Noise pattern

In this test we are pushing the camera near to its limits to examine the type of noise produced at high sensitivities and whether the noise is random or has any noticeable pattern to it. Pattern noise is far more objectionable than random noise because the human eye tends to pick it out more easily.

During our test of the Canon EOS-1D we noted horizontal pattern noise combined with two different 'sides' to the image, this an effect of the camera's two readout buffers (one on the left of the sensor, one on the right). While not visible in all shots it may be noticeable in any high ISO image with a dark background.

We decided to look at this potential issue with the D2H, the images below were taken deliberately underexposed by two stops (-2.0 EV) they were then loaded into Photoshop and level adjusted to be four times brighter (equiv. of +2.0 EV). This was done to make any background noise more visible. These adjusted images were then resized and are shown below. The original unchanged images are linked to these images.

My overall feeling is that there is some faint horizontal banding in pushed high ISO images from the D2H but that this wouldn't be visible in 90% of everyday shots. Additionally the noise pattern is more grainy than the EOS-1D and so looks more like film grain.

ISO 800 (pushed to ISO 3200 equiv.)

As you can see at ISO 800 (pushed to ISO 3200 equiv.) the D2H does appear to exhibit some horizontal banding pattern although it is not as strong as the EOS-1D and does appear to be more 'grainy'. Vertical banding isn't noticeable here, but certainly is on the EOS-1D image.

Nikon D2H @ ISO 800, levels adjusted to 4x brighter (ISO 3200 equiv.)
Canon EOS-1D @ ISO 800, levels adjusted to 4x brighter (ISO 3200 equiv.)

ISO 1600 (pushed to ISO 6400 equiv.)

At ISO 1600 (pushed to ISO 6400 equiv.) a couple of faint vertical bands do begin to appear but again the actual pattern isn't as noticeable as on the EOS-1D.

Nikon D2H @ ISO 1600, levels adjusted to 4x brighter (ISO 6400 equiv.)
Canon EOS-1D @ ISO 1600, levels adjusted to 4x brighter (ISO 6400 equiv.)

White Balance

I had high hopes for the D2H's automatic white balance, the new external white balance sensor does take a rather prominent position on the front of the camera and Nikon's literature about the camera talks a lot about the fact that the camera uses three difference data sources to calculate white balance (the external sensor, the metering CCD and the imaging sensor). So I was obviously disappointed to see that although the camera appears to be getting close to the correct balance (it's in the ballpark) it still exhibits a pink cast in incandescent light and a yellowish or green cast in fluorescent light. I had expected better.

Settings: Sharpening: Auto, Tone: Normal, sRGB (I), ISO 200, 50 mm F1.4 D, Medium / Fine JPEG

Outdoors, Auto Outdoors, Cloudy (or Sunny) Outdoors, Manual
Incandescent, Auto Incandescent, Incandescent Incandescent, Manual
Fluorescent, Auto Fluorescent, Fluorescent Fluorescent, Manual

White Balance Fine Tuning

The D2H provides the ability to fine tune each of its preset white balance settings by an arbitrary value -3 to +3. As you can see from the samples below -3 produces a warmer white balance (less blue), +3 produces a cooler white balance (less red).

Settings: Sharpening: Auto, Tone: Normal, sRGB (I), ISO 200, 50 mm F1.4 D, Medium / Fine JPEG

Incandescent -3 Incandescent +0 Incandescent +3


For this test we used Nikon's SB-50DX Speedlight, unfortunately we didn't have the newer SB-800 to test at this stage. With the Speedlight set to 'D-TTL auto flash' we got good results with direct flash shots and some slight underexposure bouncing the flash off the ceiling (this could of course be fixed with a little flash exposure compensation).

Settings: Sharpening: Auto, Tone: Normal, sRGB (I), ISO 200, 28-70 mm F2.8 @ 50 mm, SB-50DX Speedlight

SB-50DX direct - good results SB-50DX bounced - slightly underexposed
Color patches - even illumination, well metered, accurate white balance, good color response  

Night exposures

The D2H has an optional dark frame subtraction noise reduction feature for long exposures. The shots below are fairly typical long exposure night shots, 15 and 30 seconds. Close examination of the images with noise reduction off shows no hot pixels at all and very low noise characteristics, enabling noise reduction appears to improve contrast slightly but beyond that doesn't appear to be required at these shutter speeds. I have deliberately brightened the 100% crops below (by adjusting their gamma) to emphasize the fact that noise even in virtually black sky is non existent.

Settings: Sharpening: Auto, Tone: Auto, sRGB (I), ISO 200, 17-35 mm F2.8 D, Large / Fine JPEG

Noise Reduction off Noise Reduction on
ISO 200, 15 sec, F11 (two 100% crops are gamma adjusted)
ISO 200, 30 sec, F16 (two 100% crops are gamma adjusted)