Hmm, that is strange. I found that you can actually pull a lot out of the shadows especially from Velvia. On any other film that I have recently used this shot would have really looked very very bad (sorry, it is oversharpened, have to re-process):
Velvia has thicker shadows than any other slide film. I see something like 4+ stops of shaddow detail in my NCPS Velvia scans before black while I might get 1 or two out of any Kodak film.
I think your 'underexposed' areas look bad in that scan. They look muddy and lack contrast. It looks like the black point wasn't sent correctly.
I'm afraid I can't get at the shadows. I don't know if it is a limitation of the scanner or something else. Here is a second example with high dynamic range, I redid the scans at different settings and I also happen to have a digital example taken at the same time. I converted the images to grayscale in order to just focus on the exposure.
Plain scan with Epson Scan:
Epson scan, "Backlight Correction" to "High":
Epson scan, custom curves to open up the shadows:
Silverfast AI, plain scan:
Silverfast AI, maximum exposure:
Silverfast AI, three exposures blended together using Enfuse:
Epson Scan, TIFF imported into lightroom and fill light set to maximum:
Digital, plain from raw (default settings Lightroom)
Digital, fill light set to maximum in lightroom:
The latitude I'm getting from the Velvia in the shadows is something like 3-4 stops behind what I can get from the digital file...
Again, Luka, how do the slides themselfs look? That still looks like underexposed.
In addition I just don't believe that any of your scanning software leaves the scans alone, except maybe the first scan plain Epson scan. What you definately want is switching ANY automatic corrections off when scanning slides. Many corrections in scanning software is 8bit, don't get fooled that your settings tell you is 16bit. It only really is if any image corrections are REALLY switched off. Off course you loose a lot with 8bit.
Have a look at the structure of the street between the cars in your first set. There is definately some contrast adjustment going on in the scanning software, except in the first scan, and also some blackpoint adjustment. You do not want that be done by the scanning software.
Also you are scanning with some borders of the film. Don't do this. It fools the scanner for sure.
Heinz, again, the slides look underexposed. The scan is 48bit, not 8, not 16. Of course all but the "plain" ones are with corrections - that was the point. Borders make no difference except to set a reference for an absolute black point making sure there is no clipping.
Anyway, forget those images. Look at the one you posted (fantastic shot by the way!) - it's the same thing. The contrast is high, shadow DR low. I'm starting to think there's a bit of a misunderstanding going on here - i.e that we're expecting different things. There is this "common knowledge" that film has more dynamic range than digital, but I'm starting to suspect that 1) it does not apply to slides and 2) that was a long time ago.
Let me show you what I mean by OK shadow recovery. This was shot in mid day in Egypt, inside a tunnel. The aperture was f/8 and the shutter speed 1/1000. That's EV 16.
No adjustments:
After fill light, exposure & recovery adjustment:
Now, I have a second shot that correctly exposes for the interior and it was shot at f/4 and 1/125 - that's EV 11. So we're talking about an effortless 5 stop recovery in the shadows here with digital.
Highlights with the M9 are a different story - there negative film does better. Not compared against the 5DII though which has a really large latitude in the highlights.
So I'm thinking that either my scanner is insufficient for the task or the whole story of large dynamic range of film is an anachronism that has not been true for a bunch of years - or both.
We have only one missunderstanding,Luka: with 16bit I mean 3x16bit = 48bit. I repeat that most corrections inside scanning software are done at 3x8bit = 24bit.
Second, trying to pull something out of the shadows from an underexposed slide is no fun at all, it will simply not work.
I could have pulled a lot more out of the shadows of my example above (thanks for the nice comment ) but I did not want to. The scene was just like that in reality.
Heinz, how many bits doesn't matter anyway - we're not talking color accuracy but max and min luminosity values. Larger number of bits only means greater accuracy and finer separation - it won't affect dynamic range.
The slide was correctly exposed for the sky, underexposed for the shadow regions. The scene was simply outside the dynamic range of the slide. My first example shows that it would not have helped to expose it for some average or I would have blown the sky. I really don't see how anything could have been done about it. A scanner with a stronger light could have perhaps penetrated the dark regions of the slide better and pushed the shadows, but that would have killed the highlights. Possibly multiple scans at different light levels and then a HDR combination. I did try that in the example above and it didn't work - although that could be because the V700 has a too weak light to begin with.
For me with my gear it comes down to this:
Highlights important-> digital with large headroom in the highlights (i.e 5DII) or negative film
Shadows important->digital with large headroom in the shadows (i.e M9)
Limited dynamic range->slide film
So if I'm likely to work with scenes with high dynamic range I should avoid using slide film (or at least Velvia 50).
Luka, this is the image above, directly from the scanner, no adjustments at all, just converted to DNG and then resized through LR Flickr publishing service without sharpening.
I pulled fill light to 100 (maximum setting) in LR:
Yes, I'm not surprised - that would be about what I expected. Perhaps a bit better. You do recover some things but to a large degree it's simply moving up the black levels, losing contrast. I think it will be easier to do a comparison if you set the black point so that the overall contrast is good.
This is definitely better than what you get with negative film, but not compared to what you can get from a RAW file produced by a modern digital camera.
denoir wrote:
Heinz, how many bits doesn't matter anyway - we're not talking color accuracy but max and min luminosity values. Larger number of bits only means greater accuracy and finer separation - it won't affect dynamic range.
well, as you loose about 80% of the tonal values with 24bit vs. 48bit because the scanning software cuts off left and right of the tonal value curve you definately loose the headroom for manipulations to a large extend. At least this is my understanding.
denoir wrote:
The slide was correctly exposed for the sky, underexposed for the shadow regions.
Well, you never expose a slide for the highlights, but the shadows. That is how it is and has always been.
I think we both make the same mistake: we are looking for the same image files from analogue and digital. This is not possible, at least I have not managed to achieve that myself. I am not talking about better or worse here, film is just different. For example film is much better at the highlights, as others have pointed out on my images in the Leica M thread. Sure, they also blow, but burned highlights just look so much nicer on film
denoir wrote:
Yes, I'm not surprised - that would be about what I expected. Perhaps a bit better. You do recover some things but to a large degree it's simply moving up the black levels, losing contrast. I think it will be easier to do a comparison if you set the black point so that the overall contrast is good.
This is definitely better than what you get with negative film, but not compared to what you can get from a RAW file produced by a modern digital camera.
This has never been a perfectly exposed image - the opposite is true - and it was shot against the sun, around midday with a 50 years old camera. So this is a worst case scenario. I just wanted to show how much detail can be recovered from the shadows.
There is not much sense to tweak that any further. On a decently exposed slide, yes. But I am not in front of a calibrated screen now. I will try to find a better example later.
Morfeus wrote:
well, as you loose about 80% of the tonal values with 24bit vs. 48bit because the scanning software cuts off left and right of the tonal value curve you definately loose the headroom for manipulations to a large extend. At least this is my understanding.
The difference is if you use 2^8 = 256 or 2^16 = 65536 values to describe each channel. It does not however affect the gamut - that's the color space for nor does it affect the dynamic range.
Well, you never expose a slide for the highlights, but the shadows. That is how it is and has always been.
I've heard exactly the opposite - that you should expose for the highlights and let the shadows fall where there may - that a slide has more headroom in the shadows than in the highlights.
^Just first google hit, I'm sure there are better sources.
I think we both make the same mistake: we are looking for the same image files from analogue and digital. This is not possible, at least I have not managed to achieve that myself. I am not talking about better or worse here, film is just different. For example film is much better at the highlights, as others have pointed out on my images in the Leica M thread. Sure, they also blow, but burned highlights just look so much nicer on film
Yep, at least when comparing M9 with film, but the M9 has poor headroom in the highlights. A 5DII for instance handles it much better. How it compares to negative film, I don't know. Definitely better than Velvia though.*
To me film or digital are different tools - analogous to different lenses. I have to do comparisons to know which to use to get the look I want in a specific case. My obsession with doing tests is not because I particularly like doing tests but in order to know what tool to use in what situation.
*= I'm not speaking of absolute terms, but what I can get with the equipment that I have.
denoir wrote:
I've heard exactly the opposite - that you should expose for the highlights and let the shadows fall where there may - that a slide has more headroom in the shadows than in the highlights.
^Just first google hit, I'm sure there are better sources.
Luka, I cannot believe that I posted such a stupidity, sorry. I should not post during work
Off course the opposite is true, when saying that so generally how I did it above.
What I meant to say is, that you never expose slides for the highlights in a high contrast scene when the information in the shadows are important because of the small range in exposure compared to print film.
Edit: found the link I was searching for:
"The schema in Fig. 1 shows the method I use for determining color slide film exposure... if I do have the time. The first step is, to decide whether the scene presents one of the following two cases:
Only bright tones (highlight areas) are relevant in the picture; shadow detail can be sacrificed.
The scene does not contain any important bright areas; detail and color saturation in the highlights can be sacrificed. "
There is this "common knowledge" that film has more dynamic range than digital, but I'm starting to suspect that 1) it does not apply to slides and 2) that was a long time ago.
Luka, that 'common knowledge' is pretty accurate. For negative film. Slide film typically only has 5-7 stops of usable range (speaking in 'scene' brightness range here). Which is also common knowledge. #2 is probably more wrong than it was before. Current negative film has even more range range than older negative films.
What you are finding is very typical of slide film. You meter and expose for the highlights to ensure your highlights that you want to keep are about 2 stops above middle gray. The shadows fall where they will. Very similar to what you do on digital. Unlike digital, with slide film, there's only so much you can retrieve out of those shadows. They get too dense to scan at some point, though there might be a little more detail in there visible to the eye. But past a certain point, it's just black...
The Dmax on your scanner gives you some indication of the scanner's ability to penetrate dense parts of the negative. Even though your scanner (V700/750?) says it's 48 bit and probably has a Dmax of 4.2, that's just found by taking the bit depth and converting it straight to Dmax, which is wrong. I'm sure it's Dmax is a good deal lower, which means you'd be able to get a bit more out of the shadows on a scanner with a higher real Dmax, like a Coolscan or a Flextight. Definitely on a drum scanner. Some scanners can really change the exposure, letting you make a multi exposure blend of two scans which actually rescues a bit more detail out of the shadows.
As far as films go though, if you feel like being technical (and I know you do ), go download some spec sheets for the films you are shooting from Fuji and Kodak. Take a look at the 'Characteristic Curve' for each film. That tells you the relationship between scene brightness and negative density. Each 0.3 is a stop. You can see where the film gets non-linear in it's behavior and where it's linear. You can also see where the film becomes insensitive to exposure changes. For Velvia 50, as I read it, you've got about 6 stops of range in there - the usable part of the curve goes from about -1.75 to 0, so 1.75/0.3 = 5.8 stops. Go compare that to a good negative film like Portra 400 or Fuji 400H (which should give you on the order of 11 stops with no upper end of the curve plotted, i.e. there's even more up there).
Negative film works a lot differently. The densities are much lower than on reversal AND they capture more scene brightness range. Which leads to much lower contrast images in post. While you can (and should) jack up the contrast in Photoshop if it pleases you, you probably will not be able to replicate the exact slide tonality.
While looking at characteristic curves, keep in mind that they are an indication of what you might get. Changing development will change those curves. Also, on many negative films, the upper end of the curve isn't plotted, indicating they really do have a large range in the highlights.
) but I did not want to. The scene was just like that in reality.
