After various requests, I have re-done the non-technical noise comparions once more (I'll be doing technical noise measurements with DXO Analyzer later).
This time I have carefully equalized exposure and white balance (though this ignores the fact that there are wide exposure discrepancies between the cameras). I have also re-checked focus and re-shot where necessary. As before, I have posted 1:1 (1 image pixel to 1 screen pixel) crops from the centre of the frame.
The ISO values tested now include 400, 800 and 1600. The sharpening I originally applied has been moderated from 50% to 30%. People have asked me why I have applied sharpening and it's simply to reflect the fact that these cameras, in general, do need sharpening up a little and it does make any grain present show up better. But the sharpening rate applied is the same for all the images so the relative effect is constant. I have also not used RAW because in-camera JPEG performance is very important to these cameras.
Canon EOS-350D (Digital Rebel XT) The 350D is remarkably noise-free, even at 1600 ISO, considering it has an 8MP sensor is smaller than Sony 6MP sensor used in most of the other cameras. Only a bit of luminance noise shows up at 1600 ISO, but there is a sharp reduction in detail at this level. Despite calibrating the white balance, the result was a bit cold and lacking in saturation.
ISO 400:
ISO 800:
ISO 1600:
Konica Minolta Dynax 5D (Maxxum 5D) The Konica Minolta 5D did a good job of limiting noise, but sharpness was lacking even at ISO 400. Chrominance noise is almost enbtirely absent, even at ISO 1600. Without compensation, the 5D produced very dark images. The colour balance, despite white balance calibration is rather too yellow.
ISO 400:
ISO 800:
ISO 1600:
Nikon D50 Without doubt, the D50 delivered the best overall performance in this test. Although sharpness does get eroded as you climb the ISO scale, it's a moderate decline and the ISO 1600 result is still very acceptable. The images were very well saturated, if too warm, but on balance this is preferable to being too cold.
ISO 400:
ISO 800:
ISO 1600:
Olympus E-500 (EVolt) The E-500 at ISO 400 is very usable, but at ISO 800 luminance noise has started to show and there is a drop in sharpness. The E-500's ISO 800 performance is similar to ISO 1600 with some ofthe competition. At ISO 1600, some unpleasant chrominance noise has also made an appearance, making this ISO level best avoided for all but the most pressing uses. With calibrated white balance, the E-500 delivered the most faithful colours.
ISO 400:
ISO 800:
ISO 1600:
Pentax *istDL The *istDL didn't exhibit much noise, but this appears to be explained by very limited image sharpness compared to the others, even at ISO 400. The test was repeated 3 times for fear of a focus error, but this has been ruled out. Also, there is a lack of blue and surplus of yellow in the images. Like the Konica Minolta 5D, considerable exposure compensation was required to prevent the images from being very dark.
ISO 400:
ISO 800:
ISO 1600: