Moon Maid's sensor comparison

Started Feb 29, 2012 | Discussions thread
Shop cameras & lenses ▾
Flat view
Marianne Oelund Veteran Member • Posts: 6,596
Moon Maid's sensor comparison

Alright, I'll admit it: Every time I see someone claim that camera A is "X stops" better than camera B for high-ISO use, it makes me cringe. Generally, it will fail to be true, and there isn't a single number that can characterize sensor performance - or the difference between sensors. In fact, one of the points of this exercise is to demonstrate how close the current-generation cameras are, at least across the mid- to light-tone range.

Here, I would like to share with you how my engineering mind visualizes sensor performance with respect to noise, considering a few cameras which are presently quite topical. These are SNR (signal-to-noise ratio) curves at some selected ISO settings, which give an assessment of freedom from noise over a wide tonal range.

The vertical scale is in dB. At 40dB, SNR is 100:1 and noise is barely perceptible. At 30dB, noise is fairly visible at large print size, but would not be considered objectionable by most users. At 20dB, SNR is only 10:1 and you will want to use noise-reduction software to obtain a quality print, and at 10dB, results may still be considered only fair after careful NR. The horizontal scale is EV, corresponding approximately to what your camera's exposure meter would indicate on a blank wall; sensor saturation is typically just above +3EV.

Since this includes cameras which are not yet available in production form, please keep in mind that the data is preliminary and subject to change. However, the analysis was performed with great care, and I am not expecting significant changes.

The curves represent only shot noise (photon noise) and read noise. Pattern noise, which is only of practical import across the upper part of the tonal range at low ISO settings, has not been included. This is partly due to the difficulty of making the measurement - the sensor must be extremely clean and my cameras are rarely in this condition.

Often, cameras are compared on a full-image basis, i.e., as though they would all be used to make the same size prints. To support this particular point of view, 12Mp-normalized curves are included. Personally, I also believe that per-pixel performance is a valid comparison approach for certain output sizes, so those curves are included as well. You may interpolate between the two curve types to find how the camera would perform at various downsampling ratios.

One final point I would like to emphasize: Recent marketing claims regarding high-ISO improvements are referring to the camera's JPEG output, not the basic sensor performance, and rely on camera processing throughput and NR algorithm improvements. This is great news for JPEG shooters, but such claims have little bearing on post-processed results from RAW files.

Canon EOS 10D Canon EOS 20D Canon EOS 30D Canon EOS 40D
If you believe there are incorrect tags, please send us this post using our feedback form.
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