What with the
Fuji X100 bagging all the headlines, I've decided to have a look at one of Fuji's older cameras, the
Fujifilm Finepix S3 Pro. It was one of the most interesting digital SLRs of the last decade or so, with a sensor that emphasised dynamic range rather than resolution. It was released in 2005, at a time when other manufacturers were loading their cameras with more and more megapixels. With the S3, Fuji showed the world a new way, although in practice the world didn't listen.
Back then, the market for digital SLRs was booming, but starting to stagnate. The quirky, odd designs of earlier in the decade had been replaced by a mix of broadly identical six megapixel, eight megapixel APS-C models that shot at 5 fps or thereabouts, with competent autofocus, decent high-ISO performance, and pleasing colours. Each model was replaced by a new model that had a few more megapixels, a slightly larger screen, a few more menu options. And a button that could send images directly from the camera to a printer. A button that nobody wanted and nobody pressed, which is a shame because
buttons love to be pressed. They love it. They love to be touched. Stroked, pinched. The anticipation of being touched.
An early photographer tries to persuade a woman
to take off her clothes and put on a mask
Fuji S3 / Samyang 85mm f/1.4
Buttons that nobody wanted. They weren't born,
so much as they fell out. Most of the class of 2005 have since made their way to eBay, stripped of their packaging and sold for pennies. Cheap credit was all the rage back in the middle of the decade. People bought digital SLRs that they didn't really need, to take on holidays that they had booked so that they had something to photograph with the digital SLR they had bought. They bought the cameras to take pictures of things that they bought in order to have something to take pictures of. A bigger television to watch a newly-remastered DVD that had been purchased in order to show off the bigger television. I'm digressing here. Before I write about the S3 I will have a strong cup of tea, and then write a brief history of Fuji's digital SLR range, which blazed a trail that no-one followed. They took the road less travelled by, and ended up cold and hungry, and so they retraced their steps. They saw some sights, all right, but there's no sense dying for a dream.
But,
before even that, here's an example of the Fuji S3's big trick. The British Isles contain some pleasant scenery, marred by an abundance of unpleasant things, including drunk thugs and ugly people and overcast skies and indeed almost all of the people in general, with the exception of David Attenborough. The S3 can't solve Britain's social problems, but it can help with overcast skies. On the left, an image shot with my Canon 5D Mk II, a modern digital SLR. On the right, the same image, different focal length, similar exposure settings - if anything, a bit brighter - shot with a Fuji S3, which came out six years ago. It's a rubbish photograph, but it's
true, because the weather really was like that:
Here's the exact same pair of images run through Photoshop, in an attempt to bring back some detail in the clouds. 5D on the left, S3 on the right. The S3's image isn't perfect, and the colours are different - although that's a Fuji thing, rather than a clipping thing - but it's a lot more pleasing to the eye. You can see texture in the clouds. The 5D's sky is a kind of neon wash:
The 5D's image is much larger, of course. Parts of it are excellent. But the image as a whole is completely ruined. In theory I should have used an actual real-life physical graduated neutral density filter, or I should have underexposed and brought up the shadows - albeit that the 5D, for all its solid high-ISO performance, doesn't cope too well with that - or I should have bracketed the shot and layered a couple of exposures, but, dammit, I ain't got the time. The S3 lets me get loose, and gives me a file that I can mould. Albeit that it's quite a small file by modern standards. Here's the difference between six and twenty-one megapixels (the S3's file has been tweaked to match the 5D's colour balance):
Still, the Fuji S3 is a bit of a paradox. It was aimed, in part, at the kind of photographers who
would have been prepared to bring along graduated filters and take multiple exposures and so forth, and yet its big gimmick was a kind of EZ training wheel exposure mode that laughed at all that. And the kind of run-and-gun photographers who would have benefited most from the S3's looseness would have been irritated by its slow speed and plastic body.
A kind of metal thing, at the V&A
I've written before about
Kodak's range of digital SLRs, the DCS series. Kodak essentially invented the digital SLR back in the 1990s, but struggled to commercialise their invention successfully. Although the company knew how to make sensors and imaging electronics it didn't know how to make cameras, and so it had to rely on bodies provided by Nikon and Canon. Kodak offset the expense of all the R&D with astronomical prices - the $28,000 tag for a six megapixel
DCS 460 makes the recent Sigma SD1 seem like a cheap load of tat, especially given that those were 1998 dollars - but ultimately the Kodak DCS range was doomed once the major camera manufacturers started to sell their own digital SLRs.
Fuji faced the same problem. The company is almost a Japanese mirror image of Kodak, but
green instead of yellow. They both have a rich history dating back to the film era, and they both spent a lot of time and money developing digital imaging solutions when it became apparent that the film era was not going to be around forever. Kodak won that race by a few years, although Fuji was not far behind. Ultimately Kodak frittered away its head start, trying to hold onto the high profit margins of high-end digital SLRs and film, but that's another story.
The one-and-a-bit megapixel
Nikon/Fujix E2 was launched in the mid-1990s, at about the same time as Kodak's second-generation digital SLRs. It combined Fuji's electronics and sensor technology with a new body designed by Nikon, loosely derived from the Nikon F4 film SLR. Although it wasn't as successful as Kodak's contemporary DCS models - which had cornered the press market - it nonetheless deserves a place in history as the very first wholly-original digital SLR, rather than a custom-modified film body with digital components*. It was also the first digital SLR with
full-frame lens coverage, although it wasn't strictly speaking a full-frame camera. Nonetheless it was a technological dead-end that remains fairly obscure nowadays. Nikon eventually went their own way, with the extremely successful
Nikon D1, although the company seemed to retain a soft spot for Fuji, and for Fuji's money.
* As pointed out by Ross Alford, the clever man.
From that point onwards Fuji followed Kodak's model, adding their own sensors and electronics to existing Nikon bodies. Kodak based their cameras around the hefty, professional Nikon F5, and their electronics package was humongous, which resulted in a series of large, weighty, expensive models aimed squarely at professional photojournalists and magazine photographers. For their first modern digital SLR, the
Fujifilm FinePix S1 Pro (2000), Fuji instead opted for a consumer-level Nikon F60 body, and their digital module was relatively compact. As a consequence the S1 Pro undercut the competition by quite a bit. It sold for roughly $3,000 - the list price was $3,499, but Fuji cut some aggressive deals, bundling the camera with one of the hot new
IBM Microdrives at one point - which was two thousand dollars less than the Nikon D1 and five thousand dollars less than Kodak DCS 620.
As is the case with anything that has PRO in the name, the S1 was actually aimed at the high-end amateur market, and thus its most direct competition was the Canon D30, which was released a short while after the S1 and was, of course, not compatible with Nikon lenses. As a consequence the S1 seems to have sold well, although I have
absolutely no idea if it did or not. Digital SLR sales figures circa 2000 are, unsurprisingly, hard to come by. No doubt one-man wedding, events, product photography businesses lapped it up, and Nikon's own prosumer digital SLR, the Nikon D100, didn't come out until mid-2002. On the other hand, the S1 was three times more expensive than a typical high-end digital compact of the day, and it was a whopping eight times dearer than an actual Nikon F60. In fact it was more expensive than a top-of-the-line Nikon F5. So I surmise the audience was limited to those who really, really needed a lot of good-quality images quickly, and thrill-seekers, early adopters, technophiles, the
Wired crowd, Apple cultists. You know, back in 2000 the good old CRT iMac was still selling like hot cakes and Steve Jobs
still looked young. The company was just launching the G4 Cube, remember that? APS film was all the rage. Truly, 2000 was a century ago.
I'll describe the S1 in a bit more detail, because the S2 and S3 were very similar. Essentially the S1 was a Nikon film body with a Fuji sensor and Fuji back wrapped around it. The design was compact but ugly, with lots of shiny cheap-looking plastics and an odd square bulge on the front that housed a USB port, video out, a power socket, and an empty hole that did nothing. The F60 never had a portrait grip, and neither did the S1, even though the lower half of the camera was Fuji's plaything, to modify and abuse for their pleasure.
The back was Fuji's plaything, too. It had a pair of memory card slots - one CompactFlash, one SmartMedia - and a clever dual-screen / hotbutton interface thus:
FUNC beyond the call of duty
That's actually the back of an S3 Pro; the S1 was almost the same, although the FUNC button was slightly lower. The buttons had no set function; you cycled through the orange menu by pressing the FUNC button, and on each page the buttons had a new set of functions. In general they were used to change resolution, contrast, colour saturation, and sharpening, whereas the conventional menu system was used to manage image files and set the camera's custom settings. Unfortunately the all-important ISO setting was only accessible from the mode dial, on top of the camera, which meant that changing the ISO setting broke the flow of shooting.
The lovely Ms Orion
demonstrating the S3's dynamic range
The S1 needed two sets of batteries in order to work properly. Four AAs for the digital components and a pair of Lithium CR123's for the F60, plus a CR2025 button cell to hold the date. This was technically inelegant although some photographers enjoyed being able to use commonly-available AA batteries rather than the chunky bricks favoured by Nikon and Kodak. The F60 itself was one of Nikon's low-end bodies, with a flash sync speed of only 1/125, top shutter speed of 1/2000, half-stop exposure compensation steps. Tracking autofocus was only available in "sports" mode, and centre weighted metering was only available in manual mode, and there was no spot meter at all. It didn't autofocus with AF-S lenses and it couldn't use Nikon's then-new image stabilisation technology and
there were no rocks in it.
Why did Fuji use an F60? At the time the general assumption was that they were being cheap, but it's also possible that the company had no choice. The S1 was developed just as Nikon was moving from the F90 to the F80, and I assume the company didn't want to divert production of their hot new product to Fuji. It's hard to imagine a time when companies wanted to stockpile film SLRs, in anticipation that large numbers of ordinary people would buy them. To take pictures with. Ordinary pictures of everyday things.
Fuji's big thing, and the sole reason for the S1's existence, and indeed for the company's foray into the digital SLR market in the first place, and I apologise for my staccato writing style but that's the way I'm made, with up so floating many bells down, was their Super CCD sensor. Conventional CCD sensors use a grid of square photosites to gather light, whereas Super CCD sensors
wore tights and a cape and were powerful enough to fly through the sun used octagonal photosites in a tightly-packed grid tilted by 45°, which apparently improved horizontal and vertical resolution at the expense of jagged-looking diagonals. For [insert technical reason] cameras equipped with a Super CCD sensor mapped their files into a conventionally-aligned rectangular grid at twice the sensor's resolution thus [insert technical solution]. And so the S1's three megapixel sensor produced six megapixel files, although they didn't actually have six megapixels of resolution. Three and a bit. By law, every article about Fuji's cameras has to have a paragraph about Super CCD, and you have just read mine. I hope I never have to write it ever again.
The claims of six megapixel resolution annoyed a lot of people at the time, in the same way that Sigma's resolution claims for their Foveon sensors annoyed people a few years ago. In the same way that people get annoyed when women wear really short skirts, or young men wear baggy trousers that show off their boxer shorts. The next trend, mark my words, will be baggy boxer shorts. Young men will show off their bottom cleavage. And then moustaches will make a comeback, and long hair.
As mentioned in
my nostalgic look back at the Nikon D1x, crack raw developer DCRaw can bypass Fuji's remapping, which reveals that Fuji's cameras stored their pictures thus:
That's a sample image taken with a Fuji S2. Now, if you imagine the image divided into pixels - in this case really big ones for the sake of illustration, voila:
I've rotated the second image by 45 degrees, and as you can see the pixels are, er, they don't work any more. Wrong shape. And on reflection I think I should have swapped those two pictures around. Yes. The camera's sensor sees the bottom picture, with the grid sort of skew-whiff. There's no way to directly map those pixels onto a conventional pixel grid, and so the camera tilts the picture 45 degrees counterclockwise. The result is a square image, twelve megapixels large, albeit that a lot of those pixels are just blank. Nonetheless the picture... I'm overwhelming my mental capacity here. Fuji's in-camera routines didn't just tilt the image by 45 degrees, you see, they had to tilt the image and then, er, do some clever stuff to transform one into the other, image, with more pixels in between (waves hands).
And that's how it works!
The S1's high-ISO performance was at least as good as the competition, although it had a limited range, starting at ISO 320 and maxing out at ISO 1600. ISO 320 plus a sync speed of only 1/125th made daylight fill-flash problematic; ISO 320 plus a top shutter speed of 1/2000 made shooting in daylight with very wide apertures tricky. Neither of those things impressed actual professional photographers, and for some reason Fuji didn't include RAW. The camera could output a variety of JPGs and 8-bit TIFFs, but not RAW (or .RAF, in Fuji's case). Reviews praised the camera's
perky nipples and
pleasing colours - much less magenta than the D1, more saturated than the D30 - and this became a recurring theme with subsequent Fuji SLRs. On the one hand raw files from the D1 and D30 could probably be tweaked to match those of the S1, but on the other hand this was 2000, and raw files were strange magic.
Still, in summary, the S1 was a solid sensor and image processing system married to a slightly shoddy camera. It won the
2000 DIMA Shoot-Out Award in the prosumer / professional category, although this seems unimpressive given that it was the
only prosumer / professional digital SLR at the time. The S1 was much cheaper than a D1, especially given that Fuji bundled some software that could browse images and do tethered shooting. Fuji's subsequent SLRs were sold with
Hyper-Utility, a simple but functional RAW developer. In contrast, Nikon's equivalent -
Nikon Capture - was an optional purchase that originally sold for $500.
The S1 was launched a few years before the sudden boom in digital SLR sales, at a time when the idea of an individual owning a digital SLR just to take holiday snaps was a bit extravagant, and an awful lot of people who might have bought an SLR a few years later instead opted for a high-end compact instead, such as a Nikon Coolpix 950 or the semi-SLR Olympus E-10. Nonetheless the S1 must have sold enough to justify the
Fujifilm FinePix S2 Pro, which was launched in 2002. Or perhaps Fuji felt that they had unfinished business.
The S2 Pro was broadly similar to the S1, although it had a six megapixel sensor and was based on a Nikon F80 body rather than the F60. The F80 was considerably slicker than the F60 and was a popular choice as a digital chassis - Nikon built their D100 around it, and Kodak used F80 bodies for their full-frame 14n and DCS Pro SLR/n and SLR/c models. It's interesting to ponder whether sales of the original, film F80 outweighed the combined sales of its digital offspring; probably, but nonetheless it was the very last of Nikon's advanced amateur film SLRs. The company never replaced it. The F80 was a bit controversial amongst Nikon people when it was introduced - some people felt that the body was a step back from the semi-pro F90. I've used an F90 body, attached to a
Kodak DCS 460, and it definitely felt denser and more solid than the S3's F80 underpinnings. On the other hand, the F90 doesn't activate the vibration reduction system on modern Nikon lenses, and it can't properly set the aperture on G-type lenses, which don't have an aperture ring. The F80 can do those things.
The F80 had a flash sync speed of 1/180th, which was better than 1/125th at least, and the S2's lowest ISO was dropped to ISO 100. The hotbutton interface was much the same, although the LCD screen was slightly smaller, and the ISO setting was still on the mode dial. Boo! Unlike the S1, the S2
could autofocus AF-S lenses - in addition to Nikon's traditional body-driven screw drive autofocus - and it
could use VR.
As with the S1, the S2 produced double-size output, transforming its six megapixels into a twelve megapixel file. Back in 2002 six megapixels was good, twelve megapixels was very impressive; the general consensus was that the S2 churned out slightly more than six megapixels' worth of information, although the difference was tiny. The camera was praised for its colours, which were pleasant. I used to own an S2, and it seemed to boost the reds and greens at the expense of the blues. Foliage and human skin and autumnal leaves had a pleasing glow; the sky often looked a bit greenish. Here are some of those colours, with a straight-from-the-camera JPG (HARD contrast, HIGH colour, auto white balance) BUT with the complicating factor that I used a Sigma 15-30mm, which is
a bit yellowish:
Fuji's own Hyper-Utility software apparently got the most resolution out of Fuji's RAF files, and here's a very unscientific comparison, Adobe Camera Raw at the left and Hyper-Utility at the right. First the full frame:
Now a 100% crop from a twelve megapixel rendering, ACR on the left, Hyper-Utility on the right, similar sharpening and noise reduction settings, click for the full thing:
God, what a waste of time that was. ACR's rendering looks a bit sharper, but it has what appear to be moire artefacts within the tree branches that Hyper-Utility has eliminated. Notice that the branches are black in the Hyper-Utility rendering but multi-hued in the ACR version, with blue and yellow patches.
Ironically for a camera that originally sold at a slight price premium, the S2 Pro is one of the cheapest models on the used market today, selling in the UK for £150 or thereabouts. In fact it has now become a kind of entry-level beginner's Nikon digital SLR, along the same lines as the later Nikon D40. Its performance overall is inferior to the D40, although as mentioned it will autofocus screw-driven lenses, something that the D40 cannot do. As with the S1, the S2 requires two sets of batteries - four standard AAs and a pair of CR123 Lithium cells - and uses standard CompactFlash cards, although it completely fails to work with cards larger than 2gb. It also has a second memory slot for the long-dead SmartMedia format. The largest SmartMedia cards were only 128mb, and the S2's RAW files were 12mb each, 'nuff said. The AA battery tray is a flimsy thing, and replacements are expensive and hard to find on eBay.
There were also some reliability issues. One batch of cameras was fitted with a faulty sensor that conked out after a while, although Fuji kindly replaced it at no extra cost. It is possible that your used example has not had this modification, however. S2s pop up on eBay every so often with non-functional screens. On a physical level the camera was made of slightly cheap-feeling hard plastics, but it seemed tough enough. The F80 shutter unit was designed in the days of 36-exposure rolls of film, and presumably isn't as durable as modern digital SLR shutter units; but then again I haven't seen droves of people complaining about broken shutters in their D100, which also used an F80 body and shutter. Ultimately if a used S2 breaks it is probably going to stay broke. No-one fixes them any more, and it would be cheaper to buy another unit.
What else? Although the D100 had a portrait grip, the F2 didn't, although from a distance it
looked as if it had one. It's possible to modify the F80's MB-16 portrait grip to fit the S2, although the vertical shutter release doesn't work. It will however power the camera. In common with most non-pro bodies, the F80 doesn't meter with Nikon's pre-autofocus lenses. It can't use pre-AI lenses, either - they apparently snap something off the lens mount, so I can't try out my
Vivitar 200mm f/3 on my S3, sadly. The S2 has an odd, irritating post-capture histogram display, whereby you can either automatically save the image to the card, or display the histogram, but not both. Camera compatible with SB-24, popular Strobist choice. ISO 800 fair in good light, ISO 1600 not too good. Banding noise. Responsiveness fine. Dog carcass in alley this morning. Tyre tread on burst stomach.
Must investigate further.
The S2 also won a DIMA award. I don't know whether a DIMA award is prestigious, or not; it seems that
everybody has won a DIMA award. Still, the S2 was launched at a time when the digital SLR marketplace was in a state of rapid change. When the camera was announced there were no entry-level digital SLRs, for the same reason that there are no entry-level jet planes. I mean, yes, there
are entry-level jet planes, but they still cost millions of dollars. You have to be extremely committed or wealthy or both in order to justify the expense of a jet plane, and so it was with a digital SLR back in 2002. At launch the S2 was roughly $500 more expensive than the contemporary D100, but they were still both very expensive and outside the reach of all but the most committed photographer. However, by the end of 2003 the mid-range market was rapidly coming down in price, and Pentax, Minolta, Olympus and so forth were on the verge of launching their own SLRs which were bound to tempt photographers who had no investment in Nikon lenses. For photographers who
did have a bunch of Nikon lenses, there was the D100, which was better-built than the S2 and technically very similar, superior in some respects, cheaper. And this was going to continue; Nikon and Canon could still charge a tonne of cash for their top professional cameras, because they were the best, but the rest of the market was coming down in price,
fast.
Fuji therefore needed to pull a rabbit out of a hat if they were going to continue their SLR line, and the
Fujifilm FinePix S3 Pro (2004) was that rabbit. Rather than increase the megapixel count or go full-frame, or vastly boost the ISO rating, or all of those things, Fuji instead added a high dynamic range sensor that I have mentioned already. Was it clever enough to dominate the market? No. Didn't even make a dent.
But more about the S3 in the next post. It was essentially an S2 with a built-in portrait grip, a 14-bit imaging pathway (rather than 12-bit), plus a new "Super CCD SR II" sensor that could record several extra stops of highlight range. Not just to fill in grossly-overexposed overcast skies, but to smooth out lightly-burned clouds, white clothes, reflections on shiny skin and shiny metal, snow, sunlight coming in through a window, teeth, etc. It was a clever idea that worked well, although it really required some Photoshop tricks in order to get the most out of it. It worked best when the photographer exposed for the shadows and developed for the highlights, just like film, in stark contrast to the digital convention whereby it is best to expose for the highlights at all costs.
Here's another example. On the left is the original exposure - shot so that the histogram just touched the left of the bounding box, with the right side of the histogram flying off into space - and on the right the same exposure processed with software, using a graduated filter effect, with the contrast boosted:
It's cartoony and unsubtle but it was the work of twenty seconds, and you get the idea. The camera was apparently aimed at the wedding and portrait photography market, but a combination of slow write speeds, tiny buffer, and a general sense that it should have had a tougher body and more resolution led to some unenthusiastic reviews. The extra highlight range was an esoteric feature that was hard to explain in simple terms - even some of the supposedly professional reviewers didn't understand it properly, cough
Luminous Landscape cough - and of only marginal use in most shooting situations.
With the extra dynamic range turned off, the S3 was barely more advanced than the S2. It even used the same F80 body, but with a built-in portrait grip. The S3 apparently sold quite poorly, even though Nikon's own-brand equivalent - the D100, still - was getting long in the tooth. At a launch price of $2,499, which was the same as the S2 Pro of two years previously, there was a sense that Fuji had not kept up with the marketplace. The S3 was a thousand dollars more expensive than Canon's faster, eight-megapixel 20D, which sold well to neutral parties and was even tempting for Nikon shooters.
And the days when
every digital SLR cost a fortune had gone; a whole new generation of photographers who had never before imagined owning a digital SLR were handing over comparatively modest sums for a Canon 300D or Nikon D70. Neither of those cameras were officially aimed at the same market as the S2, but they offered just as much functionality at less than half the price, in bodies that were not really any flimsier. Still, the S3 won a
2005 TIPA Award in the midrange category, presumably beating the 20D (the professional-level award went to the sixteen megapixel, full-frame Canon 1Ds MkII).
In 2006 Fuji launched the
Fujifilm FinePix S3 UVIR, which was an S3 modified to record quite a bit of infrared and a little bit of ultraviolet. It was aimed at the professional forensic market. I don't know what it was with Fuji and niche markets. Perhaps they had some S3 bodies left over, and were unwilling to just junk them, and the head of the digital SLR division was a big fan of
CSI.
There was never an S4 Pro, because four is an unlucky number in Japan and Fuji didn't want to tempt fate. The
Fujifilm FinePix S5 Pro of 2007 was a radical departure from Fuji's previous models. This time Fuji spent money on the body and saved on the sensor, rather than the other way around. Fuji couldn't use a midrange Nikon film chassis any more, because Nikon had stopped making them, and so the company stuck the S3's sensor into a top-quality digital Nikon D200 body. In the process they abandoned the hotkey interface of previous S models in favour of a more straightforward menu-driven arrangement.
The S5 was faster and better-built than the S3, but the lack of progress on the sensor front and the relatively high price did it in. By 2007 six megapixels was becoming old hat. Even twelve megapixels no longer had much of a wow factor. To compound the problem, the S5's output was unusually soft, even compared to other six megapixel cameras. And the S5 was more expensive than the D200, which was faster and had four more megapixels. Back in 2007 people wanted megapixels, the more the better. The finer points of tonality and dynamic range were less important, and so ultimately most photographers opted for a D200 instead. And to be fair, the D200 was by all accounts a very good camera.
I have no idea how many S5s Fuji managed to sell; probably not many. The camera still has a following today. It's a bit like one of those films that flops at the cinema but becomes a fondly-remembered classic once it comes out on home video. The megapixel race seems to have fizzled out - at least, outside the world of medium format digital - and nowadays photographers are more appreciative of the S5's qualities, especially given that they don't have to pay $2,000 for it any more. Fuji also launched a UV/IR version of the S5, called the
Fujilfilm FinePix IS Pro, which is no doubt fantastically rare. Fuji continued to list their remaining stock of S5s up until 2009, at which point the camera was quietly discontinued. Roll over, death.
Overall the company's adventure in the land of digital SLRs started strongly but faded out into anonymity and obscurity. The ground shifted and Fuji either could not keep up, or were not willing to spend enough to do so. They faced the same problems as Kodak, and ended up the same way, chasing the not-quite highest end of the market with products that were often good, but not good enough. The cameras were generally on a par with the competition, at least to start with, but for each killer feature there was a quirk, and some things that turned professionals off entirely.
Fuji also had a knack of publicising aspects of their technology that were understandable but arguable, alongside things that were inarguable but incomprehensible. The Super CCD sensor might have had a slight edge in resolution over its contemporaries, but visibly not twice as much; the fact of the photosites being tilted was true, but so what? The extended dynamic range was most definitively there, but not every picture requires twelve stops of dynamic range. Along the way Fuji's good high-ISO performance, solid Nikon compatibility - no reverse-engineering for Fuji, they paid Nikon good money to use their technology - and the clever interface seemed to be get lost in the noise. The S3's implementation of Live View beat the competition by several months, but passed without comment, and nowadays most people assume that Olympus invented that feature.
As for the Super CCD SR II sensor that appeared in the S3 and S5, it remains unique to those cameras. The company's subsequent EXR range of compacts used a similar system, but with pixel binning rather than an entirely separate set of miniature photosites. Fuji's most recent release, the APS-C fixed-lens compact X100, has an extended dynamic range mode, but this uses a conventional technique whereby the camera underexposes the image and then boosts the shadows. This apparently works quite well - the X100 uses a new sensor that has very low shadow noise - and it could well be that this system has rendered the Super CCD SR arrangement obsolete. In fact the X100 doesn't actually uses a Fuji sensor at all, it uses a conventional unit made by Sony, and as of 2011 it seems that Fuji has given up on Super CCD.
Which is be shame. A full-frame, twenty megapixel, D700-bodied FinePix S6 would have been a thing to behold. Although digital cameras nowadays come in an enormous range of sizes and formats, they mostly use the same sensor technology, and indeed a large proportion use the exact same sensors, a 14.2 APS-C unit from Sony and a 12.1 megapixel 4/3 unit from Panasonic. The only other
third way - fourth way? - was Sigma's Foveon technology, but that seems to be dying a protracted death, periodically revived like the torture subjects in
I Have No Mouth and I Must Scream so that it can be tortured and killed off again. Presumably none of the other camera manufacturers wanted to use Fuji's Super CCD sensor, but it's a shame that someone else didn't borrow the sub-pixel idea.
Nowadays the S1 is available cheaply on eBay but doesn't make a lot of sense as a used buy; the S2 is about £150 or so and is a decent entry-level choice. The S3 and S5 remain unique in their field and command a small premium, because there's nothing quite like them. In the UK, Fuji sold off the last S5s for the bargain-basement price of £499 - which seems sad and wrong - and they have only depreciated by a hundred pounds or so since then, because the S5 was the most highly-developed of the breed and had a stronger body, like top figure skater Katarina Witt.
Next, the S3. Although firstly I need to actually photograph something with it. The weather has been crap. It's summertime in England, the weather is crap.
"You'll find there's so much else to think about, to remember. Our lives are different to anybody else's. That's the exciting thing, because nobody in the universe can do what we're doing."
V1.2
"War", sang Edwin Starr. "What is it good for?" By coincidence the Fuji S3 Pro is pretty good for high-key portraiture, although as a means of winning wars it would be terrible. Still, the sensor doesn't blow out the highlights unless you're a complete klutz, and so you can expose to the right to your heart's content, and claw back anything that lies outside the histogram by using Photoshop's Recovery slider, as per the image at the top of the page, of the lovely Bleeka 
The reflections on the lion's nose are blown out; most cameras would render this as a patch of blank whiteness with some ragged, monochrome edges, but with careful processing the S3 file looks natural, and the shadows are nice and clean, too. Here's another example, of a knight:
Again, the brightest highlights have gone, but the transition into clipping is gradual and pleasing to the eye. That was shot at ISO 1600, at which level the image is grainy but holds up well, with good colours. It was further processed with NeatImage. On a more practical level, here's a shot of a metal microphone standing next to a white computer, which would have been a tricky shot with a conventional camera:
And here's Bleeka again, just because, with the histogram mostly crushed into the top stop, shot like almost everything on this page with a Samyang 85mm f/1.4 at f/1.4:
So, the S3 has no problem with women, lions and knights. Put like that it would seem to be the ultimate camera, but there are problems. Mostly little ones, but lots of them. The biggest one is speed. It's not a fast camera, at least not if you have the wide dynamic range mode turned on (and if you don't, the S3 becomes a run-of-the-mill six megapixel SLR, no better or worse than any other). The problem stems mostly from the huge raw files, which are 25mb each, not much smaller than the files from my 5D MkII. The camera takes ages to write them to the card, and you can't review the shots until the very last byte of the last image has finishing writing. If you're photographing a person and you want to show them the results you'll spend a lot of time waiting for the little red light on the back of the camera to stop flickering.
In common with Fuji's other SuperCCD cameras, the imaging pathway produced a double-sized, twelve megapixel output file, with apparently slightly more resolution than a six megapixel file. In practice the advantage was undetectable and buried amongst the other variables involved in converting a raw file into visible output. Compared to a twelve megapixel camera (such as the popular Canon 5D), the Fuji S3's files were obviously less detailed. Here's a little example. First, the lovely Helen Diaz, straight from the camera, shot with a Samyang 85mm at f/5.6, at which aperture the lens should be more than a match for the sensor:
And here's a mouseover. I used Adobe Camera Raw to render the file at both resolutions, and then sized the six megapixel file up to match the twelve megapixel file. Move your mouse over it to see the twelve megapixel file:
I surmise that some photographers bought the camera, tried it out, tried to live with it, failed to do so, and then got rid of it as quickly as they could when the Nikon D200 came out. Fuji offered a buffer upgrade as a paid-for modification, but this would have probably made things worse, stretching the post-capture review delay into minutes.
I'll tell you about how the S3 does its magic, by the way. The sensor is divided into two six megapixel arrays. One group of pixels are of conventional digital SLR size, and record the image in the usual way. With the dynamic range enhancement turned off, the camera only uses these pixels, and as a consequence the file sizes are more manageable.
Buried between the big sensors is a second array of smaller pixels which are much less sensitive to light. They record the same image, but roughly four stops darker. For in-camera JPGs the S3 uses this information to fill in the highlights that the larger sensors have blown out, and most raw developers allow you to do the same. DCRaw, the top-notch free raw developing tool, can even split the S3's raw files apart, and this is what the component images look like:
I should, er, probably have picked an image with blown-out highlights. Here's a better example, normal shot at the top, sub-pixel exposure at the bottom:
Shot on a rare sunny day in England during the middle of summer. The S3 doesn't meter with pre-autofocus, AI-type Nikon lenses, and that includes the Samyang 85mm f/1.4 that I used for most of the shots on this page. As G-SRUM flew by I focused correctly, but got the exposure very wrong. Here's a 100% crop of the sub-pixel image:
It's relatively noisy - the noise resembles film grain, oddly enough - because the pixels are relatively small. This isn't a huge problem, though, because they're only used to recover highlight tones rather than useful detail, although in practice and with a bit of noise reduction this image could be turned into a technically correct albeit very boring photograph. Unfortunately, although the two sets of pixels are slightly offset, they can't be layered so as to increase the camera's resolution. At least, I haven't found a way of doing so. They used the same microlenses, and so were exposed to the same packets of light.
What about the other, smaller problems? It's irritating beyond belief having to switch the ISO value from the mode dial. There's no auto ISO feature, so you have to leave A or M or P or what-have-you, turn the dial to ISO, rotate the appropriate control wheel, and then turn the dial back to A or M etc. Which is a bother. The camera's automatic white balance really loves the colour green. In A, the meter seems to want to overexpose by half a stop. Perhaps Fuji tweaked the meter in order to take advantage of the camera's highlight range, although having said that the meter doesn't read any differently when the extra dynamic range is turned off. I'm unwilling to make too much of this, because it could be the lens (my only autofocus Nikon lens is an old, full-frame, screw-driven Tokina 20-35mm).
The S3 has a pair of card slots, one for Compact Flash and a second for xD cards, which you swap between by diving into the menus. Unlike some old cameras the S3 will work with cards larger than 2gb. You can write to one or the other, but not both at the same time, and you can't copy between them or do anything clever like that. Nowadays the xD format is defunct, although the cards are still on sale. The largest you can buy is 2gb, which will set you back about a tenner. 2gb is enough for 81 of the S3's full-fat raw files. xD cards are about twice as expensive per gigabyte as Compact Flash - this was one of the reasons the format died off - and so there's no point buying one unless you just want to fill up the slot. Perhaps you have an obsessive-compulsive tendency that compels you to fill holes. Obsessively. That obsessively compels you to fill holes. Really, though, when you're a man, everything is a hole.
You'd have to be careful to select the right card, though. Get it wrong and you'd be in trouble. Do you remember that bit in The Good, The Bad, and the Ugly, when thingy puts on the Confederate uniform in order to blend in with a column of approaching soldiers, but it turns out that they were actually Union soldiers whose outfits had been covered in dust? That bit always impressed me. It's cartoonish, but not so cartoonish that it's silly; it feels just ridiculous enough to have actually happened, and it fits the film's audacious, giddy tone. I've always liked Spaghetti Westerns, because they were loud and disrespectful and aggressive. Like punk. They exhibited an active lack of respect for their genre, their field, the world as a whole. It wasn't a passive, uncaring lack of respect; it was an obnoxious, extroverted contempt.
The extroverted contempt of the oppressed, confronted by the arthritic fists and uncontained drool of the senile, elderly oppressor. John Ford had an emotional connection with the old west, and with the Western; in contrast, the Italian film industry didn't give a flip about 19th Century America or cowboys or indeed films at all. Only fools, and their money. And yet, paradoxically, it produced some of the most iconic films of all, and some of the most iconic Westerns, at a time when the genre was dying hard in its native land. The big bang of Fistful of Dollars and its heirs reinvigorated what remained of the American Western, energising one last push that gave us The Wild Bunch, Little Big Man, McCabe & Mrs Miller - a revision of revisionism - and High Plains Drifter, The Outlaw Josey Wales, The Long Riders.
When the Western died for the second time - Heaven's Gate probably did it in, but the genre was running out of steam before that - there was a sense that it was a good death. No unfinished business. No more to be said. Young Guns attempted to mash the Western with the mid-80s action film and the Brat Pack, but fizzled out. Unforgiven was another revisionist take on a revisionist take, a superb film, but a throwback to a bygone age. And yet there is so much from ages gone by that still excites the mind, the liver, the loins. No shame in that.
Fonda dies, too. There was no good in him. He was a throwback to a bygone age, although this isn't set up very well; he seems a lot more urbane than the grizzled heroes of Leone's earlier Westerns. Ultimately Once is a superb ten-minute short film followed by a good five minutes of set-up followed by an hour and ten minutes of nothing and then an unspectacular build up to an unspectacular gunfight, with a central mystery that isn't as poignant or macabre as it should be, and a major action climax that actually happens off-screen.
It's striking how much George Lucas borrowed from Once for Return of the Jedi. The relationship between the two chief villains - the decrepit mastermind and his brutal thug - is similar, but more openly antagonistic; the basic thread of the plot, in which a group of heroes assemble in order to help a pretty, feisty lady from the baddies, and one of the heroes has a particular bone to pick with one of the baddies... and they both have the in the title. The assault on the Death Star is basically the same as the assault on Morton's train, except that (a) it has spaceships instead of horses, and (b) in the immortal words of Mark Prindle, I decided not to come up with an ending.
