Sony A900: A Closer Look at 24.6MP Resolution and Noise
by Wesley Fink on October 27, 2008 2:00 AM EST- Posted in
- Digital Camera
Looking Forward
The Sony A900 is a truly revolutionary camera. As the highest resolution full-frame DSLR you can currently buy it distinguishes itself with astonishing resolution. As discussed in our preview it is also the first full-frame with body-integrated Image Stabilization (add 2 to 4 stops to hand held shooting) and the viewfinder is the best, brightest, sharpest that we have ever seen in a DSLR - or perhaps in any camera.
It is important to keep those huge pluses in perspective because image noise at extended ISOs, while just one factor in the total image quality equation, is definitely not a strength of the A900. Compared to the Nikon D700, noise in the A900 is about two stops worse. That means the D700 at ISO 6400 exhibits about the same noise as the A900 at ISO 1600. In the case of the D700 that applies to both the actual pixel comparisons as well as the double pixel samples downsized to reflect the same image view in both crops. Despite the huge noise advantage, the D700 outputs nothing near the resolution and detail of the A900, which should not come as a surprise.
The Canon 5D is now three years old, but noise control is still about a stop or two better than the A900. However, we noticed for the first time how much Canon softens images at higher ISOs in an effort to control noise. Looking at the pixel level, high ISO Canon images are extremely soft at higher ISO settings, though they are quite sharp in the lower ISOs most sites use in evaluating the 5D output. Taken in total the A900 is again a clear leader in detail compared to the 5D. The 5D has perhaps a one stop advantage in noise when the A900 and 5D are equalized for sharpness.
Finally we come to the biggest surprise of all, which is our comparison of the A900 noise to the APS-C sensor in the A700. This is the same sensor used in the Nikon D300 and the most recent Nikon D90. When the A700 was first released, some in the press trashed it because of its heavy-handed noise reduction techniques. Sony has worked hard to answer those complaints, and most users were pleased with the v3 firmware fixes. With the release of the A900 Sony also released a v4 firmware for the A700, which allows complete disabling of NR plus a range of adjustments in noise reduction. V4 incorporates in the A700 all Sony had learned in developing the A900.
With sensor density of the A900 less than the A900 (2.9 vs. 3.3) we really thought the A900 would shine compared to the A700. In fact the A700 is about two stops better in noise control looking at actual pixel crops and one to two stops better looking at crops equalized for image view (A900 down-sampled to equivalent 12MP). This performance difference is a complete surprise and it is either good news or bad news depending on your perspective. The bad news version is that the A900 sensor is inherently a high-noise sensor and the high ISO noise performance is as good as we will see. The good news version is that the A900 was just released and, like the A700, Sony will continue to improve the noise performance in future firmware/software releases. We certainly saw that in the A700, but we did have the Nikon D300 as the constant reminder of what that sensor was capable of.
Our hands-on test of the A900 reveals a mixed bag of performance. At Lower ISOs, 100-400, nothing on the market comes close to the Sony. A little is given up at 800-1600 and if that were as bad as it got the A900 would be something of a Holy Grail. However, noise at ISO 3200 and 6400 are truly not competitive. High detail is still there but it is seriously marred by high noise. In the end our expectation is that professionals will stick to ISO 100-400 for the highest resolution shots you can get from a camera in this class. Action shooters and the rest of us will also be happy with ISO 800 and 1600. However, dial in above ISO 1600 only if you will be happy with stunning detail in smaller prints.
The last page contains a few sample images taken in the couple of weeks the Sony A900 has been in the office. As always, we chose the images to show both the strengths and weaknesses of the camera we are testing. The A900 is capable of capturing amazing detail and if you do a lot of pixel-peeping on these images you will see exactly what we mean. There are also a couple of shots that show the potential impact of poor noise performance at higher ISO, which is not always as bad as it sounds.
The Sony A900 will likely be long remembered for its record-setting resolution and its utility as a camera that is a true working tool instead of an amalgam of gadgets. The stunning 100% viewfinder is one example of that, but so is the logical and simple control that is a part of every aspect of the A900. It is not without its flaws but in total the A900 is definitely greater than the sum of its parts or any one area of measured performance.
Facebook
Twitter
RSS
45 Comments
View All Comments
kalow - Monday, November 17, 2008 - link
Hi Wesley,Can you confirm if your a700 has firmware 4? I looked into the exif and it says v03
Gantlett - Friday, October 31, 2008 - link
The photographed subjects in the photo reviews pose 0 challenge for my VGA camera, let alone a multi-megapixel multi-thousand $ professional DSLR.With all due respect, I think AnandTech.com should focus on computers and not photography (even if they really like digital photography...)
oldscotch - Wednesday, October 29, 2008 - link
Hi there, thanks for the update. Nice to see informed and unbiased evaluations of the s900.I just had a couple of questions about the test conditions, or maybe I missed them? Mostly, are these test results raw conversions? or are they jpegs out of the camera? Also curious what lenses were used for each camera.
I find the a900 vs. a700 test most interesting, I've been looking for a controlled comparison of the two at higher isos. The fact that the a700 actually takes cleaner shots at higher isos leaves me believing that the noise out of the a900 is largely due to its processing - meaning it can be improved upon (and there's definitely room for improvement, even with raw).
Do you know what might have caused the different colours on the a700 though?
Thanks again,
Wesley Fink - Wednesday, October 29, 2008 - link
All test shots used each manufacturer's prime 50mm f/1.4 lens (Canon, Nikon, Sony/Minolta). Test conditions are in the test introduction, but all shots at at f/4 with shutter speed varied as ISO changes. This provides a constant depth-of-field. The tripod location was the same in all shots and a shutter timer was used to minimize camera shake.As for the samples images on the last page we used the famous Minolta beer can (70-210mm f/4), a Sigma 105mm f/2.8 EX Macro, a Minolta/Sony 24-105mm f/3.5-4.5, and a 50mm f/1.4 for the various shots. You can determine the lens from this list by looking at the EXIF data in each full-sized image.
We used the Tungsten WB preset for the test shots on all tested cameras. It seems clear that the A900 and A700 have a different color temperature default for Tungsten. ALL shots used camera defaults and in-camera JPEG processing. RAW is very useful in many situations, but with a new camera like the A900 which RAW converter is the best choice for processing is very much a question mark. A couple of PROs we know who are shooting the A900 in their studio use Capture One RAW conversion software.
jamesbond007 - Wednesday, October 29, 2008 - link
As a semi-pro, I find Sony's images to be very soft. For a lens with an aperture capable of f/1.4, one would at least HOPE it would be sharp by f/4 or especially f/5.6. It is appropriate to offer a client a variety of images, including various crops. I would be ashamed if a $3,000 body and added glass could not deliver such capable images. Additionally, Sony's noise performance leaves much to be desired.Granted, while lighting is by far the most important element of photography, sharpness/AF accuracy is among my top priorities as well. Although I am sure no Unsharpen Mask or High-pass sharpening was applied to any of the sample images, Sony's still appear far softer than what I'm accustomed to.
Wesley Fink - Wednesday, October 29, 2008 - link
The results would also likely improve in sharpness if we had used the microadjust feature to tweak AF for each lens. I have seen results of the 70-210mm f/4 before and after microadjust and it certainly did improve performance, so that may also be the case with the f/1.4.We have been working with the A900 just over a week and tweaking the microadjust can be very time-consuming with lots of samples to carefully evaluate. Sony makes this even more difficult since there is no zoom or magnify feature in Intelligent preview. That feature alone would have made the micro-adjust faster.
jamesbond007 - Thursday, October 30, 2008 - link
Really, even at f/4? I've seen some slight variations with faster lenses (f/1.2-f/2.8) using the microadjustment tweaks inside the custom settings, but I really have my doubts at f/4 or even f/5.6 if a noticeable difference would be visible. Perhaps the lens you had was simply a lemon. :Pstrikeback03 - Thursday, October 30, 2008 - link
Remember, the full-frame sensor has a more shallow DOF at a given aperture value than APS-C, so if you are used to seeing shifts at f/2.8 on APS-C, that is ballpark what you can expect on FF at f/4. Check the DPreview images from the Sony and Canon 1DsIII, both sensors are high enough resolution to show parts of their test scene moving in and out of the plane of focus which all looked in focus in previous tests of lower-resolution sensors.forest23 - Friday, August 14, 2009 - link
Oh no , not someone else who believes that putting a lens ona half frame camera gives greater depth of field. But see Coxes Optics
1974 pages 68-97
Wesley Fink - Thursday, October 30, 2008 - link
The Minolta 50mm f/1.4 is the same lens used in noise tests of the Sony A700 and the past review of the Sony A200. We have 50mm f/1.4 lenses available for all tested brands except Olympus/Panasonic. Because of the 2X lens factor we use an Olympus 35mm Macro lens as the test lens for 4/3 cameras.