D90 high ISO advantage is real compared to D300 40D and 50D

Started Nov 19, 2008 | Discussions thread
Flat view
OleThorsen Veteran Member • Posts: 3,111
D90 high ISO advantage is real compared to D300 40D and 50D

As a follow up on the discussions on this forum about the D90 sensor compared to the D300 sensor, and the latest sensor data from DxOmark, here is my contribution comparing D90 D300 50D and 40D RAWs using RawTherapee, a high quality free RAW converter. Both CaptureNX, DPP and to a lesser degree ACR applies some NR and exposure compensation at 0 settings without user control. In my opinion comparing cameras must be as neutral as possible, otherwise we are just comparing RAW converters and not the cameras.

The RAW samples are downloaded from Imaging Resource's (IR) Still Life test

Lenses and files used:
D90: Nikon AF 105mm f2.8D micro, @f8 12bit NEF
D300: Sigma AF 70mm f2.8 EX DG macro, @f8 14bit NEF
50D: Sigma AF 70mm f2.8 EX DG macro, @f8 14bit CR2
40D: Canon EF 100mm f2.8 macro, @f8 14bit CR2

RAW conversion:

I used RawTherapee v2.4beta1 (dcraw 8.88 build), EAHD demosaicing algorithm processed to 16bit sRGB TIF,

neutral profile, no contrast NR and sharpening, In-camera WB (custom made by IR), Exposure compensation between +1.10 to +1.30EV with all cameras to give colour checker's white patch a RGB value of 220. Presentation made in Photoshop CS3 with no processing except 100% cropping and conversion to 8bit JPEG for web presentation.

I added Photoshop pixel statistics standard deviation values (red numbers) for the colour checker's black patch to give some impression of the noise levels.


When IR tested the 40D the Proportional scale (at right) was nailed to the wall, while it was raised from the wall to the same focus plane as the bottles when they tested 50D D300 and D90. Even though the DOF should cover the whole scene, it could cause a little less sharp Proportional scale on the 40D crops.

As I described in my above I used RawTherapee's EAHD demosaicing algorithm processed to 16bit sRGB TIF. This algoritm creates some cross hatch pattern or needle artefacts in noisy areas and a little more visible red noise, mostly visible if you zoom to 200%+. I may try to use the HPHD algoritm in RawTherapee in a new comparison, since I have discovered it creates a little less visible artefacts than EAHD (almost not visible at 100% crops but only at higher zoomlevels), but it will be very similar for all cameras. I'm sure there is no NR applied at all, which should be very visible at the higher ISOs, which shows a lot of luminance and chrominance noise and details.

Comparison ISO100 to ISO6400 (5 x 100% crops per ISO) (click on picture to expand)

My evaluation:

First of all I want to say I would be very satisfied owning any of the tested DSLRs from an image quality point of view. I have the D300 myself and haven't used any of the other cameras, but I have tried to be as neutral as possible in my evaluation.

Resolution: 40D has clearly a little less resolution than the other cameras at all ISO values, which is really not surprising having a 10MP sensor. The difference between D90 D300 and 50D is very small.

Sharpness: They are all very close and any differences could just as well be caused by different lenses. But if I should rank the cameras with the camera with less pixel sharpness first it would be 40D 50D D90/D300.

Noise: Remember the RAWs were underexposed from 1.1 to 1.3EV. This is typical using in camera converted JPEG histograms to evaluate exposure. 50D and D300 has very similar noise levels at the most ISO values, while 40D has a little less noise than 50D and D300. But the difference is really minor. To my surprise it seems like the D90 has the best noise performance of all the cameras in this evaluation, around 1/2 to 2/3 stop advantage. Even though it's almost invisible in these crops, both 40D and 50D shows some linear noise patterns in dark areas (not showed here) from ISO1600 and up, but it is most visible at ISO3200 and ISO6400 (50D).


Imaging Resource has an advanced laboratory-grade HMI lighting system in their studio, which is used for i.e. the Still Life test I used for this comparison, so I'm pretty confident that the lighting conditions are as equal as possible in the test of the cameras in this comparison. All cameras in this comparison are tested with the new studio light implemented in 2006. File names with a lower-case h after the camera name is using the new HMI light. See this interesting article at IR describing their studio including pictures.

Regarding real RAW exposure, here is Rawnalyze screen dumps for the ISO800 RAW files of the compared cameras. As you can observe both camera settings (aperture/shutterspeed) and the RAW histograms (before WB and demosaicing) indicate that the cameras are very close in exposures. Judging from the green channel in the histograms all cameras are underexposed between -1.1 to -1.3EV from saturation, which was the exact Exposure compensation values I used converting the RAWs with RawTherapee to get identical RGB 220 values in the colour checker's white patch, as I described above.

Ole Thorsen


Knowing what
thou knowest not
is in a sense
(Grook by Piet Hein)

 OleThorsen's gear list:OleThorsen's gear list
Nikon D70 Nikon D200 Nikon D300 Nikon 1 V1 Nikon 1 V3 +32 more
Flat view
Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow