printing it as that and you should see that your image is very soft when printed at 300ppi, mine looks pretty good. Next I did one more test, I resize the above down to 50% and back to 100%, your image show no real change whereas mine has clearly lost a ton of detail. http://www.pbase.com/konascott/image/87798435/original So if your image has the same level of detail that mine does how is it that when I down size and back up your image is untouched by mine is pretty much trashed? Scott > And this begs the question: why is David > getting a "no change in size"? [quoted text clipped - 25 lines] > the jpg settings, but you do get SOME > loss. > I did not see where David said there was no difference in the size of > the file, what matters is not if there is a difference in the size of > the image after down sampling and back up, what you are looking for is a > visible loss of detail. So, the criteria has changed yet again, hasn't it? You both claimed that you down-upressed and got the same size file and that therefore there wasn't any detail in it past 4MP. Which is blatantly wrong in view of the results I just showed you and using the criteria that you both put forward first. When are you going to agree on an objective criteria that doesn't change? > I took your original image and did the down sample and back up and could > not see a change in detail. In other words I can produce a file that > has about 2.5MP that has all the visible detail that is in your image. That is a complete lie. > I have resized one of my images from the 350D to 4000 pixels wide, you > can see it here.http://www.pbase.com/konascott/image/87798130/original ah yes, another stitched image to "prove" that digital is "sharper"and "cleaner". Starting to sound very tired, Scott. Veeeeery tired indeed... When will you learn that I KNOW what a dslr can produce BECAUSE I have one? What you are showing there is a stitched image, which is NOT comparable to any single image, be it dslr or film! > This is a far sharper image then your 4000 pixel wide image, here is a > crop from both side by side.http://www.pbase.com/konascott/image/87798324/original Of COURSE it is! It is a STITCHED image made of quite a few other images. There is no way in the world a 10MP dslr can take that sharp a SINGLE image. > The resolution is set at 300ppi, try printing it as that and you should > see that your image is very soft when printed at 300ppi, mine looks > pretty good. Because my image is made out of a single original, instead of a collage of many others like yours is. Perhaps when you learn to compare apples with apples you'll be able to reason properly? > Next I did one more test, I resize the above down to 50% and back to > 100%, your image show no real change whereas mine has clearly lost a ton > of detail.http://www.pbase.com/konascott/image/87798435/original Really? That's funny, because when I take mine down and up, I can see the difference in detail. But then again, I'm working from tiff, with the original... > So if your image has the same level of detail that mine does how is it > that when I down size and back up your image is untouched by mine is > pretty much trashed? Because your image is a false one, made up of many stitched other images and therefore in no way representative or comparable to a single sensor image. I challenge you openly to post a single image, NOT stitched, from a 10MP sensor, taken with a wide lens of a reasonable depth scene that can even approach the level of detail of mine. And I'll be showing mine at that rez with full detail and full sharpening, something I haven't done yet. Simple as that. Until you do so, you are simply shooting air, Scott. And I'll show not just one but both of the images I noted initially although of course you both picked on the one with less detail. >I challenge you openly to post a single image, >NOT stitched, from a 10MP sensor, >taken with a wide lens of a reasonable depth >scene that can even approach the level of detail >of mine. Canon 1Ds vs. scanned 645: http://www.jjsviewbox.com/Articles/6451ds.htm You're wrong, Noons. Give it up. This has nothing to do with TIFF vs. JPG or the fact that Scott creates composite digital images. rafe b www.terrapinphoto.com >>I challenge you openly to post a single image, >>NOT stitched, from a 10MP sensor, [quoted text clipped - 5 lines] > > http://www.jjsviewbox.com/Articles/6451ds.htm Another good one. It's too bad that the bloke at Luminous Landscape was so wrong on the 1Ds vs. 6x7 test. Sigh. Here, whatever I do, 6x7 has no trouble womping the 5D, as long as there's some fine detail in the subject. Of course, that difference can't be seen on a 12x18 print, which is why said bloke messed up so badly. (Also, the Imacon scanners are dogs (or at least his was) and just don't get the detail.) I like the comment at the bottom of this page to the effect that "The significant difference between the Hasselblad and the 35mm film surprised me as well." ROFL. I started out in 6x6, and every time I've tried 35mm I've been aghast at how poor it is. http://www.ales.litomisky.com/projects/Analog%20versus%20Digital%20Shootout%20(H asselblad,%2035mm,%20Canon%205D).htm > You're wrong, Noons. Give it up. > > This has nothing to do with TIFF vs. JPG or the > fact that Scott creates composite digital images. Yep. It's about how soft film is when printed at 13x and how good digital looks at that same 13x. David J. Littleboy Tokyo, Japan > >I challenge you openly to post a single image, > >NOT stitched, from a 10MP sensor, [quoted text clipped - 5 lines] > > http://www.jjsviewbox.com/Articles/6451ds.htm since when is that one of my images? And which part of: "I don't accept an unprocessed film scan against a hyper-processed dslr image as a fair comparison" can't you folks understand? > You're wrong, Noons. Give it up. I'll do so when I see a relevant and consistent comparison made that proves dslrs can handle detail better than MF. No, NONE of the examples on the net are of any use. For the very simple reason stated above, so please don't bother posting more urls that I've seen umpteen times. > This has nothing to do with TIFF vs. JPG or the > fact that Scott creates composite digital images. Oh yes, it very much is. Once someone comes up with a criteria that doesn't change every 5 minutes. Recently, Noons <wizofoz2k@yahoo.com.au> posted: >> Canon 1Ds vs. scanned 645: >> >> http://www.jjsviewbox.com/Articles/6451ds.htm This is a curious choice to "prove" quality differences... A top-of-the-line DSLR compared against a Bronica with a zoom lens? Better would be a top-of-the-line camera and lens. The "film" was scanned in all likelihood without a glass film holder, as it would have been worth mentioning if it was. My ArtixScan 120tf is essentially the same scanner, and there is a *huge* difference in scann quality when the glass film holder is used. Still, comparing top-of-the-line options would suggest a professional drum scan (and not a "fake" drum scan e.g. something like the Imacon). Not terribly objective and is quite typical of the comparisons I've seen on-line, but it "confirms" what some people want to believe. Neil > Still, comparing top-of-the-line options would suggest a professional drum > scan (and not a "fake" drum scan e.g. something like the Imacon). OK. http://www.ales.litomisky.com/projects/Analog%20versus%20Digital%20Shootout%20(H asselblad,%2035mm,%20Canon%205D).htm David J. Littleboy Tokyo, Japan Recently, David J. Littleboy <davidjl@gol.com> posted: >> Still, comparing top-of-the-line options would suggest a >> professional drum scan (and not a "fake" drum scan e.g. something >> like the Imacon). > > OK. http://www.ales.litomisky.com/projects/Analog%20versus%20Digital%20Shootout%20(H asselblad,%2035mm,%20Canon%205D).htm I do like this comparison much better than Rafe's, David, but, it still has issues. As presented: * Scanning details would be a good thing, rather than just stating the type of scanner. What resolution and optimizations were used to make the scan? One can purchase "cheap" professional scans, and one can do poor scans on a good scanner, too. ;-) * Post-processing adds too many variables for the both digital and film workflows. I can see no reason to involve these manipulations in an objective comparison. Finally, I don't think that the results support Litomisky's conclusions. Even accepting the above issues and just looking at these shots as presented raises questions. The image from the 5D has more contrast, but loses detail (a condition that is favored by Photoshop's bicubic algorithms, btw), as can be seen by comparing interior portions of the branches, faces of the leaves, and mortar between the bricks to the Hasselblad shot. So, I still think that what one thinks is "best" all depends on what one wants in an image. Neil > Recently, David J. Littleboy <davidjl@gol.com> posted: > [quoted text clipped - 26 lines] > So, I still think that what one thinks is "best" all depends on what one > wants in an image. What I see is pretty typical of film vs. digital. Film can get fine detail, as long at the contrast in the detail is very high. But more subtle parts of the scene the film misses all together, in the 5D image I can see that the leafs are not flat but have a ripple to them, in the Hasselblad image this is mostly missing. If you can look at both the Hasselblad image and the 5D and say that you think the Hasselblad has better image quality then you would watch a naked man walk down the street and say that "yes the emperors new cloths are mighty fine", IMO. Scott Recently, Scott W <biphoto@hotmail.com> posted: >> Recently, David J. Littleboy <davidjl@gol.com> posted: >> >>>> Still, comparing top-of-the-line options would suggest a >>>> professional drum scan (and not a "fake" drum scan e.g. something >>>> like the Imacon). >>> OK. http://www.ales.litomisky.com/projects/Analog%20versus%20Digital%20Shootout%20(H asselblad,%2035mm,%20Canon%205D).htm >> I do like this comparison much better than Rafe's, David, but, it >> still has issues. As presented: [quoted text clipped - 24 lines] > image I can see that the leafs are not flat but have a ripple to > them, in the Hasselblad image this is mostly missing. Neither the Hasselblad nor 5D rendering of the leaves look flat on my monitor. > If you can look at both the Hasselblad image and the 5D and say that > you think the Hasselblad has better image quality then you would > watch a naked man walk down the street and say that "yes the emperors > new cloths are mighty fine", IMO. I wouldn't say either one of them is "better". They differ, and I've merely pointed out some of the differences. Let's not overlook the fact that whether this particular image "works" or not is not the issue. If details are lost, as they are with the 5D rendering, that can be just as much a problem as a film image being less contrasty. Which camera might do a "better" job comes down to what is important to present in the image, and that would be the basis for my picking up one or the other. Neil > Recently, David J. Littleboy <davidjl@gol.com> posted: >>> [quoted text clipped - 8 lines] > I do like this comparison much better than Rafe's, David, but, it still > has issues. It demonstates that 645 and the 5D are rather similar beasts, and that 35mm is sick by comparison. Which of 645 and 5D is "better" is an argument that splits hairs far finer than I'm interested in. All these make it abundantly clear that 645 and 5D are _roughly_ equivalent. If you find one of these not good enough for a particular application (e.g. a 30 x 40 advertising poster that people can walk up to), what you really need is a larger format or a lot more pixels. The other won't be enough of an improvement. That is, 645 is, for all practical purposes, equivalent to 12 or 16MP or so. 645 is "worth" a lot more than 8MP and a lot less than 24MP. (With decent technique, 645 is always better than 8 and worse then 24MP.) So 6x7 (having 1.66 times the information of 645) is "worth" a lot more than 13MP and a lot less than 40MP. Which is rather different than QG's 100MP. QG is off by a factor of four. > As presented: > > * Scanning details would be a good thing, rather than just stating the > type of scanner. What resolution and optimizations were used to make the > scan? One can purchase "cheap" professional scans, and one can do poor > scans on a good scanner, too. ;-) Quibble quibble. Those are quality, professional drum scans. > * Post-processing adds too many variables for the both digital and film > workflows. I can see no reason to involve these manipulations in an > objective comparison. Those are the standard, agreed techniques for making high-quality enlargements. But I see your point. He's playing a similar game to this guy. http://www.shortwork.net/equip/review-1Ds-SQ-scantech/ That is, they're both making prints that are (in my opinion) way beyond reasonable for either format. But that's actually a good way to compare, because the differences are more visible. At 12x18 (where I think 645 and the 5D womp 35mm), some people think 35mm looks great. At 16x20 (where I think 6x7 womps 645/5D), some people think 645 and the 5D look great. > Finally, I don't think that the results support Litomisky's conclusions. > Even accepting the above issues and just looking at these shots as [quoted text clipped - 3 lines] > branches, faces of the leaves, and mortar between the bricks to the > Hasselblad shot. He's using Velvia, and the grain is seriously ugly at that enlargement. Provia 100F wouldn't look so bad, and TMX100 would probably hold a bit more detail. > So, I still think that what one thinks is "best" all depends on what one > wants in an image. We're in some sort of agreement here: I don't think there's enough of a difference to call a winner in the 645 vs. 5D fight. David J. Littleboy Tokyo, Japan >> Recently, David J. Littleboy <davidjl@gol.com> posted: >>>> [quoted text clipped - 28 lines] > than 13MP and a lot less than 40MP. Which is rather different than QG's > 100MP. QG is off by a factor of four. First, I don't want to argue with someone as knowledgable as Mr. Littleboy. (suck-up mode off) The figures I've heard were that 35mm (24x36mm) film was roughly equal to 32-36 Mpixels. This would put 645 equal to about 100 Mpixels, based on pixels per square millimeter. Where does my math fall apart? I also wonder about making comparisons between a digital photo and a film negative online. After all, the film neg has to be scanned to be posted online. It seems to me that the only valid comparison would be to view both photos in person. What time should I come over?! > The figures I've heard were that 35mm (24x36mm) film was roughly equal to > 32-36 Mpixels. This would put 645 equal to about 100 Mpixels, based on > pixels per square millimeter. Where does my math fall apart? Well, 32-36 Mpixels is unrealistic for any normal 35mm film. It would be enough for a really sharp 16x24 inch print. (Something like Eastman 5360 is easily sharp and grainless enough, but it doesn't qualify as a normal film.) 16*300*24*300 = 34560000 or about 33 megapixels. Medium speed 35mm film can make a really nice 8x12 in good cases. That means 8.25 megapixels is more realistic. Peter.  Signaturepirwin@ktb.net >>> http://www.ales.litomisky.com/projects/Analog%20versus%20Digital%20Shootout%20(H asselblad,%2035mm,%20Canon%205D).htm >>>> [quoted text clipped - 24 lines] > 32-36 Mpixels. This would put 645 equal to about 100 Mpixels, based on > pixels per square millimeter. Where does my math fall apart? There's nothing wrong with your math. The problem is with the meaning of "roughly equal to 32-36 Mpixels." In real life, 12MP dSLR images simply look way better than 35mm. In prints. Every time. It's not even close. So what does it mean to say 35mm is "roughly equal to 32-36 Mpixels"? I don't know. It seems to be complete nonsense. What does it mean to you? David J. Littleboy Tokyo, Japan >>>> http://www.ales.litomisky.com/projects/Analog%20versus%20Digital%20Shootout%20(H asselblad,%2035mm,%20Canon%205D).htm >>>> I do like this comparison much better than Rafe's, David, but, it still [quoted text clipped - 36 lines] > > What does it mean to you? This is my take on it. There seems to be a lot of confusion between how many pixels are needed to get all the information off of film vs. how many pixels a digital camera needs to match the image quality of film. To get ALL(I don't use cap often) the information off of film you need to consider scene contrast that is 1000:1 and recorded on the film at the 5% point on the MTF curve. This is only useful if you are trying to read something like high contrast text, say a license plate. But for image quality the ability to turn 1000:1 contrast into 100:95 contrast has little benefit. The best test, IMO, it to photograph a typical scene and make prints from both and compare the prints. With the wonders of the internet anyone with a somewhat decent printer can make prints from my files and I can make prints from theres. I have done this once with a friend, it was the last 35mm film he ever shot. Scott >>>> http://www.ales.litomisky.com/projects/Analog%20versus%20Digital%20Shootout%20(H asselblad,%2035mm,%20Canon%205D).htm >>>>> [quoted text clipped - 42 lines] > David J. Littleboy > Tokyo, Japan What I understand it to mean is that a 35mm frame contains 32-36 Mega-'grains'. If the quantity of pixels is an indication of the amount of picture information that can be captured by a dSLR, then the quantity of grains should be an indication of the amount of picture information that can be captured on film. Of course, I'm not considering other factors: lens quality (either shooting or printing), subject matter, etc. If one is to compare the 'potential' quality of the method of image capture (film or digital sensor), you would have to compare the individual light sensing elements-- pixels or grains. Comparing the final print seems to me to be a subjective method that introduces too many other factors. If I shoot with a Holga and evaluate the prints, does that mean that all 120 size film is soft focus with light falloff toward the edge of the frame? That same 120 film is going to yield better quality shot with a Hasselblad or even a Kiev, both of which may or may not use the full potential of the film's grain. >> There's nothing wrong with your math. >> [quoted text clipped - 17 lines] > can be captured on film. Of course, I'm not considering other factors: > lens quality (either shooting or printing), subject matter, etc. But in real life, if you look at the images, 35mm (ISO 100, decent lens) produces prints that are slightly better than those produced by 6MP cameras, about the same as those produced by 8MP cameras, and really bad compared to prints produced by 12MP cameras. > If one is to compare the 'potential' quality of the method of image > capture (film or digital sensor), you would have to compare the individual > light sensing elements-- pixels or grains. The geometry of how they form images is completely different, so that doesn't make sense. > Comparing the final print seems to me to be a subjective method that > introduces too many other factors. Hmm. I thought the object of photography was to make photographs. Silly me. It's really to brag how many photosensitive elements were used. One learns something new every day. David J. Littleboy Tokyo, Japan >>> There's nothing wrong with your math. >>> [quoted text clipped - 29 lines] > The geometry of how they form images is completely different, so that > doesn't make sense. This is where my brain falls down and goes boom. I realize that there is a difference in how the image is formed: light strikes film grain, causes latent change, yada, yada, or light strikes image sensor, causes an electrical charge, yada, yada.... But whether it be grains of silver or silicon pixels, all the little 'bits' add up to make a photo. And if one system has more "little bits" than another, that one should be able to capture more detail, potentially, right? >> Comparing the final print seems to me to be a subjective method that >> introduces too many other factors. [quoted text clipped - 3 lines] > > One learns something new every day. OK, you scored big points with that line! I've got to go next door to the fire station and have them use the jaws of life to get my head out of where the sun don't shine! Of course the object of photography is to make photographs (perhaps even 'create a work of art'!). But I'm asking about the potential. Can a digital image sensor potentially capture more detail per square area than a piece of film? Recently, Ken Hart <kwhart@fullnet.com> posted: > [...] But > I'm asking about the potential. Can a digital image sensor > potentially capture more detail per square area than a piece of film? One of the problems that I have with this kind of discussion is the terminology. I have the impression that we are not all using terms such as "detail" in the same way. However, your question about _potential_ rather than real-world examples is problematic for other reasons. For digital technology, the number of sensor sites per square area is in constant flux. The 12MP sensor of today is no larger, and in many cases smaller than the 8MP sensor of a few years ago. So, the _potential_ of digital sensor technology is so far unrealized. OTOH, film is a mature technology, where recent improvements are not so much in the area of increased resolution or detail, but in better performance at higher ISO ratings. The resolution of currently available films is below such films as discontinued Panatomic-X and Kodachrome 25 from decades ago. However, films such as TMax and Portra 400 far outperform ISO 160 films of that period. Another consideration is that while there is a higher count of film grains per square area than sensor sites in today's digital sensors, not all of those grains contribute to the captured image (for various reasons) while most of the digital sensor sites do. Add to that losses of resolution and detail when printing optically or by scanning the film, and the final image is considerably more limited than the number of grains per square area might suggest. Perhaps the best way to establish equivalence is to compare the final printed image rather than trying to find a mathematical solution. And, IMO that's going to be a pretty subjective determination for some time, yet. Neil >> The geometry of how they form images is completely different, so that >> doesn't make sense. [quoted text clipped - 6 lines] > system has more "little bits" than another, that one should be able to > capture more detail, potentially, right? Apparently not. >>> Comparing the final print seems to me to be a subjective method that >>> introduces too many other factors. [quoted text clipped - 10 lines] > the potential. Can a digital image sensor potentially capture more detail > per square area than a piece of film? In real life, the 5D makes a rather nice 12x18 print. Something I find impossible to do from 24x36mm of film. And I've tried very hard, many many times over the years. You really need medium format to make a nice 12x18 print. Seriously, film turns to mush at 13x. The 5D doesn't. And despite Neil's quibbling, this isn't about scanning (and whether or not the operator was as perfect as Neil), it's about the better professional films that people actually use not standing up to a 13x enlargement. If your theory doesn't explain that, it's problematic as a theory. David J. Littleboy Tokyo, Japan Recently, David J. Littleboy <davidjl@gol.com> posted: > Seriously, film turns to mush at 13x. The 5D doesn't. And despite > Neil's quibbling, this isn't about scanning (and whether or not the > operator was as perfect as Neil), it's about the better professional > films that people actually use not standing up to a 13x enlargement. Please, David. I am not claiming that making 13x enlargements is a reasonable practice if one is as allergic to grain as you. Nor am I talking about 35mm film being competitive with top-of-the-line digital OR MF w/r/t producing "clean" images. OTOH, I am not one to discount the opinions of people who like to emphasize the grain in their images. To claim that grain always destroys image quality is analogous to claiming that Monet didn't know how to paint because you could see his brush strokes. My contributions to this thread have been limited to the on-line comparisons between digital and MF that you and Rafe posted. I wrote to address the complete lack of objectivity and/or technical controls in those so-called comparisons between formats. In short, as I've written several times now, those comparisons prove nothing, and even your own comments about Litomisky's choice of film validate that quibble. Neil snip > In real life, the 5D makes a rather nice 12x18 print. Something I find > impossible to do from 24x36mm of film. And I've tried very hard, many many [quoted text clipped - 5 lines] > as perfect as Neil), it's about the better professional films that people > actually use not standing up to a 13x enlargement. I guess this is where my confusion sets in. I have many times made 11x14 and 16x20 prints from 35mm. I'll grant that the negative must be needle sharp and properly exposed, but still, the prints look good. I've got a 20"x30" enlargement of a sunset over the Gulf of Mexico taken on board a cruise ship with 35mm, I don't remember anymore what film, but it would have been a C22 process consumer film. You can clearly see grain when you get up close (within inches), but then again, you don't normally view a photo of that size from 6" away. (Isn't there a guideline that viewing distance should be equal to the diagonal?) The detail is sharp: lights around the railing of the ship, the wave tops. The color has good contrast and saturation: it's a very warm tone photo with shades of orange, red, and yellow. This is my confusion: you say a 35mm won't go to a wall print. And I print them frequently. Is my Canon FX and my Omega D2V with the Schneider lens that much better than whatever you use for 35mm (that's not intended to be a put-down, but an actual question)? > snip >> In real life, the 5D makes a rather nice 12x18 print. Something I find [quoted text clipped - 15 lines] > get up close (within inches), but then again, you don't normally view a > photo of that size from 6" away. Landscape work assumes the viewer will come in for a closer look. And if you watch visitors in a gallery, you'll see them coming in for a closer look. If you want to argue that 35mm is adequate for 20x30 prints and one doesn't need MF, that's really a different discussion. Just as Holgas and Neil's complaining about my dislike of grain are. Also, given a strong image, whatever the technology used to produce it, much of the time, it'll have more impact (and look "better") the larger you print it. So the digerati find that they make insanely large prints and are very happy with them. That doesn't change the fact that larger film (and more MP) produces better images. > (Isn't there a guideline that viewing distance should be equal to the > diagonal?) The detail is sharp: lights around the railing of the ship, the > wave tops. The color has good contrast and saturation: it's a very warm > tone photo with shades of orange, red, and yellow. Well, you don't need medium format or higher MP count digital, then. But that doesn't change the fact that you'd have a better print had you used MF. And that you'd be at a disadvantage if you were showing prints in a gallery next to MF and LF folks. > This is my confusion: you say a 35mm won't go to a wall print. And I print > them frequently. Is my Canon FX and my Omega D2V with the Schneider lens > that much better than whatever you use for 35mm (that's not intended to be > a put-down, but an actual question)? I've used a variety of 35mm over the years, most recently Olympus OM. But my comments about 24x36mm film are based on what I see from the center 24x36 from various MF cameras (Mamiya 7, Fuji GS645S, Mamiya 645, Rolleiflex MXV3.5 (Tessar)), scanned on a Nikon 8000 and checked with a 60x microscope to verify that the scan got what was there. David J. Littleboy Tokyo, Japan >snip > [quoted text clipped - 3 lines] > Rolleiflex MXV3.5 (Tessar)), scanned on a Nikon 8000 and checked with a > 60x microscope to verify that the scan got what was there. As soon as you mentioned a 60x microscope, I decided that your standards are a bit different from mine! I respectfully thank you for your viewpoint, you have given me some food for thought. For the record, I shoot 35mm for my own pleasure. In the studio, I shoot 6x6 mainly, sometimes 645, 6x7, or 4x5". I'll make you a deal: when you bring me to Tokyo, I'll shoot with my KoniOmega Rapid M! >> snip >> [quoted text clipped - 10 lines] > mainly, sometimes 645, 6x7, or 4x5". I'll make you a deal: when you bring me > to Tokyo, I'll shoot with my KoniOmega Rapid M! He is of course using the microscope to look at the negatives not the prints. The reason is to check if the scanner is capturing all the detail that is on the film. It really has little to do with standards. Scott Recently, David J. Littleboy <davidjl@gol.com> posted: > I've used a variety of 35mm over the years, most recently Olympus OM. > But my comments about 24x36mm film are based on what I see from the > center 24x36 from various MF cameras (Mamiya 7, Fuji GS645S, Mamiya > 645, Rolleiflex MXV3.5 (Tessar)), scanned on a Nikon 8000 and checked > with a 60x microscope to verify that the scan got what was there. Looking at the center 24x36 of a MF is a good way to compare the differences between 35mm and MF lenses. It can be presumed that the same area of the same type of film will have similar characteristics. I also have some Olympus OM systems, but also a Leica R system. The differences between these two w/r/t image qualities are pretty easy to see (not "better or worse", unless talking about a particular parameter). The Leica will image a 24x36 area of film with much better resolution than either my Rolleiflex MX (3.5 Tessar) or 6008i with Schneider or Zeiss lenses, but it obviously won't make a better 20" and larger print than either Rollei. Neil > snip >> In real life, the 5D makes a rather nice 12x18 print. Something I find [quoted text clipped - 23 lines] > that much better than whatever you use for 35mm (that's not intended to be a > put-down, but an actual question)? A very good way to get an idea of the effects of resolution on a print is to take your best shoot with your lens of choice, say a good 50mm prime lens. Take this shot with the camera on a tripod and then change to a longer lens, say 200mm or so. Now print a crop from the 50mm shot that would be printed to say 20 x 30 inches if you printed the full image. Print the shot from the 200mm lens so that it has the same magnification and compare. We talk about detail, and detail is nice but what really bothers me in large prints that don't have enough resolution is the lost of texture, in some images this does not matter but in a lot it makes a big difference. For a given size print you reach a point where more resolution does not improve the quality of the print, for 35mm this is about an 8 x 12 and about the same from a 8MP digital. But go up to a 12x18 inch print and MF will produce a noticeable better print then 25mm and a 12MP camera will produce a noticeable better print then a 8MP camera. When you do the side by side viewing of the same scene shot at different resolution you find that what looked good by itself no longer looks all that good, at least this is what I find. Scott > We talk about detail, and detail is nice but what really bothers me in > large prints that don't have enough resolution is the lost of texture, in > some images this does not matter but in a lot it makes a big difference. But that's where grain comes in. Film doesn't need to image textures since it provides its own. David J. Littleboy Tokyo, Japan >>>> There's nothing wrong with your math. >>>> [quoted text clipped - 49 lines] > potential. Can a digital image sensor potentially capture more detail per > square area than a piece of film? For any normal film the answer is digital can capture way more detail / sq inch. A 1/2.5 inch sensor that has 8MP has about 14,000 pixels / inch. Whereas the image may not be real share at the pixels level on these sensors a good camera is far sharper then film scanned at even 4000 ppi, an example can be seen here. http://www.dpreview.com/reviews/panasonicfz18/page11.asp A full 35mm frame at that pixel density would have about 280 MP, of course not many lenses could use a fraction of that resolution. If you put that many pixels per inch on a 4x5 inch sensor you would have a bit over 4 billion pixels. There are reason why making your pixels that small is not a good idea, much lower noise and better dynamic range can be had with larger pixels. So take the case of my 1.6 crop factor cameras, they have very close to 4000 ppi, and capture far more detail per pixel then a 4000 ppi film scan. As a way to detect photons film is incredibly inefficient, the only real advantage is that you can use a huge area of film fairly cheaply where as the same area for digit sensor would cost a fortune. Scott >>> Recently, David J. Littleboy <davidjl@gol.com> posted: >>>>> Still, comparing top-of-the-line options would suggest a [quoted text clipped - 36 lines] > online. It seems to me that the only valid comparison would be to view both > photos in person. What time should I come over?! If you want to come to Hawaii I would be delighted to compare prints. Scott Recently, David J. Littleboy <davidjl@gol.com> posted: >> Recently, David J. Littleboy <davidjl@gol.com> posted: >>>> [quoted text clipped - 3 lines] >>> >>> OK. http://www.ales.litomisky.com/projects/Analog%20versus%20Digital%20Shootout%20(H asselblad,%2035mm,%20Canon%205D).htm >> I do like this comparison much better than Rafe's, David, but, it >> still has issues. > > It demonstates that 645 and the 5D are rather similar beasts, and > that 35mm is sick by comparison. Was 35mm relevant to this dicussion? > Which of 645 and 5D is "better" is an argument that splits hairs far > finer than I'm interested in. Unfortunately the issues I brought up regarding the technique can not assure that conclusion. One would need to address those issues to know one way or the other. >> As presented: >> [quoted text clipped - 4 lines] > > Quibble quibble. Those are quality, professional drum scans. How do you know anything about either the options that Litomisky requested (output intent), or even just the resolution of the scan? I'm willing to give the benefit of the doubt to the scanner operator, but it still warrants mentioning that there are differences between operators on the same piece of equipment at the level of a Tango. >> * Post-processing adds too many variables for the both digital and >> film workflows. I can see no reason to involve these manipulations >> in an objective comparison. > > Those are the standard, agreed techniques for making high-quality > enlargements. But I see your point. Making high-quality enlargements involves optimization for a particular output device, medium and method. Performing post-processing operations is inappropriate to use as an objective comparison of captured raw image information. > He's playing a similar game to this guy. > [quoted text clipped - 5 lines] > But that's actually a good way to compare, because the differences > are more visible. I don't think it is a good way to compare, because it obliterates the original content and injects a lot of personal decisions. >> Finally, I don't think that the results support Litomisky's >> conclusions. Even accepting the above issues and just looking at [quoted text clipped - 7 lines] > enlargement. Provia 100F wouldn't look so bad, and TMX100 would > probably hold a bit more detail. So, these are some other factors that raise questions about whether the comparison supports the writer's conclusion. One can't even be sure whether the 5D image is as good as a 5D can do because of the post-processing, but for certain the Hasselblad shot isn't as good as a Hasselblad could do on the scene because of film choices *and* post-processing. >> So, I still think that what one thinks is "best" all depends on what >> one wants in an image. > > We're in some sort of agreement here: I don't think there's enough of > a difference to call a winner in the 645 vs. 5D fight. Yes, we're not that far off, and in my opinion, both the Leica DMR and M8 blow the 5D out of the water. My point is that this comparison doesn't prove the notion that digital beats MF any more than some of the others. It seems that these people are more interested in supporting their notions than providing objective proof one way or the other. Neil > Recently, David J. Littleboy <davidjl@gol.com> posted: >>> Recently, David J. Littleboy <davidjl@gol.com> posted: [quoted text clipped - 14 lines] >> > Was 35mm relevant to this dicussion? Yes. 35mm and 6x7 are critical to this discussion. >> Which of 645 and 5D is "better" is an argument that splits hairs far >> finer than I'm interested in. >> > Unfortunately the issues I brought up regarding the technique can not > assure that conclusion. One would need to address those issues to know one > way or the other. I think that we're on different wavelengths again. For |
Enable EMail Alerts
Start New Thread
Reply to this Message