Canon EOS 40D Dark Noise Tests

This page allows you to view raw histograms of dark frames shot at 1/100s with all possible ISO settings of the Canon EOS 40D. Move your mouse over the list of ISO sensitivities on the left to view the corresponding histogram. You can get the original file (csv) with the histogram data by clicking the link. If you want to know how these were made see the original rawhistogram page.

The intermediate (1/3) ISO settings of the 40D are "fake" just like in the 30D, and as far as I am aware all other Canon DSLRs that have them, apart from the EOS 5D and the EOS 1D. There are two groups of intermediate ISO settings, the "125, 250, 500, 1000" group is derived by multiplying (upscaling) the previous whole ISO and the "160, 320, 640, 1250" group is derived by dividing (downscaling) the next whole ISO. This is clearly visible from the graphs - since the data in the first group is in effect "stretched" the histograms exhibit gaps. The histograms in the second group are "compressed" and therefore have "spikes". If you compare this to the EOS 30D test you will observe the same effect. Because of this "stretching" and "compressing" ISOs like 160 appear very good in the following noise graph and 125 and the like appear bad. This is however deceptive - it is a purely mathematical manipulation of the data and therefore when shooting raw is useless - the same effect can be achieved by shooting at whole ISOs and pushing or pulling the image during raw conversion. When shooting JPEG on the other hand the intermediate settings can be useful.

The graph below represents the signal/noise ratio as a function of ISO speed. If you prefer dB units to "counts" move your mouse over the image. Since this is a ratio of maximum signal to dark noise it essentially represents the dynamic range of the camera. Keep in mind that there are other sources of noise in addition to the dark noise measured here. A Google query should give you some quite good introductory information.

To put the 40D performance in perspective:

It is significantly better than the 30D for all ISO speeds up to ISO 500, above that the 30D is better.
It is worse than the 1D Mark III for all real ISO speeds. In fact:
- ISO 800 on the 1D Mark III is slightly better than ISO 400 on the 40D.
- ISO 400 on the 1D Mark III is almost as good as ISO 100 on the 40D.

The values were calculated as follows:


SNcounts = (2^14 - Mean)/Sd

SNdB = 20 * log(SNcounts)

Where "Mean" is the mean of the histogram (very close to 1024 for all ISOs) and "Sd" is the standard deviation - which is what increases with noise.