If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Display Modes |
#512
|
|||
|
|||
Hi Stephen, ;-) quoting you: No. 1,000:1 input produces some sort of discernible pattern on the film. That's not the same as saying that that whole range can be reproduced. Look at it this way: if you need 1000:1 input to get any output at all, that means that an input with 500:1 contrast produces no pattern whatever. endquote: Film is analog; if a film can support a 1,000:1 range, then it similarly means it can sustain lesser ranges like 500:1 etc. I have various density step wedges that I can use in the darkroom and testing, and I've never seen a consumer film or paper produce no pattern at lower levels??? Similarly, there are lots of tricks with developing and printing to compress or select the desired range recorded on the film for printing within the lesser (~5 stops) range of prints (or rather more for illuminated slide films being projected). I agree with you that digital sensors are now in the same range of most films as far as dynamic range goes (i.e., 10 to 12+ stops). That's good, because of the claims I've seen (e.g. 5 stops at Dr. Clark's site URL I cited) which seem to ignore the full dynamic or density range of film. Here is another cite, and recap, which is independent and allows you and others to follow the analysis used to derive that 12 stop value etc.: 12 stops for film, here is a cite to tables showing some (older) examples, plus a recap with cites and explanation from a past posting (archived on my film pages) on the 12 stops or EV range. quoting Richard Knoppow on Wed 03 Jun 1998 in Rollei user group* The latitude for modern B and W film can be as large as 12 stops. Dr. Richard Henry, in his book _Controls in Black and White Photography_ shows measurements of some films showing this. endquote: *see http://medfmt.8k.com/mf/filmp.html for full posting from the same site, quoting MikeWhy post in rpe-MF of 05-jan-2004: For discussion, the characteristic curve for Porta 160NC is at http://www.kodak.com/global/en/profe...009_0153ac.gif The horizontal axis is log exposure. It relates directly to EV, about .3 log exposure to 1 EV, or f-stop (Log10 of 2). The toe for all three layers in Portra begins at about log exposure -2.7; the slope is linear from about log exposure -2.1. The curve is still linear at the extreme right of the plot, at about +0.9. The dynamic range is thus from -2.7 to +0.9, or 12 EV. To recap: dynamic range is the log exposure range where the film exhibits useful density changes. EV and f-stops relate to log exposure as 1 EV for every 0.3010 change in log exposure. The characteristic curves for Portra 160NC shows a useful dynamic range of 12 f-stops, with as much as 15 not being unreasonable. E100G has a dynamic range of 9 f-stops. Conservatively, then, all ten of Ansel Adams's value zones fit easily on Portra, with one stop or more latitude above and below, while slide film barely squeezes in nine zones plus white. Negative films, here exemplified by Portra 160NC, have much broader dynamic range than slide films, and also more readily accommodates errors in exposure. end-quote: Again, I have cited tables of films with 10 to 12 stops of dynamic range and various examples of 12 stops or more ("as much as 15 not being unreasonable" above ;-), and my own tables showing many films specifications for 1,000:1 contrast ratio tests (see tables for dozens of films at http://medfmt.8k.com/mf/lenslpm.html etc.). I find it surprising that digital folks would even challenge that film has 10 to 12 or more stops of dynamic range (EV ranges)... regards bobm -- ************************************************** ********************* * Robert Monaghan POB 752182 Southern Methodist Univ. Dallas Tx 75275 * ********************Standard Disclaimers Apply************************* |
#513
|
|||
|
|||
Hi Stephen, ;-) quoting you: No. 1,000:1 input produces some sort of discernible pattern on the film. That's not the same as saying that that whole range can be reproduced. Look at it this way: if you need 1000:1 input to get any output at all, that means that an input with 500:1 contrast produces no pattern whatever. endquote: Film is analog; if a film can support a 1,000:1 range, then it similarly means it can sustain lesser ranges like 500:1 etc. I have various density step wedges that I can use in the darkroom and testing, and I've never seen a consumer film or paper produce no pattern at lower levels??? Similarly, there are lots of tricks with developing and printing to compress or select the desired range recorded on the film for printing within the lesser (~5 stops) range of prints (or rather more for illuminated slide films being projected). I agree with you that digital sensors are now in the same range of most films as far as dynamic range goes (i.e., 10 to 12+ stops). That's good, because of the claims I've seen (e.g. 5 stops at Dr. Clark's site URL I cited) which seem to ignore the full dynamic or density range of film. Here is another cite, and recap, which is independent and allows you and others to follow the analysis used to derive that 12 stop value etc.: 12 stops for film, here is a cite to tables showing some (older) examples, plus a recap with cites and explanation from a past posting (archived on my film pages) on the 12 stops or EV range. quoting Richard Knoppow on Wed 03 Jun 1998 in Rollei user group* The latitude for modern B and W film can be as large as 12 stops. Dr. Richard Henry, in his book _Controls in Black and White Photography_ shows measurements of some films showing this. endquote: *see http://medfmt.8k.com/mf/filmp.html for full posting from the same site, quoting MikeWhy post in rpe-MF of 05-jan-2004: For discussion, the characteristic curve for Porta 160NC is at http://www.kodak.com/global/en/profe...009_0153ac.gif The horizontal axis is log exposure. It relates directly to EV, about .3 log exposure to 1 EV, or f-stop (Log10 of 2). The toe for all three layers in Portra begins at about log exposure -2.7; the slope is linear from about log exposure -2.1. The curve is still linear at the extreme right of the plot, at about +0.9. The dynamic range is thus from -2.7 to +0.9, or 12 EV. To recap: dynamic range is the log exposure range where the film exhibits useful density changes. EV and f-stops relate to log exposure as 1 EV for every 0.3010 change in log exposure. The characteristic curves for Portra 160NC shows a useful dynamic range of 12 f-stops, with as much as 15 not being unreasonable. E100G has a dynamic range of 9 f-stops. Conservatively, then, all ten of Ansel Adams's value zones fit easily on Portra, with one stop or more latitude above and below, while slide film barely squeezes in nine zones plus white. Negative films, here exemplified by Portra 160NC, have much broader dynamic range than slide films, and also more readily accommodates errors in exposure. end-quote: Again, I have cited tables of films with 10 to 12 stops of dynamic range and various examples of 12 stops or more ("as much as 15 not being unreasonable" above ;-), and my own tables showing many films specifications for 1,000:1 contrast ratio tests (see tables for dozens of films at http://medfmt.8k.com/mf/lenslpm.html etc.). I find it surprising that digital folks would even challenge that film has 10 to 12 or more stops of dynamic range (EV ranges)... regards bobm -- ************************************************** ********************* * Robert Monaghan POB 752182 Southern Methodist Univ. Dallas Tx 75275 * ********************Standard Disclaimers Apply************************* |
#514
|
|||
|
|||
"Bob Monaghan" wrote: Lots of us have paid major $$ to buy pro quality lenses to get this extra level of performance too. I seem to remember someone doing some testing that demonstrated that people can't tell the difference between cheap and expensive prime lenses in actual prints. Oh, yes, that was none other than you yourselfg! So my point is that we already have the experience (at least as far as resolution is concerned) of the current digital vs. film quality debate. Shooting digital cameras with maximum resolutions of 55 lpmm (8.2 to 11 MP dSLRs) versus film cameras with resolutions of circa 88 lpmm is easy to setup experiment for comparison. And, as I explained in detail, you can't put 55 lp/mm of useful detail on the film with real film, normal to wide lenses (longer lenses do perform better at the edges), real subjects. Actually getting 88 lp/mm on the film at a usable contrast level is physically impossible with film; you are talking about barely detectable responses to artificially high contrast targets: most films have zero response to low-contrast targets above 60 lp/mm. By the way, has anyone seen measured MTF curves for real lenses at over 40 lp/mm? Again, people who look at the prints, find 11 MP better than 35mm. Simply because 11MP actually captures more perceptually valid detail than even the best currently available films. I think it also answers Stephen's query - does it matter? is this fine contrast and high resolution detail even seen in the image? The answer is evidently YES!, Well, you're simply wrong. it matters a lot, and many of us have spent kilobucks to buy the better lenses The good news for people willing to spend the money is that the better lenses perform better at 20 and 40 lp/mm, and so actually do look better in real prints. While the limiting resolution numbers are irrelevant to pictorial photography, they do _imply_ higher MTFs at the frequencies that are actually significant to real photography. which can deliver not just a paltry 55 lpmm but over 80 lpmm imagery. There are visible differences in crispness of images, esp. at boundaries of (facial, tree branch..) lines and complex imagery, where a high quality photograph is clearly superior to one shot with a cheapy zoom on fast film (i.e., lower resolution) ;-) I don't have a reference at hand, but one of the standard things you see in signal processing textbooks is the improvement in edge rendition as further terms in the Fourier transform are added in. You can't attenuate those terms by 90% and still see improvement in edge definition, it simple doesn't make a difference; the terms have to be close to their full amplitude to make a difference. Furthermore, those terms constitute an infinite series at fairly wide (2 octave) spacing, so moving from a sharp cutoff at 55 lp/mm to a long under 10% tail out to 100 or 120 lp/mm doesn't contribute to improved edge rendition. (All you get is one more term in the series at a tiny fraction of the amplitude necessary to affect the shape of the edge.) Similarly, shooting with a fog filter or softening filter or anti-aliasing (aka low pass filter) will all give similar experience with loss of resolution ("smoothing" or smearing) of the image details. Again, this is an easy "experiment" to run, and will show to each individual how such losses would impact their particular types and subjects of photography... This is, of course, completely silly. Softening filters reduce contrast throughout the frequency range. You could scan a frame at, say, 8000 dpi, and then bicubic downsample to 5600, 4000, and 2800 dpi, and print all four at A3 (upsampling the low res images appropriately), and compare to an A3 wet projection print and see how they differed. I'd bet that all four scanned prints look better than the wet projection print. That would be an interesting test. David J. Littleboy Tokyo, Japan |
#515
|
|||
|
|||
Brian C. Baird posted:
In article , says... Quite the opposite. Sorry. Then howcome you're completely unable to recognize totally unscientific 'testing' and writing? Because he isn't publishing papers in a major journal, he's providing some testing to show that digital can, and often does, exceed film quality. I've been avoiding this thread somewhat, but I'll try to put a different spin on things for once. I've glanced over Roger's site, and without getting into methodological arguments which I'm not qualified to make, I can say this - From the POV of someone who has done prepress work and is relatively familiar with the limitations, needs and parameters of printing and prepress equipment, the scans from the unnamed drum scanner look really bad. This needs to be investigated, since the noise and posterizing is definitely unusual for 3300 dpi drum scans. Again from a DTP perspective, which I believe applies here since we're disussing high end imaging and cameras that cost thousands, not family vacation snapshots, the golden rule is 300 dpi with no interpolation. If anyone plans on submitting photos for print in an album or magazine, digital will be limited in size. This isn't addressed on the page, although I feel that it's worth mentioning. The 20D @ 3504 x 2336 pixels will yield a printable image no more than 10" wide (color, 5" wide in grayscale) and the EOS 1Ds Mk II @ 4992 x 3328 gives an image measuring 16" (color, 8" grayscale) in width. Unless the customer takes responsibility for how it turns out, most prepress people won't take the risk of upsampling a digital image. After sharpening, even 10" and 16" may be too big with artifacts becoming visible. So albums and magazine advertising are rather out of bounds since you either lack image size or pixels for cropping. So it's stretching the truth to claim that digital can compete with 35 mm. In mid to high-end prepress it can't. I realize that my perspective is not that of the fine art photographer, who may enlarge his/her work and be personally accountable but in DTP the equipment is much less forgiving, and no one has three days to manipulate one image to tweak it for printing. In any discussion of such equipment, I believe that the DTP angle is valid, and needs to be addressed. -- Petros Ap' ola prin ipirche o Logos |
#516
|
|||
|
|||
Hi David, Can you take just the actual (non-interpolated) data from the digicam and print just the actual data values recorded by the sensor, without allowing interpolation in the imaging software or printer? Now if you use the same data and keep making larger prints, without allowing any interpolation, you get pixelated and bad looking prints, even at 8x10", right? Right? ;-0) I know you should, again from simple math. If we take 6 MP of data, and split it up amongst 80 sq. inches (an 8x10" print), we get 75k pixels per sq. in, or about 275 pixels per inch, which is just below the 300 dpi needed for a full frame print. As I've noted, this is a less than Leica standard critical sharpness print, as it is only 6 lpmm or so on the print vs. the 8 lpmm of the Leica standard (for an 8x10" print held at 10" etc.) Now if you do an 11x14" print, your actual data only permits 197 pixels/in. Some high end lenses and films can still provide that 8 lpmm on the print; for example, Zeiss tests of real world photos shows up to 200 lpmm is possible, well above this level, yes? ;-) At that level, your smoothed interpolated images start to break down and become obviously full of artifacts, while the zeiss lenses and fine grained films are showing us what the real world looks like ;-) And of course, the DSLR image data is actually interpolated from a Bayer pattern in the first place, with various effects on color, resolution etc. At some point you end up dealing with 75% or 90%+ interpolated values; it may be that such very smoothed images look good to many users, but the real world has a lot more contrast and details which is lost in such manipulated and interpolated imagery. grins bobm -- ************************************************** ********************* * Robert Monaghan POB 752182 Southern Methodist Univ. Dallas Tx 75275 * ********************Standard Disclaimers Apply************************* |
#517
|
|||
|
|||
Hi again David, quoting you again: No, it can't _in actual practice_. It can record barely detectable hints of a pattern for 1000:1 test charts using a lens that costs more than my annual income. That's a laboratory phenomenon completely unrelated to pictorial photography. Even 50 lp/mm _at enough contrast to be useful in quality imaging_ is essentially impossible in actual practice. end-quote: To which I say "nonsense" ;-) Look again at the cited URL in this thread (http://www.wlcastleman.com/equip/rev..._ccd/index.htm) - he shows 80+ lpmm for the canon EF 85mm f/1.8 lens, and shows the lens+digicams are delivering at least 50% MTF at 50 lpmm. In my own pages, I cite a 1970s photo lab study of cheapy 50mm normal lenses, which delivered 100 lpmm resolution on 1970s filmstocks. I have a number of these lenses, some of which were only $17 (w. s/h), $25 etc. ;-) see http://medfmt.8k.com/mf/limits.html#100, and quoting Pop Photo: October 1978 Modern Photography by Bennett Sherman and Al Gordon titled "How Sharp Can You Get" [p.112-3, table from p. 178, 182, 186]. They noted: "The first thing which can be inferred from the tables is that the 100 lines per mm level of image sharpness can be reached by many normal focal length lenses when the aperture is set at f/4 or f/5.6, and the "best" film is used." endquote: For a number of years, Modern Photography published lens tests with both lpmm and %contrast values at various f/stops. Lots of lenses had scores above 50% MTF. Many lenses can deliver 70% MTFs at 50 cps, see Hans cited URL at http://www.photodo.com/art/35_m9.shtml top chart ;-) The system resolution charts for film+ lens MTF% show system resolution values at 50 cps above 40% MTF (35mm) and 60% MTF for some of my medium format lenses ;-) Incidentally, I have found some of the zeiss planar 80mm hassy lenses for as little as $175 used, so even such high end lenses are affordable ;-) So I think thee doth protest too much - there are lots of affordable photo lenses which do a very good job at delivering high contrast imagery and high resolution details. grins bobm -- ************************************************** ********************* * Robert Monaghan POB 752182 Southern Methodist Univ. Dallas Tx 75275 * ********************Standard Disclaimers Apply************************* |
#518
|
|||
|
|||
Stephen H. Westin wrote:
It's way more complicated than that. We can ask: 1. What is the maximum range of exposure that will still produce a difference in output? (That's what I've been talking about) With modern B&W negative films, the range is huge. With older films like Ilford FP2 of the 1940s, the top of the shoulder was reached around 12 stops above the Jones point. You would still get a difference in output when you went past the top of the shoulder, but you would start to get decreasing density. If you overexposed FP2 by 12 stops, you would get a positive instead of a negative. This doesn't happen any more. With modern films you will get a very dense negative from which you can get a reasonable looking contact print. 2. What is the range of exposure that results in some sort of acceptably usable image? (A smaller range) With B&W negative film, it depends how much you want to enlarge the negative. Overexposure causes increased graininess (due to the extra density of the negative) and a loss of sharpness (due to irradiation, or light spreading within the emulsion.) Both of these effects will be hard to see on a contact print, but will become obvious with only a few stops overexposure on an 8x enlargement. Even one stop overexposure will show up when doing lens resolution tests. Number 2 is probably the most useful measure, but hard to come by, if only because of the range of definitions of "acceptable" and "useful". We can guess that film and digital capture will be limited by different factors, so comparing E vs. D curves will only get us a general idea. You almost never see curves of negative material in B&W or colour where the graph is continued up to the shoulder. The only ones I've seen that show a definite shoulder are very old. With B&W film the loss of accutance due to irradiation can give the appearance of blocked highlights even though the exposure is nowhere near the shoulder. This appears to be the reason that the larger formats have much more usable overexposure latitude. I certainly noticed this when I first started using 6x6, though it took me some time before I understood why this was the case. Peter. -- |
#519
|
|||
|
|||
"Mike Kohary" wrote in :
First off, I don't think anyone who brags about having a 65,000 drum scanner gets to use "lack of funding" as an excuse. I admit I've been kind of skimming the thread...he has access to one, or he owns one? Don't know. Again, he doesn't tell us. So far he's just thrown out that number, seemingly as a means to justify his choice... something along the lines of "it costs lots so it must be perfect" reasoning. Secondly, he's claiming to have a background in scientific publishing (editing if not writing). so it's not unreasonable to expect at least a degree of accountable methodology. A degree, yes, but you're holding him to the standard he'd be held to if he was being published in a journal or something. He's not; it's just a web page he hosts. That's fine, then my original statement still stands. He doesn't do much for his credibility if he's only citing himself. If he wants to be able to make and support the claims he's made, then he needs something better than "oh just trust me!" which is essentially what it's come down to. I agree of course that truth is important no matter what is being published where, but I do think the context of the publication makes a difference in how hard to criticize someone. As long as his information is accurate, I think it's not such a big deal if his sample size is small, for example. Except there are too many questions being raised, especially later in the thread by people with much more experience than I, as to the quality of his scans, his methodology, etc; this makes it impossible to know whether his information -is- infact accurate or not. That may be important, but it's also understandable - how is he supposed to sample the way a funded scientific study would? I doubt he has the resources for that, so it's just something to be lived with (though it's a good caveat to bring up when arguing against his position). Well, how much does twenty rolls of film cost? How much would it have cost him to include agfa and ilford? I know i wouldn't be raising quite the same questions if he -had- tested five different rolls of the same speed from each major manufacturer, but as it is he tested a roll of 50iso slide, and a kodak gold 200. No matter how you slice it, that's simply not sufficient to draw the claims he's making. This all started out with me pointing out that referencing yourself (what he was doing) doesn't do a lot for your credibility. That he's unwilling to even listen to legitimate criticism about the problems with his tests only reinforces my position. I understand where you're coming from, and it would be helpful if his research was backed up by the research of others. Still, for what it is, it's not bad at all and provides a lot of food for thought. To me, it seems like his 'research' or 'testing' is just another (exceedingly long winded, albeit) way of saying "I think this is better than that!" and we've probably all grown tired of arguments that follow those lines. If someone wants to go and do reproducable tests following and adhering to a strict methodology with -justified- choices, then we might be able to get somewhere. Trees and mountains, poorly scanned (so say others, not me), just don't cut it. -- http://www.neopets.com/refer.phtml?username=moosespet |
#520
|
|||
|
|||
"Mike Kohary" wrote in :
Jon Pike wrote: Brian C. Baird wrote in news:MPG.1c18290eae79331c98a3f3 @news.verizon.net: You're such an ignorant jackass, Jon. You seem to have your panties in a bunch. That's not very objective of you. *sigh* typical usenet... just keep trolling. That's funny. See what I mean? Typical usenet troll. When you run out of intelligent things to say, you resort to weak insults. I think everyone here sees you now for what you really are. Not really, Jon. You've kind of lost people on this argument, not because you're literally incorrect (you're not), but as I said, you're applying a very rigorous standard to what is obviously an informal project. You also constantly overstate things - resizing is "digital manipulation" to you, but of course one term sounds rather devious while the other sounds perfectly reasonable. Resizing is only the visible proof of what he's doing; we don't know what else he might be doing because, as I've said over & over, he doesn't tell us. Do you have a point in any of this? State it, or let it go. Point was stated with the very first post I made in this thread. Citing yourself doesn't give you much credibility, especially when your 'tests' are so poorly conducted. His claims and conclusions are, thus far, unsubstantiated. -- http://www.neopets.com/refer.phtml?username=moosespet |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
35mm on grade 3 explained | Michael Scarpitti | In The Darkroom | 240 | September 26th 04 02:46 AM |
Digital quality (vs 35mm): Any real answers? | Toralf | 35mm Photo Equipment | 274 | July 30th 04 12:26 AM |
Digital quality (vs 35mm): Any real answers? | Toralf | Digital Photography | 213 | July 28th 04 06:30 PM |
Will digital photography ever stabilize? | Alfred Molon | Digital Photography | 37 | June 30th 04 08:11 PM |
Which is better? digital cameras or older crappy cameras thatuse film? | Michael Weinstein, M.D. | In The Darkroom | 13 | January 24th 04 09:51 PM |