A lot is riding on the new sensor. How far will it take you?. | Add a comment or send Thom feedback on this article. This is a long review. I'm sure there are rough spots and things that need fixing. Bear with me... | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Disclosure (well, okay, technically a non-disclosure): Unlike the S2 Pro, my S3 Pro body was purchased through B&H Photo with my own dollars and I have no financial or other interest at the present time with Fujifilm USA. Indeed, I've had no contact with them since the camera appeared. They'll be as surprised about this review as you will be.
The S3 Pro marks the third in a series of Fujifilm DSLRs based upon Nikon bodies. The S1 was based on the Nikon N60, while the S2 Pro and S3 Pro are based on the Nikon N80. The S1 came out in 1999, the S2 Pro appeared in 2002, and the S3 Pro appeared in late 2004, so it seems that Fujifilm is shooting for about two-year product cycles. Overall, the Fujifilm digital bodies have tended to be a bit under specified for a pro user, their apparent target, but their image quality has been quite good. It's the sensor, baby! We'll see about that in a bit, but we've got some other things to get out of the way, first. The S3 Pro is enough like the S2 Pro that it's useful to have an understanding of the older model before we move on to the newer. If you haven't read my S2 Pro review, you might want to take a gander at it before continuing. Once again the questions are these: Is Fujifilm's unique sensor approach better than the others? Can Fujifilm compete with Nikon using an older film body as a basis for the camera? Is Fujifilm still in the digital SLR game? I won't keep you in suspense. The answer to those questions are maybe, maybe, and yes. (Note this is a slight step back from where they were with the S2 Pro, where the second answer was yes.) But there are plenty of other questions to be asked and answered, and this review will try to address them. Plus you'll want my reasoning behind those big three questions, so let's get under way.
Fujifilm's approach to creating a digital SLR from the Nikon N80 body differs a bit from Nikon's design for the D100 or D70 (also loosely based on the N80). Where Nikon has put everything under close scrutiny and essentially left only a few parts from the donor, Fujifilm uses considerably more of the film SLR the S3 Pro is based on. The AF and metering systems are unchanged from a N80, as is the shutter. From the top plate up, everything is exactly the same as N80 (except the viewfinder relief, which is a little deeper on the S3 Pro). S3 Pro Fujifilm did more reconstruction below the waist line than they did on the S2 Pro. While we've got the same number of buttons and controls, the back has been completely redesigned. It's a little cleaner, the door to the card slots is better made (though still cancels a write if opened), and we've got rubber material in key places, which dulls the "plastic feel" the S2 Pro had. Another touch: the LCD cover is a little better in terms of clear viewing of the screen, plus it now covers the second LCD, too. On the downside, the new cover does seem a little prone to popping off. The S3 Pro has grown a vertical grip, too, though you'll be less enthused when I tell you that neither of the Command Dials made it to the second release, let alone the AE-L/AF-L button. The reach to the autofocus sensor direction pad is also a stretch from the vertical release, making it pretty much an inconvenient release. The vertical release does have a lock collar so that you don't accidentally trip that release while carrying the camera. And the rubber grip material has been extended around to the hand position you'd have while using the vertical grip. The lack of vertical grip controls is big miss in my book. The left side of the camera has a couple of new goodies, most notably a 10-pin socket for remote releases (the shutter release is still threaded, though; as I wrote: no changes from the N80 above the belt). You get a somewhat awkward cable organizer with the camera, essentially to keep the AC and IEEE cables from pulling out and getting in the way while shooting in studio-type situations. The same battery tray as the S2 Pro graces the bottom of the left side, and now it's the only battery tray (uses four AA 2300mAh rechargeable NiMH batteries, which are supplied, along with a charger and AC supply). There are plenty of subtle below-the-belt body changes, but remarkably, none of the them other than the 10-pin and vertical grip are going to change the way you use the camera over the S2 Pro. That's mostly good news, as it means that if you've used an S2 Pro you can pick up an S3 Pro and start using it immediately. Indeed, Fujifilm's use of a second LCD with four buttons below it remains, and if you know how to use the S2 Pro, you'll pretty much know how to use the S3 Pro out of the box (more on that in handling). In terms of camera features, not a lot has changed since the S2 Pro. Everything's pretty much just as it was, including half stop exposure adjustments, a choice of six ISO values, 10-zone matrix metering, PC sync socket, 1/4000 maximum shutter speed, and all the other N80-based things you'd expect. One place where a significant change was made is flash. The S3 Pro is now a D-TTL camera, which means significant restrictions for multiple and wireless flash (no TTL), and a pre-flash that's not cancelable. Flash sync is 1/180. So if there aren't a lot of big changes in the basic camera specifications, what did change? The sensor and the digitization routines, basically. The big deal with the S3 Pro is a new SuperCCD SR II sensor. Like the S2 Pro, it has 6.17 primary megapixels that the camera by default interpolates to 12.34mp. What's new is that there is a second, smaller set of photosites positioned smack in the middle of the pattern of the primary sensor (think pitcher's mound in a baseball diamond, where 1st and 3rd base are green primary sensors, 2nd base is red primary, home plate is blue primary). If you look at it closely, the secondary sensor also has a Bayer pattern orientation (pitcher's mound has the same Bayer filtration as home plate, if you're following along with my analogy). So why the extra photosites? Fujifilm uses the secondary photosites to boost dynamic range capabilities of their capture. But it doesn't work like some reviewers seem to be thinking. Basically, the primary photosites do most of the work, and indeed work pretty much the same as in the S2 Pro. The secondary photosites are much more resistant to recording light due to their smaller size, thus they really only gather data about bright points of the scene. Properly integrated, this means that you tend to expose more for the shadows and the highlight data is produced by the secondary photosites. The result is that you don't blow out highlights. Say that again? Take a normal outdoor scene. I often encounter 10 stops of dynamic range (or more) and the usual impact is this: I have to expose for the highlights (i.e. make sure they all get measurable exposure), lest they all blow out to 255,255,255. Since most digital cameras only have about seven stops of dynamic range, at best, that means that the bottom three stops in the scene is all effectively rendered as black, and even the midrange is pushed down towards the shadows. Perhaps a post processing curve can pull out a bit more from the shadows and fix the midrange position, but more often than not that simply increases the noise properties of the image. With the S3 Pro everything flips (though you still have to watch the histogram to make sure you're not blowing out highlights). Expose to keep the shadows above black. This would normally blow out highlights, but we have that second set of photosites that is more light resistant, which means that they retain highlight detail that is lost by the main photosites. The trick is this: how the heck do you integrate the secondary sensor data with the primary? In camera, the S3 Pro can do exactly that with JPEGs, with settings for Auto, Wide1, and Wide2 in a new function called D-RNG (for dynamic range). Fujifilm's documentation and explanation will confuse you (what does "[set] the dynamic range...to 400%" mean?), so let me try: in theory you can get up to two stops more dynamic range (we're going to have an argument over the exact amount when we get to the performance section, below). No other current DSLR has anything like this ability, so, if it works, the S3 Pro gets bragging rights in at least one imaging area. We'll talk about this more later in the review. Another new ability linked into the sensor and digital bits is something Fujifilm calls "film simulation." Ironically, Fujifilm fails to specify which film it's emulating, giving the settings only the generic F1 and F2 names. F1 holds more dynamic range and in theory is targeted at people shooting portraits with print film. F2 is said to be for landscape and nature photography, with more "vibrant reproduction of natural colors [sic]". Sorry, they're not natural any more if they're enhanced, are they? Reading between the lines, F1 is Realia, F2 is Velvia. Or so they want you to believe. The S3 Pro has an interesting preview mode where the mirror flips up and you get a monochrome preview of what the sensor sees (for up to 30 seconds). You can zoom in on this live preview and check focus, but that's about all I can find that it's good for--it can't act as a mirror lockup, as the shutter release is disabled while the preview is active, and the mirror flips back down when the preview is complete. This seems like a miss on Fujifilm's part: if the shutter release is pressed during live preview, the camera should shut the preview and take a picture before returning the mirror. Amazingly, we've run out of features to talk about. Add it all up and Fujifilm is pushing image quality as the key improvement on the S3 Pro. We'll get to that in a moment, but first we have to wade through a couple of other items. | Recommended
Image quality is the high point of this camera, especially for JPEGs, and is the one thing that makes the S3 Pro compelling to some. Table of Contents
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The Basics[Note: this section is mostly copied from my S2 Pro review, with only those things that are different being changed. The two bodies are pretty close in specification.] The Fujifilm S3 Pro has a feature set that sounds almost identical to the D100. Indeed, most of the "camera" features of both are derived from the N80 body. In appearance the S3 Pro closely resembles a bulked up N80. The digital controls look and feel just like the S1 and S2 controls, a benefit to those upgrading from those bodies. The autofocus system is decently fast, and features five sensors (CAM 900) that can track rapidly moving objects, or direct autofocus to a specific area of the frame. Autofocus detection works from EV �1 to EV 19 (specified at ISO 100). The metering range extends from EV 0 to EV 21, plenty wide for virtually any shooting you might do. Note that the spot metering range is slightly lower, from EV 3 to EV 21. Unless you make a habit of spot metering in unlit situations at night, you're not likely to encounter that limit. Shutter speeds can be controlled in 1/2 stop increments from 30 seconds to 1/4000. All shutter speeds are handled mechanically on the S3 Pro. Indeed, the shutter sounds pretty much like the N80's in action (a very quiet "clunk" sound). Single shot and continuous firing at 2.5 fps is supported, though I have my doubts as to whether the S3 Pro actually achieves 2.5 fps (the best I seem to be able to achieve is around 1.5 fps, and using wide dynamic range seems to slow it further). It's relatively easy to (inadvertently) turn on features that would make motor drive even more leisurely in pace. The S3 Pro manages matrix metering the old-fashioned way, with a 10-segment metering pattern. The matrix metering also incorporates information about the focus point you�re using if you�ve mounted a �D-type� lens. Nikon also lists �subject positioning,� �overall scene brightness,� and �scene contrast� as factors in the matrix metering calculations. In short, it�s hard to second guess the camera as there are so many factors being considered. If simplicity suits you, the spot meter option is a better choice. (Center weight is 60/40 and sees outside the CCD frame size, making it not overly useful.) As befitting a serious camera, Fujifilm [Nikon] keeps the mode selection simple: Program, Shutter Priority, Aperture Priority, and Manual are the full lineup. Program mode is easily adjusted by spinning one of the control dials. There's also an automatic adjustment of the program due to focal length of the lens used; thus there is no �Program High� or other special automatic modes as there are in some other Nikon bodies. Like all Nikon bodies, you can override the programming using the rear command dial, something Nikon calls "Flexible Program." An exposure compensation button gives you a +/- 3-stop range in 1/2-stop steps. A built-in bracketing system allows two or three shots at half-stop values up to 2 EV apart. Exposure (as well as autofocus) can be locked by on-camera controls (though this takes some reading of the custom functions to understand completely). ISO values can be set to 100, 160, 200, 400, 800 or 1600. Flash sync works to 1/180 of second. Flash metering uses five TTL sensors and can be balanced, slow-synced, and synchronized with the rear curtain. The flash system is now D-TTL, which requires an SB-28DX, SB-50DX, SB-80DX, SB-600, or SB-800 Speedlight, and which always performs a preflash. This is a step backward from the S2 Pro, which could use any of the TTL Speedlights and allowed its preflash to be cancelled. In the viewfinder, you'll see 93% vertical, 95% horizontal of the full frame, which means you're not seeing between 50 and 75 pixels worth of information at every edge (more if you set 12-megapixels as the size). Shutter speed, aperture, exposure mode, metering method, focus indicator, exposure compensation, and flash ready are all visible in the viewfinder, even to eyeglass wearers like me. Unfortunately, frame counter and/or buffer remaining indicators, as the D100 has, are not present. The viewfinder is still a bit small and dim compared to some DSLRs (.8x magnification), though it does appear a tiny bit better than the S2 Pro. The camera information still appears quite far from the image area (i.e., there's a large space between the two). This takes a bit of adjustment, and it's impossible for me to pay attention to framing while looking at camera settings (or vice versa), even though they're both visible at the same time, even with my glasses on. On the color LCD on the back of the camera, you'll see 100% of the image. Probably of most use on the color LCD is the ability to see a histogram of any picture you take, allowing you to tinker with exposure to get every last bit of dynamic range out of the sensor. You can also zoom in on your image's pixels to check focus. The color LCD on the back can display a monochrome preview of image before you take the shot (called Live Image in the menus). Frankly, I don't find this to be a plus at all, since you can't take a picture while it's active (it would have been a nice faux mirror-up function). It does allow you zoom in, which is nice for checking focus, I suppose. The regular viewfinder is just fine for determining composition and focus, though the image is a bit smaller and darker than the N80's. You have three choices for review of images. You can set the Image Display function to On or Preview (or do it all manually if you've set Off). Preview mode has been changed significantly and now records the image unless you press the Trash Can button. Note that Preview is disabled if the camera is Continuous shooting mode. In On mode, the image briefly flashes on the display, then disappears while it is being written to the storage card. Compared to any Nikon DSLR, this is annoying, to say the least, especially if your goal is to examine the histogram (Histogram Preview is no longer available, by the way, but Preview remembers whether you selected histogram or highlights display, so Histogram Preview isn't technically necessary). And while we're on the subject of the color LCD, one other niggling detail bothers me: if the camera has gone inactive, you have to reactivate the camera to get the color LCD to display something. Pressing any button should reactivate the camera (Nikon bodies do this). Finally, the backlight source (actually four sources) is very visible on this LCD, making the bottom edge of a photo look a bit brighter than it should. The Fujifilm S3 takes any Nikon F mount lens (well, lenses earlier than the AI manual focus Nikkors damage the mount if you try to put them on the S3, and a few specific lenses won't work on the S3, usually because they have elements that stick into the mirror box and require mirror lock-up). Non-CPU lenses don't allow metering and must be used in Manual exposure mode. When you mount a lens on the S3, the effective focal length is increased by about 1.5x (e.g., a 20mm lens shows roughly the same angle of view as a 30mm lens would on a 35mm body [Note: the actual magnification impact is closer to 1.56x, which is somewhat higher than the Nikon DSLRs, enough so to be visibly apparent with very wide angle lenses.]). Apertures aren't really affected by this change. Because only the central portion of the lens is used, if you've been making any exposure adjustments at maximum aperture to account for light falloff, you should not do that on an S3. The SuperCCD SR II sensor the S3 Pro uses is currently unique to the S3. It consists of two 3024 x 2016 arrays, one with large and and another with small photosites. I wish Fujifilm would supply more information about exactly how they're integrating the two arrays, especially since this is a key feature of the camera. But basically you've got relatively normal large-sized photosites doing the heavy lifting, with a second set of smaller ones helping with expanding dynamic range by holding highlight detail better. The offset between arrays is geometric, so there's a small chance that someone will figure out how to get (almost) real 12mp images out of RAF files at some point, though I'm not sure how that will work in shadow areas (it will obviously work in highlights and maybe down through the mid-range, but the small photosites just don't get enough photons to hold any significant shadow detail). Thus, when Fujifilm says the effective pixels are 12.34 million, that's a bit of a stretch. From an image standpoint, it's 6.17 and the camera interpolates out to 12.34 million. The S3 Pro produces two types of files: JPEG and RAF. The JPEG options work as you'd expect, but you pay a small penalty for using that format: the files are compressed and lose a bit of detail, plus they are converted to 8-bit format, losing some of the S3's tonal range in the process. Fujifilm's JPEG engine tends to produce very sharp-looking images. The RAF format is the only one that retains the full 14-bit data the S3 is capable of acquiring. Indeed, it doesn't do any compression, so the files turn out to be about 25MBs in size when the expanded dynamic range is used (13MB when you use the standard dynamic range, which tells you that the small photosite data is thrown away). The camera comes with Hyper-Utility software, Fujifilm's converter, which still lags behind Nikon Capture in terms of both ability and UI. A few other converters already handle S3 Pro RAFs, most notably C1 and Photoshop CS. Like the S2, the S3's CCD has a wide landing area alongside it, which means you can get a Sensor Swab (or Sensor Brush) completely off the imaging area with ease, unlike the Nikon bodies. Also like the S2 Pro, my S3 Pro's CCD was quite dirty when I received it, requiring an immediate cleaning. In the US, the S3 Pro comes with an initial set of batteries (four 2300mAh NiMH AA batteries, with charger). Battery life is quite dependent upon a number of factors, and can range from a few hundred shots to over a thousand, at least in my observation so far. Note that IBM Microdrives use more power than Compact Flash cards, and thus, exhaust the batteries faster. Overall, battery life seems better on the S3 Pro than on the S2 Pro. The camera weighs in at 28.7 ounces, and that's without a lens or the battery, so it's a lot lighter than the D1 and D2 bodies, a bit heavier than the D100, but significantly heavier than the N80. Overall, it's only two ounces heavier than the S2 Pro, but somehow it feels more substantial than that small change indicates.
HandlingThe S1 was often dubbed a man-made monster, and even called a "Frankencamera" [actually, the Germanic spelling and pronunciation seems more appropriate: Frankenkamera (FRONK en kom uh rah)]. The S2 still had that separation of digital and camera shooting controls and the battery situation also still reflected a digital/camera divide. The S3 finally seems to better integrate the camera functions together, but the digital controls are still mostly done by the FUNC button and row of smaller buttons under a second that the S1 and S2 also had. As it turns out, most of the Frankenkamera objections weren't due to the digital/camera divide, but to the fact that the internally the integration wasn't complete, especially the power management system. The S3 finally seems to rid us of the main integration problems. In my review of the D100 I called it's aesthetics and build quality that of a Dodge RAM pickup. In my review of the S2, I said it was in that same realm. Well, since writing that, the Dodge RAM has gone through an upgrade, and so has the S3 Pro. Fujifilm's paying a bit more attention to small details now (the rubber grip, redesigned contours for better hand holding, position of controls, etc.), and overall this makes the S3 Pro feel like an model-year improvement over the S2 Pro. But in reality, I don't think the S3 Pro is any more reliable or rugged than its predecessor. Under the new rubber grips the body materials are the same, and some of the vulnerable parts (Mode dial, metering dial) are still the same. Compared to the D100 (and even the S2 Pro), the S3's right hand position is more refined and well thought out. You can actually hold the camera and shoot with only your right hand, should the desire strike you (the technique I've memorialized now as the Galen-Rowell-hanging-off-cliff-while-shooting position); the D100 doesn't have that feel to it. The rubber grips and more pronounced thumb grip and notch feel very comfortable to me. Indeed, the first day I shot with the S3 Pro I didn't put a neck strap on it and carried it for several miles using only my right hand. Not a lot of bodies I'd try that with, but it worked just fine and my hand was comfortable the whole time. Most camera controls with multiple settings (flash mode, metering mode, exposure compensation, etc.) are set by holding down the appropriate button and rotating one of the knurled command dials. Unlike the Nikon digital cameras, most of the unique digital controls aren't buried, but appear as a four-button/LCD combo on the back of the camera. Press the FUNC button, then press one of the four buttons (the LCD immediately above each shows the function and current setting) to change a setting. Press the FUNC button again and you get a different set of digital controls. White balance, focus area mode, image size, image quality, color, tone, sharpness, dynamic range, and film simulation are all set this way (with a couple of problematic exceptions I'll get to in a minute). ISO uses the Mode dial and command dials, and this seems pretty natural and straightforward, though not as convenient as a button-based system. Better still, you can "lock" these FUNC settings so they aren't accidentally changed. The four button + LCD interface is also used for other things, such as previewing and browsing functions, while the small LCD itself shows camera settings, date and time when not being used for something else. From a functional standpoint, the four button + LCD interface works well and gives you fast access to a number of useful functions, with two key exceptions: dynamic range and film simulation. The FUNC button allows you to set D-RNG (dynamic range), but only if you've set Wide in the main menu system on the color LCD. Why the D-RNG doesn't have STD, AUTO, W1, and W2 instead of AUTO, W1, and W2 seems like a major faux pas in UI design. Film Simulation isn't quite so bad, but it too interacts with the D-RNG setting in the main LCD menus--you can't set Film Simulation with the FUNC buttons if you've set Standard dynamic range in the Set Up menu. (Film Silmulation also doesn't work with AdobeRGB as the color space.) These things, coupled with the fact that the camera has to be active in order to access either the FUNC settings or the main menu system, will frustrate someone who makes a lot of changes to their settings. I can tell you that it was frustrating the heck out of me as I tried to go back and forth between the dynamic range values for testing purposes. The S3 Pro would be so much better in handling with just some minor tweaks to the UI. In particular, any button press should reactivate the camera. Beyond that, more thought needs to be given to the interaction between the Set Up menu and the FUNC menus. Other minor things abound: while waiting for those huge RAW files to finish writing you can't change to a JPEG image quality, even if there might be room in the buffer to shoot a few. This is one of the disquieting things about the size of the raw files created and the slow write speed (see performance, below). The S1 felt like a very modal camera, the S2 less so, but some aspects of the S3 are feeling slightly more modal again. This is just one example of the many things in the S3 Pro UI that feel Ogden Nash-like ("You Can't Get There From Here..."). The 10-pin, video out, Firewire, USB, and DC in connectors are all on the left side of the camera. The PC Sync is on the front. Other than the video and AC connectors, the rest all are enough out of the way so that you can hold the camera with it connected to something and not feel that the cables are getting in the way of your hands. Fujifilm supplies a small plastic cable organizer that hooks into the neck strap hook that helps keep multiple cables together and from pulling out of their sockets. A nice touch, I suppose, but a part that's easy to lose if it's not always connected. Once again, changing custom settings on the Fujifilm S2 is a breeze. Move the mode dial to CSM, then use the rear command dial to change between CSM numbers and the front dial to change values. Text for the the current CSM you're working on is shown on the color LCD. Getting to a CSM and changing it takes mere seconds, and has none of that repeated button pressing you find on the Nikon digital bodies. Fast and convenient. Flash use still overrides all but Custom White Balance settings, something that some, including me, find problematic. That's because it effectively negates the use of fill flash unless you set a Custom WB, which slows you down more (sensing that slow theme again?) I'm less enthused by the menus, which are modal (different menus when you're playing back images versus not); fortunately the design of the two sets is different, so you won't mistake one from the other; still, I don't get the need for the modality. The playback menu options are identified by somewhat cryptic icons at the bottom of the color LCD. The Set Up menus consist of five not-very-well organized pages of four options each, and some, such as the USB and 1394 Mode menus use icons for settings that'll have you scratching your head. Things you use most often, such as Discharge, should be on the first page instead of things that you set once, such as Image Disp., especially since the first page always comes up (i.e. the menu system doesn't remember where you last visited, another poor UI touch). Overall, there is a disjointedness to setting things on the S3 Pro. ISO and CSM on the Mode dial, most of the image quality stuff via the FUNC button (but not all), more settings on the Set Up menu on the color LCD, and some things apparently in multiple places (there's a Format on the Set Up menu and an Erase Images on the PLAYBACK menu--can you figure out which one you should be using from those words? Pysch! They're the same).
| Did Fujifilm do Enough? Some of you might remember that after the S2 Pro appeared I came up with a list of suggestions for the S3 Pro. You'll find those, along with my conclusion about whether the S3 Pro has them, here. Short answer: no.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
PerformanceWell, here we go. Nothing I've said so far should stir up much in the Fujifilm brethren. But I bet you that something in the next sections will. Let's start with the easiest to measure: shot-to-shot performance. Let's start with the really bad news: set the S3 Pro to shoot raw with wide dynamic range and you'll get three shots off in continuous mode and then be waiting for at least ten seconds for the buffer to clear. [Note: Fujifilm announced an optional doubling of the buffer size in September 2005 in Japan. As I write this, it's not yet available in the US and unclear what the price will be, though I expect it to be about US$199 and require the camera to go to Fujifilm for the update.] Worse still, those three shots are produced at only a clip of slightly higher than one frame per second (Fujifilm claims 1.4 fps, but I can't quite achieve that on my camera). There are three bits of bad news in that last paragraph:
Now, for my shooting, none of those things cause me a great deal of concern. I'm not a shutter release masher when confronted with action (which usually means an animal has wandered into my scene). I often take one shot at a time and evaluate it before taking another. But for someone who shoots events or contemplates shooting action with the S3 Pro, be aware: it is not a particularly snappy camera. What does bother me, even for landscape and scenic shooting, is the slowness of playing back images. Depending upon the image quality settings, it can take over two seconds for an image to appear on the color LCD after pressing the Play button, and scrolling between images can be an exercise in frustration. Moreover, the S3 Pro doesn't remember the last image you were looking at, so you may find yourself scrolling back through them, and you can actually press the direction pad buttons too fast, meaning that the S3 Pro ignores some of your presses. Yuck. These things, taken together with the need for the camera to be active for some of the buttons even to have an impact, rob the S3 Pro of much of the spontaneous performance and access you associate with, say, the D70 or D2h. This represents a step backwards from the S2 Pro, which wasn't a very spontaneous camera itself. Another thing that can be measured relatively easily is ISO noise tendencies. I use Imatest for this these days coupled with visual evaluation of the quality of the noise. I was actually a little surprised by what I found:
Notice something? There's substantially less noise in the S3 Pro at ISO 100 (that's what both these tests were at). The exposures are slightly off, with the S3 Pro's being "brighter", but even when I went back and retested the S3 Pro with what turned out to be about a half stop lower exposure, the results were still that the S3 Pro was significantly lower in noise. (You might wonder why I didn't show those--the S2 Pro test I've got actually needs to be redone because it's underexposed; the S3 Pro test is properly exposed. At some point I'll go back and retest my S2 Pro to get apples versus apples charts into this review, but you've all been asking me "when's the review coming?" and I got tired of answering "soon.") What's most impressive to me is that the blue channel noise, which was often the Achilles' heel of the S2 Pro, seems to be much, much better. In some ways, these are pretty impressive noise figures, so you're probably thinking "what about ISO 1600?" Glad you asked:
Amazingly, the luminance channel noise (Y-noise) is better than my old S2 Pro test at ISO 100! When I readjusted for a lower exposure to match the pixel values, I was still getting values in the red and blue channels that were lower than the S2 Pro, though the luminance channel was now slightly higher for some of the gray patches on the ColorChecker chart. More important, we've pushed the sensor up four stops (100->200->400->800->1600) and not had ruinous results. In general, Std Devs of 2 or less don't produce significant visible noise in images. You can see this just in file sizes alone: the ISO 100 image produced a 4.37MB image and the ISO 1600 image produced a 4.76MB image on identical subjects and at the same settings. This is an unusually low increase in size for that much ISO bump. All isn't perfect with noise, though. When you use the dynamic range enhancements, you'll get a bit more noise:
(Ah, foreshadowing! Note the pixel values. This is with the camera set to ISO 100 and AUTO D-RNG [compared to ISO 100 and Standard D-RNG in the first test]. But those pixels values are changing, even though the exposure didn't). First, note that all the noise values went up slightly while the pixel values went down--that makes noise more visible (though at ISO 100 it's just not really visible at all without drilling down to something like 200% view). At ISO 1600 this becomes somewhat more problematic, as the noise will start to come more into visible range. Overall, noise handling on the S3 Pro is state-of-the-art and an improvement over the already quite good S2 Pro. Since many people use the S3 Pro as a JPEG camera and JPEG artifacts can get produced by and interact with noise in very bad ways, this is very good news. Indeed, I'll just simply say it here: if you only shoot JPEG, I don't know that there's another DSLR available that can outdo the S3 Pro in image quality (assuming, of course, that you like the color rendering; see next sections). White balance was a mixed bag. The S3 Pro consistently missed the mark on Auto White Balance in every setup I tried (even after the supposed firmware updates that addressed the problem). More often than not it was significantly to the high side of the actual color temperature, producing warm results. I think this is why everyone says the Fujifilm bodies do such great automatic white balance--they like warm images. But it's not accurate by any stretch of the imagination. The more severe I made the test, the more the S3 Pro missed by. In my testing alcove, there's usually a mix of diffused ambient, diffused incandescent, diffused flash, and (if I let it in) a bit of extra warmth due to reflections off of off-white walls. In this condition, middle gray was off by 1387K, a considerable margin, and more than I've seen from any other camera I've tested this way. If accurate color is your goal, I strongly suggest that you use Custom White Balance on the S3 Pro. It seems to work well and produce little miss, even in weird situations. Color accuracy, while we're at, is not an S3 Pro strong point. The magenta channel seems to be too pink, and cyans come out a bit green. Greens, purples, and reds all seem to pop, though, which some people like. In Imatest's color analysis, the saturation for STD settings is 115%, which is too high, in my opinion (ORG settings reduce this, as does using any of the D-RNG settings other than Standard). All the blues drift towards green; all the reds drift towards orange; greens are the most accurate. Compared to the Canon 20D, the S3 Pro has the same red drift, but much more blue shift and slightly more green shift. Put another way: in the ColorChecker chart the S3 Pro produces a red channel for most colors that tends to have too high a value, the green channel is relatively close to ideal, and the blue channel value is low. Don't get me wrong, the S3 Pro produces pleasing color, but not necessarily accurate color. Even before we get to the dynamic range stuff, we can say this: portrait and wedding photographers probably like all the things I've described up to this point. Warmth in portraits is good, low noise at high ISO values is good, color doesn't need to be accurate if it's pleasing and pops. For a landscape photographer like myself, none of these things are particularly exciting. Too much warmth can make the red channel saturate in places such as Southern Utah, and the blue color shifts make skies somewhat artificial (too cyan, a trait the S2 Pro also had). A final thing that is relatively concrete to measure is battery performance. Here, there's good news. The supplied NiMH AA batteries power the camera for hours and hundreds of shots, even under the worst case conditions (I just shot 100 shots in three hours outdoors at 10 degrees Fahrenheit, and the battery indicator still shows full). While I'm not a fan of NiMH batteries, which lose about 2 to 3 percent of their charge each day they're unused, a fully charged set in the S3 Pro puts it in the same category as a D70: a full day's shooting probably won't trigger a battery change. (Perhaps when I've had more time shooting with the S3 Pro I'll be able to pinpoint the number of shots better. Fujifilm says an average of 400 frames using Image Review, and I'd say that is actually low based upon my initial use, but it usually takes months of use before I can sort things out and say anything specific. Suffice it to say that I perceive a significant improvement over the S2 Pro, even though Fujifilm's specifications in the manuals would indicate otherwise. This just in: I've done a couple of worst-case tests on the batteries now. Basically, continuous shooting, continuous autofocus, some use of image review, some use of backlighting the LCDs provides slightly over 900 images per charge [and takes over 45 minutes to accomplish due to the slow write speeds--this actually may be the more critical aspect, as the camera was active for the entire 45 minutes and the Nikon bodies have high discharge rates when the camera is active]. Even at that, resting the battery for a long period afterwards allowed taking many more pictures.) Note that since the S3 Pro uses NiMH rechargeable batteries, Fujifilm has also provided a Discharge option in the Set Up menu. Since very few NiMH batteries come conditioned to full capacity from the factory, you should use the Discharge option the first two or three times you use your new batteries. After that, don't use the Discharge option unless you notice that the battery set is losing significant shot capacity (remember, they'll lose 2-3% a day just sitting, so make sure that your assessment is from a freshly charged state). I'm sure many of you are also wondering about resolution. For most subjects, the S2 Pro performed about at the level you'd expect an 8mp camera to, and the S3 Pro is right in the same range. Why 8mp? Well, it has to do with the orientation of the Bayer pattern. Most cameras orient the Bayer pattern in a horizontal/vertical manner, which puts the highest resolution recording capability on the diagonals. Fujifilm's trick with the SuperCCD has been to put the photosite orientation on the diagonals, thus putting the highest resolution recording capability on the horizontal/vertical axis. In the real world, that's important, as most of what we perceive as "resolution" in our pictures is detail in the horizontal axis (vertical less so). Thus, if you take a picture of something with an S3 Pro and a Canon 20D and compare, the Fujifilm body actually seems to resolve things that the Canon doesn't. That's because the S3 Pro is resolving at the level of an 8mp camera and the Canon's default settings (and even highest settings) tend to produce slightly softer sharpening effects. As I'll point out in my 20D review, the 20D produces images that are a little less detailed than you'd expect from 8mp. The S3 Pro produces images that are more detailed than you'd expect from 6mp and about the level you'd expect from an 8mp camera. Thus, the S3 Pro gets an ever-so-slight (and I do mean slight) nod in the resolution department over the 20D. Not enough to worry about, but S3 Pro users will be wanting to brag about everything they can, since there are so many things they won't want to brag about (like price). Exposure and Dynamic RangeIt's time for me to debate myself. Most of the things I wrote about exposure with the S2 Pro apply to the S3 Pro, as well, since none of the basic metering components or algorithms have been changed. What has changed is, according to Fujifilm, dynamic range capability, which interacts with exposure. I guess I should step back for a moment and talk about the interaction issue. You can't talk about exposure accuracy without talking about dynamic range. If you're talking a picture of a scene with 10 stops of dynamic range in it with a film that can capture 10 stops of dynamic range, most people would probably say that the "proper exposure" would be the one that fits Tab A in Slot B (scene's dynamic range in film's dynamic range). Ansel Adams disciples will of course point out that this isn't 100% true, but it's close enough to the reality for most photographers that we'll consider it as a straw man. The problem is that many scenes--especially the outdoor scenics and nature shots I take--often have more dynamic range in them than our film or sensors can capture. State of the art for digital bodies prior to the S3 Pro was about seven stops (possibly higher if you tolerate some noise), which means that in our 10-stop scene we either have to let the top three stops blow out (yuck!) or the bottom three stops fall to black (still yuck), or some variation of the two. The big innovation in the S3 Pro is centered around this issue: the S and R photosite split in the S3 Pro's SuperCCD is designed to provide additional dynamic range. S photosites have one sensitivity to light, R photosites a different (lower) sensitivity. The big issue is how much difference is there between the two and how well are the two integrated together into images. Trying to get my head around that produced a self-debate. Fujifilm doesn't help things by using percentages, as in "400% improvement." 400% improvement of what? Taken literally, that would mean that if the S2 Pro managed 7 stops of range, the S3 Pro would manage 56 stops. No, that can't be what they meant. A diagram on the Japanese Fujifilm site produces the answer: a "normal" exposure is 100%, so 400% represents two stops of dynamic range increase (100->200 = 1 stop, 200->400 = 2 stops). The diagram itself would fail all the Tufte standards for representing data, and percentages are expressed against a known value (technically the way Fujifilm represents 400% is 400% of no exposure), but we'll let that go. Here's where things get fuzzy. Most of us have learned to set exposures by punching our histograms as far "right" as possible without blowing anything out. The thing is, what does the S3 Pro histogram show? There's no question in my mind that the histogram integrates data when the expanded dynamic range is active, but remember that the camera is consolidating two exposures into one, so I think we're seeing the results of the integration, not necessarily the change in dynamic range. That's a subtle difference, but it is an important one, I think. You can see this by running a test with dynamic range off, on, and at different levels--the overall histogram will change if there's enough dynamic range for the camera to start integrating.
Again, Fujifilm's hint seems to imply three stops are possible. In reality, pushing a lot of wide dynamic range scenes at the camera (it conveniently snowed just after I got the camera ;~), I see one to maybe two stops of overall improvement, which puts us within shouting distance of Fujifilm's claim (once you figure out what 400% represents ;~). There certainly is an improvement, but trying to put an exact value on it for any scene has proven to be a very difficult proposition. Like I said, the main photosite values tend to get shifted during the integration, which complicates things. Overall, the expanded dynamic range shots tend to look "flatter" than the non, which is as you would expect. (An aside: the ColorChecker chart fits within the dynamic range of every digital camera I've tested. Thus, you can't simply take the chart and shoot it with both dynamic range on and off to determine what is going on. What happens in that case is that values get pushed downwards and the bottom to top value is compressed. In the case of the ISO tests, the expanded dynamic range had a pixel value range of 175 while the standard dynamic range had a pixel value range of 189.2. So first, you have to find subjects that'll extend outside the standard dynamic range. Then you have to figure out how to reset the exposure.) Let's try tackling this from another angle. One thing I noticed right away was that areas of blowout were reduced dramatically at the same exact exposure when I switched to wide dynamic range (at any of the three settings, actually) in high contrast scenes. But at the other end, I'm not seeing much more or less shadow detail; sometimes I see a bit of compression of shadow detail. This led me to a few new experiments. My conclusion: the S3 Pro joins the D1x and SLR/n as excellent high-key cameras. (If you don't know what high-key style means, it's a type of photography where most, if not all, of the elements are middle gray or brighter; i.e., very little deep shadow. Some proponents go so far as to only shoot against bright backgrounds that often just blow out.) In scenes that would challenge a D70, the S3 Pro does just fine once you find the magic exposure. Which brings me to a new Achilles' heel: getting the maximum benefit of dynamic range (WIDE2 setting with the right scene) is a trial and error production that's hampered by the histogram. Darned if I can tell what's happening at the right edge of that histogram (though I only have half stop settings to work with, so I'm not going to be able to maximize things, anyway). I've found that I rely more no the highlights display than the luminance channel histogram. In extensive testing in snow scenes I'm finding that I can sometimes push a WIDE2 exposure 1.5 stops beyond what I can push a Standard D-RNG exposure. This is real dynamic range improvement that's usable and shows up in images. But maximizing the benefit (and is 1.5 the most I can get?) is frustratingly hard. I'll keep playing with the S3 Pro to try and lock down what I think the real improvement is. Suffice it to say there is a significant dynamic range improvement available if you want or need it. It's not huge (I've yet to find a three-stop improvement), but it's also not insignificant (I can routinely get at least a one-stop improvement, and often more). Overall, this tends to help JPEG exposures in tough situations that would ordinarily produce blowouts--the camera seems to a very good job of integrating the two sets of data. (Since the Fujifilm cameras appeal most to JPEG shooters, I've been shooting almost entirely in JPEG this first month with the camera. At some point I'll come back and add RAF information to the Performance section.)
| top | home | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
AutofocusI've seen posts that claim the S3 Pro's autofocus is better than the S2 Pro. I don't think so. While it's difficult to test apples against apples and get precise measurements, I don't see any speed increase in the S3 Pro (nor did I expect any when I read the specs). Remember, if you use dynamic autofocus, turn off Closest Subject Priority to get better autofocus in low light. What I do think I see is a slightly better performance in low light, low contrast situations. I can lock focus onto some things with my S3 Pro that the S2 Pro seems to hunt for. This, however, could be something as simple as dust on my S2 Pro's autofocus sensors that need cleaning (though I didn't see anything obvious and did do a quick maintenance cleaning before doing the tests). Using the central autofocus sensor only, the S3 Pro is as fast at autofocus as any Nikon body. AF-S lenses "snap" to focus using the central sensor. The only times you'll see a difference between the S3 Pro and the D1 or D2 series is in low light and off center subjects. Fuji uses a slightly different autofocus sensor pad than the N80 (and the S2 Pro; feels a bit more like the F5 pad than the N80 pad). It's nowhere near as stiff as the Nintendo pad some of the Nikon bodies use, and the location is not burdened by nearby ridges or controls. Unlike the S2 Pro's concave pad, the S3 Pro's is a smaller concave area within a convex button, which makes it easier to push to one side or another.
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Before we get to specific drawbacks and positives, a general comment is in order. The S3 Pro is a love it or hate camera. Overall, if you're willing to give up some camera convenience and performance for superb JPEG image quality, you'll probably be in the "love it" camp. If you shoot only raw files and need a responsive camera, you'll probably be in the "hate it" camp. More so than most DSLRs, I think the S3 Pro is going to create a very bipolar response amongst photographers. Personally, I'm not yet sure which camp I'm in. I tend to only shoot raw, but I also don't need a responsive camera. I'm looking forward to taking the S3 Pro down to the Smoky's in April--it seems like the perfect spot to see if that "green punch," low noise, and extended dynamic range works in a really tough landscape environ. Until then, I'm not sure where the S3 Pro fits into my shooting gear choices (especially since a D2x should arrive shortly).
PositivesWhile I wouldn't take my drawbacks lightly, they do get knocked down a bit in importance when you consider the positives. The S3 Pro takes beautiful, color-saturated photos, and when used correctly can give any digital SLR that costs less than US$2495 a real run for the money. Yes, that means everything from the Nikon D70 to the Canon 20D. It's only when you get to the pro bodies that the S3 Pro starts to suffer in comparisons, though even then its image quality holds up quite well. The S3 Pro loses a small bit in the resolution race, but not nearly as much as you might think; it performs about like an 8mp camera. In short, expect to produce good looking images out of this camera.
What Fujifilm Really Needs to FixIt's all about one thing: need for speed. Almost every flaw of the S3 Pro body relates to it:
I could go on, but fixing just these things would bring the S3 Pro up to the level of most pro's expectations in terms of operating speed. I urge Fujfilm's engineering team to address these things ASAP, even though it probably would require coming out with a new model, ala the S3 Pro+ or S3 Pro II. Firmware UpdateIn early March 2005, Fujifilm announced a firmware update (version 2.1.2) that does the following:
The update is available here. Since I wrote the above, two more updates have appeared, 2.1.6 and 2.1.8. Amongst other minor things, these addressed white balance and color issues some users encountered. Personally, I still rest by my comments about white balance: it's slightly unpredictable in Auto and generally off by quite a bit. I also stand by comments on color: the S3 Pro isn't an accurate camera; the changes fix a bug that caused the camera to produce color values differently than Fujifilm intended, but it's still clear that Fujifilm didn't intend to produce a neutral color result. | | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Sunday, January 9, 2011
Fujifilm S3 Pro Review
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment