It all comes down to the print

[mauro_franic]

<p>I must admit after printing all the crops from Daniel, that he has the best upscaling technique I've seen and that current DSLRs have come long ways in the last several years.</p>

<p>Contrary to the popular belief that 6MP DSLR were a suitable replacement for 35mm (to produce actual prints), they were far from it. Current ~20MP DSLRs, for color, can produce prints almost on par with the best 35mm color films.</p>

[danielleetaylor]

<p>Wow. I thought this thread was dying and Mauro and I were going to end up in a never ending debate. I can't believe it's still going and popped back onto the front page.</p>

<p>I can't possibly catch up with everything in the thread since I bowed out. But I've skimmed the last few posts and I hope this helps answer some questions, and helps move the debate forward.</p>

<p>* The 7D RAW file is here: https://docs.google.com/leaf?id=0Bz1sfHfXHVDCYTgzYmZhMGMtYmVkOS00MTI1LWI5ZWEtYzg5ZGIzNzEzZDNm&hl=en</p>

<p>* In another thread I've been posting new comparisons with some improved processing. Participants in the thread also discussed the comparisons further. http://www.photo.net/casual-conversations-forum/00WHC0</p>

<p>* In my experience, you cannot resize a previously resized/processed crop and expect optimum output. Quite the contrary, the output falls apart with the second resize and subsequent sharpening. You must start with a fresh crop from the RAW file, resize to the desired target size, and apply any processing after the resize. (I realize the thread proceeded without the RAW file and that's why people were playing with existing samples.)</p>

<p>* I've used Bicubic Sharper for all scaling. I've been playing with another image application which has a lot more options than PS for scaling, but I'm not finding a significant difference yet.</p>

<p>* While the 7D crop obviously needs to be scaled up to the Imacon crop size for comparison, I see no point in scaling them further or observing at greater magnification. When you do this you only test the interaction of your chosen scaling algorithm with the samples. Past a certain point you're not actually testing or observing the camera systems themselves. (At a scale like Mauro posted on Apr 23, 2010; 12:27 p.m., the vast majority of pixels are interpolated.) Along those lines if you scale to some insane level, then scale back down for print, you're just introducing artifacts. IMHO scaling the 7D sample to the native Imacon sample size and then observing and printing is the proper way to compare.</p>

[mauro_franic]

<p>Daniel,</p>

<p>Can you please post an image from the 7D to match the crop above (in size and sharpness) so I can print it next to the Velvia scan at 300 dpi. (24x36 print)</p>

<p>You can produce a better upscaling from the raw than what Stuart and I were able to do from the jpegs. My upscaling of the 7D may not do justice to it.</p>

[danielleetaylor]

<p>With those notes out of the way...</p>

<p>The last, best Imacon 35mm Velvia 50 vs 7D image I posted in the other thread is here: http://static.photo.net/attachments/bboard/00W/00WJ26-238625584.jpg</p>

<p>As I mention in that thread, I tried optimizing both samples to produce best possible crops. When I print this at 259 ppi (30" x 45" equivalent) on an Epson R800 using Epson Premium Glossy paper I observe the following:</p>

<p>* The Velvia sample still has a bit more noise (this is after Noise Ninja was applied to the Velvia sample).<br>

* The 7D sample is a touch sharper, but also looks a bit harsh in some areas.<br>

* High contrast detail looks about the same.<br>

* The 7D resolves a bit more low contrast detail.<br>

* Some halos are evident in the Velvia sample due to the aggressive sharpening and LCE used to increase its clarity to a level comparable to the (newly processed) 7D sample.</p>

<p>Overall there's not much to choose between them. I would say the 7D out resolved Velvia based on the low contrast detail. But I also believe that Velvia would resolve a bit more very high contrast detail on a target with such detail (i.e. B&W), scanned on an Imacon or drum scanner. Either way, they're pretty darn close to one another, though it did take more processing on the Velvia side to get there. I think if anyone wanted to split hairs further than this they would have to use a more precise resolution target, printed and photographed at varying contrast levels. I'm not interested in doing that at this time.</p>

<p>While both put in a good performance, they are starting to break down at this scale. For prints at this size you would want to stitch frames or shoot a larger format for optimum print quality.</p>

<p>The only other note I'll make is that Velvia 50 on an Imacon is head and shoulders above any other color film / scanner combination. These results don't speak for "color 35mm film", but for Velvia only.</p>

[danielleetaylor]

<p><em>Can you please post an image from the 7D to match the crop above (in size and sharpness) so I can print it next to the Velvia scan at 300 dpi. (24x36 print)</em></p>

<p>What crop size are you looking for?</p>

<p>Going with the above link: the 7D crop has to be scaled by roughly 2.29x to match the Imacon sample I have on hand. (I got it from another thread that Les posted in.) So...</p>

<p>(5,184 x 2.29) / 24 = 495 ppi (rounded)</p>

<p>I actually made a minor error in my above post where I printed the samples to 30x45" equivalent. The ppi should have been 264 ppi.</p>

<p>I feel the link above represents my best processing of Les Sarile's scan sample and my 7D sample to date. I can list the processing steps if you like.</p>

[mauro_franic]

<p>I pretty much agree with your observations on Velvia and the 7D.</p>

<p>I think the sample posted from Ektar is not optimal. Looks too light and a bit fuzzy.</p>

<p>On black and white, even from a 4000dpi of 35mm film clearly outresolves current DSLRs. Not sure why; other than films like TMAX provides resolution oversampling on top of the 4000dpi that result in a crazy crisp scan. </p>

<p>All and all, current DSLRs are definitely an alternative for prints up to 16x20. At this size, observing resolution alone, it is difficult to tell it apart from a print from 35mm film. At 24x36, films like Velvia and TMAX, scanned at 8000 dpi will have a visible edge.</p>

 

[mauro_franic]

<p>Here, please replace the right side with your own 7D upsample. Mine is not as good as from the raw.</p>

<p> </p><div></div>

[danielleetaylor]

<p>I have to retract an earlier statement. Earlier I said <em>(At a scale like Mauro posted on Apr 23, 2010; 12:27 p.m., the vast majority of pixels are interpolated.)</em></p>

<p>I misjudged the area Mauro posted and thought he had scaled even further than the native Imacon scan size. He did not. With that out of the way...</p>

<p>Mauro - I took both sides from the test sample linked above.</p><div></div>

[mauro_franic]

<p>For some reason your Velvia shot looks softer than the one above I sharpened directly from the TIFF.</p>

<p>Here is the one I sharpened next to your 7D.</p><div></div>

[mauro_franic]

<p>Like I said before the 7D holds very well up to prints 16x20. Printed at 16x20 or larger I can see the edge of film showing on the print. (You can print at 300 dpi for a rough 24x36 comparison).</p>

[mauro_franic]

<p>Here it is the above at 50% (bicubic downsampled) which looks very close on the monitor to the print I'm holding in my hand:</p>

<div></div>

[danielleetaylor]

<p><em>For some reason your Velvia shot looks softer than the one above I sharpened directly from the TIFF.</em></p>

<p>We can keep sharpening all day, but I think we're moving past the point of what's realistically possible with normal subject matter. For print I would use the sample I link to above, not these super sharpened samples.</p>

<p><em>Like I said before the 7D hold very well up to prints 16x20. Printed at 16x20 or larger I can see the edge of film showing on the print. (You can print at 300 dpi for a rough 24x36 comparison).</em></p>

<p>330 dpi would be equivalent to 24x36. At 259 ppi (30x45) I feel the 7D has a slight edge (less noise and fewer artifacts), though they are pretty darn close to being equal. Have you tried printing the sample I linked to?</p><div></div>

[mauro_franic]

<p>I am printing it now.</p>

[mauro_franic]

<p>They look almost indistinguishable on print from the link you posted.</p>

[mauro_franic]

<p>This has been a very rewarding thread with better and finer comparisons than previous ones I read. Also a better discussion and exchange of experiences - at least from my point of view.</p>

[danielleetaylor]

<p><em>This has been a very rewarding thread with better and finer comparisons than previous ones I read. Also a better discussion and exchange of experiences - at least from my point of view.</em></p>

<p>Agreed.</p>

[StuartMoxham]

<p>It has been a very good thread . Nothing got out of hand which makes a nice change with these kind of threads. The 7D and 35mm Velvia seem to be fairly well matched when making prints up to 16x20 inches.</p>

[StuartMoxham]

<p>They are huge magnifications though Les, 16x20 prints seem to be quite equal. At 100% crop they are both looking rather rough or rather excelent depending on how you look at it. For life size repros of maps with fine print I don't think I would choose either though.</p>

[danielleetaylor]

<p>The 7D resolves all the details you point out, but you are saying that the Velvia scan resolved them better.</p>

<p>You need to keep in mind that some of my processing decisions optimized the print at the expense of the screen display. As I said in the other thread, "rough" at 100% on screen comes off as "sharp" in print. LCE in particular makes the black ink appear thicker and obscures some fine detail within the letters and stars.</p>

<p>That said, I would agree that Velvia on the Imacon exhibits slightly higher resolution of high contrast detail. But the Canon 7D exhibits higher resolution of low contrast detail, as is clearly evident in the relief portions of the map. This is consistent with what we know about film and digital.</p>

<p>These differences are not easily observed outside of a 100% screen view.</p>

[robert lee]

<blockquote>

<p>... the Canon 7D exhibits higher resolution of low contrast detail</p>

</blockquote>

<p>That's precisely the point isn' t it.</p>

<p>The premise of this thread is that the it's the print that matters in the end. This means a few things:</p>

<ol>

<li>If you're following the discussion up to this point, then you're likely printing digitally. So, the maximum number of source pixels needed is simply what's necessary to fill the target print size at around 300ppi. This is a good number practically because it's at the limit of unaided vision (and thus no surprise that it's around the native output capability of most printers.)</li>

<li>The cover all bases target print size should be no larger than where the recording media starts failing in delivering high fidelity pixels. Pick 50% MTF generally; 70% MTF if you're picky (I am.) For many color film, spatial fidelity as a function of increasing image detail noticeably degrades far sooner than its digital sensor peer.</li>

<li>A DSLR sensor at low ISO settings has a much lower noise floor than many 135 format film.</li>

</ol>

<p>This is why taken together an obsolete, 8MP Canon XT can more easily produce a better looking 8x10 than 135 Reala, and why "a 6MP DSLR is the equal of 6x7."</p>

<p>I'm an avid film user; it's an enticing medium for many, many reasons. A singular insistence on resolution as the defining characteristic, however, is puzzling to me.</p>

 

[mauro_franic]

<p>Robert,<br>

I print routinely (with this I mean almost every day) at 16x20, 24x30 and 24x36 from film (both 6x7 and 35mm). The results are excellent.</p>

<p>In contrast, I do not print from my 10MP 40D larger than 11x14 or 12x18 as a stretch because it break down.</p>

<p>Those are the results. A numerical theory that shows the opposity is wrong because it does not match reality.</p>

<p>I will try to address the flaw in the theory next but the best would be for people to print the test sheet with the pictures of the two ladies (see toward the beginning of the thread) at home and look at the print instead of relying on theories.</p>

<p>Set the resolution of this image at 360dpi and print it ltter sized using your printers maximum resolution:<br>

http://shutterclick.smugmug.com/Other/Standard-Lenses/9379726_tZVrK#838499551_wYtLb-O-LB</p>

<p> </p>

[mauro_franic]

<p>Scenes are regularly made from a combination of details (depending on print size and capture they will be small, medium or large) and contrast.</p>

<p>The reason why the theories of comparing digital at MTF50 do not match the results on the prints is because MTF50 is just one subset of several found in the scene.</p>

<p>This chart could help explain:</p>

<div></div>

[mauro_franic]

<p>Now here is the real revelation. You cannot look at a single detail column or at a single contrast row since THE PRINT IS THE COMBINATION OF ALL DETAIL AND CONTRAST LEVELS.</p>

<p>Resulting in:</p>

[mauro_franic]

<p>OVERALL RESULTS FROM EACH SYSTEM ON A 24"X30" PRINT:</p><div></div>

[StuartMoxham]

<p>While I do agree with Mauro that the 35mm film does show more resolution often what I found for myself is that I often got better 8x10 prints when I went from 35mm film color neg films to a 6mp D70. 35mm films color I was using at that time Fuji press 800, superia 400 and 800, kodak elite 200 (was a print film in Europe). Those films would show some grain at 8x10 but the D70 prints would be grainless. Many other also found that they prefered the grainless digital prints which is what I believe caused many to say the prints were rather like medium format prints. It was not about the resolution but the smoothness of the prints.</p>