UPDATE, July 8 2010

In the past month I’ve been using Silverfast multi-exposure almost every day, in a re-archiving project, and it has worked flawlessly.  I also have to admit that my sometimes harsh comments about Lasersoft I always regret later. Sometimes they can be infuriating, but often as not I suspect that it is mainly language issues. They’re actually a great bunch of people doing a great job of keeping film scanning alive for mere mortals who can’t afford Hasselblad’s luxury good price tags.  Oh, and my comments about excessive pricing ? I’m wrong.

So for the sake of consistency, I’ll leave this article up, but take it with a VERY large pinch of salt

---

Lasersoft’s Silverfast has long been considered the best scanning software around, although fans of Ed Hamrick’s VueScan would disagree. I’ve had a love/hate relationship with it for about a decade. I love the results it is able to deliver (once you’ve got over the learning curve) but I really dislike the user interface, and I have little time for the company itself, with its cranky staff and very exaggerated prices. I don’t believe I’m alone in this.On the positive side I have to recognise their continued support for a large range of scanners, many obsolete and/or orphaned by their makers. They play an important role in keeping film alive. I also realise that it must be getting harder and harder to maintain their business, especially sales of their higher range products such as Silverfast AI Studio. Which leads me to the point of this article, revisiting a topic discussed some time back.Around about version 6.0 Silverfast was pretty much complete. There wasn’t really much to add, which is a problem for a software company. Nevertheless things were added, often hyped to the heavens but actually delivering very little. For example, the “Studio” version of 6.5 added things with clever sounding acronyms (e.g AACO, Auto Adaptive Contrast Optimisation) which actually didn’t seem to do anything useful, although they spent a long time doing it. Ever desperate for upgrade revenue, a more recent attempt was Multi Exposure. As opposed to Auto Adaptive Contrast Optimisation, Multi Exposure is supposed to, er, optimise contrast, auto-adaptively. It does this by making two scans, the first at normal exposure, and the second deliberately over exposing to pull out shadow detail. It then combines the two into a final image. Initially I seem to recall there was an option to make 4 exposures, but this seems to have been quietly dropped.

Some film scanners, like my Minolta Scan Dual Pro, have the ability to multisample, taking a number (between 2 and 16 in the Minolta’s case) of samples at each point and averaging them out to improve the signal to noise ratio, especially in the shadows. Many Silverfast users were puzzled about the difference between “Multi Exposure” and “Multi Sampling”, especially as they are mutually exclusive in Silverfast, even for scanners like the Minolta where the film doesn’t move. An interesting discussion took place here. The drawback of Multi Sampling is that scan times are increased by the same factor as the sample count. Lasersoft promised that Multi Exposure would not only be faster, but would deliver better results.

Well, Multi Exposure went through a few iterations, and the 4x option vanished.  My experience is that it does not offer any significant dynamic range advantage over multi-sampling, at least as far as scanning slides is concerned. It is quicker, slightly faster than 4x multisampling. However, it has a serious flaw, which others have noted: the results are considerably softer than standard or multi-sampling. This may be due to misalignment, or due to flare or bloom in the over-exposed scan. The result can clearly be seen in the 100% crops below:

Top: 4x Multi Sampling - Bottom: Multi Exposure

Sometimes Multi Exposure works fine, but it is just too unreliable to use routinely. In most cases I find that 4x multisampling gives excellent results, with diminishing improvements (if any) at 8x and 16x. And in extreme cases, you can make two multisample scans and different exposures and blend them in Photoshop. So, in conclusion, another pointless feature.

I’ve said it before, and I’ll say it again: if Lasersoft really feel there is a future in this product, then they should concentrate on repackaging the technology in a completely new, modern user interface. Unfortunately, I would guess that the codebase is ancient, and I’ve never seen any evidence that Lasersoft have any interest in genuinely improving the Silverfast user experience. Since the competition is at best no better, and in general considerably worse, I suppose there’s little commercial incentive in doing anything.

It works.

That’s it really. Despite strong fears to the contrary, the venerable Minolta Dimage Scan Multi Pro (I always have to look at the faceplate to make sure I got that right) works fine under Mac OS X 10.6 “Snow Leopard”, with Rosetta installed.

Tested configuration:

MacBook Pro 2.5Ghz, OS X 10.6.1
Dimage Scan Utility v1.0.0 (Dimage Scan Installer v7.4, 5 Oct 2005)
Silverfast AI Studio v6.6.0r5

No noticeable issues of any kind with either application.

That’s a relief….

I’ve been scanning slide film for ages. I’m on my 3rd film scanner (and probably the last, the way the market is going), but I only very recently decided to try working with negative film. I was tempted to do this after hearing about Kodak’s new Ektar 100 film. It sounded like it combined the best of both worlds: the extended exposure latitude of negative, and the definition of slide, and threw in excellent colour density into the bargain. Unfortunately, I haven’t used it yet, although I do have 5 rolls waiting for me to put them in the camera. Obtaining anything even vaguely esoteric in this sleepy cut-off corner of Switzerland is never easy, and one week was not a long enough lead time for my order to arrive before I left for the Eolian Islands. However, I did manage to find a 5 pack of Kodak Portra 160NC, which I was assured would be great for landscape (a lucky coincidence, since it was the only thing the shop had). So when I finished the few rolls of assorted Velvia 100F and Provia 400X I had found at the back of a drawer, I tried using it in my Hasselblad XPan.

The next challenge was getting it processed but I’ll gloss over that (try explaining to Hicksville Cameras that no, you don’t want prints, and NO you do NOT want the film to be cut ... ). Finally comes the moment of truth when it has to be fed into the scanner. I daresay that with experience you can judge the merit of a frame by eyeballing a negative, but I certainly can’t. So scanning is the only way to reveal what I’d actually photographed.  Turns out I’d spent a few days on Mars. Or had accidentally used colour infrared film. At least that was my conclusion looking at the results of SilverFast’s much vaunted (by them) NegaFix tool. This is supposed to characterise the stated film - and yes, it does included Portra 160NC in its list - and produce a beautifully rendered inversion to true colour. Well, to be blunt, it doesn’t.  I spent a frustrating day trying endless combinations of settings, fooling around with SilverFast’s arcane user interface, trying to convince myself that colour calibration was indeed disabled as it should be, but all to no avail. Everything ended up looking like a faded 1970’s Kodacolor snapshot.

I searched around on the web to find out if I was actually attempting the impossible: maybe Portra 160NC doesn’t work outdoors ? But instead I found a few examples of beautiful landscapes taken using it - and a reference to something I’d never heard of, C F System’s ColorNeg Photoshop plugin.  So I tried it.

I hate to say it, but the installation process for ColorNeg, at least on the Mac, is something of a challenge. And the user interface is, to put it politely, unconventional. But the results soon made me forget any superficial objections. I’ll say straight away, ColorNeg renders SilverFast totally pointless. In fact, it even introduces a suspicion that most of the vast array of correction tools available in SilverFast are mainly there to overcome glitches it inserts itself. With ColorNeg, all you need to do is a make a “raw” 16-bit linear scan, open the file in Photoshop, and point ColorNeg at it. Five times out of ten, ColorNeg’s default rendition is perfect, and the rest of the time a minor tweak of the lightness slider gets the rendition I want. Actually, since there is no “correct” way to convert a negative scan to display space, ColorNeg would need to be equipped with a mind reading interface to get it right every time. So problem solved: I can scan Portra 160NC.

But it gets better: ColorNeg has a sibling, ColorPos, which does the same magic on slide film. Again, scan at 16bit linear, point ColorPos at it, and hey presto, perfect result. This is unbelievable, compared to fiddling around in SilverFast or VueScan or whatever. I’m not going to even try to understand the mathematical thinking behind these plugins, which David Dunthorn, their author, describes on the C F Systems website, or the plethora of advanced adjustments available to the initiate, but what I do understand is that they work.  I have immediately adopted a new workflow - scan and archive raw 16bit linear scans, and convert them to display space in Photoshop using these plugins. This is a workflow I’ve been aware of for ages, but trying the various implementations of it using SilverFast HDR or VueScan has never convinced me.

It is really amazing that at this point in the evolution of digital imaging an individual could go back to basics and reinvent the whole process with such effective results. How come companies like Adobe and Lasersoft, with all their resources and experience, cannot do as well ? Probably because there is no commercial benefit to them. It is easy to bury non-optimal or even mediocre processes under layer upon layer of feature creep, which marketing has a far easier job with than selling the message “hey guys - guess what ? we got the basics wrong, so we’re starting over”.  David Dunthorn deserves far, far more recognition that he’s got.

Fumaroles on the crater of Vulcano. Colour reproduction by ColorPos

Registration is $67, which covers both versions, as well as GamSat, a colour integrity-preserving saturation adjustment plug-in which I’ve only peeked at so far. Excellent value for money.

(Read more about ColorNeg over at the Auspicious Dragon. Somehow I missed it when these posts were published last year)

Moving from one RAW workflow tool to another is always going to be painful, a fact that few reviewers ever touch upon. This is the downside to non-destructive editing. Until now, nobody has come up with a way to translate RAW settings from one converter to another, and even if they ever do, it is likely to be an imprecise art. For example, while both Lightroom and Aperture have local contrast enhancement tools (Clarity & Definition, respectively), they behave and respond differently, and it is unlikely that these differences can be easily characterised. So, moving from Lightroom to Aperture, or the other way, is going to be complicated and potentially involves a lot of work. A RAW converter is not just for Christmas. Things are a bit better on the cataloging front. IPTC and keyword metadata written into DNG files in Lightroom should import into Aperture, although some workarounds are required, in particular where ratings are concerned. One thing I dislike about Lightroom is way that keywords are edited and managed, and especially how this is all mixed up with search. I especially dislike the way that I have to switch to the Library module to work with keywords. Well, with Aperture, you don't need to switch to anything, but I have to admit it took me a while to work out how I could make bulk edits to keywords. When importing keywords with certain characters, for example "á", Aperture mangles the keyword. So "Jökulsárlón" became "árlón". To fix this I tried to do it the "Lightroom way", which obviously didn't work. You can't edit keywords in a multiple selection using the metadata panel, at least as far as I can see. But you can use the Keyword HUD: This can be quickly used not only to edit, add or remove keywords, but also quickly apply them to images, whatever you're doing to them, be it editing, retouching, arranging for print, for web, anywhere. And for my scattered mind, this is way, way better than Lightroom's rigid approach.

I'm in the process of moving, or trying to move, from using Adobe Lightroom to Apple Aperture. The reasons for this I'll get into later, but I'm getting the feeling it may not be a fully satisfying experience. It seems that Aperture 2.1 still has serious performance issues. I'm running it on new MacBook Pro, with 2.5Ghz CoreDuo processor, GeForce 8600M GT graphics card, and 4Gb RAM. Should be enough, really. But I'm beginning to think it isn't.

Aperture 2.1 zips along fast enough in image browsing mode, but as soon as I start adding adjustments, things start going downhill fast. The loupe, for example, starts staggering around like an intoxicated tortoise, and strange video artifacts show up, such as half the image blanking out, or the image disappearing altogether when I move a slider.

Most irritating, the histogram in the levels "brick" doesn't display, and frequently the main histogram doesn't either.

Levels adjustment histogram missing in action...

The histogram itself is very sluggish, and cannot be used to evaluate the effects of adjustments in real time. All in all, it is quite worrying, and also a bit baffling. This is the top end Mac laptop. It is used in PR shots for Aperture 2. And yet it performs at a level which, honestly, is barely adequate. Do I really need a Quad Core Mac Pro to run this thing ?

UPDATE: deleting Aperture's preferences has restored the histogram, and, it seems, performance. I suppose the fact that I repeatedly created and deleted a lot of projects whilst trying to get Lightroom metadata to come across may have had some side effects. Hopefully they won't return.