Sony Alpha A900

Noise Reduction (6.68)

Noise reduction processing didn’t kick in until ISO 800, at which point it provided a substantial improvement over results with noise reduction turned off. The difference between the effects of the three settings, though, was quite small.

The noise trends among individual color channels and luma noise (the noise in gray areas of the photo) are consistent, though noise in the red channel is clearly more a problem for the A900, while yellow and green are better controlled. More on how we test noise.

The Sony A900 is a middle-of-the-pack performer when it comes to image noise. We found that with the noise reduction system turned off, the A900 started off well at ISO 200, with only the noise-busting Nikon D700 posting better figures, but as ISO settings increased, noise rose more quickly than the other cameras in our test group until, at ISO 3200, only the Canon 50D produced higher image noise. The noise reduction system is effective, keeping average color noise below 1% through ISO 1600 at maximum level, but not as effective as the system found in the other cameras: the A900 displayed the highest average color noise levels among the tested cameras across the ISO range when all were set to their maximum noise reduction settings.

Image noise is to digital photography what film grain was in the old days — spots and dots most noticeable in areas of flat color, caused in this case by random electrical activity in the image capture process. To test how well each camera keeps this phenomenon under control, we shoot the X-Rite color chart under bright 3000 lux illumination, at each official ISO setting (low-light noise performance is measured separately, along with low-light color reproduction, in our Long Exposure testing). When there are multiple high ISO noise reduction levels, we shoot at each. For the Sony A900, this means four sets of data, one with noise reduction off and then three at different noise reduction settings.

ISO Options

Available ISO settings range from ISO 100 to 6400, all at full resolution, a more modest range than some competing cameras. Both the Canon 5D Mark II and Nikon D700, for example, offer extended-range ISO settings out to 25,600.

Sony doesn’t follow the Canon / Nikon route of clearly labeling ISOs that produce less-than-optimal results with non-standard naming conventions (dubbing extended-range settings as L for Low or H for high and leaving ISO numbering off). Sony numbers the complete range but highlights ISOs below 200 and above 3200 with rules above and below in the menu screen. The manual says, ‘The available luminosity limits for an image (dynamic range) are a little narrower in the range less than ISO 200. When ISO 3200 or higher is selected, the range is treated as an expanded range and the noise is more noticeable.’ As shown in the actual-size image crops below, we don’t see a major drop-off in dynamic range below 200, but there’s no question image noise explodes when we pass the ISO 3200 mark, and has already jumped upward long before we get to that point.

When set to Auto ISO, you can set a limit to the acceptable ISO range. Five settings are available: 200-400, 200-800, 200-1600, 400-800 and 400-1600.

Focus Performance

The Sony A900 autofocus system uses 9 primary, user-selectable focus points plus 10 supplementary ranging points. Autofocus sensitivity ranges from 0 to 18 EV (at ISO 100). An autofocus illuminator is positioned between the grip and the camera lens. According to Sony, its working range is approximately 3.3 to 23 feet (1 to 7m).

There are three choices when selecting a focus area: Wide, Spot, and Local.

Long Exposure (10.68)

In our long exposure testing, which evaluates both color accuracy and image noise performance at shutter speeds from 1 second to 30 seconds, the Sony A900 scored very well, coming in second only to the Nikon D700 in our comparison group.The camera handled both tests with good results, maintaining a color error below 4 until the camera tipped into overexposure at the 30-second exposure, and average noise well below 1% across the board.

Dropping the lights in our lab down to a level found in a darkened room (20 lux, which is barely bright enough to read by), we set the camera to ISO 400 and shoot the X-Rite test chart at a range of exposure times, from 1 second to 30 seconds. These images are then analyzed using Imatest to measure image noise and color accuracy; for cameras offering long-exposure noise reduction, we run the test twice, with this function on and off.

Both color error and saturation changed very little between shutter speeds when shooting with the Sony A900 until we reached a 30-second exposure, when the shots were overexposed even with the aperture fully stopped down. More on how we test long exposure.

Long-exposure noise reduction seems like a great idea, but it is minimally effective in most of our tests, and sometimes actually makes matters worse. This was the case with the Sony A900; except at the slowest shutter speed, long exposure noise measured higher with noise reduction on, by as much as 20%. The explanation for this counterintutive result? Image noise is by its nature random. Long-exposure noise reduction systems attempt to recreate the noise in the original image by taking a second, shutter-closed exposure and then removing the noise found in the black exposure from the original. Problem is, removing random bits can quite logically do more harm than good.

As shown below, the Sony A900 and Nikon D700 produced substantially superior scores in this test, while the other full-frame camera in the field, the Canon 5D Mark II, scored slightly lower than even the APS-C sensor Canon 50D and Nikon D90.