An in-depth case study of the use of DxOMark dataTuesday June 09 2009 Sensor Insight
Results below are in print mode to provide a fair comparison of the two sensors.
The difference between these two cameras on the Sensor Overall Score is not explained by their RAW noise measurements. Slight advantage for Nikon in low ISO, slight advantage for Canon in high ISO.
1. Pixel quality — screen modeAs the amount of photonic noise depends on the surface area of the pixel, smaller pixels will theoretically have lower SNR than larger. To compare the pixel quality of the two sensors available on the DxOMark Database, pixel pitch has to be taken into account, as the variance of the photonic noise will be proportional to the area of the pixel. So between two pixels with different pitches, (,) their respective SNRs () have the following relationship:
So the theoretical difference of SNR between two pixels of the same sensitivity and different size (5.5 µm and 4.7 µm):
Theoretically, as the Canon’s pixel surface is 1.37 times smaller than the Nikon’s, the difference between the two SNR should be 1.36 dB.
In order to compare fairly each pixel, the SNR 18% is computed for each real ISO in the chart below. SNR 18% values are deduced from the SNR 18% (mode screen) values, as only the behavior of the pixel has been evaluated.
|SNR 18% Nikon D5000 (dB)||37.23||35.37||32.42||29.04||25.70||22.03||16.04|
|SNR 18% Canon 500D (dB)||35.97||33.46||30.59||27.69||24.65||21.43||16.79|
|Theoretical Delta (dB)||1.36||1.36||1.36||1.36||1.36||1.36||1.36|
|Measured Delta (dB)||1.26||1.91||1.83||1.35||1.05||0.59||-0.75|
If Measured delta is less than Theoretical delta, the Canon pixel has a better SNR; if Measured delta is greater than Theoretical delta, the Nikon pixel has a better SNR.
As a difference of 3 dB represents a difference of 1 stop, we can see here that the maximum difference observed is only 2/ 3 of a stop, and globally inferior to 1/6 of a stop. Except for ISO 6400, the gap is very close between the Nikon and Canon pixels. (At ISO 6400, the Canon’s pixel performs better by 2/3 of a stop.)
2. Global sensor quality — print mode
To fairly evaluate and compare two sensors with different resolutions, we have to normalize the sensors so that both have the same resolution, and so compare SNR 18% after normalization (see more about DxOMark Sensor normalization).
Larger sensor surfaces will capture more photons than smaller surfaces and should provide the better SNR. As for pixel quality evaluation, it is also possible to evaluate the theoretical difference of SNR due to sensor surface differences.
The relation between two SNRs 18% for a sensor 1 with a surface area of and sensor 2 with an area of is :
So the theoretical difference of SNR between two sensors of the same sensitivity but different sizes is equal to 372.88 mm² and 332.27 mm²:
To compare global sensor quality, we used SNR 18% (print mode) and defined a new delta with the same method as described above. The only change is that the ratio is now computed for the entire sensor surface, with the results obtained shown below:
|SNR 18% Nikon D5000 (dB)||39.12||37.26||34.31||30.93||27.59||23.92||17.93|
|SNR 18% Canon 500D (dB)||38.71||36.20||33.33||30.44||27.39||24.17||19.53|
|Theoretical Delta (dB)||0.50||0.50||0.50||0.50||0.50||0.50||0.50|
|Measured Delta (dB)||0.41||1.06||0.98||0.49||0.20||-0.26||-1.60|
If Measured delta is less than Theoretical delta, the Canon sensor has a better SNR; if Measured delta is greater than Theoretical delta, the Nikon sensor has a better SNR.
The Canon EOS 500D sensor is a little smaller, with a ratio between the two sensor surfaces of 1.12. Consequently, the Canon should have an SNR 0.5dB smaller than the Nikon D5000.
As previously observed for the pixel quality analysis, the global sensor quality of the Canon EOS 500D is quite good with respect to its SNR results, especially at high ISO settings.