Jump to content

Film vs Digital - Color Rendition


Recommended Posts

<p>Looking at the resolution charts above I wonder why you guys are not using the slanted edge method and imatest to determine the resolution. Doesn't it work well with scanned film?</p>

<p>With digital I found that I get very repeatable results.</p>

Link to comment
Share on other sites

  • Replies 387
  • Created
  • Last Reply

Top Posters In This Topic

<p>Yeah I've been doing some more reading & I think most of our disagreements stem from what Norman Koren calls the 'old definition' for resolution vs. more modern methods of reporting MTF10, MTF50, etc. values.</p>

<p>Pete would you suggest 'Imatest Studio'? Some of those versions are ridiculously expensive ($2500!).</p>

<p>Thanks,<br>

Rishi</p>

Link to comment
Share on other sites

<p>Richi, I have an old version that is somewhere between Studio and Master but I think the $99 Studio version will do just about everything you need it to. You can download it and try 20 times before you need to buy it.</p>
Link to comment
Share on other sites

<p>Also, thanks C. Sharon for the reference to the Fujifilm Dataguide. </p>

<p>For those still interested:</p>

<ul>

<li>Fuji's 'resolving power' rating is determined by looking at film under a microscope & determining the smallest spacing of lines still discernible</li>

<li>The above technique is what Mauro & I were doing to determine the 'resolution' of Velvia when we quoted ~24 megapixels</li>

<li>The above technique is rather outdated, according to Norman Koren, b/c there's a lot of subjectivity in determining what spacing is still resolved in the medium. Typically, the MTF in these limiting regions is somewhere between 2-5% (probably 2% if you're Mauro :), and generally the term 'resolution' was reserved for MTF < 10% (since that's what was considered near the limit of the system's capability)</li>

<li>MTF50 (the spatial frequency that yields 50% contrast in the final medium) is more closely related to perceived sharpness (according to Norman Koren). For Velvia, MTF50 is around 50 lp/mm, which'd give it approx 8.6MP (not the limiting resolution!). Of course, this does not suddenly make it comparable to a 8.6MP digital camera; it makes it more or less equivalent to a digital camera to shows 50% contrast to a 50 lp/mm chart. <em>I'm still confused as to what the contrast of the original chart was when determining this MTF50... b/c it should change based on the contrast of the chart being photographed.</em></li>

<li>Using Norman Koren's formula for calculating MTF (<a href="http://www.normankoren.com/Tutorials/MTF.html">here</a>), I get a MTF of 8% for that Imacon scan of Velvia 50 at 4.25 (corresponding to 150 lines/mm or 75 lp/mm). Hope I did that calculation right... it seems a bit optimistic. </li>

<li>The 1/Rsystem = 1/Rlens + 1/Rsensor + 1/Rscanner equation <strong>IS</strong> correct... so I'm not even sure how we're getting 75 lp/mm... unless Mauro's lens really does have a limiting resolution of 2400 lines/mm or Fuji underrates their Velvia film (i.e. 160 lines/mm actually yields a formidable contrast in the film). I'm more inclined to believe the latter, though I just don't know anymore. I have to do the test myself.</li>

</ul>

<p>Fully ready for someone to now shout out 'Who cares?' :)<br>

-Rishi</p>

Link to comment
Share on other sites

<blockquote>

<p><a href="../photodb/user?user_id=4958594">C. Sharon</a> , Apr 03, 2011; 12:13 a.m.</p>

</blockquote>

 

<blockquote>

<p>Rishi: Is that equation right?<br>

Yes the equation is right. And to get 150 l/mm from 160 l/mm film you need a lens that can give you 2400 l/mm. Of course such a lens does not exist.<br>

Google for "Fujifilm Dataguide" and download the pdf. On page 128 of the pdf document you can read how Fuji determines the resolving power of film, or the combination of film+lens.</p>

 

</blockquote>

<p>Did you mean to type 2400 or 240? As I've seen resolution on film tests upwards of 125 to 150 show up, I know you don't need 2400 at the lens level.</p>

 

Link to comment
Share on other sites

<blockquote>

<p>Did you mean to type 2400 or 240?</p>

 

</blockquote>

<p>I meant 2400. It was in response to one of Rishi's previous posts in which he wanted to know if the number 2400 that he had derived was correct and in my post I confirmed that he was right.</p>

Link to comment
Share on other sites

<p>Well since 2400 lines/mm seems insane, it's possible we're seeing 150 lines/mm b/c at that point the MTF is something really low (like 2-5%) & perhaps Velvia, processed the way Mauro processed it, actually has something like 240 lines/mm at its 'limiting resolution' (MTF 2%-5%). Then, the lens would only have to have a resolving power of 400 lines/mm to result in a final resolution of 150 lines/mm.</p>

<p>I'm just making up numbers at this point, but it's probably some combination like that & the fact that we're allowing really low MTF percentages to pass as 'acceptable' in determining the resolving power of the film.</p>

Link to comment
Share on other sites

<p>Dave, care to back that up with any evidence or speculation as to how we're getting 150 lines/mm on film? I hypothesize something in my last post -- does that sound reasonable to you?</p>

<p>This is purely for academic interest at this point...</p>

<p>Cheers,<br>

Rishi</p>

Link to comment
Share on other sites

<p><strong>Rishi</strong>: I believe the inconsistency you think you are seeing in your numbers is due to confusion between lines/mm and line pairs/mm. You say you could see 150 lines/mm on the test shot. That is 75 line pairs/mm. Fuji mentions Velvia's resolving power as 160 lines/mm but I have always taken it to mean 160 line pairs/mm. So you if you get 75 lp/mm in your image then you would need a lens that can resolve approx 140 line pairs/mm which is entirely reasonable.</p>
Link to comment
Share on other sites

Wait, C. Sharon, you're saying that Velvia resolves 320 lines/mm with a 1:1000 chart? Doubtful... Seriously there's so

much misinformation out there, as I'm starting to realize. Really the best way to quantify this is via MTF curves that

map the output contrast of the medium based on spatial frequency. As I understand it, these curves don't depend on

input contrast (though I imagine you use charts of varying contrast to fit a global set of data to derive the MTF curves).

 

Rishi

Link to comment
Share on other sites

<p>Rishi, </p>

<p>When Roger Hicks tested the Voigtlander Bessa R3a in Shutterbug mag, he tested his setup between 125 and 150 with Ilford Delta 100. The lens does not need to 2400 before you can obtain 150 on film. </p>

<p>If you're trying to tell everyone that 2400 is needed at the lens level, then the misinformation is your. The math is readily available.....just google it. I'm not wasting any more time with this.</p>

Link to comment
Share on other sites

<p>Dave that is all fine and good but Rihi was talking about Velvia 50 not Ilford Delta 100.</p>

<p>The point is pretty simple, if Velvia has a resolution of 160 then you need a really great lens if you are going to get 150 line pairs/mm.</p>

<p>In looking at the MTF curves Fuji only plots out to around 75 cycles/mm so it is hard to know what it would be at 150 cycles/mm. But it sounds like there is more data other places so maybe that is where the 160 cycles/mm number has come from.</p>

Link to comment
Share on other sites

<p><strong>Rishi</strong>: Resolving power of 160 lines/mm means the number of lines that can be distinguished from one another per mm. For every black line to be distinguished from the next black line there needs to be a white line between the two. That is why resolving power of 160 lines/mm is interpreted to mean 160 line pairs/mm.</p>

<p>The test targets used by dpreview or imaging resource give the resolution in line widths/mm which can be converted to line pairs/mm by dividing by two. That is why the 150 lines/mm you saw is 75 lp/mm. Actually I find 75 lp/mm a bit high. I would have expected it be around 65-70 lp/mm but there is no need to quibble over that.</p>

 

Link to comment
Share on other sites

<p>That's why I prefer Fuji's nomenclature of cycles/mm. No chance of ambiguity here.</p>

<p>Practically, a film is useful out to about 50% response on the MTF curve - particularly so for landscapes where both image detail and tonal subtlety are important. Differentiating between two leafs on an oak at 100m and 1/3 EV apart means preferably working at the 70% MTF of the film regime. Good luck digging details out at the other end of the MTF curve.</p>

<p>For Velvia MTF 50 is about 50 cycles/mm. Of course there's more detail beyond this. If using film is important, however, it's generally easier to just shift up in format when the media itself runs out of steam.</p>

Link to comment
Share on other sites

<p><strong>Scott:</strong> The 160 lines/mm figure comes from Fuji's own data sheet for Velvia 50. You can download it <a href="http://www.fujifilm.com/products/professional_films/pdf/velvia_50_datasheet.pdf">here</a>.<br /><br /><strong>C. Sharon</strong>: Wow, thanks for that clarification. So basically Mauro & I are resolving something like half of what Fuji rates as the limiting resolution of the film for very high contrast stuff? I can believe that. Actually, that's much more believable to me than Mauro actually achieving 94% of Fuji's rated resolution of the film with high contrast subjects... So in a nutshell, our calculation of the resolution of the 35mm frame of Velvia film was fine, but my calculation for the lens required for that resolution was off b/c my number for Rmedium was off by a factor of 2 (I shoulda been using 320 instead of 160)... Actually, as long as we're being thorough, we should be using the cycles/mm figure for the 1:20 or so contrast ratio Mauro's chart actually had under his lighting, right? But anyway, like I suggested earlier, say Velvia's resolution is 120 cycles/mm for a 1:20 contrast subject (not unreasonable I don't think given that it's 80 cycles/mm for 1:1.6 contrast subjects). Then:</p>

<p>1/75 cycles/mm = 1/120 cycles/mm + 1/Rlens; Rlens = 200 cycles/mm</p>

<p>Then you'd only need a lens with a resolving power of 200 cycles/mm (400 lines/mm or 400 line widths/mm) to get the 75 lp/mm Mauro & I see. Not a lens with 2400 lines/mm (1200 cycles/mm) resolution... <strong>Dave</strong>, perhaps you shoulda read my earlier post that said exactly this.</p>

<p>Still, I feel queasy about these numbers. Is this final interpretation correct? Anyone else want to chime in?<br /><br /><strong>Robert</strong>: Actually, in Fuji's data sheet above they say '160 lines/mm'. Am I to take it that they actually mean 160 cycles/mm (i.e. 160 line pairs/mm)? I guess I'm willing to buy this b/c if you extrapolate the MTF curve past the highest spatial frequency Fuji reports MTF for, a ~5% MTF corresponds to about 160 cycles/mm (assuming the falloff continues with the same slope as the right-most part of that MTF curve... which of course is a big assumption).</p>

<p>That would make Fuji's 'resolving power' for high contrast subjects (1:1000) 320 lines/mm (160 cycles per mm), which then if you apply the philosophy of 1 line resolved = 1 pixel, gives you:</p>

<p>320pixels * 24mm = 7680<br /> 320pixels * 36mm = 11,520</p>

<p>Effective resolution of a 35mm frame of Velvia for 1:1000 subject contrast: 88,473,600 (~88MP).</p>

<p>Kinda crazy, but I guess irrelevant in the real world since, like you say, MTF50 is much more important in evaluating the relative sharpness of the film format (I'm assuming the useful MTF, MTF50 or MTF30 or MTF20, for any medium depends upon its rolloff at higher spatial frequencies... so digital may be useful past MTF50?). For Velvia, MTF50 of 50 cycles/mm equates to ~8.6MP for a 35mm frame of film. Interestingly, that's exactly the MP resolution of a full-frame digital sensor that Norman Koren suggested back in like 2005 would be needed to rival 35mm film.<br /><br /><strong>Dave</strong>: My math isn't wrong (1/Rsystem = 1/Rlens + 1/Rmedium is well established). And I'm not trying to tell anyone anything. I'm just pointing out that if you use this formula, you get a ridiculous resolution requirement for the lens. Hence what I was trying to say was that <em>something was wrong </em>with our calculations or our interpretation. I.e. at the onset I <strong><em>agreed</em></strong> with you that it seemed ridiculous to require a lens capable of resolving 2400 lines/mm to get the resolving power of Velvia we calculated from Mauro's experiments. You're not contributing to the discussion by just repeating 'you don't need a 2400 lines/mm resolving lens to get 150 lines/mm on Velvia film'. The other guys here who posted after you are actually helping by offering a reason for why my calculation might be off...</p>

<p>Cheers,<br /> Rishi</p>

Link to comment
Share on other sites

<p>Wait a minute, but if we take Fuji's rating of:</p>

<p>1:1.6 contrast : 80 lines/mm</p>

<p>To mean 80 line pairs/mm... then... does that mean that with our observed resolving power of 75 lp/mm for something like a ~20:1 contrast ratio subject (the printed chart) we're getting the resolving power that Fuji gets with a 1.6:1 contrast ratio subject?</p>

<p>I guess this could make sense, since Fuji doesn't use a lens when calculating these numbers, right?</p>

<p>-Rishi</p>

Link to comment
Share on other sites

<p>Let us not go overboard with cycles/mm and confuse it with lines/mm. Cycles/mm is used with MTF charts because MTF of a film is determined by contact printing a sinusoidal wave pattern on a film, without using a lens. The sinusoidal wave, being a wave, does not have lines but cycles so cycles/mm is used in this case. The sine wave can be approximated with a square wave so line pairs/mm is used interchangeably with cycles/mm.<br>

In the case of resolving power of a film or lens, a resolution test target is used that has lines hence it is more appropriate to use lines/mm or line pairs/mm in this case, instead of cycles/mm. It is useful to maintain the distinction.</p>

Link to comment
Share on other sites

<p>Wow the discussion went on quite a bit.</p>

<p>What is the question? Velvia's resolution with different lens combinations?</p>

<p>I can answer this. Both my MF and 35mm lenses resolve 160 lines per milliliter with Velvia 50. 220 lines per milliliter with TMAX. (on a b&W resolution chart).</p>

<p> </p>

Link to comment
Share on other sites

<blockquote>

<p>Effective resolution of a 35mm frame of Velvia for 1:1000 subject contrast: 88,473,600 (~88MP).</p>

</blockquote>

<p>Sure. Why not. </p>

<p>It's also not practically useful. I mean think about what 5% MTF response for a 1:1000 test target says. You've got a source test pattern that's super contrasty; it's also just barely visible when duplicated onto the film.</p>

<p>Since we're on this track. There was a color film that was both much higher resolving and much sharper than Velvia. The following is the MTF curve for the previous generation (GA6) Kodak Gold 100. Man, it was my favorite for landscape-ish scenes with a small format camera.</p>

<p><img src="http://www.photo.net/bboard/uploaded-file?bboard_upload_id=35523084" alt="" /></p>

<p>Notice two things. First, it is a honest 50% MTF at 100 cycles/mm. Second, it was 150% MTF at 20 cycles/mm: this emulsion had extremely high acutance, i.e., it was super-sharp. This film gave some scenes an almost etched quality.</p>

<p>This Gold 100 Gen6 is gone unfortunately. The current generation Gold (GA7) trades the above qualities that made it truly interesting for "me too" lower grain.</p>

 

<blockquote>

<p>I'm assuming the useful MTF, MTF50 or MTF30 or MTF20, for any medium depends upon its rolloff at higher spatial frequencies... so digital may be useful past MTF50?</p>

</blockquote>

<p>It's not that, but rather digital capture maintains high MTF fidelity much deeper into the high spatial frequencies. Of course, the downside is also that digital has a hard stop; it falls off the cliff right past the particular high spatial frequency limit</p>

 

<blockquote>

<p>For Velvia, MTF50 of 50 cycles/mm equates to ~8.6MP for a 35mm frame of film. Interestingly, that's exactly the MP resolution of a full-frame digital sensor that Norman Koren suggested back in like 2005 would be needed to rival 35mm film.</p>

 

</blockquote>

<p>That's pretty consistent with my experience.</p>

<p>Just until recently, I shot with an 8MP Canon XT. At 8x10 print sizes, it was mostly better than 135 film and indistinguishable from 6x7 really. Past 8x10, the digitally originated image fell apart quickly and without recourse. The 135, however, generally had more to give. It was just that each successively larger print sizes required more and more heroic efforts in post.</p>

Link to comment
Share on other sites

<p><strong>C. Sharon:</strong> thanks for the clarification. I'd forgotten about cycles/mm being derived from a sinusoidal pattern... makes sense now. In that case, one full cycle can be approximated by a black + a white line.</p>

<p><strong>Robert:</strong></p>

<blockquote>

<p>It's not that, but rather digital capture maintains high MTF fidelity much deeper into the high spatial frequencies. Of course, the downside is also that digital has a hard stop; it falls off the cliff right past the particular high spatial frequency limit</p>

</blockquote>

<p>Right, that's exactly what I was trying to say -- thanks for wording it much more elegantly :) I believe this is also what Daniel Lee Taylor was trying to bring up earlier.</p>

<p>I'm glad we finally seem to be in agreement! Thanks to all for humoring my often academic exercises. I still think they're useful b/c sometimes they help (me, anyway) push more out of a medium than may have been thought possible... which can certainly have its advantages given the convenience & portability of, e.g., 35mm systems. I have to admit I may have never bothered to try a 16x24 print from my 135 Velvia shots had I not been dragged into the film vs. digital debate years ago by Mauro. And I'm extremely glad I did. But perhaps I enjoy pushing more out of mediums than others might :)</p>

<p>FWIW, just today I made the jump & invested in a MF system (it was at the right $$), that's easily upgradeable w/ a digital back in the future. It's the Phase One 645AF (Mamiya AFD III) so it's comparable to my EOS system (sort of, for MF anyway). I'm looking forward to having some fun shooting film in it until I can afford a digital back. Of course this now means that my modified 35mm scanner won't be enough... we'll have to see what I can do about that ;)</p>

<p>Cheers,<br>

Rishi</p>

<p>P.S. <strong>Mauro</strong> my question that led to all this discussion was over what the resolving power numbers Fuji reports actually means & why simple math demanded that we have some ridiculously high resolving power lens to get 150 lines/mm. Turns out our 75 lp/mm measurement was just below Fuji's 80 lines/mm (actually cycles/mm or ~lp/mm) rating for 1:1.6 contrast subjects. This lower number makes sense, even though your resolution chart was considerably higher than 1:1.6 contrast, because our measurement is also affected by the lens.</p>

Link to comment
Share on other sites

<p>I see. Well out of all people you know since you have scanned the film yourself (even though the test chart was a printed version lower contrast generated at home) and measure 160 lines per mm.</p>

<p>Resolution is just a small part of the equation. Acutance from digital cameras plays in your favor while you are within the limits of its resolution, but it is a big enemy once once the print breaks down in resoluton making it very obvious. When it comes to real life applications, (in general I shoot TMAX) film looks great even at 24x36. My 40D (10 MP) maxes out at 11x14. This is by my subjective threshold of quality.</p>

 

Link to comment
Share on other sites

<p>Something that might be of interest:</p>

<p>It seems that sharpening with film is important for maximum high frequency detail while with digital sharpening has only minimal effect on high frequency detail.</p>

<p>Norman Koren (author of imatest) measured 103 lp/mm @ MTF10 using Velvia with a 8000 dpi scanner and sharpening. And only 64 lp/mm without sharpening.<br /> <a href="http://www.normankoren.com/Tutorials/Scan8000.html">http://www.normankoren.com/Tutorials/Scan8000.html</a></p>

<p>Erwin Puts made some tests using Leica M9 and Nikon D3X and with the D3X he got 1750 lp/ph @ MTF10 both without and with sharpening. 1750 lp/ph is 73 lp/mm.<br /> <a href="http://www.imx.nl/photo/leica/camera/page155/m9part2.html">http://www.imx.nl/photo/leica/camera/page155/m9part2.html</a></p>

<p> </p>

Link to comment
Share on other sites

<p>I like those numbers much more. Why?</p>

<p>Because MTF50 of 56.5 lp/mm for Velvia yields an effective resolution of 11MP for 35mm Velvia at MTF50. Which'd explain why I see more detail in my 35mm Velvia scan (scanned @8000 dpi on an Imacon & aggressively sharpened) than my 5D images (also aggressively sharpened) when both are upscaled for a 360dpi 24x36 print (i.e. such that the final pixel x pixel count is the same between the two).</p>

<p>The MTF10 figure a resolution of 36.7MP for 35mm Velvia... maybe... but I guess it's not that useful b/c 10% response for lower contrast stuff probably would be hardly discernible.</p>

<p>Isn't Norman Koren's stuff all simulated & not actual scans/data?</p>

<p>-Rishi</p>

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...