Purple Fringing (Chromatic Aberrations)

It's a huge pity but the DSC-F828 has a fringing problem. We'll call it 'purple fringing' but it's clearly a combination of chromatic aberrations caused by the interaction of the lens and the microlenses on the sensor and enhanced by a certain amount of blooming which carries the color out further than the original artifact.

While worse at full wide angle and maximum aperture (28 mm equiv., F2.0) it is visible throughout the zoom range and at smaller apertures. It's clearly possible to take images which will never suffer from this effect, it does require either a strong backlight or sunlight reflection to become apparent. The issue really is that in a normal shooting situation you would have no real idea if the image had suffered until you return to your computer and view the image at full size.

It's also possible to reduce the visiblity of these artifacts with careful post-processing, using specific CA reduction tools or simple color replacement filters. However there is no getting away from the fact that these artifacts should not be there.

Image thumbnail 100% crop
28 mm equiv, F2.0 (maximum aperture)
28 mm equiv, F2.5
28 mm equiv, F2.8
200 mm equiv, F4.0

Foil "torture test" - wide angle (28 mm equiv.)


Foil "torture test" - mid-zoom (80 mm equiv.)


Overall Image Quality / Specific Issues

There has been a huge amount of debate around the DSC-F828's image quality. Iniital results have certainly been mixed, with some excellent images exhibitting superb resolution to others with strong chromatic aberrations or visible noise. The DSC-F828 is certainly capable of delivering considerable resolution, much more than we have seen from any previous prosumer digital camera. And in ideal circumstances the camera will produce superb results free from any artifacts.

However it's clear that Sony are pushing the boundaries of what is currently possible with the eight megapixel sensor and so a certain realistic expectation existed on that front. However hopes were higher for the 'Carl Zeiss T*' lens, hence the disappointment on the CA front. Is this a good example of marketing ahead of quality?

The DSC-F717 was a class leader, and continued to be for much longer than most prosumer digital cameras. The DSC-F828 is a first-of-a-kind and appears to be suffering from those 'version 1' problems which may well leave it as second best to any new eight megapixel digital cameras coming up in 2004.

Green hue shift at over-exposure

One thing the RGBE sensor does compared to the RGB sensor the DSC-F717 is improve color reproduction, notably in blues and yellows. This can be seen on previous pages in this review and is certainly welcome. That's not to say the RGBE sensor hasn't introduced a strange issue of its own.

Quite early on I noticed that the tip of the green crayon in our standard 'crayons' shot had a strange cyan tint to it, as you can see below.

Correct color Highlight areas have cyan color shift

I decided to explore this phenomem a little further by shooting a standard Gretag ColorChecker chart at progressively higher exposures, from the metered exposure (+0 EV) to three stops over the metered exposure (+3 EV). The color value for six patches was measured and can be seen below. First column for each color is DSC-F828, second column is DSC-F717.

As you can see the F717's colors simply increase in luminosity and eventually become white. The DSC-F828 however exhibits a very different behaviour for green, cyan and to a lesser extend red. Most noticeably green becomes bright cyan at +3 EV.

Exposure Neutral 5 Red Green Blue Yellow Cyan
F828 F717 F828 F717 F828 F717 F828 F717 F828 F717 F828 F717
+0 EV                        
+1 EV                        
+2 EV                        
+3 EV                        


To be fair to the DSC-F828 this was the only real life sample we could find / produce in all of the several hundred sample images we shot. After taking the first image I noticed that the background color on the road sign wasn't correct, knowing about the hue shift problem I applied a -1.0 EV exposure compensation and took the shot again. As you can see, the second image is actually the correct color.

1/60 sec, F2.8 (metered) 1/60 sec, F4.0 (-1.0 EV compensation)