A Flawed Jewel
A Shot in the Foot
In mid-August, 2003 Sony announced the DSC-F828. It captured the imagination of many photographers, myself included. Initially promised for November, it began to ship in mid-December, and I received one for testing at that time. This is possibly one of the most eagerly anticipated digital cameras in some time — and because of the 5 month time lag from announcement to shipment I do mean some time.
Sony shot itself in the foot by not providing web reviewers with production samples before cameras arrived in consumer's hands. This has meant that well meaning but uninformed amateurs have been almost hysterical in their on-line discussions about their newly delivered cameras, citing all sorts of concerns. This would not have happened in Sony had simply FedExed a dozen cameras from the first production run to the leading online review sites. This would have meant that knowledgeable reviews and properly produced and processed sample images would have been available around the time that cameras reached early buyer's hands, and much of the sillyness would have been avoided.
The Sony F828 has raised the bar, and calling it a digicam (which has indeed become something of a pejorative term) is to underestimate its significance. The camera has several specs that place it apart from the mainstream and that make it more of a camera than what one might want to use just to take family snapshots. Among these are...
— an integrated 28-200mm (equivalent) f/2-f/2.8 Carl Zeiss Vario-Sonnar T* zoom lens
— the availability of RAW mode (the first in a Sony camera)
— the use of Compactflash cards, including Microdrives up to 4GB (again, a first for Sony)
— a 4 colour RGBE Super HAD CCD sensor, that uses Emerald Green as the 4th colour in the Bayer matrix — again a first
But, these are just the raw specs. As impressive as they may be they tell us nothing about how fit the Sony 828 may be as a productive photographic tool.
I don't often test or use many digicams. During the past couple of years they appear at the rate of about one a day (almost literally) and keeping up with the latest crop is a pointless task. Few professionals or serious amateurs pay much attention to the ever changing model parade, though just about everyone owns at least one. Over the years I've owned several Fuji and Canon pocket digitals, simply for travel use and family snapshots, but I rarely write about them. (My reviews of the Nikon 5700 and Canon S50 are exceptions).
There are two reasons why I will not be comparing the F828 to other digicams. The first is that I don't have a large reference base to compare it to. There are many sites on the Net that frequently test digicams, and I refer you to these so as to be able to compare the nitty gritty laundry list of specs, knobs and features that they provide so well. My expertise is with regard to the vast range of film-based cameras, digital SLRs, and medium format digital backs. Equipment which for lack of a better pigeonhole I describe as either professional, or suitable for the serious fine-art photographer.
Why review the F828 then? Because it appeared to me to be the first non-SLR digital camera that could provide photographers with image quality, image size, lens quality and range, and build quality that might meet more than the needs of the amateur market. But does it?
And, as for what to compare it to I am going to use the Canon 10D, for a few reasons. The first is that I have one handy and I am quite familiar with its image quality. The second is that it is typical in image output to the current generation of 6MP DSLRs from other manufacturers. Thirdly, it is several hundred dollars more expensive (without lens) than the Sony F828, but its image quality is almost identical to that of the Canon Digital Rebel 300D, which is comparably priced to the Sony.
One thing to bear in mind about the F828, and that's its size. In Fig. 1 above you see it compared to a Canon 10D with Canon 70-200mm f/2.8L IS lens attached, and the Canon 24-70mm f/2.8L zoom beside it. The Sony is about the same size and weight as the short zoom alone, and is positively tiny compared with the DSLR with long zoom attached. Yet, the F828 has a higher pixel count imaging chip and a high quality zoom lens covering almost the same focal length as the two Canon zooms combined!
Is this a fair comparison? Read on...
A Review in Two Parts
This review consists of two parts. The first is found on this page, and consists of initial impressions and tests conducted during the third week of December. The second part will appear in a month, in the third week of January and will be based on 10 days of use and testing on a photo safari and expedition in Africa. I will be shooting there primarily with the Canon 1Ds and Canon 10D. My lenses will be the Canon 500mm f/4L IS, Canon 100-400mm f/5.6L IS and Canon 70-200mm f/2.8L IS. I had planned of course to bring some wider lenses, but instead their place in the camera bag will be taken by the Sony F828 which will be my primary camera for casual shooting and when something wider than 70mm is needed.
This type of testing is in my books the most valid way of coming to grips with the handling and imaging possibilities of a new camera — in the field, with dust, heat, hard knocks and the pressure to perform. Stay tuned.
Pixel Size and Noise — Understanding The Problem
Next to resolution (and part of the same equation), the size of the imager, and the size of each pixel plays a vital role in determining image quality. Signal to Noise Ratio is one of the most neglected yet most important aspects of imager performance, and it is determined in large measure by these two factors.
Let's start by trying to visualize the size of the imaging chips used in various camera types, and see how they compare to the F828.
No, Figure 2 isn't some new country's flag. Rather, it shows the relative sizes of full frame 35mm, reduced frame 1.5X such as the Canon 10D, the Sony F828, and a typical pocket-sized digicam's sensor (white).
Full Frame 35mm = 24 X 36mm
Reduced Frame 1.5X Digital = 15.1 X 22.7mm
Sony F828 = 6.6 X 8.8mm
Typical pocket digicam = 3.5 X 4.5mm
So, since everyone is familiar with the size of a piece of 35mm film, you now have some idea of how really small digital sensors can be. Remember, the above illustration is much larger than actual size. Visualize a piece of 35mm film and now you'll see how really tiny these sensors are.
The next thing to consider is the size of each individual pixel or, as they are most commonly called each photo site. Larger imaging chips, such as those used on 1.5X and full-frame DSLRs have photo sites that are about 7 — 8 microns in size. A micron is one millionth of a meter. The Sony F828 has photo sites that are 2.7 microns in size.
Now we see how the F828 is able to put 8 million pixels into a chip that is significantly smaller than those in a 6 megapixels DSLR. They've made the pixels smaller — about 2.5X smaller.
So, good for them, you say. What's wrong with that? Potentially the problem is what we used to call grain (when dealing with film), which we now call noise, and which is better described as signal to noise ratio (S/N). S/N is exactly what it says — the ratio of the signal (that which carries information) to noise (that which carries no information but which is there regardless).
The law of physics are such that the smaller the individual pixel the fewer the number of photons that can reach it. 2.7 microns is incredibly small, and this means that except in brightly lit situations (in fact even in such situations) the number of photons that can reach the bottom of the photo site can be such that the noise overwhelms it, and we end up with unacceptably noisy images. Where does this noise come from? Part of it from the system's electronics, part from the random motion of atoms within the silicon, part from cosmic rays and so forth. So, the lower the signal the higher the S/N, and the noisier the image. This is the challenge that Sony faced when designing an 8 megapixel chip with 2.7 Micron pixels. A considerable challenge. Let's see how well they met the challenge.
Build Quality & Handling
The first thing that strikes you when you take the F828 out of its box is that the size is just right. It's not as small as many digicams, which can feel almost toy-like. But, it isn't so heavy as to start to feel like a burden after a few hours of walking around with one in hand. Over the shoulder it weighs hardly anything.
Build quality is exemplary. The body and lens are all metal — beautifully textured and finished. The main setting knob on the top of the camera looks like it came from a precision scientific instrument. It is not out of line to say that subjectively the build quality on the camera is matched only by the Olympus E1 out of those cameras that I have tested during 2003, and it leaves every other digicam that I have seen or handled feeling chintzy by comparison.
Sony's swiveling"the camera is the lens" design is not new to the F828. It was seen before in the predecessor F707 and F717 models. But though these cameras spring from the same genetic roots, the F828 looks and feels by comparison with these earlier models like its been taking steroids to bulk up and a full semester course at a finishing school.
Controls, Screens and Handling
Anyone looking for a DSLR will be disappointed. This is a digicam — unrepentantly so. There is no pentaprism, no mirror and no optical view finder. Instead there are two screens; a rear panel LCD and an electronic viewfinder. You can, through sliding a rear panel switch, change at any time from one to the other, and both show almost identical images and data displays (though not simultaneously).
Both screens are good — about as good is this technology gets in late 2003. But, an electronic viewfinder is not an optical SLR, and even though it has some advantages (which I'll describe shortly) it pales by comparison with the clarity and you are thereness of the view though even the poorest Single Lens Reflex. The rear LCD is actually quite viewable in bright daylight, something that many others aren't.
I have become enamoured of the real-time histogram which Sony offers on the F828. Being able to fine tune exposure with a live histogram is a joy, and I believe also a window into the future. One day soon all cameras, including DSLRs, will offer this. All it would take on a DSLR would be to siphon off a bit of light (as the metering system does now) and use it to drive a small chip that has similar sensitivity characteristics to the main imaging chip. If a manufacturer were to provide such a live real-time histogram as an adjustable heads-up type display within the optical viewfinder of a DSLR photographers would flock to their door — err, camera.
The live viewfinder histogram would be even better is it was accurate — which it sometimes isn't. I'm still exploring what the issue might be and will have a further report on this in next month's update.
Joystick and Menu Navigation
As do most digicams, the Sony F828 has a miniature 4-way joystick for navigating screen menus. This is the best such design that I've yet seen. Canon's version on their digicams is very poorly executed by comparison as I'm constantly finding that my fingers are making the wrong moves, even though I have small hands. With the Sony I rarely execute the wrong setting. Nicely done.
A word as well about Sony's menu graphics. Again Sony is the class of the industry with their simple yet effective colour choices, crisp text and intelligent animations, which are more than gimmicky. They really do aid comprehension on a small screen.
Which Screen? Oops. Wrong One
But now to the part that doesn't please me anywhere near as much. Again, like many other digicams the F828 uses its screen for both scene display and selection menus. Just about any menu information that can be displayed on the rear LCD can also be displayed in the viewfinder, superimposed over the live or playback image. You even have three levels of information — little — moderate, and the kitchen sink, including histogram. Well done. But the Achilles heel for anyone who has worked with a DSLR is the necessity of manually changing back and forth between viewfinder and rear LCD, This is a total pain. Here's what I mean, and as well a suggestion for Sony on how to fix it.
Let's say you are about to use the camera in the usual eye-level viewfinder manner to take an photograph. You can have as much or as little info on-screen as you wish. You take the shot. Now you decide that you need to set a higher ISO because the aperture is wide open, the shutter speed is slow, and there's the risk of camera shake. So you take the camera down from your eye to look at the rear setting screen to make the ISO adjustments, but of course it's blank.
Oh yes, the switch is set for the viewfinder LCD. You make the necessary changes (which can't be done with the camera at eye level in any event because the joystick is right where your forehead would be), and so now you put the camera back to your eye to shoot, but the viewfinder is blank! Why, well because you have to manually switch back to it. With a DSLR the viewfinder is always live. With a two CRT digicam it's either one or the other.
The solution? It seems to me that a sensor that detected the camera being placed up to the eye would do the trick. This is not new technology. Some cameras have done this for years. The viewfinder LCD would be live when the camera was at eye position and the image would instantly switch to the rear LCD when it was removed from contact with the eye.
The fact that this camera doesn't feature such a capability simply says to me, once again, that like so many other cameras I have reviewed this past year they seem not to have been tested by real photographers in the real world. It look me all of 10 minutes to discover how awkward the current design is, and to come up with the solution. And I'm not a camera designer, by any means — though I have been using the damn things to make a living for the past 40 years.
Other than this, and the placement of non-exposure related items on the exposure knob, the design of the F828 is excellent. All controls, doors and latches have a solid feel. All-of-a-piece, as the saying goes.
The Sony's articulated body design appears to be a bit strange at first, but one quickly discovers its logic and strong ergonomics. It also has a remarkably solid and smooth movement, with just the right amount of tension. The quality of the engineering and execution are remarkable given the camera's price point.
Sony F828 @ ISO 400
The image above, titled Sullen Smokers, shows the advantage of being able to work with the Sony's articulated body and rear panel viewing screen. Of course many digicams have the ability to tilt the LCD screen for waist-level shooting, but the F828 is the first camera of this sort that I've used this way, and I must say that for certain types of street shooting it really does the trick.
These two women, smoking in the December cold outside an office building were very wary of my presence with a camera. But facing slightly away from them I simply looked like I was reviewing images or playing with the controls because the camera was at waist level rather than up to my eye. A slight turn sideways and the shot was made. This was a lot like shooting with a twin lens reflex and its pop-up finder, something I haven't done in decades. Lots of fun.
Shooting Priority — Not
One of the hallmarks of a fast handling digital camera is that regardless of whatever else you may be doing, such as reviewing a previous image or making a technical setting, the instant you press the shutter release the camera should be ready to take a photograph. This is the case for example with Canon DSLRs. Unfortunately it is not the case with Nikon lens mount DSLRs such as those based on the Nikon F80 design. This includes the Fuji S2 Pro, and the Kodak DCS 14n. It even includes the Pentax *st D, which doesn't even share the same gene pool.
The flaw in these designs is that certain settings need to be made by placing the shooting mode knob onto a non-shooting position. This means that for the camera to be able to function the knob has to be manually turned back to one of the shooting modes, such as P, A or S.
The Sony F828 shares this design flaw and even doubles it by placing both the main settings control window activation and image review positions on the top control knob. Sony needs to rethink this design if they want to produce a camera that serious photographers will find responsive.
Image Quality — Noise
Noise (grain) is a subjective evaluation. I could provide screen after screen of comparisons, but what would be the point? With products like Noise Ninja, Neat Image and Grain Surgery available at moderate prices, removing all but the worst noise is quick and simple.
Nevertheless, it needs to be said that at high ISOs the Sony F828 is noisier than some other digicams, and quite a bit so compared to most 6MP DSLRs. Here then is my subjective comparison based on some tens of thousands of digital exposures with a great many digital cameras over the past few years.
Very smooth. Nearly invisible
Visible, but not too objectionable
Visible and annoying. Definitely needs improvement with 3rd party software
Blotchy. Sometimes unusable even with software noise reduction
By way of comparison with the 6 Megapixel Canon 10D; the Sony and the 10D are both essentially noise free up to ISO 100. At ISO 400 noise just starts to become visible on the 10D, though it's not objectionable, and at ISO 800 it looks like ISO 200 film used to look. ISO 1600 (which the Sony can't do) is much cleaner than the Sony's top sensitivity of 800, but still needs software help. Overall I'd say that the Sony F828 has about 1.5X more noise at speeds above ISO 100 than the better current 6MP DSLRs.
But, since nothing beats making ones own evaluation, here are test frames taken at a focal length of 50mm, with an aperture of f/5.6. The 100% enlargements below are shown at 100%, as shot, and also as noise processed via Noise Ninja software (a review of this program will be forthcoming in late January, '04).
Processed with Noise Ninja
ISO 64 — Filtered
ISO 100 — Filtered
ISO 200 — Filtered
ISO 400 — Filtered
ISO 800 — Filtered
Noise — A Further Look (This section is an update to the original review)
A member of this site's Discussion Forum raised an interesting point a couple of days after this review first appeared. He asked why, when I soundly criticized the Kodak 14n's noise as compared to that of the Canon 1Ds, I didn't discuss why it would be less of an issue in prints, as I have indicated is the case here with the F828. Here's what I replied:
The type of noise displayed between the F828 and the 14n are of completely different types. On the Sony (with the very clear exception of ISO 800) the noise takes on the appearance of "grain", similar to that of film. Using this as an analogy, on the Sony at ISO 64 it's very slight, maybe like Panatomic-X. At ISO 100 it looks like Plus-X. At ISO 400 it's like Tri-X developed in Rodinal.
The Kodak 14n on the other hand displayed blotchyness similar to the Sony at ISO 800, even at ISO 320 and above.
So, while the noise "grain" of the Sony is ranges from almost insignificant
to moderate with images from the Sony, the 14n's noise characteristics (at least
from when I tested it) was much coarser, and therefore more visible and annoying.
Long Exposure Noise Reduction
5 seconds / f/5.6 @ ISO 64
Because digital sensors are susceptible to thermal noise during long exposures, the F828, like many cameras features long exposure noise reduction. What this does is to take a dark frame right after the initial exposure and then compare them, removing everything that isn't "image" and which is therefore noise from the difference between the two. This is automatic on exposures of longer than 1/25th of a second, and can not be turned off on the F828. It works very effectively, with the only downside being that total exposure time is doubled. In other words, a 5 second exposure takes a total of about 10 seconds.
The Macbeth colour chart seen above was photographed under tungsten halogen lamps at 3200 deg. It was shot in RAW mode and the gray point was set in the Sony RAW converter software. 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.
Browns and oranges and yellows are slightly redder than the actual chart. Reds and magentas are very accurate and so are greens. The skin tone square (second from the upper left) is reasonably reproduced, though a little bit magenta. Tracking across the neutral gray squares is quite consist ant with a bias towards the red channel being weaker at lighter tonal values and stronger at darker tonal values.
Overall very decent performance, and nothing that should raise any red (pardon the pun) flags. In fact there's a very natural look to the camera's colour palette which some may find a bit soft, but which I find to be very appealing. Something to do with the new HAD 4 colour processing filter? No idea — but if so it's doing its job well.
It should be noted that the camera has two colour "modes", Real and Standard. Real mode gives the greatest accuracy. Standard mode is punched up colour — sort of digital Velvia — if you like that sort of thing, which I don't. Unfortunately it's the default mode, and a lot of people will end up with slightly garish colours when shooting anything but RAW, and wonder why.
Chromatic Aberration, Purple Fringing, Blooming and Other Hobgoblins
Every since the first F828 cameras started to appear in user's hands there was much wringing of hands and fretting in some of the online communities over CA (Chromatic Aberration). The bandwagon had left the station, and many people jumped on, downloading JPGs of shots of people's cats, lawn ornaments and Christmas decorations, and bemoaning what they saw.
I won't belabor the point. Here's what I see. Yes, there is some "blooming" when overexposed areas are blown out. This is easily seen in the water above in Fig. 11. Below is an enlargement at 100% that allows you to more clearly see the effect.
That's the bad news. The good news is that the Sony 828 isn't the only camera that displays this characteristic. I've seen it in every single digital camera that I've ever tested or used, including my Canon 1Ds and all of the 16 and 22 Megapixel backs that I've reviewed — some of them costing $30,000. Sometimes you see it and sometimes you don't. Sometimes it's cased by lens CA, and sometimes by overloaded photo sites bleeding into adjacent darker frames. I've often seen CA on $6,000 lenses when using film. It's a fact of life.
Below in Fig. 13 is a detail showing some flare, where CA is often seen, taken from the frame seen in Fig. 14 in the next section below. There is no CA or blooming whatsoever. (1/800 sec @ f/6.3. ISO 64)
Is the Sony F828 worse that other digital cameras in this regard? Yes, a bit. But not so much as to cause all the the teeth gnashing that I've heard.
In any event, this is something that can easily be cleaned up with software, and there is in fact a new product coming in the Spring of 2004 that will handle CA as easily as programs like Noise Ninja handles high ISO noise.
The F828 comes with a very remarkable lens. It is a Carl Zeiss T* Vario Sonnar of 7.1 — 51mm focal length. This translates to 28 — 200mm in 35mm terms. The widest aperture is F/2 at 28mm and reduces to f/2.8 at 200mm. The 35mm equivalent focal length is what's displayed on the manual zoom ring, which makes life much easier than having to take the actual focal length and multiply by 4X to understand what it is in more common terms.
Did I say manual zoom? Yes, indeed. I'm quite happy to let autofocus handle its task 95% of the time with a camera like this, but I hate the time lag and lack of precision of motorized zoom controls. This lens' manual zoom feels very smooth, and allows for precise framing and focal length selection with a minimum of hassle. The size of the lens is also almost perfect as a platform for support by the left hand.
How good a lens is it? While there's nothing magical about the Carl Zeiss name, it is usually an indication of first quality optics, and this Vario Sonnar certainly appears to live up to its reputation. In examining several hundred frames taken during my testing period I don't see any reason to believe that this optic won't produce excellent quality images. While I have not done any rigorous testing, I will soon have an optical bench set up that will allow me to put this lens through a battery of tests. This camera will be one of the first for which I publish results, some time in the late Winter or early Spring.
One typical defect that I've seen with this lens is barrel distortion at the wide end of the zoom range. This is not unusual with wide angle lenses, and zooms in particular. Fortunately it is easily corrected on the computer if necessary.
In the meantime, combining the ultra-high resolution of an 8 Megapixel chip with the excellent resolution and contrast of this Zeiss optic leads to some first-rate images from a technical perspective.
Something to bear in mind is that though the lens barrel is marked with focal lengths in 35mm terms, from 28mm — 200mm, in reality the lens is a 7.1 — 51mm optic and therefore has the depth of field of such focal lengths. Like most digicams with such short focal lengths depth of field is necessarily very deep, which can make selective focus almost impossible. Because of the 7X zoom range though, it is possible to shoot with somewhat narrow depth of field when the long end of the range is used.
Keep in mind though that the minimum aperture possible is f/8, and so the range of F stops available is limited. The reason for this is to restrict diffaction effects which appear at smaller apertures.
Lens Performance Compared to the Canon 10D
It took two large and very expensive Canon L series lenses to match the coverage and aperture of the Zeiss lens on the F828, the Canon 70-200mm f/2.8L IS and the Canon 24-70mm f/2.8L. (I had to alter focal lengths used on the Canon to match that on the Sony, because of course the Canon 10D has a 1.5X cropping factor over standard 35mm focal lengths and the Sony is 8MP resolution Vs. the Canon's 6MP).
I took a great many frames with both cameras, at 200mm, 100mm, 50mm and 28mm focal lengths. The scene that I photographed was the skyline of Toronto on a sunny day, seen above in Figure 16 at 100mm focal length. Here are a few selected comparison enlargements, with some general observations. All of these comparisons were done both on-screen at 100% and also in large prints. There simply isn't time or room to post them all, so I have just selected a few typical frames.
Sony F828 at 200mm @ f/5.6 — ISO 64
Canon 10D at 200mm @ f/5.6 — ISO 100
At 200mm the lenses are almost identical in performance. There's little to choose between them except for the fact that the Canon 70-200mm f/2.8L (in non-IS version) retails for over $150 more than the entire Sony F828 camera! Something to ponder.
Sony F828 at 28mm @ f/5.6 — ISO 64
Canon 10D at 28mm @ f/5.6 — ISO 100
Please note when looking at the above 100% crops that these
are tiny segments of much larger images. The full size of
these frames at web resolution is almost 3 feet by 4 feet across. Therefore
while there is some noise and other
artifacting visible at these magnifications, it is essentially irrelevant at
normal print sizes up to about 13X19".
My comments about the camera's noise characteristics refer to what can be seen
in normal sized prints
of the full frame, not with ones nosed pressed up against the screen, counting
In any event, what is being evaluated here is lens performance, not noise.
One of the reasons that I am only showing a few comparison frames, in addition to trying to keep the size of this report within reason, is that there really isn't a significant difference to be seen at almost any focal length that lies within the range of the F828 and the two selected Canon lenses. In some cases the Canon lens has a slight edge in resolution and contrast, and in others it's the Sony's Zeiss lens.
Frankly, I'm astonished at the Sony's performance. When you consider that the non-IS version of the Canon 70-200mm f/2.8L sells for about $1,150, and the Canon 24-70mm f/2.8L retails for some $1,300, the Sony offers astonishing value, especially since these two L series Canon lenses are considered among the finest in their respective focal lengths.
To answer the obvious question — did you test the lenses wide open at f/2.8 at each focal length — the answer is no, I didn't have time. I'll leave that one for another day. And neither have I done anything other than empirical testing of other lens characteristics. As mentioned elsewhere in this report, I have a new lens testing system that will be entering service here some time in the late winter and at that time I'll be able to produce more objective test results.
In the meantime I know that many people are going to find these results hard to believe, but I assure you that they are what I see, and are quite repeatable.
Update: Within hours of first publication I was taken to task by a good friend for not publishing comparisons of photographs taken with the lens wide open at f/2.8. The point that he made was that most lenses do well stopped down — but it's when used wide open that the good are separated from the not so good.
Of course he's right, and it was my intention to do more thorough tests when my optical test system in installed in a month or so. But it would not be fair to leave the question unanswered, even preliminarily. My time is limited though, as I am preparing for conducting a two week workshop abroad, but I spent an hour or so doing some comparisons with the Canon 70-200mm f/2.8L IS and the Sony F828, with both lenses wide open at f/2.8
In Figure 19 you see the full frame, while in Figure 20 below are crops of both the center and edges of the image from each camera.
Sony F828 @ F/2.8
Canon 70-200mm L @ f/2.8
Edge of Frame
Edge of Frame
I see virtually no difference at center frame, but while the edges are extremely close the Canon may be slightly sharper, though the Sony seems to have a bit more contrast. In all, considering these frames and a few others done at different focal length comparisons, it seems that what I've seen with the lenses stopped down also holds true with them wide open as well. (There is a bit more JPG artifacting on the Sony image, which also needs to be taken into consideration. I didn't have time to do the tests in RAW mode, which would have eliminated this differential).
Autofocus / Auto exposure / Auto White Balance
The F828 excels in each of these areas. Frame after frame is well exposed, properly white balanced and (usually) properly auto focused. I found little need to play with exposure compensation as in multi-segment meter mode the system seems on a par with Nikon and Canon's multi-zone metering systems. In other words — first rate. There are also two more restricted meter modes, including center-weighted averaging and spot metering, but I haven't yet needed them.
Auto White Balance is uncanny. Even in tough mixed light source situations it rarely does anything other than a good job. This is a blessing because (as will be seen below) RAW mode (which allows post-exposure white balance setting) is almost an unusable option for a number of reasons.
Autofocus is flexible and on par with larger and more expensive cameras. There are three modes — Multipoint using five sensors, Center AF and Flexible Spot AF, which allows the use of the joystick to move the focus sensor to any part of the screen, something most appropriately reserved for tripod mounted applications.
There is also what Sony calls Hologram AF. The camera has a low power laser built in (totally safe — they assure us) that projects a patterned beam on subjects up to about 6 feet away, which aids in establishing focus in low light levels. It works very effectively.
High Speed Shooting
In addition to the normal single shot mode the F828 has two high speed shooting modes, something that to my knowledge is unique among digicams, and which can rival many DSLRs in shooting speed. These modes are called Speed Burst and Framing Burst. They are essentially the same thing, except that in Framing Burst you briefly see each frame as it is taken, while in Speed Burst the screen blanks out while shots are taken. Obviously Framing Burst is to be preferred if you are panning or tracking a subject.
Framing Burst can take 2.4 frames per second, while Speed Burst is capable of 2.6 FPS, in both cases for up to 7 frames. Not at all shabby for a digicam.
But, there is a fly in the ointment. The F828 has a small buffer, and unlike many DSLRs isn't able to read and write data at the same time. This means that whether you take one frame or 7 frames in these modes, the moment you take your finger off the shutter release the camera starts writing to disk, and isn't available to take another frame until the write process has completed.
I was also confused at first when I tried to use burst mode at slow shutter speeds (admittedly, not something that one would often need). What I saw was that the shutter speed shifted from whatever it had been set to 1/30th second, severely underexposing the image. I was baffled until I realized that exposures longer than 1/25th second require dark frame subtraction noise reduction, and obviously this was being turned off by the camera, as it was clashing with the high frame rate setting. Not a big deal, but thought you'd like to know.
Auto Bracketing & Exposure Compensation
There is a three frame auto bracketing mode which allows for automatically bracketed exposures with a range of +/- 2 stops. It is activated via the same button on the side of the lens that selects high speed shooting modes.
Just below the shutter release there is an exposure compensation button, which when used in conjunction with the control wheel that falls under ones thumb allows for +/- 2 stops compensation in 1/3rd stop increments.
Start-up time (from turn-on to first frame shot) is 2-3 seconds, depending on the type of card used. The camera will auto-power-down after 3 minutes of none use. This is not user changeable— but should be. While the camera is powered there is a small green light visible just next to the shutter release. When the camera goes to sleep it takes the full 2-3 seconds to come back to life.
Autofocus is what detracts from the speed of shutter responsiveness on most digicams. Sony has come up with a clever mode that lets the camera constantly autofocus even when your finger is not on the shutter release. They refer to this as Monitoring AF. This makes the camera much more responsive than other digicams I have used. It isn't an M series Leica, but it also isn't that far behind most autofocus DSLRs. In addition there are the usual Single AF and Continuous AF modes.
Batteries and Charging
The F828 comes with a rechargable Lithium Ion battery, one of Sony's InfoLithium series, as used in many of their other video and digital cameras. These are high capacity and of excellent design, as they work on concert with the camera to inform you of the amount of time left (in minutes) and also, when recharging, how many minutes of use the battery has at any point in the charging cycle. First rate.
But, the charger that Sony supplies does not come with a battery holder. Instead the battery is charged in the camera. This is fine as far as it goes, but if you then buy a second battery (which you will), you also need to buy an external charger as well since it makes no sense to tie up the camera while charging the second battery. All of this adds about another $125 to the cost of the system.
What I have done is to buy the very small BC-TRM charger and I leave the supplied much more bulky charger and cord behind.
As for battery life, it looks to be very good. I shot a couple of hundred frames one day in temperatures slightly below freezing and still had about 1/3 capacity left. I will keep track of usage during my African trip and report afterwards on the typical number of frames and review usage that one can expect in the field in warm conditions.
Movie Mode — Don't ditch your video camera just yet
Like many digicams the F828 allows the recording of live video as well as stills. But what sets this camera apart is that it can record full-screen 640X480 PPI at 30 FPS, with audio. The data rate at this resolution is 1.2 MBPS on disk. It takes a 1GB memory card to hold about 12 minutes of video, but if you have one of the new 4GB Hitachi Microdrives this can hold over 45 minutes. According to Sony either a Memory Stick Pro or Microdrive is a must to record video in this highest resolution mode because of the fast transfer rates required.
That's the good news. The bad news is that the image quality is fairly poor — at least in comparison to Mini-DV. Mini-DV video is 720X480 at 3.6 MBPS. Specs aside, it also produces an essentially broadcast quality image. Footage from the F828 on the other hand is loaded with artifacts; strobing, moire on movement, and constant pumping from the auto exposure mechanism. This is particularly annoying since Movie Mode can not be set to operate with manual exposure.
This again points out the error Sony made in putting both exposure modes (P, A, S) and non-exposure-related setting positions on the same dial. This is the same mistake as made by Nikon in its N80 based cameras, and other manufacturer's models based on this design. I'm sure that if Movie Mode was accessible as a menu selection Manual exposure mode setting would be possible.
Movie Mode has a basic editing capability built in. This allows you to take any segment and divide it into two pieces — for example clipping the poor end off a sequence. Once it's a separate segment it can then be deleted the same way that any image can.
So — what's the F828's movie mode good for? It seems to me that it's simply a reference tool for occasional use. I'm pleased that it's there, and I expect that there will be the odd time that it comes in handy. But it really isn't a substitute for even the lowest-end camcorder. Even High-8 video would be better.
I expect to shoot a bit of video footage with the F828 when I am on Safari, and will put a few medium sized clips on disk on Issue #10 of The Video Journal. This issue, due in late March, 2004 will contain my full field review of the F828, shot following the shoot in Tanzania. That way you'll be able to evaluate results for yourself. Maybe a shot of a charging bull elephant? We'll see.
Infra-Red Night Shot and Night-Framing Modes
In addition to its Laser Holographic focusing assist mode the F828 features a separate infra-red transmitter, which when combined with an automatically retracting infra-red blocking filter inside the camera allows for some rather neat tricks. In Night Framing mode the emitter allows for focusing in total darkness combined with exposure using the built-in flash, while in Night Shot mode the infra-red emitter allows exposure without flash in total darkness. This is a very weak beam, and so a tripod is recommend, and it can only be used at short range — but it's a nifty feature.
More to the point for many photographers is the fact that when in Night Shot mode the camera's infra-red blocking filter (which all digital sensors have) is retracted from the light path, and so along with the use of a suitable infra-red filter on the lens itself special effects photography is possible. I did not explore this feature to any great extent.
Update: I have discovered what appears to be an undocumented "feature" of Night Shot mode. When the camera is set to P, A or S exposure modes the currently set ISO is applied. But, if the camera is set to the green "dummy" mode, wherein the camera does almost all settings automatically, when the light gets low enough the ISO gets automatically bumped up to 1600. This, combined with the built-in Infra-red transmitter, permits shooting with hand-holdable shutter speeds in very low light levels. The results are greeny-gray and very noisy, but usable for some purposes.
A RAW Deal From Sony
The F828 is the first camera from Sony to provide RAW files. This means that the files recorded have not been modified in any way by the camera's processing electronics. You are free to change White Balance, Sharpening etc., all in the RAW conversion software without penalty, which is not the case with JPGs. The F828 also records a JPG at the same time so you don't need to make the decision of which type of image to shoot, though recording both does increase storage demands.
That's the good news. The bad news is that Sony has encrypted the image data in their RAW format for the F828. This will make it difficult to impossible (technically and/or legally) for third party software to decode their raw files without Sony's permission and cooperation. This means that terrific RAW programs like Adobe Camera RAW or Phase One's Capture One, nor any other third-party RAW conversion software will be available for the F828 unless Sony permits it.
What can Sony be thinking of? Throughout the history of technology this type of proprietary approach has proven to be counter-productive, and neither in the best interest of the consumer not the company that decides to be controlling and secretive. Two thumbs down on Sony for this anti-competative, anti-consumer and ultimately misguided attitude.
To make matters worse they have done one more thing to cripple using RAW mode on the F828. That is the camera's...
Insy Tinsy Buffer
The camera appears to have approximately a 25MB buffer. This allows it to shoot seven 3.5MB Fine JPG images in just a bit over 2 seconds. Amazing performance for a digicam. But, when shooting in RAW mode it takes some 12-15 seconds for each combined RAW plus JPG file to be written to disk, and because the camera can't write to disk and record at the same time (which many DSLRs can) the camera is essentially "frozen" until the disk write is completed.
This is simply unacceptable. If Sony had upped the buffer to just 100MB and put in some smarter code they would have been able to record at least 3-4 RAW files before the camera slowed down. This is yet another area where Sony has missed the plot.
Clearly Sony wants to be a "player" in the digital camera marketplace. Adding Compactflash card capability as well as a RAW recording mode shows their serious intentions. But why cripple the system so that using RAW mode becomes difficult to use, or annoying at best.
Sony Image Data Converter V1.0
That's the name of Sony's RAW converter software. In late December cameras are shipping with a Windows only versions. There is a flyer in the box that says that Macintosh RAW software will be available in February, 2004. If it's anything like the Windows version I wouldn't get too excited.
V1.0 looks to be exactly that, a first generation approach. It's not really terrible, (Canon's own RAW software is only marginally better after 3 years) but it's also not very good. One of its real flaws is that files can only be exported in 8 bit mode. What are they thinking of? The camera must be recording in either a 12 or 14 bit colour space, so why remove image quality by not allowing export in padded 16 bit mode?
Also the files are in the sRGB. Camera makers have been using the wider Adobe RGB colour space for a couple of years now. Common Sony, get with the program.
And while you're at it, remove the encryption or at least provide major companies like Adobe with access to your file format. You'll benefit, your customers will benefit and you will be better equipped to join the ranks of companies that are "serious" about photography. Your current approach to support of RAW format photography is simply mickey mouse.
Update: February 2004
Adobe's Camera Raw 2.1 became available as a free download for current users of Photoshop CS mid-month. It contains support for a number of new camera's Raw formats, including (hooray) that for the Sony F828. It seems that the Sony encryption was broken within days of appearing on the market, but more to the point Sony apparently decided to work with Adobe and make its Raw format available. Sony is to be congratulated for doing the right thing, and as an F828 owner myself I say — thank you! Now, could you please add enough RAM buffer so that RAW format became really usable?
Price & Competition
When it was announced in the summer of 2003, the Sony F828 was to have had a list price of U.S. $1,200. By the time it shipped in December that price had dropped to $995. Of course this was not the results of Sony's altruism but rather the appearance of the Canon Digital Rebel 300D, which had a $995 price on introduction, with lens.
This of course begs the question — which should one get for one's thousand dollars?
I won't answer this because it isn't really a sensible question. It very much depends on ones photographic needs and wants. Looking strictly at build quality, the Sony wins hands down. It is of all-metal construction and built like a brick (though nowhere near as heavy). The Canon, on the other hand, has a decent feel, but is quite plasticy. The Canon's provided kit lens isn't a patch on the Sony's Zeiss glass, either in optical quality, maximum aperture or focal range. On the other hand, the Canon lens is removable, and any of Canon's 50 odd lenses, including all of their superb L series glass is available.
The Sony has an 8 Megapixel chip, while the Canon is 6 Megapixels. This means that all other things being equal you can make somewhat larger prints from the Sony at the same output resolution. Then again, the Canon's files will have considerably lower noise levels at anything other than 100 ISO.
The Sony's dual screens can provide greater shooting versatility in some situations, while the Canon's optical SLR viewfinder is much better for framing and judging composition.
See what I mean? Not quite apples and oranges, but with enough major differences to give one pause, and to make a simple A or B choice very tough to recommend.
No! This is not an example of green Chromatic Aberration.
The abandoned brickworks where this was shot has green tinted windows.
There are still a number of the Sony F828's features that I haven't covered. This includes it's built-in flash, (which works very well, and covers 28mm smoothly), to red-eye reduction, the elegant motorized elevation of the flash pop-up, available remote control unit, AE lock mode, date imprinting, self timer, audio recording, and built in coffee maker. Actually, the F828 doesn't have a built-in coffee maker, but I wanted to see if you were still paying attention. I'll leave it to other sites that list every knob and feature to provide you with the run down on these, and possibly a few others that I may have forgotten. Hopefully I've covered the most important ones.
I began this review by referring to the Sony F828 as — A Flawed Jewel. Though after reading the above review you may find that I have been harsh in many of my critisms of the camera, I think you'll also find that I have given praise where due. No one conclusion is possible, and I think I am most frustrated by the fact that so many aspects of this camera are truly excellent that its flaws are made all the more evident.
The bottom line though is that I know of no other camera at the price, or even close to it, that offers the small size, low weight, superb build quality, excellent and versatile optics and fine image quality of the Sony F828. Now if only Sony would address and fix some of the camera's glaring flaws it would really be something.
The Good News
— Fantastic lens
— Great resolution
— Very good colour reproduction
— Superb build quality
— Very good handling
— Live histogram in viewfinder
— Use of CF cards, including Microdrives (rather than ridiculously expensive Sony Memory Sticks)
The Bad News
— Noisy above ISO 100
— Non-exposure functions on main dial prevent making exposures unless dial is reset
— To small a buffer and inability to record images and write data at the same time
— Useless RAW mode due to the above
— No 16 bit files or Adobe RGB colour space
— Primitive RAW software
— Encrypted RAW software
So, should you wait for the F838 and some of the needed fixes. Hell no — you'd be missing out on one of the most enjoyable digital cameras yet available. There's always going to be something new on the horizon. But in the meantime the Sony F828 has found a permanent place in my camera arsenal— it's the ideal digital travel camera, and I'll be back here in a few weeks with some real-world illustrations from my Tanzania Wildlife Workshop and Safari.
Now that production cameras are shipping full reviews will soon appear, as well as the "previews" that have been online since last summer. Below are links to several sites which have previews available, and which will soon have full reviews. These are also an excellent source for product photos and detailed specifications.
The second part of this review will appear in mid-January, after I return from Tanzania. It will contain my field experience, a portfolio and additional tests. A comprehensive field report, full resolution sample images and more will appear in Issue #10 of The Video Journal, due for publication in late March, 2004.
I should point out, for the benefit of those that see noise in the ISO 64 and ISO 100 F828 frames, that to the extent that it's visible in 100% on-screen enlargements, it's invisible in A3 or Super A3/B sized prints. The same applies to very small amounts of colour fringing. These just can't be seen in real world photographic prints.
Unfortunately there are some folks who see a bit of noise texture in a 100% crop and somehow jump to the conclusion that the image is crap. It shouldn't be necessary to point this out, but if the entire image were shown it would be equivalent to an on-screen image almost 3 feet high by 4 feet wide. Of course you're going to see some noise and other artifacts. But the important thing to keep in mind is how does it look in a typical sized print — say up to 11X17" image size on paper? (Look at 42" Plasma HD TV from 6 inches away and you'll see dots. Watch the football game from 9 feet away and you'll enjoy the show).
The answer is, as I've stated several times in the above review — at ISO 64, and to a slightly lower extent ISO 100 — noise levels on prints are excellent. Almost indistinguishable from images shot with a Canon 10D. Does it matter that on-screen, at actual-pixel magnification, there are more flaws on the Sony image that in the 10D image? Not one bit. If it can't be seen on a print it's relevance to the creative artist is essentially non-existant. To what some have called the "pixel-paparazzi" though, it is a call to arms. So be it. But readers who chance upon these on-line discussions (and the Sony Forum on DP Review is a hornet's nest) should take a pause after reading some of the vitriol.
Long-time readers will recall that there was a similar hubbub three years ago when I published the first review of the Canon D30, claiming that on modest sized prints it surpassed 35mm scanned film. "No way", cried the film faithful. But, within a year though there was almost universal agreement that the low noise of Canon's first DSLR was a major factor in its imaging capabilities, regardless of pixel count.
Then, just a bit over a year ago I published the first review of the 11 MP Canon 1Ds and found that it surpassed medium format scanned film. Again there was an outcry. "How can this be"? Yet, it wasn't more than a few months later that mainstream magazines around the world and countless pro photographers agreed that indeed images from the Canon 1Ds were the equal if not superior to scanned medium format film.
So, now I make the claim that on prints up to 11X17" the Sony F828 produces images that are the equal of those from a 6MP DSLR and a high quality zoom lens. Disagree? Go do the test for yourself if you can. Do an honest appraisal of well-shot, well-printed images. When you do, if your findings are different than mine — fine. Let me know and we can discuss it. But the rabid "Oh no, it can't be" Cassandras should, in the meantime, make the test for themselves rather than spouting their diatribes based on the way they think things should be, rather than on empirical tests.
Now that I've got that off my chest — back to photography.
There are None So Blind as Those That Will Not See
I suppose I should have seen it coming. The inhabitants of the world of digicams are different than the professional and advanced amateur photographers who are the typical visitors to this site.
I rarely read on-line Forums (other than the ones on this site) because I find the signal to noise ratio so abysmally low. But when my statistics package showed me that the source of a great many of the 15,000 people that had visited this page within the first 24 hours was from one Forum, I spent some time reading what was being written there.
I decided not to join the on-line fracase because it's a no-win. The more rabid denizens of these boards are interested in only one thing — scoring debating points and proving how smart they are, and joining them in this activity is a mugs game.
But I did write the One More Thing paragraph above as a partial rebuttal. Nevertheless there are those now that are "seriously" accusing me of having a drug problem, a drinking problem, having lost my eyesight, being in the pay of Sony, being in the pay of Canon, or being in league with the devil. (I made the last one up).
I'll just have one more shot at a reply, and then it's my intention to close the subject.
I leave just before New Years Eve for a two week long trip to Africa where I'll be leading a photographic Safari and workshop. I'll be shooting alongside the dozen participants, and I expect that there will be some once-in-a-lifetime photographic opportunities to be had.
I'll be bringing my Canon 1Ds as my main camera, with a Canon 1D as backup. A range of L series Canon lens will be in my kit. I'll also be bringing my new Sony F828, and I plan on using it for the "travel" aspects of the shoot, as well as for portraits of people and landscape — wherever short to medium long focal lengths are needed.
Will I be doing this because I want to come home with inferior, noisy, CA littered and otherwise technically inferior images to sell to stock agencies and to use in future gallery exhibitions and books? Of course not! I'm bringing the Sony because my testing has convinced me that I will be able to produce prints of a high enough technical quality level to meet all of these needs.
Am I bringing the F828 because, as some on the boards have slanderously accused, Sony gave me the camera, or because they are somehow sponsoring my Safari? No, again. I purchased the camera with my own money from Sonystyle Canada, and the only Sony employee that I have ever spoken with is the telephone salesman who took my order last month. If the camera had not been to my satisfaction I would simply have returned in within 30 days for my money back. I certainly wouldn't be bringing it along as a primary tool on a once-in-a-lifetime shoot, now would I?
Here's the thing that the pixel peepers don't get. My testing has shown that photographs produced by this camera can be of professional grade — suitable for stock use, professional publication and fine-art exhibition. The fact that it is somewhat noisier than other cameras at any given ISO, and that there is some occasional CA really has little to do with its usefulness as a photographic tool.
Those who spend their time peering at computer screens will be able to find fault. Those whose livelihoods and artistic passions are related to producing the best images that they can with the best tools they have available will find the Sony F828 to be an estimable camera because of its fine qualities — which include low bulk and weight.
I'm taking it as one of my main photographic tools on my trip to Tanzania, and that's why, at least as far as I'm concerned, the subject is now closed.
Update: Dec 29, '03: Uwe Steinmeuller at Digital Outback Photo has now published his initial diary review of the Sony F828. "Holy Pixels Batman", his conclusions are fairly similar to mine. Imagine that. He must be on Sony's payroll as well :-)
Jeff Keller at Digital Camera Resource Page has also published his review of the Sony F828 today.
Steve's Digicams now has its review of the Sony F828 online.
DPReview's final review is now online.
Feb 5, 2004: Imaging Resource has just updated their Sony F828 review.
Update: January, 2004
The Sony F828 has been subjected to a comprehensive optical evaluation using DxO Analyzer.
These results are found here.