dslr.ws

The Online Photographer has a very interesting post up comparing the Sony A900, the Nikon D3 and D700, and the Canon 5D Mark II. TOP usually doesn’t get into such potentially controversial territory — at least as far as camera brand loyalists are concerned — and it’s a very interesting post.

Luckily, to avoid pissing off any one faction on the Internet too much, Mike is able to recommend each of the cameras in one way or another. He says the Sony A900 wins for “the ultimate in image quality,” although he says that it’s not perfect for every kind of photography.

But between its staggering resolution and very good dynamic range, its willing response to the Exposure and Recovery sliders, and its more “photographic” image quality and lack of digital artifacts—and despite its less-than-accurate color—it’s the IQ emperor for now, among these four (i.e., the three under discussion and the D3).

But he also mentions that if you don’t need the A900’s whopping megapixel count, then it shouldn’t really be in the running.

Up next is the “most recommendable” camera, which he says is the Nikon D700.

Given its sensible size compared to the D3, robust build, fast autofocus, overall responsiveness, superior ergonomics, unmatched high-ISO performance, and perfectly sensible file size, it’s going to be the most bang for the buck for more photographers than either of the others. The Nikon is flat-out a better camera than the Canon, a point exemplified by its clearly superior autofocus performance…. my feeling is that it would help more photographers take better pictures in more situations than either of the other two.

That leaves the Canon 5D Mark II as the “best compromise” between the A900 and D700. The 5DMkII doesn’t have the almost-ridiculous resolution and detail of the A900, but it’s up there. It also doesn’t have the high ISO noise performance of the D700, but it’s good. And, while it isn’t as good as the two leaders in those categories (in this comparison) it’s overall much better balanced:

And here’s the thing: [the 5D Mark II] has much more resolution than the Nikon, and much better high-ISO capability than the Sony. So its win over the Nikon where resolution is concerned is bigger than the margin by which it loses to the Sony in the same department, and its win over the Sony in high-ISO performance is much more decisive than the margin by which it loses to the Nikon on that score.

So if you giving each camera a score in both image quality/detail and high ISO capabilities, the 5D Mark II wouldn’t win either of those categories, but if you add them up to give you a total overall score, it would win. He makes sure to point out that he’s ignoring the strong video capabilities of the Canon 5D Mark II, so if you have a use for HD video in your DSLR, this becomes a much easier decision to make.

Then come the image quality issues with the 5D Mark II, which include some unusual chromatic aberration results that I haven’t seen mentioned in other reviews so far:

It’s been eight or 10 years since “purple fringing” (also called “CA,” not entirely accurately) first heaved into our collective consciousness as a peculiarly digital anomaly, and since then, other artifacts have been dealt with in their turn. I don’t see much in the way of purple fringing at all from the 5D Mark II, but there’s what Carl Weese calls “blue replacement,” by which narrow objects imaged against a brighter background change from their own color into a darkish pastel hue. You see it most often in twigs and telephone lines. The 5D Mark II isn’t particularly bad, but it shows up a lot more than it does from the D700. And its susceptibility to blue replacement makes it a candidate for a lovely lens aberration that I’d never actually seen before in a picture I’ve taken myself—longitudinal chromatic aberration (LoCA), which shifts objects in front of the plane of best focus to magenta and those in back of the plane to green.

There’s a sample photo from the 5DMkII showing all sorts of ugly purple and green branches, which makes me feel a little bit ill inside. I’ll try not to lose my lunch all over your shiny new cameras.

Mike also feels that Canon’s noise is blotchier than other cameras, and says that the highlight clipping is harsher and “less fixable” on the 5DMkII than on other recent cameras, and describes it as the kind of thing we would have seen a few years back. And, of course, the much-discussed black dot problem to the right of blown highlights. Mike somehow manages not to turn into a raving photography forum inhabitant with his reasonable downplaying of the black dot issue:

This doesn’t bother me at all—you’d never see it in prints and you probably wouldn’t notice it if it were visible—but hey, I’m just a reporter, I gotta report what I see.

I think we can be reasonably certain that Canon will fix this in a future firmware update.

Overall, a very interesting comparison between three excellent cameras, and I recommend everyone go visit the site and read the full review.

Yes, that’s a little weird, but it’s true. Michael Johnston of The Online Photographer used to be a big lens connoisseur. However, then they became so consistently good and less unique and interesting, and he lost some of that interest. Luckily — for him, not so much for everyone else — Nikon sent him the AF-S VR Zoom-Nikkor 24–120mm f/3.5–5.6G IF-ED lens when they shipped him a Nikon D700 to review. And he sure has some choice words about the quality of the lens. Check out his post, or just enjoy these highlights:

• It’s a piece of shit.
• Despite its fancy specs, this is for all intents and purposes a perfect throwback to the days when even good zooms couldn’t aspire to the performance of ordinary garden-variety primes. Its performance is for all the world like an early-’80s mid-level zoom—smack dab in the middle of the era in which zooms earned—and deserved—their still-lingering bad reputation.
• It has flagrant amounts of linear distortion not only at its wide setting but well into the middle range, and apparent perspective distortion even near the middle of the frame(!).
• The D700 could hardly focus the thing—I got more out-of-focus shots than I have with any AF lens in years
• Its sharpness is lackluster. At 120mm, I don’t think the thing gets sharp. At least, not without stopping down further than I was able to.
• The deterioration in performance toward the corners is often marked—and not just at the extreme corners, either.
• Color transmission borders on sucky (I know this from having recently used the 24–70mm f/2.8 on the D3).
• This is a very inexpensive lens that is not worth half of what it costs.
• If you innocently purchased one of these and are not lucky enough to be using it on a DX sensor, try to get your money back if you possibly can. Otherwise, stop down and avoid the extremes of the zoom range, even though they’re probably why you bought the thing in the first place.
• The VR doesn’t even work very well. It works, but it’s the least effective image stabilizing I’ve yet experienced.

So, there you go. Avoid the lens like something super scary that you should run away from, and be happy that Mike has renewed his interest in lens quality!

Phil Holland has posted a nice, in-depth review of the Canon 5D Mark II DSLR, including a number of RAW files for people who want to do at-home pixel peeping and comparisons. Phil is a longtime 1DsMkIII user, and recently replaced his backup 1Ds with a 5DMkII.

He likes the viewfinder from the 1Ds Mark III better — for usability issues more than image quality or anything like that — but says the 5D Mark II does have a nice viewfinder. I’m sure anyone upgrading from an APS-C sensor camera to the 5D would be much more blown away by the viewfinder than he was.

Phil says that Canon’s new automatic brightness feature for the rear LCD display drove him nuts, and he quickly turned it off:

Canon implemented a new automatic brightness feature that uses a light sensor on the back of the camera to adjust image playback brightness based on the ambient lighting conditions. This is a neat idea, but drives me crazy as the image will literally change before your eyes. For instance, if you are shooting a sunset, the image you look at on the screen will look different if you are facing the sun or facing away from it.

With auto focus, Phil says the 5D Mark II is an improvement over the original 5D, but finds that the 1Ds Mark III is leaps and bounds ahead of either of them. He does mention that this will probably only matter to you if you do a lot of photography in very low-light situations, since there’s not much to distinguish them in well-lit autofocus performance.

As far as image quality goes, Phil finds that images from the two cameras are pretty much indistinguishable, as would be expected from the same sensor. He does note that images from the 5D Mark II were very slightly brighter and with deeper reds, and images from the 1Ds Mark III very slightly bluer, when compared. It’s a teeny tiny little difference, though, which you can see if you look at the sample photos he posted in his review, but otherwise doesn’t really seem worth mentioning (other than to point out that it’s there and then forget about it.)

Phil does seem to confuse a Bayer filter with an antialiasing filter, based on this paragraph:

In this “real world” example you can see that the 5D Mark II does a nice job of capturing detail. I chose a subject that has both subtle colors and lot of highlight, middle, and shadow detail. If you have ever shot with any of the higher end PhaseOne backs you’ll notice the still rather strong effect of the bayer filter getting in the way of true pixel sharpness on the 5D Mark II, but this is the way of digital SLRs. What I’m saying is the sensor itself and the lens used for this photograph could likely squeeze more detail out of this scene, but the bayer filter (which removes moire patterns) is softening up the image a bit. That said, these are amazing results for a camera at this price point.

I read through that paragraph a few times, and the only way it makes sense is if you assume Phil is talking about the antialias filter. The Bayer filter will get in the way of true pixel sharpness, but that’s because it’s averaging the brightnesses detected at neighboring photosites to determine the correct color, it doesn’t affect moire patterns. I’m also pretty sure that the Phase One backs use a Bayer filter as well, although I could be wrong — some Googling didn’t turn up anything definitive. The only cameras that wouldn’t use a Bayer filter are the Foveon sensor cameras and the early digital backs that did three separate passes — in a long, slow process — one for each color they were scanning. The antialias filter prevents moire patterns, but that’s separate from the Bayer filter.

As far as high ISO noise, it looks just a touch better than the Canon 1DsMkIII based on the samples he provides, and is one of the few reviews to admit that there’d be a use for the highest and noisiest ISO setting:

I think with more aggressive noise processing you can certainly use the ISO 12800 and 25600 images, if nothing else, for web reproduction. For me I think the highest I’ll be going is ISO 6400 to maintain print quality.

I’m always surprised by reviews that talk about how useless the highest ISO settings are and why would the manufacturers even include them. Why wouldn’t you want the option available, just in case you need it? I’d much rather have the option of shooting at ISO 61,200 and get something with a minuscule signal-to-noise ratio, than to wind up with no image at all (or, at least, a blurry image because I was forced to use a longer shutter speed than I should have.)

He spends several paragraphs discussing the 5D Mark II’s very interesting HD video capabilities, which are very cool and also have some severe drawbacks. There’s the 30 fps limitation that we heard about early on, but there are some other ones that are even odder. Apparently you’re limited to using the camera’s automatic exposure mode when you’re in video mode. Really? Huh? Canon is trying to turn their $3000 5DMkII into a point and shoot digicam? You can try to trick the camera and work around the limits (apparently a common way is to use an old manual Nikon lens with a mount adapter so that you at least get manual aperture control on the lens itself) but it doesn’t seem like there’s any technical reason why Canon couldn’t let you do video in manual exposure mode. It just doesn’t make sense. You do get two stops of exposure compensation in automatic mode, and exposure lock still works, so you might be able to trick it into the settings you want that way, but it’s still a very weird limitation on such a groundbreaking feature.

You’re also limited to 12 minutes of video per file, which shouldn’t be a big deal for most people. You’ll also get that CMOS “jello” effect, although that’s something you’re just stuck with if you’re doing video with a CMOS camera.

That said, he does offer this bit of awesomeness about the 5D’s video:

In reality though at this price point, even with the limitations of the HD video mode, it’s still an amazing deal. I’ve already shot footage that I could not shoot with rigs that cost me 3x the price of this camera and in higher quality no less.

This was posted a little while back, but I hadn’t gotten around to reading through it until today: it’s DPReview’s Sony A900 review.

They point out a few areas where the Sony A700 has better specs than the A900, which are all related to burst shooting capacities. For example, the A700 can shoot 18 frames of RAW in burst mode before pausing, while the A900 only does 12. This is no big surprise given the resolution increase between the two cameras (12.2 megapixels to 24.6 megapixels) and the much larger amounts of data that the A900 needs to move around with each picture taken. In every other specification measurement, the A90 is a clear improvement over the A700, except for the slightly larger size and heavier weight (895g with battery for the A900 versus 768g with battery for the A700.)

DPReview brings out a big wheelbarrow full of sarcasm with the choice quote, “Once again Sony has included that professional photography essential, the Memory Stick Duo slot.” They also helpfully describe the Memory Stick slot as “a small, narrow place to keep dust.” Who knows what Sony was thinking with putting Memory Stick Duo slots on their “professional” level cameras, other than their usual “we’re going to keep pushing our useless proprietary format long past the point where we become the laughing stock of the industry.” See also ATRAC.

Like other reviewers, DPReview loved the A900’s viewfinder:

Aside from the headline-grabbing sensor there’s no doubt that the single feature Sony decided to concentrate on with the Alpha 900 was the viewfinder (even the design looks like the entire camera was built around the prism). And what a viewfinder it is; with 100% field of view and an incredibly bright, clear image it sets a new standard for this class of camera that I can’t see being improved upon in the near future. The eye relief is a little tight (you need to get pretty close to see the entire engulfing view), but boy is it worth it. Start using the Alpha 900 ‘in the field’ and you soon forget about fripperies like live view and simply enjoy the experience of a truly involving photographic experience.

They also pick on the new top panel LCD display as being relatively useless, although they combine that with praise for the “superb interface” on the main LCD screen. But it does seem like a silly feature. People who complain about cheap DSLRs that don’t have a separate status LCD screen want one because it provides constantly available information at a glance while shooting, and the A900’s status LCD just doesn’t manage to pull off that function (since almost no information is always there, forcing you to press a button to temporarily display one of the limited bits of information they let you use the display for.) Using such a tiny display also means the information is much more confusing than any other DSLR I’ve ever seen:

The fact that the display appears to be based on a dime store digital watch also means that the values are displayed in what is occasionally rather crude and far from immediately obvious manner (High speed continuous, for example, appears as ‘oooH’) – there are very few icons

That sounds more like the LCD display you’d get on one of those $15 keychain digicams than what you’d expect to find on a top-of-the-line digital SLR.

DPReview managed to get tremendously better battery life than Luminous Landscape did intheir A900 field review; LL was getting around 250 shots per charge, while DPReview says they were right in the ballpark of the CIPA standard-rated 880 shots that Sony’s specs claim. I’m not sure what would cause such a huge discrepancy, unless Luminous Landscape had a faulty battery, or using the battery in slightly colder outdoor temperatures (they reported using it between freezing and 65 degrees Fahrenheit) has a much larger impact than expected on battery life. My unfounded guess would be the former.

One extra-interesting element of their review is a look at the Image Data Converter SR 3.0 software that’s included with the camera for RAW conversion. This looks like one horrible piece of software. A quick glance at 100% crops of a resolution chart shot shows that in-camera JPEGs look much better than RAW files converted with the IDC software; their software results in tons of moire artifacts and blurring of the highest frequency lines on the test chart. The JPEGs show much milder moire artifacts. They also compare it with Adobe Camera Raw, which, as expected, beats IDC and JPEG at the resolution chart. Other tests show that IDC applies more sharpening than plain JPEGs or ACR, which generally isn’t what you want out of your RAW converter.

For high ISO noise, they find slightly higher noise levels in the A900 than in the cameras they compared it to (the Nikon D700, Canon 5D, and Canon 1Ds Mark III) up to ISO 400. At higher ISOs the difference becomes more marked, with Sony throwing more and more noise reduction at the image to try to reduce noise, sacrificing detail to the ISO Gods in the process. “By ISO 3200 the result is a blurry mess with little fine detail – with the added insult of visible chroma noise in the shadow areas. I think it’s fair to say that ISO 3200 and 6400 are firmly in the ‘emergency use only’ bracket (of course with 24MP to play with you shouldn’t have many problems at small print sizes).” They also point out that the A900 just has a very noisy sensor, and so the difference between NR turned off and NR turned to the lowest level is huge; there’s not nearly as much difference between low NR and high NR.

Thankfully, it looks like Sony hasn’t done the same thing that they did with the original A700 firmware, where they were doing noise reduction in RAW files to try to deal with how noisy the A700’s sensor was. They can’t be completely sure, but DPReview’s testing makes it look likely that the A900 gives you a true RAW file, at least from the standpoint of unwanted noise reduction.

Comparing noise to other cameras, in RAW mode:

Unsurprisingly the D700’s larger pixels give it a distinct advantage here, and show clearly that at higher ISO settings it’s not all about pixel counts — the ISO 6400 output retains noticeably more detail than the Alpha 900 with visibly lower noise. But it’s also worth noting that at ISO 1600 the three cameras are broadly comparable (Sony has more chroma noise but also the most resolution). At ISO 3200 Canon shows that its in-camera JPEG processing does a far better job at removing noise without destroying information (and that the actual noise levels in raw are lower than Sony’s).

And wow, 6400 ISO is really, really bad on the A900. It would be an interesting experiment to figure out how large you could print a 6400 ISO photo before the noise became visible. At some small photo size, the noise should be able to be averaged together with surrounding pixels so that you can’t see it. The question is, how small would it need to be? This should even be able to be simulated in Photoshop by shrinking the image until noise become invisible, but you’d have to work out the DPI and size in inches that give you a nice, sharp photo at that resolution. Maybe when I get a free hour I’ll play around with it.

DPReview joins the other reviews I’ve seen in stating that the A900 gives you an absolutely amazing dynamic range: “All the lab results tied in with our observations when using the A900 in the field; JPEG dynamic range (and particularly highlight range in the region over ‘mid gray’) is excellent, and there’s even more in the raw files if you need it. Impressive stuff.”

Overall, they really liked the image quality of the A900:

Overall there’s really little to complain about here – and much to like – and I’d only caution concern if you regularly intend to shoot at ISO 1600 or higher. 100% comparisons are a valid and useful way to check out the absolute limits of a camera’s resolving power, but in my book the superb tonality, appealing color and surprisingly reliable metering / focus systems count for a lot and mean the Alpha 900’s output at lower ISO settings is amongst the best we’ve seen.

Interestingly, they come to a different conclusion than the Luminous Landscape review. LL predicts that the A900 will put a huge dent in the Nikon D3x sales, and be one of the top cameras in its target market along with the Canon 5DMkII. DPReview, however, doesn’t think it’ll break out of the Konica/Minolta faithful:

In conclusion this is, more than anything else at this end of the market, a true photographer’s camera, with at least one totally unique feature (the Super SteadyShot stabilization) and one that offers the best viewfinder and highest nominal resolution (and the lowest ‘cost per megapixel, incidentally) in its class. It’s capable of stunning results at up to ISO 400 (and is fine at ISO 800-1600 as long as you’re not printing posters), and it is incredibly fast and responsive in use. If Sony had managed to keep the price nearer to the $2000 mark (even if this meant fewer megapixels) I think it would be flying off the shelves. As it stands it will, I fear, struggle to make a serious impression on anyone other than the Sony/Minolta faithful. One thing is clear, however: anyone who thinks a consumer electronics giant can’t make a heavyweight photographic tool is seriously misguided.

As long as you take into account our reservations about the high ISO image quality (which we’d more easily forgive on a camera that wasn’t the best part of $3000), the Alpha 900 is a camera that just, by the skin of its teeth, offers enough to gain our highest award.

It’ll be interesting to see what actually happens with this camera, which is certainly a unique and groundbreaking one for Sony.