FujiFilm FinePix S3 Pro
A First Look
This isn't a full review of the new Fuji S3 Pro. It isn't even a partial review. It's the result of my spending a few days shooting with a full production camera just days after the S3 started shipping in the U.S. and Canada in mid-December, 2004.
Why bother then? Because I, as well as many who follow development in digital technology, have been fascinated to see what Fuji would deliver with their new Super CCD SR II dual sensor design. (More on this soon). This potentially makes the S3 more than just another new DSLR, and therefore of considerable interest.
But, inveterate camera tester that I am, I can't resist looking as well at what kind of camera Fuji has chosen to put this sensor into. So, here then is a brief hands-on look at the FujiFilm FinePix S3 Pro.
The S3, as I'll call simply it from now on, is based on the S2, which has been Fuji's DSLR offering for the past couple of years. I reviewed it on this site and was highly complimentary about the camera's image quality, while being a bit less than enthusiastic about its build quality and handling.
The S3 was announced and first shown at the PMA show in February of 2004. So, about 10 months have passed between introduction and first shipment. The reason for the delay appears to have been a redesign by Fuji of the Super CCD SR II sensor. The original design had two photosites; one regular one for normal exposure (called an S-Pixel), and a second one for highlights, called an R-Pixel, located in the same unique Fuji-design hexagonal shape.
The new SR II design, finally incorporated in the S3 camera, has the R-Pixels physically separated from the S-Pixels. There are apparently two reasons for this. One, is that it allows the S-Pixels to be larger (as can be seen in the diagram above), and also it allows for separate micro-lenses for both sets of pixels, providing improved light concentrating capability. Larger pixels, of course, translate into a superior signal-to-noise-ratio.
The net of all of this is to create a sensor chip with greater dynamic range. Fuji claims as much as two stops greater. Does it deliver? In a word – yes, sort of, but with some caveats.
But first, let's look at the S3 camera itself.
Though superficially the S3's design is much like that of the S2, it's a substantially different body both inside and out. There's a lot less of a cobbled-together Frankenstein look about it. Material quality has been improved with the use of higher quality plastics, and a more tactile and hand-friendly rubberized grip and finish.
While the S2 was really a Nikon N80 chassis (film spool compartment and all), with Fuji electronics bolted on, the S3 is a fresh digital design. For example, there is now a vertical release on the lower right of the body, and the silliness of two separate sets of batteries powering different functions has been done away with. The camera uses 4 rechargeable AA batteries (provided) to power all functions, and Fuji claims (and my tests show) very good battery life. Not as long as some dedicated Lithium Ion batteries in competitive models, but quite respectable nonetheless, and with other advantages, as we'll soon see.
I won't do a detailed button-by-knob review of the S3, since it's actually fairly similar to the S2 Pro. Though the body is new, and no longer based on the Nikon N80 film body, it still is almost completely built from components from the Nikon parts bin, and it shows.
ISO Knob – Still an Issue
One of the areas that I continue to be less than pleased with is the placement of the ISO control on the top Function knob. This makes changing ISO quite awkward, requiring two hands. And, more importantly, it means that the camera can be left in the ISO setting mode accidentally, and then not be able to shoot until it's moved to one of the normal shooting modes. This is utter silliness in a contemporary digital SLR design.
Otherwise the control layout is in the standard Nikon idiom, and anyone coming from virtually any Nikon camera will not find it at all unfamiliar. (Needless to say the Fuji S3 takes Nikon mount lenses).
Speaking of ISO, the S3 turns in very good performance at high ISO settings. Like most current DSLRs, ISO 800 is now a quite usable speed, and even 1600 can produce decent results, though with the use of some noise reduction preferred.
Fuji S3 Pro with Tamron SP 28-75mm f/2.8 @ ISO 400
The S3 takes both CompactFlash cards (including Type 2 and Microdrives), and also Xd cards. It can shoot to one or the other, but not both simultaneously, and you can't copy between cards in camera.
The S3 also features both Firewire and USB 2 connectors, for file transfers as well as tethered shooting.
There are two monochrome LCD screens as well as a colour LCD for image review and some menu settings. It may be simply a matter of time till one gets used to which settings and controls are found where, but I'm afraid that I found this to be one screen too many. I'm sure though that owners will become familiar with what's where soon enough.
As mentioned, the S3 now just uses 4 rechargeable (or disposable) double AA batteries. As a photographer who frequently works in remote locales, I like this, because in a pinch I can always buy AA batteries, almost anywhere – including the most remote places on earth. AA batteries are ubiquitous.
The provided Fuji Ni-MH batteries are 2300 mA, but there are now even longer lived ones available from third parties. The battery tray is available as an relatively inexpensive accessory, and so additional ones can be carried in ones pocket and a fresh set inserted in moments when needed.
I have to admit that I am annoyed at the camera industry for foisting off on consumers a never-ending succession of almost identical L-Ion batteries that are obviously out of the same factory, but which are unique in their fitting to a particular brand of camera. Shame on them. And Kudos to Fuji for bucking the trend.
Otherwise, the specs tell much of the story, and I won't repeat them here. You can get all the details from the Fuji web site, or visit your local dealer to handle a camera and obtain a brochure (which can be downloaded as well).
I don't have every competitive DSLR on had to do a comparison, but with the Canon 20D as my reduced frame benchmark I was a bit surprised at the S3's viewfinder. It is quite a bit smaller and appears further away than that of its Canon competitor at least. It's therefore worth a visit to a dealer, and to having a look though the viewfinder of the various models that you may be considering, to see how the viewfinders stack up.
Overall shooting speed and responsiveness with the S3 is just average. Simply not as fast as some of the latest generation of cameras. Part of the overall feel of the camera that causes this assessment is that the camera can't multitask. By this I mean that while files are being written to disk it can't do anything else. Since even some digicams can now do this (the Minolta A2 comes to mind) it simply means that Fuji hasn't endowed the S3 with enough processing power. At this price point that's no longer acceptable.
I'm not going to get into the whole issue of Fuji hexagonal photo sites. The S3 is a six megapixel camera, but because of the shape of its pixels Fuji extrapolates the data into larger files. Combine this with the extra R-Pixels of the S3 and you have a camera that is either a 6 Megapixel camera or a 12 Megapixel Camera, and that produces various file sizes on the card, and also different sized ones on hard disk of the computer depending on how the camera is set and which Raw software is used. JPGS also have a great many different settings as well. This isn't really a criticism, but it is confusing.
On a subjective basis though I would judge the S3 as being the equivalent of a 10-11MP camera when it comes to resolution. This only goes to show that Megapixels are no longer a measure of a camera's true capabilities.
Unique among DSLRs, Fuji has implemented a live video preview. Frankly, it's of limited usefulness, since it's in B&W only, is only accessible from a menu (not a shooting setting), and last for only 30 seconds. But, it's a portent for the future, and some user's may find it of use.
The rear colour LCD is very nice, large and bright, and competitive with anything else out there.
Update: I originally stated on this page that the camera was not able to show a histogram on instant review. I was in error, and regret any inconvenience that this may have caused.
Multiple Exposure Mode
To my knowledge Fuji is the only DSLR maker that allows for multiple exposures. These can be done manually, with the same exposure or with bracketed exposures. I wasn't able to do more than experiment briefly with this features, but many photographers may find this to be of value and fun to experiment with.
Raw Software – Oh Oh!
There's good news, and there's bad news. The good news is that Fuji's Hyper-Utility Software (HS-V2 Vr. 3.0), which used to be sold separately for about $149, is now included with the S3. The bad news is that compared to third party offerings like Adobe Camera Raw and Phase One's Capture One Pro, it isn't very good when it comes functionality.
This puts Fuji in the same camp as most Japanese camera makers. For whatever reason, they simply don't produce very good computer software. Never have.
I tested the HyperUtility software on a Mac – a Powerbook G4 with 2GB of RAM. My first concern was when I installed the software and noted that 2,289 files were installed. Now, I have no desire to explore the directories installed, but doesn't this seem to be a little excessive?
The problem isn't that the software is slow when doing a Raw conversion. It's no slower than many manufacturer's offerings, and not that much slower than Capture One. The real problem is in usability. It's a scattered program, without a unified interface. It almost appears to be created from glued together components written by different groups. It also makes the various steps of fine tuning an image tedious and time consuming, as it doesn't appear to be working with a proxy file the way CR and C1 do, but directly on the Raw file itself. This means that with every adjustment the file has to be reprocessed so that an up-to-date preview is available. Can you spell s-l-o-w.
What do I know about software design? Well, I was head of software development for two major public software companies over a ten year period, so I have some small experience with user interface. Bottom line – Fuji's Raw software just isn't a program that the pro user can take seriously.
But, there's an alternative. The new version of Phase One's Capture One, V2.6 supports the Fuji S3. Frankly, the support isn't refined yet, and there are some bugs (and as this is written on mid-December, 2004) only a Mac version is shipping). But my tests show that it does a decent job on S3 files. (See below for more on C1 and the S3). It's also safe to assume that the S3 will be supported in an upcoming version of Camera Raw.
The Macbeth colour chart seen below was photographed under tungsten halogen lamps at 3200 deg. The gray point was set in Photoshop. Otherwise no changes have been made to the file. Comparisons are based on viewing the chart under a controlled light source. Please go by what I write, not by what you see, since web conversion, your monitor set-up, and a dozen other variable make it impossible to know whether what you're seeing is the same as what I see. Evaluation of the computer file was done on a Sony Artisan reference monitor.
The Fuji S3 turned in a less than ideal performance when it comes to colour accuracy. This is surprising, because over the past six months or so almost every camera that I've tested has produced nearly perfect colour reproduction results. I thought that the manufacturers had this nailed.
The S3's Orange and Orange/Yellow reproduction (second row, far left and far right) are undersaturated, as is Yellow. Greens as well seem to be somewhat muted. Unless my sample camera proves to have been off spec, my suggestion to professional users is that they perform their own tests, and if needs be create a custom profile for when critical colour work is being done.
The Fuji S3's claim to fame is its enhanced dynamic range through the use of its proprietary Super CCD SR II sensor, with two separate photo-sites – one to record the normal exposure range, and the other just the highlights.
The camera can be set to record just using the S (normal) pixels, or also with the R (highlight) pixels as well. These are then combined in the Raw file, and you have the choice within the Fuji Hyper Utility II Raw conversion software, or via in-camera settings if generating JPG files, to blend these either automatically, or in varying degrees.
Now, these extra highlight sensors don't come without a price. The size of the Raw file is doubled, from some 13MB when just the normal S sensors are used, to 25.5MB when they are integrated. Just by way of comparison, an 8MP Canon 20D's Raw files are about 8 MB in size on disk, while those form the 16MP Canon 1Ds MKII are about 18 Megabytes.
So what we see is that because of the characteristics of Fuji's S3 hexagonal sensors, combined with the secondary R pixels, the S3, which is nominally a 6MP camera when it comes to resolution, generates a Raw file that is larger than that from a 16MP camera. In fact they are even larger than the typically 24MB sized Raw files from the Phase One 22 Megapixel back.
What's going on here? Well, part of the story is that both Canon and Phase One losslessly compress their Raw files, achieving roughly a 3:1 compression ratio (depending on the subject matter), while it appears that Fuji does not do any lossless variable compression on their files.
The bottom line is that a 1GB card doesn't hold that many shots; about 40. This needs to be factored into your cost equation when shooting Raw. Either a handful of 4GB cards, or a device like the Epson P2000 or a laptop are going to be necessary when shooting extensively on location with this camera in Raw S+R mode.
But. we've digressed. What about Fuji's claim to more dynamic range?
This is very hard to evaluate. As I've written elsewhere, there is no standardized test for DR. It is highly subjective, and depends very much on ones tolerance for noise, as well as ones ability to discriminate detail in shadow areas.
In comparing shots taken with and without the R pixels active, there is no question that I can see an improvement in DR, and depending on overall exposure, whether it's extended into the highlights or the shadows. But, the question which I was curious to know is how it would compare to a competitive camera from another manufacturer; one that doesn't use Fuji's technology.
I conducted two such comparisons. The first against a Canon 20D, and the second against a Canon 1Ds Mark II. The reason for these two particular cameras is because I had them at hand, and they are representative in the first case of another contemporary mid-priced camera, and in the second to what is arguably the state-of-the-art in DSLRs at the moment.
Vs. Canon 20D
These shots were taken seconds apart under a uniform overcast sky, using each camera's automatic metering. They are within a fraction of a stop of each other, which I suppose says something about factory calibration tolerances at both Fuji and Canon. Both were processed with C1. (I also processed the Fuji file with Fuji's Ultra software, and could hardly see any difference when it comes to dynamic range. I did a comparison including Fuji's software below, so you can see how small the difference is between them when it comes to DR.)
What we see here is that with the snow on top of the fence post normalized to a reading of 235, the visible and measurable difference between the two cameras in the shadow areas isn't all that great. On the diagonal fence board just beneath it, the green channel reading is 45 on the Fuji file and 35 on the Canon file. This is a real, and visible DR improvement, but in my opinion not a terribly significant difference.
Fuji S3 Pro Histogram
1/180sec @ f/6.7. ISO 100
Canon 20D Histogram
1/200sec @ f/6.3. ISO 100
Above are the original histograms of both raw files. Draw what conclusion you will from them.
Vs. Canon 1Ds Mk II
The test below shows three samples rather than two. The reason for this is that I wanted to see how Capture One 3.6 would handle the Fuji's files vs. Fuji's own Hyper software. What I wasn't prepared for is that C1 produces files that are half the size of those from the Fuji software. Does this mean that they don't use the R pixels? No, I don't think so, because otherwise while the file would be smaller the output resolution would be the same. Instead, what I think is happening is that C1 isn't processing the hexagonal pixels in the same way that Fuji software does, which performs interpolation and thus larger files. Whether this interpolation actually produces higher resolution is a debate that I don't want to get into. It's been raging for several years already with no common conclusion.
This shed has corrugated aluminum siding that is extremely reflective. I exposed both camera so as to just avoid clipping, using the rear LCD histograms. In the Raw software each was reduced so that the highlights were no higher than 250. The shadows were allowed to fall where they may, and no other processing was done to the files, either in the Raw software or in Photoshop. Photoshop was eventually used to convert the files to sRGB for use on the web, and they were reduced from 16 bit to 8 bit mode and resized to screen resolution at 100%.
Fuji S3 processed with C1 Pro 3.6
Fuji S3 Processed with
Fuji Hyper Utility II
Canon 1Ds MKII processed with
C1 Pro 3.6
What you see above is what you can't see. That is, shadow detail.
On my Sony Artisan calibration monitor I differentiate shadow levels beginning at 3, and can see individual level differences in single unit increments. Most monitors can't do this. So, please don't go by what you see above. Trust my description. Unless you're looking at the original files on a monitor with equivalent capabilities you won't be able to tell much at this level of subtlety.
And subtle it is. On the C1 processed file I can see the most shadow detail. There's ever so slightly less shadow on the Fuji Hyper software file. A quibble of a difference rather than anything significant. The Canon 1Ds MKII shows less shadow detail; noticeable yes, but not a difference worth getting excited about. Don't judge resolution or accutance. No USM has been applied, and as we know different cameras need differing amounts.
I need to add two more observations. Firstly, Phase One software brings Fuji files in with very strange looking colour balance. It apparently isn't reading the meta-tagged White Balance correctly. Once you click on a gray point the colour snaps into place and is very good indeed. After speaking to Phase about it I expect that this will be addressed in an upcoming revision.
The Fuji Hyper software got the white balance right, but then when the gray point selector was used on the upper part of the shed, (as it was for all three images and conversions) the file came in with pink highlights. Beats me.
What is my conclusion then about the S3's dynamic range capabilities? Yes, the camera does indeed appear to be able to capture somewhat more highlight detail when the exposure is biased toward capturing good mid-tone and shadow detail. Or, conversely, a bit more shadow detail can be seen if the highlights are kept at the same level.
My DR Conclusion
My problem is that I'm a photographer, not a technologist. But I have been an equipment reviewer on the web for some 7 years, and in magazines for some 30 years, so I have a bit of practice at the game. For me what things boil down to, is, can you see a difference? Not, can you measure a difference – but can you see it in real world situations? Sure, the measurements are important, but ultimately it's what appears on a print that counts.
There is no question that Fuji's dual sensor technology produces a slightly / somewhat (choose your preferred adjective) bit more dynamic range. But, when it comes to working with the files in Photoshop I find that even minor use of levels and curves, and especially exposure compensation in the Raw converter, tends to dilute the distinctions.
I know that I'm going to roast in hell for these conclusions (or at least on some Fuji oriented discussion boards), but the reality is that I simply don't see a huge advantage from the Fuji over competitive cameras in this area. Maybe my test methodology is flawed. But I know that my eyes aren't, and no matter whether I process the Fuji's files in Fuji's Ultra software (which I'm not that familiar with) or in Phase One's Capture One Pro 3.6 software (which I'm very familiar with) I get very similar results. The files are very nice, and have good dynamic range, but are simply not that much better than what I see from either the Canon 20D or the Canon 1Ds Mark II to get all that excited about.
It's when one shoots in JPG mode with the R pixels turned on that the expanded dynamic range benefit of the S3 is most clearly seen. It's just not that significant when shooting in Raw mode, which if the highest image quality is the photographer's goal, will be what's used.
I didn't do any rigorous resolution testing. But, what I have seen from the field shooting that I did is that the S3 is capable of producing some astonishingly high resolution images, and this isn't just a result of their hexagonal interpolation. I used several lenses, and in every case was more than impressed with image clarity and sharpness.
As I review what I've written, I see that one could draw from it an overall negative conclusion about the Fuji S3. That certainly wasn't my intent. I always was impressed with the image quality of the S2 Pro. But, I have to say that the extra year that Fuji took to release the S3 because of the chip redesign appears to have somewhat blindsided them in the competitive marketplace.
List prices for mid-range DSLR in early 2005 are around U.S. $1,500 from Nikon, Minolta and Canon. With a MSRP of a thousand dollars higher than this for the S3, one has to ask what it is that one is paying a premium for. That's hard to see. The shooting speed isn't the fastest, the size is larger than competitive bodies, the viewfinder isn't as large or bright as some, and the colour rendition isn't as accurate as that from several competitors.
Yes, there is a measurable benefit to the dual sensor technology that Fuji has hung its hat on, but when one sweeps away the marketing hype and technical measurements what one is left with is an advantage that isn't all that compelling, especially in the light of the camera's inability to best the competition in other areas, and with concerns for the premium pricing.
It should be noted though that this slim margin is only the case when dealing with Raw files. If one is going to be shooting in JPG mode and have the R Pixels turned on, then the S3 will definitely produce files which have a noticeable dynamic range increase over the competition. It's just that with a modest amount of processing in Raw, most contemporary camera are close to being competitive in this one area.
My feeling is that the biggest danger in today's marketplace is to focus on just one competitive attribute at the expense of all-around performance, and this I'm afraid is what Fuji has done. During the year that Fuji spent redesigning the S3's Super CCD SR II dual sensor, the world moved on. Nikon introduced the D70, offering tremendous value and very fine image quality, and these are selling now for well under $1,000. It would therefore appear that at $2,500 the Fuji S3 is going to have a tough row to hoe, except among Fuji's most loyal fans.
The bottom line is that the Fuji S3 produces top quality images, of very high resolution, and with extended dynamic range (but only of real significance when shooting in JPG mode). But, this is achieved at a significant cost premium over Nikon's offerings, let alone some non-Nikon lens mount alternatives. It's therefore going to be hard for potential purchasers to know whether the substantial premium that one has to pay for the Fuji S3 is ultimately worthwhile.
Copyright 2004, Michael H. Reichmann
As usual, when a review like this appears there are those who question the results. In this case there have been a few people who have written to say that the error in this test is that I underexposed the shots of the barn, and therefore the S3's "R" pixels didn't have a chance to do their work.
No, that's not the case, and that's not the way the digital photography works. When we talk about the ability to capture more highlight detail (which the S3 does), it's not because the chip can record data above 255 (assuming, for the sake of clarity 8 bit mode). The way this works is that if you expose so that the brightest object in the scene is just below 255, then the extra dynamic range will be found in the shadows. In other words, everyone bumps their heads up against the same ceiling. The advantage is in how far away the floor is.
That's what I did in the shots with the barn. I exposed both cameras so that the extremely bright corrugated metal was just below burn out (according to the camera's histograms). And, I also bracketed the shots, so that I could judge them afterwards to choose the best one, since I didn't want to be fooled by an inaccurate in-camera histogram.
I've seen this mislead people in other situations. For example, a high key photograph of a bride in a wedding dress, if exposed properly, will appear to have tremendous highlight differentiation. If there are no deep shadow areas, of course this can be done. It's in the shadows that we see ultimately the extra dynamic range, not the highlights, though it's the R pixels which capture extra highlight data that make this possible.
Thom Hogan, in my opinion the finest Nikon reviewer out there, his published his review of the Fuji S3 Pro. I was very eager to read Thom's impressions, because we had some correspondence when my mini-review appeared, and Thom had expressed concern that my DR methodology and results were suspect.
Well, Thom was right. My methodology for testing DR was nowhere near as rigorous as his. But, in the end, it seems to me that we came to much the same conclusions. The S3 Pro produces a slight increase in dynamic range (though you have to work for it). We also agree on the slowness of the camera, and on that fact that the resolution is very high. We disagree on the colour accuracy (I rated it much higher than Thom did), but then this could simply be a sample variation.
If you're interested in the Fuji S3 Pro, either as a potential purchaser, or simply to know what's going on in the current world of DSLRs, you can't do better than Thom's S3 Pro review.