Is Slide Really Better?
Is Slide Really Better?
There's something undefinably nice about the way a correctly exposed slide frame looks. I often get a special thrill out of them -- even scanned on my low-end Minolta Dimage Scan Dual II scanner. This is one reason I went through a rather a lot of trouble to investigate what, exactly, it is. Unfortunately, I came up with... nothing. Here's the story -- probably the most work I've put into an article yet, and probably the least useful article on my website.
The contenders: slide and neg seen by digital.
I know I'm putting myself up on a limb here, and opening myself to all kinds of criticism about sloppy methodology, unskilled or unfair post-processing, biases, cheap equipment, and what have you. Therefore, a few disclaimers -- and I mean them.
- This experiment has any validity only when discussing roughly similar equipment to mine -- entry-level or at most midrange consumer scanners, midrange DSLR's, commercial processing.
- The experiment is especially unfair towards slide: I know for a fact that a better scanner could recover significantly more detail and at least a little more tonal range than mine.
- Nothing is "raw." I've had to post-process everything in order to get roughly comparable looks. I haven't been at all strict about my post-processing: instead, I've done it to the best of my ability for each of these media. I have not consciously tried to make any one of them look bad or another good. However, I've also not gone to any great lengths to make them look identical, except where there's a specific characteristic I'm looking at.
The thing that had been chafing me for a quite a while is this: I really like the way good, slow-speed slide looks. It's virtually grainless, delivers beautiful, natural color, and is really, really sharp. However, I find it a royal pain to work with. It's almost impossible to get anything out of it under some conditions (e.g. when there's more than average contrast), and you have to have a very precise hand with exposure to really make it click. I "lose" way more shots with slide than with any other medium, yet I've kept shooting it because of the ones that really jell. It's also a pain to post-process -- I don't use filters, so I need to white balance digitally, it takes a light hand on unsharp mask to bring it to optimal sharpness after the scan, and, of course, spotting is a chore, whatever way you look at it.
Too much contrast for slide to handle.
In particular, I like the way slide handles shadows and highlights -- when they're "right," they really look so "chewy" you can bite into them.
I've been migrating steadily to digital. The Canon 10D has become my primary medium and made film into something of a diversion. And I really wanted to know exactly what I'm losing by giving up on slide. Therefore, I went through a lot of trouble to perform this experiment -- I did it for myself, not to prove a point to anyone else; therefore, I worked with the gear I usually work with, and post-processed the results the way I usually do. My scanner is calibrated with a Provia 100F target, my monitor is ship-shape, and I think the rest of my stuff and workflow is pretty characteristic of an "advanced amateur" or "low-end pro" affair, too -- so, I've decided to share the results in the hope that someone might find them useful. What I don't want to do is start yet another film-versus-digital firefight; I find them utterly pointless.
"Tyre Boats." Canon FD 50/1.4 on Provia 100F.
What I did
I picked a day that I thought would be nice, and went to the Suomenlinna island fortress outside Helsinki, packing two film bodies, a digital body, a tripod, and 50 mm and 35 mm lenses. Then I looked for subjects I could photograph that had the combination of higlight and shadow detail that most interested me, and photographed them from the same position on three different media: Fujichrome Provia 100F professional color slide, Fujicolor 160 NPS professional color neg, and my EOS-10D for digital. Unfortunately, the day turned cloudy almost as soon as I got there, so I ended up with sub-optimally color balanced slides.
I used a Canon 35/2 on the 10D, and a Canon 50/1.4 USM on film; I know from experience and previous experiments that the two lenses are optically as close to identical, at least stopped-down (which is how I used them). I took all of the scenes very carefully off a heavy tripod, and bracketed exposure with 0.5 stop increments on slide, 1 stop increments on neg, and 0.3 stop increments on digital. Then I picked the best exposures to work on for analysis.
At home, I converted my RAW digital images with CaptureOne DSLR, and scanned the films with a calibrated Minolta Dimage Scan Dual II. Then I massaged them in Photoshop to get them looking more or less like I wanted them to look; most of the time, I used my slide frame as the reference for the others. I kept them in 16-bit until they were almost right, then applied some final touches. This is the big subjective component of the experiment, and renders it thoroughly non-scientific. However, some of you who do similar things may still find it worthwhile.
Leaves, Sky, Water, Shadow
I shot my first scene to explore dynamic range and how much of a resolution advantage (if any) my 11 MP scans from the Minolta Dimage Scan Dual II have over 6.3 MP digital. Not much, it turned out.
Obviously, at this tiny size there won't be much difference between the frames, and almost all of the differences can be attributed to the post-processing. In fact, I could tweak them to be virtually indistinguishable. However, it is significant that these "looks" came to them rather naturally, and they're consistent with my previous experiences with the media.
At this "bird's eye view" level, I like the neg version most. It has most contrast in the midtones and most detail in the sky. The slide frame has an almost completely flat sky, while the digital exposure still has a fair amount of detail there that I could recover with different RAW settings. However, I think the preference for neg may be simply because I happened to do a better job on it than the others. Now, let's look more closely.
Detail 1: Against the sky
This is the bit where film aficionados can start crying "foul," because for slide at least the factor limiting quality is clearly my scanner: a better one would certainly have recovered significantly more detail. Anyway, these are crops from the above image. The digital one is up-rezzed to match the scans, and all are sharpened to the best of my ability.
However, we'll draw a few conclusions anyway, keeping in mind the above caveat.
I had to look really hard to see any detail in the slide crop that isn't in the digital one, despite the higher "native" resolution -- after a long time of intent staring, I did find a few leaves that were defined on the slide scan but were merged together on the digital one. However, I discovered that this is not because slide doesn't have a resolution edge (in fact, it does), but rather, because the detail in this scene is large enough to be fully defined by the digital capture. With detail of this frequency, the difference is far too small to matter on prints of any size. It wasn't easy to sharpen the slide to this point, either -- it took several passes of carefully applied unsharp mask applied to separate layers and blended in Luminosity mode. This has already started to bring out the grain; more would definitely make it worse.
I was surprised at how well the digital original rezzed up and sharpened -- at a glance, it looks the most detailed and "sharpest" of the bunch. Some digital artifacts are visible, though -- in particular, a bit of stairstepping on near-verticals and near-horizontals. Nothing dramatic, though. Because of the low noise, up-rezzed digital images take to unsharp mask very well; slide and especially neg aren't as fortunate. The digital image needed much less work than either of the film ones to bring to optimum sharpness: one or two passes of USM are enough.
The neg looks a pretty clear loser in this category. Even the exceptionally fine-grained Fujicolor 160 NPS shows pretty clear grain; this is also the limiting factor of enlargeability and appears to limit the absolute resolution of the film, too -- the detail is "blobbier" and "mushier" than on slide. It does win in one respect: the branches blend with the sky completely naturally, with no signs of blooming or burn-out. I have a feeling that a better scanner would mostly dig up more grain rather than more resolution: the detail appears less well-defined and "coarser" than on the slide scan. However, it was a good deal easier to scan and process the neg -- scanner auto-focus locks on the grain every time, scanner noise and exposure don't really matter that much as neg is nowhere near as dense as slide, and you can get any level of contrast you want by wiggling Curves in 16-bit. I enjoy "digitally printing" neg, whereas slide is definitely a chore.
Detail 2: Mid-tones
More of the same: slide and digital are neck to neck, with slide possibly a tiny bit in the lead (a bit more detail in very low-contrast areas), and neg bringing up the rear. The chromatic quality of the grain is more distracting than the even grain of the slide, too; all in all, the image has a "sandpapery" texture.
Detail 3: Shadows
We'll explore tonal response in the shadows properly with the next scene. I wanted to show this one, because it demonstrates how closely the default conversion curve on CaptureOne DSLR has been tweaked to mimic slide. The digital and slide crops are very close to each other indeed. (The same goes for the first, "highlight detail" crop, by the way.) There's a lot more shadow detail present in the neg; I intentionally clipped it to get similar tones as on the slide and digital frames. However, note that the negative grain is a lot less distracting in the shadows.
Detail 4: Grass
These samples have been further upsampled by 200% with nearest-neighbor interpolation.
Ah-ha! Finally we have it: the much-vaunted resolution edge of slide. This patch of very fine, high-frequency detail catches it rather neatly. On the digital frame, the finest details of the grass and hay have been blurred into patches of even brown and green, while there's still strong detail to be seen in the slide scan. I believe that here we can see the anti-aliasing filter in action: it has simply blurred everything below the sensor's Nyquist limit.
To my eye the neg appears to be doing about as expected -- the fine detail is buried in the grain, and in fact there are some features visible in the digital and slide capture that can't simply be seen on the neg.
So, what does this mean in practice? Simply this: on scenes with fine, very high-frequency detail, such as near-infinity landscapes, slide does hold a non-trivial edge over 6.3 MP digital. This sort of thing will not be at all obvious at prints about 8 x 10 in size, but will be visible at larger sizes than that, if the image contains the type of detail that will bring out the effect. If the detail is large enough to be fully defined by the capture resolution (such as in the first crops), digital will probably enlarge better than slide, due to the lack of noise/grain. In other words, for off-the-tripod landscapes, slide resolution still can't be beat by 6.3 MP of digital.
I would also conjecture that this phenomenon is one reason why Foveon landscapes look so good in large sizes. Since there is no anti-aliasing filter, fine, random detail like this will continue down to the individual pixel level -- even if some of it is spurious. Therefore, this sort of patchy blurring will not occur.
Tonality in Rock and Shadow
My second test scene involved a low-light scene, with high contrast and a soft gradation from light to dark. Unfortunately I messed up a bit while shooting: I shot the slide ones with a larger aperture than I intended, but since I wasn't primarily interested in resolution, luckily this won't really change anything.
I took a bit more liberties at post-processing these frames individually: I used less contrast on the neg one, to show off the increased highlight and shadow detail. At this "bird's eye view" it's clear that slide struggles with the dynamic range of the scene: the shadows are completely blocked and the highlights are just on the edge of burning through. Again, the similarity of the digital default conversion curve to the slide's tone curve is remarkable.
Note: The details below are downsampled, not 100% crops.
Detail 1: Slide calls the tune
For the first detail comparison, I took the tonality of slide as a reference, and post-processed the two others to match it. I was getting a trifle sloppy at this point, but you get the picture:
Slide struggles. The tones fade to black rather abruptly, while the digital capture retains low-key detail. The neg is just there for completeness -- I clipped out a lot of the shadow end to get similar tonality on it. Look especially at the shadowed rock in the dark part of the corridor.
Here, digital is a clear winner over slide: low-key tonality continues deep into the shadows, while there was nothing but scanner noise in the corresponding areas of the slide. A better scanner may recover some more tones from it, but I'm not sure how much -- examining the slides with a loupe against a strong light doesn't really show up much more.
Detail 2: Negative versus Digital
But just how much detail is there in the digital image, and how much in the neg? Let's see: here's a low-contrast rendition of the negative, and a two stops pushed version of the digital. I didn't even attempt to make them look the same; this is simply to give an idea of what's in the shadows of each.
Hoo boy. Neg wins this round in a walk -- there's tons of detail in the shadow areas, and it's very nicely rendered, too. While digital can muster up a similar amount of detail, rather nasty banding also raises up its ugly head, and noise reduction in the RAW converter has flattened out a good deal of the fine detail. Moral of the story: there's a lot of detail in the shadows of a RAW image, but its usability is limited due to tone shifting, banding, and noise reduction. Digital surprised me positively, but it's clear that neg is still king when it comes to dynamic range. Besides which, I could easily have burned it a stop or even two more without losing the highlights.
Thoughts on a Pink Building
I shot a few other scenes as well, but I'll include only one more in this write-up, and drop neg from that comparison: there isn't a whole lot this scene can tell about neg. It's something interesting that came up when playing with a pink building.
I took my slide frame as reference -- it is from a calibrated scanner, after all. I then adjusted the white balance settings in my RAW converter until the pinks and reds of the wall were a pretty good match for the ones on the slide. Here's one result.
The interesting thing is that no matter what I did in the RAW converter, I couldn't get both the reds and the greens to match the slide. I could do the reds, but then the greens looked much bluer on the slide (incidentally, I think the digital version is closer to the way it really looked); when I targeted the greens, the wall went totally blue. I hear that Provia 100F is supposed to have especially thick blues: it could be that were's seeing its tilt that way here. In an effort to keep the greens at least somewhat similar, I left the wall on the digital version a bit cooler than the slide.
As the last crops from this shootout, yet another slide-digital comparison: this time, I'd like to point out the way each of the clips a highlight, the way high-key detail in the window frame is rendered, and (again) the larger amount of shadow detail in the digital capture.
I wouldn't draw too many conclusions about the highlight clipping on this one example, though -- highlights clip in very different ways. However, I have discovered ways of dealing with the "digital highlights" in RAW conversion and post-processing. In fact, I think that while they are a real issue for in-camera JPEG's and for some default RAW converters, if correctly approached, there's nothing inherently worse about "digital highlights" than slide highlights.
Whatever magic color slide may have, my experiment didn't catch it. The only significant advantage over digital that I could find was in the way slide rendered high-frequency detail: digital blurred everything below the sensor's Nyquist limit, while with slide detail continued until it gracefully faded into grain. However, this wasn't what I was looking for: with the exception of one or two prints, lack of detail has never been the issue with my digital work; rather, it's been some strange magical thing in light and shadow. Either my equipment was too crude, my technique wasn't good enough, or I was looking in the wrong places -- or, perhaps, there is no such magic. Whichever the case, I figure that if I can't bring it out in conditions much better controlled than my usual shooting, it's unlikely I can bring it out at all.
Therefore, I'm dropping color slide from my palette. In order to get significantly better quality out of it than my EOS-10D, I'd have to get a much better scanner -- and considering the amount of slide I shoot, that would be prohibitively expensive. However, if off-the-tripod landscape photography was my prime concern, I would be reluctant to do so -- I would really want a higher-resolution digital camera. Alternatively, I would give a Foveon-based camera serious consideration: while Foveon captures do have issues of their own, they should avoid the "blobby blurring" artifacting caused by the anti-aliasing filter on color filter array cameras, and should therefore enlarge better.
Neg is staying: the marvelous latitude and dynamic range make it an ideal medium for many things, and its characteristics make it easy and rewarding to scan and work digitally. It will eventually go, too, once digital cameras acquire the latitude of neg (and I can afford to change cameras).
Digital will continue to be my primary medium: it's so much easier to work with, both behind the camera and in the digital darkroom, has some clear advantages over slide (in particular, dynamic range and shadow detail), and most of its weaknesses can be worked around. However, for digital to put up a serious fight with slide, it needs to be the RAW deal. In-camera JPEG does often suffer from ugly clipping in highlights, has not much (if any) more dynamic range than slide, and doesn't up-rez nearly as well due to sharpening artifacts.
Doug of Volumen at the guitar. I really wish (now) that I had shot this in RAW -- those ugly highlights would've been easy to fix.
The combination of a good DSLR and a good RAW workflow does look a lot like it can put 35 mm slide out of business at the "advanced amateur" level, at least for everything other than off-the-tripod landscape photography. It matches or exceeeds slide in quality, but especially in control and convenience. I can get optimal results out of digital almost every time, while to do the same in slide requires either bracketing or more sophisticated metering than the crude in-camera meters can manage. For that, I need an ambient light meter and spot meter, plus a good head for arithmetic. Most importantly, the resolution advantage of slide remains theoretical unless I invest in a scanner a great deal better than mine -- something along the lines of an Imacon. But for that kind of money, I'd rather have an ultra-high-resolution full-frame DSLR. Not that I can currently afford either.
I've learned a quite a lot when doing this experiment, and especially in writing it up. I now understand all of the media and machines I use better, I've gotten more control over tonality in my post-processing technique, I've located some things I do like in slide and know what to do to bring them out in digital, and I've figured out a few minor ways to save effort and time. This piece of pontification has been the most work I've put into any single article on this website -- and also, probably, one of the least useful for anyone else. However, in this case at least, the work was its own reward.
Fisherman and shrine in the New Port of Tyre, 2003. Shot with a Canon 10D in RAW, post-processed in CaptureOne DSLR LE and Photoshop 6.0. I wanted the rich look to the highlights you get by underexposing slide a little. I did this by converting twice: once for the highlights, once for the rest of the frame, layering the conversions, and using a "highlight mask" action I've recorded to blend them. Yep, I find it easier than trying to finesse the exposure while shooting it.