Medium format vs digital

[steve m smith]

<p>As usual, these are all digital vs. scanner tests rather than digital vs. film tests.</p>

[jussi_vakkala1]

<blockquote>

<p>As usual, these are all digital vs. scanner tests rather than digital vs. film tests.</p>

</blockquote>

<p>I our case I disagree. Film participant of test is free to use enlargerer or any similar method if he wishes to do so to maximize results for print. Or if there is any volunteer in our area to try enlarger for comparing all three prints, I think we can arrange that also. To my knowledge our film participant has capabilities for analog method, but at this phase he has chose drum scanner for the job instead.</p>

<p>Requirement is that test plan is strictly followed and test is executed by two photographers with known and different opinions favoring either digital or film. By test plan the way to get into good quality print was not limited to a digital scan, but it was limited to the methods what the film participant of test run can use in real life. Print is the proof this time, not a claim to a second hand statement or theoretical speculation.</p>

[danielleetaylor]

<p>Jussi - I would expect the best results from an Imacon scan or a drum scan as well. Especially for a 20x30 print. I'm not surprised that the film participant is choosing that route.</p>

[andylynn]

<p>Daniel,</p>

<p>I don't like crops and I've never printed from a 7D so I'll have to take your word for it that it's magic, but I've been printing up to 36" and small sensor digital is visibly lacking in detail at that size, and even my crude attempts at scanning make better prints. I talked for a while with a guy at the Wakefield, MA camera show this weekend, who was trying to sell a Fuji 6x9 rangefinder and had brought a huge print with him - 40-something inches, scanned on his own drum scanner - and it was freakin' perfect. Well beyond any DSLR result I've seen and fantastic color.</p>

[andylynn]

<p>Les, I'm going through my film scans in Aperture looking at things like backlit trees and I can find CA's. Not as pronounced as on my DX digital shots, but I put that down to the digital shots having a greater pixel density and magnifying the CA's more. This is with 35mm film - medium format, I'm using even less pixel density and I've got nothing.</p>

[sfcole]

<p>Andrew,<br>

Isn't it also the case that for 35 mm and MF, color fringing may still be there but be less obvious due to the lower degree of magnification to any given size of photo compared to a crop sensor?</p>

[andylynn]

<p>I think that's what I said :)</p>

[stwrtertbsratbs5]

<p>I'm happy with results that I get from both film and digital. I use digital for sports, and film for most everything else. </p>

[rodeo_joe1]

<p>What seems to have been missed in this discussion is that the picture doesn't stop with the camera, be it digital or film. This should be about total workflow, and IMHO digital wins the argument here hands down.</p>

<p>Film might, just might, give you ever so slightly "peer at it through a strong loupe" better resolution. Better colour? Hmmm, not in my experience. Better looking prints? Definitely not - not unless you're willing to commit to meticulous darkroom cleanliness or to spend a long time cloning out dust or using a retouching brush.</p>

<p>Then we get to the scanning or darkroom choice. Getting that precious extra resolution out of your film requires either top notch darkroom equipment and technique or the use of a professional quality scanner. Epson need not apply! Then there's the time it takes to scan an MF negative or slide. This is not trivial if you produce more than a couple of shots a week. And to be honest, once the film is scanned you've effectively turned your precious "real" film into a digital file - so where's the advantage?</p>

<p>As for an AP-S sized DSLR producing the same results as full-frame: A few little words "control over depth-of-field" and "diffraction". Nuff said?</p>

<p> </p>

[danielleetaylor]

<p><em>BTW, did you download the same 35mm Fuji RVP film scanned with an Imacon from that folder? I first posted this scan in the thread - <a rel="nofollow" href="../digital-darkroom-forum/00S2tw"><strong>Can minilab scans rival other scanners?</strong></a> back in Jan 08, 2009; 12:28 p.m. Anyway, comparing your own "optimized" 7D crop pales by comparison at first glance and we're not talking about interpretation either . . . ;-)</em></p>

<p>I did download your Imacon sample from another thread. I think it was different from the thread you linked to because the size was much larger. Any way, doing a direct Bicubic Sharper resize from the RAW file to the Imacon sample size, then applying USM 15%/50px and Smart Sharpen 200%/0.3px yields the attached comparison. I was a little more aggressive with the local contrast enhancement and sharpening given the larger size. On the Imacon Velvia seems to be resolving the same high contrast detail as the 7D, but the 7D is still resolving more low contrast detail. In addition the overall clarity is much better in the 7D sample, though perhaps the Velvia sample could be improved here with some levels or curves work.</p>

<p>I think on screen my sharpening makes the high contrast detail a bit rough in the 7D sample, but in print I believe it would come off better with the sharpening so I applied it.</p>

<p>I don't think anyone would look at these and say the 7D "pales by comparison." In print I have little doubt that the 7D sample would be preferred by most viewers. If the clarity of the Velvia sample were improved they would probably be about equal in print. One could sharpen the Velvia sample, or not sharpen the 7D sample, to equalize that variable to taste.</p>

<p>Of course when comparing you need to resize directly from a fresh crop and apply optimal sharpening. Repeated resizing and multiple applications of USM and smart sharpen introduces too many artifacts. If you did that then yes the 7D sample would pale by comparison.</p>

<p><em>BTW, I just got your 7D RAW file - thank you. Unfortunately my fototime site has exceeded it's bandwidth limit for the month - the price I pay for posting actual full res images to share, and all links to my site will be severely limited in size. I hope the files you downloaded from my site were the actual full res version . . . ;-)</em></p>

<p>They were downloaded a while ago. I'm sorry to hear you're having bandwidth issues. You have a valuable reference there and if this continues, you should post a donation link to help with any costs. Hopefully though it will just be this month due to all of these conversations. While I have the chance, let me offer my thanks for the work you put into making this reference available to the photographic community.</p><div></div>

[danielleetaylor]

<p>Rodeo Joe - <em>As for an AP-S sized DSLR producing the same results as full-frame: A few little words "control over depth-of-field" and "diffraction". Nuff said?</em></p>

<p>Though I find DoF control with fast primes to be adequate on APS-C, shallow DoF is certainly a full frame advantage where it's applicable. But diffraction does not impact any of the formats more than the others for a desired DoF at a desired FoV. Note I said DoF and FoV, not aperture and focal length. It's important to learn the DoF of lenses on APS-C as well as optimum apertures. A lot of people apply what they know from full frame, use too narrow of an aperture, and sacrifice resolution to diffraction that they didn't have to lose. Learning hyperfocal distances also helps a lot on either format to obtain optimal resolution.</p>

[Karim Ghantous]

<p>Fascinating discussion. I'd like to see Jussi's test as well. I don't use MF but the information here is quite valuable.</p>

<p>BTW Daniel, thanks for that that Velvia vs. 7D scan. The digital file was cleaner but also harsher. Not a bad compromise, really. But there's something weird going on. Look at Lake Chad - there is a line missing in the digital file that's present in the film scan. What's up with that?</p>

[andylynn]

<p>Hm, a few things -</p>

<p>-So, looking at what Karim said, these are different editions of the map being photographed? Then how can such minute details be compared?<br>

-Since as Daniel showed, the 7D can't outresolve Les's scan of 35mm film, do we even need to have this discussion regarding medium format? 6x7 is 4.5x the size of 36x24. That's 11.6x the size of the 7D sensor. I think we can see pretty clearly what's going to happen here.<br>

-I think we've seen that even a 7D's small sensor is adequate for most situations, and with its handling characteristics there are a lot of tasks it's excellent for. But really now, a 5DII or any of the other popular full frame / FX cameras is better at many things. High ISO, for example, or wide angle. The sensor is two and a half times the size of a 1.6 crop sensor. It's not even fair.</p>

[danielleetaylor]

<p>Karim - <em>The digital file was cleaner but also harsher. Not a bad compromise, really.</em></p>

<p>It's also a compromise the photographer has some control over. Use Bicubic Smoother, don't sharpen or sharpen as much, etc. Then it's not as "harsh". I used those steps because "harsh" at screen magnification translates to "sharp" and "crisp" at print sizes.</p>

<p><em>But there's something weird going on. Look at Lake Chad - there is a line missing in the digital file that's present in the film scan. What's up with that?</em></p>

<p>The line is not present on my copy of the map. There are other small differences like that, including some notable color differences that trace back to the map photographed.</p>

[danielleetaylor]

<p><em>So are you saying that readable Libreville in the Imacon scan is resolved on the 7D - looks like Ub... can't really make it out can you?</em></p>

<p>I can make it out perfectly fine. I think perhaps it's your viewer (see below).</p>

<p>What I can't make out as well in the film scan are the mountain details. In some places they're a blur or missing completely.</p>

<p><em>What are the colors on your map for the rivers that go out to Port Harcourt anyway?</em></p>

<p>Pretty close to the color you see. The river colors on my map are not the blue that they are on yours.</p>

<p><em>Since we are now talking about the tiniest of details, I notice that your version of my crop now suffers from jpeg artifacting.</em></p>

<p>I cannot discern any JPEG artifacts switching between web and PS screens. If you feel I am mistaken please email me a comparison TIFF of the crop here along with the original you have, and circle the areas in question. I will be happy to post again with either a higher JPEG quality level, or a link to a TIFF, if indeed this is a problem.</p>

<p><em>This is evident when magnifying to see the two stars between Brazzaville and Kinshasa. They are formed in the Imacon scan and not in the 7D at 300% zoom in ACDSEE.</em></p>

<p>The star formation is roughly equal between them viewed at 100%. Magnifying to 300% in an image viewer presents a drastically lower quality rendering than resizing and viewing at 100%. Since the 7D sample has had more sharpening applied, the breakdown will be even worse. Screen magnification algorithms are optimized for speed at the expense of quality. As an example, PS CS4 uses the simple Nearest Neighbor algorithm to render screen views >100%. This looks horrible compared to using Bicubic Sharper or Bicubic Smoother to resize the image.</p>

<p>If you're studying at 300% in an image viewer it's no wonder you're having trouble making out a word or a star. If you want to view 3,000 pixel tall crops, you would need to resize the original crops directly to that size using one of the better algorithms, and then apply any sharpening or LCE. Though I would argue that such an extreme magnification would tell you more about the resizing algorithm used than about which resolved more detail. By my calculations stretching these crops to 3,000 pixels would be like viewing a 350" print! In such a view the vast majority of the pixels (>75%) will be interpolated.</p>

[danielleetaylor]

<p>Andrew <em>-So, looking at what Karim said, these are different editions of the map being photographed? Then how can such minute details be compared?</em></p>

<p>That is a valid question. I can reshoot my map and shoot MF at the same time so that there are some comparisons with the same identical map. Of course it will take some time to reshoot, develop, scan, etc.</p>

<p><em>-Since as Daniel showed, the 7D can't outresolve Les's scan of 35mm film,</em></p>

<p>On low contrast detail the 7D does out resolve 35mm Velvia 50 film, as is evident in the relief texture of the map. On high contrast detail, on the Imacon, it looks about even. If there was even higher contrast detail present, such as pure black on pure white detail under very bright lights, Velvia would probably show an edge on that detail if scanned on a drum scanner.</p>

<p><em>do we even need to have this discussion regarding medium format? 6x7 is 4.5x the size of 36x24. That's 11.6x the size of the 7D sensor. I think we can see pretty clearly what's going to happen here.</em></p>

<p>Just to clarify, I never once claimed the 7D out resolved or fully replaced MF film. In fact I specifically said that for print sizes larger than about 24" or 30", I stitch 3 frames. The debate has never been whether or not small format sensors (APS-C or 35mm) could out resolve much larger 6x7 film. But I do believe that for 24" and smaller, it's hard to tell between them in print. Someone who only occasionally prints larger than 24" might want a 7D or 5D mkII. Someone who often prints larger might want to skip stitching and get MF. MF is cheap enough now that many can afford both. Personally I only wish a good, high quality MF scanner was cheap and readily available.</p>

<p><em>-I think we've seen that even a 7D's small sensor is adequate for most situations, and with its handling characteristics there are a lot of tasks it's excellent for. But really now, a 5DII or any of the other popular full frame / FX cameras is better at many things. High ISO, for example, or wide angle. The sensor is two and a half times the size of a 1.6 crop sensor. It's not even fair.</em></p>

<p>Full frame bodies like the 5D mkII have a clear and distinct advantage at high ISO. Not everyone needs this though. I can get noise free 8x10 prints from ISO 2000 and even ISO 3200 shots. The types of things I shoot at those ISOs don't call for prints larger than about 11x14. Now for someone else, the high ISO performance of a 5D mkII could be a must. It is spectacular at those ISOs, no doubt about it.</p>

<p>But the statement "it's not even fair" implies that they're not close in IQ in any situation. Again I'll point out that at low to mid ISO, in print, they are pretty much equal, assuming all other factors are equal (lenses, technique, processing, etc).</p>

<p>While FF continues to have some advantage in WA glass (i.e. 35 f/1.4L; T/S lenses), APS-C is quite usable in wide angle applications thanks to lenses like the Canon 10-22 and the Tokina 11-16 f/2.8. The gap is just not the same as it was when the 5D was first introduced and its competition was the 20D with practically no good WA crop glass.</p>

<p>It all comes down to what you need.</p>

[Karim Ghantous]

<p>One thing which always interested me: the assumption, a wrong one, that slide film is the gold standard. It isn't. Back when Velvia was a new emulsion, Ektar 25 resolved more detail. By this time, K25 completely lost the resolution game. Negative films always resolved more than slide films.</p>

<p>I guess what I'm saying is that I'd like this test repeated with Ektar 100. Just for fun. This is, after all, about curiosity, not about proving which medium is 'better'.</p>

[swilson]

<p>

<p>Daniel's comparison between Velvia 50 and the 7D shows one of the frustrations I have had with scanning film. I get a scan that shows good detail of high contrast areas but the image still looks soft, due I believe to a lack of low contrast detail. This does not seem to bother Les, who clearly has a different criteria what was a sharp image is from my own, which is fine to each their own.<br>

　<br>

But for me film scanning has been a frustration of picking up fine detail on things like signs, but then having the overall photo not look very sharp. The conclusion I came to years ago was that sharpness is not simply a function of how many line pairs per of a 1000:1 contrast target you can resolve but is a combination of a whole lot of other factors.</p>

</p>

[Ed_Ingold]

<p><em>The conclusion I came to years ago was that sharpness is not simply a function of how many line pairs per of a 1000:1 contrast target you can resolve but is a combination of a whole lot of other factors.</em></p>

<p>For those of a mathematical bent, who would like to know why DSLR images look clearer and sharper than 35mm film (despite Les' attempts to prove it is not so when his own examples indicate otherwise), browse through Norman Koren's website (e.g., <a href="http://www.normankoren.com/Tutorials/MTF7A.html">http://www.normankoren.com/Tutorials/MTF7A.html</a>)</p>

<p>In summary, the contrast and resolution of a 6MP DSLR is significantly higher where it counts (MTF 50) than Provia (55 lp/mm vs 46 lp/mm respectively). Provia has a higher ultimate resolution (66 lp/mm vs 72 lp/mm respectively), but only at contrast too low (MTF 10) to matter photographically. Velvia tops out at about 100 lp/mm scanned at 8000 ppi, comparable to a full-frame 12 mp camera. That aside, the DSLR has lower noise than film, permitting much more sharpening, boosing resolution another 20%. Mr. Koren (as do most of us) finds that the color is cleaner and more accurate with a DSLR. This work was published in 2005 - things have come a long way since then.</p>

<p>There is more to life than 1000:1 resolution charts, newspapers taped to the wall and rows of crayon boxes ;-)</p>

[Brad_]

Without interesting/compelling content, what difference does it make?

[Ed_Ingold]

<p>For someone who demands proof positive on any statements comparing film with digital, Les seems a little skeptical about a rigorous, theoretical approach. It's obvious that examples and demonstrations never meet his approval. Korem illuminates the basis for what everybody but Les seems to have observed.</p>

<p>I'm not the first to utter the "three to one" comment (I believe it was Ellis Vener, who seems to know a thing or two), based on observation that a prints from a 6 MP DSLR look cleaner than ones from 24 MP film scans (i.e., 4000 ppi). Clarity (acuity), color and lack of grain count. The 6 MP value seems to be cited by others too, including Bjorn Rorslett. It's pointless to rant about ultimate resolution. Like complaining that Jeeps don't have spoilers (or Corvettes don't have mud flaps) - it simply doesn't matter. Hint - you would need a 20x enlargement to reap any benefit from 80 lp/mm resolution, even if you could see it in murky details (MTF < 10). Only a devout filmfan would find beauty in that amount of grain and general fuzziness, LOL.</p>

[danielleetaylor]

<p><em>I guess what I'm saying is that I'd like this test repeated with Ektar 100. Just for fun. This is, after all, about curiosity, not about proving which medium is 'better'.</em></p>

<p>Ektar 100 doesn't fair so well.</p><div></div>

[danielleetaylor]

<p>Edward - Norman Koren's web site is a must read for anyone interested in learning how different systems and work flows compare. I also find the Clarkvision web site to be very useful. Just a bit of background info for other readers: Norman Koren is the creator of Imatest, an industry standard piece of software for testing the performance of photographic systems. Roger N. Clark of the site Clarkvision.com earns his living helping NASA photograph planets. These guys know what they're talking about.</p>

<p><em>I'm not the first to utter the "three to one" comment (I believe it was Ellis Vener, who seems to know a thing or two), based on observation that a prints from a 6 MP DSLR look cleaner than ones from 24 MP film scans (i.e., 4000 ppi). Clarity (acuity), color and lack of grain count.</em></p>

<p>At smaller print sizes I would agree with this. And you're right, it's about the clarity when print sizes are smaller. For the same reason I think a Canon 7D can go toe to toe with MF at 24". The print size is too small for the resolution advantage of MF to be observable, but the differences in clarity and noise are quite evident.</p>

[danielleetaylor]

<p><em>Incidentally, the scans being displayed here are straight-up unprocessed scans.</em></p>

<p>That's a fair point. I took the Imacon Velvia 50 vs 7D sample and did a little more processing. On the Velvia side I let Noise Ninja profile and clean up the sample. I used default settings except sharpening, which I turned off so that I could use PS for that task. Then I used USM 15%/50px and Smart Sharpen 150%/2px. This went a long way towards cleaning up the Velvia scan and compensating for the softness. On the 7D side I went ahead and brightened it a bit (levels 1.1) and gave it one more round of LCE (USM 10%/50px) to try and equalize the "brightness" of the Velvia sample after LCE and sharpening.</p>

<p>This brings them closer thanks to considerable improvement in the Velvia sample. However, the 7D sample still exhibits higher low contrast detail resolution, and is still cleaner with a higher level of clarity. But the gap is not nearly so wide.</p>

<p>The only issue is that the aggressive sharpening required to compensate for the scan softness could introduce print observable artifacts with some subject matter. I'm surprised it didn't cause distracting halos in this sample.</p>

<p><em>Wouldn't you know it, no amount of post processing - save painting in the missing detail, is going to help digital capture what it cannot. Hence it makes up letters by combining "L" and "i" to produce what looks like a "U" . . . ;-)</em></p>

<p>I'm really confused by your statements here as I have no trouble recognizing the "L" and "i" in "Libreville" in the 7D sample. It's obvious even at 24" equivalent print size, where the text is quite small.</p><div></div>

[danielleetaylor]

<p><em>While we contest the difference between the top DSLR (which most people don't have) and 35mm film, </em></p>

<p>I would say that many or most people now have DSLRs which can out perform what they might have gotten with 35mm film. The only sample that even begins to challenge the 7D is the Velvia 50 Imacon scan sample. Velvia 50 is a good 20 lpmm better than any other color film made today. It's also better than most B&W films. And the Imacon has twice the scanning resolution of common dedicated film scanners. Velvia 50 on an Imacon is a far cry from, say, Portra 400 on a CoolScan, or worse, on a flatbed. Even highly regarded ISO 100 films like Provia and Ektar are now behind modern DSLRs. Judging from your own samples I would say that 12 MP was about the point where DSLR sensors pulled ahead of all but a couple films scanned using the very best equipment and technique (the D2x sample is only beaten by Velvia 50 and Tech Pan). Now that resolutions are going to 15, 18, and 20+ MP, even the very best 35mm films on the very best scanners can't quite keep up.</p>

<p>Keep in mind that the 7D's 18 MP sensor is now also found in a Rebel, and most other entry level bodies have at least 12 MP sensors. I think it's fair to say that small format digital has passed up small format film.</p>

<p>For the record I do not consider the 7D to be the top DSLR in terms of shear resolution. That title would have to go to either the Sony A900 or the Nikon D3x. Canon's next 1Ds, rumored to be in the 35 MP range, will no doubt take this title shortly. Assuming the glass can keep up, a 35 MP FF sensor is going to give MF film a run for its money at 30, and maybe even at 36 inches in print. The absolute resolution of 6x7 film will probably still be higher, but the fine detail from the next 1Ds will be stunning.</p>

<p><em>no digital today matches films ability to control the highlights.</em></p>

<p>Sure they do. Just expose for the highlights and make sure they don't clip. In terms of total dynamic range, or ability to hold detail from shadow to highlight, there are films which out perform the best DSLRs. Unfortunately those films also tend to have lower resolution and acuity. The 7D yields 10-11 stops of total DR in RAW. Top tier FF bodies yield another 1-3 stops. The very best DSLRs with optimum shooting and RAW processing are tapping on color portrait film's shoulder in terms of DR. But more importantly, they are far past color portrait film (the highest DR films) when it comes to resolution and acuity.</p>

<p>Ektar is regarded as being higher resolution than color portrait films, and looking at the Ektar sample I simply would not give up the resolution or clarity of the 7D for a couple extra stops of DR. 10 stops in RAW is more than enough any way.</p>