Just how good is your camera sensor?

[synn]

We all know that Lenses can make even the most ordinary body take breathtaking shots. But have you ever really wondered just how good (Or bad) the sensor is? I came across this website (That some of you may have already seen) called DXO Mark (http://www.dxomark.com/ ) That does objective benchmark tests for RAW outputs of all the bodies available in the market. Every camera available has been ranked acording to sensor performance (http://www.dxomark.com/index.php/eng/DxOMark-Sensor/Camera-rankings ) and each review provides some valuable insights.

Apart from that, the site has some good articles too. If nothing, it's a very interesting read, I feel.

[KrishnenduKes]

Discussion Approved

[Psycho_McCrazy]

I have also visited that site, and while it does provide some insight into the base performance of the sensor, I feel that it is not too useful a comparison site for cameras, the reasons being:

1)the different sensors of modern cameras of the same category are more or less clubbed together.

2)the image that a sensor takes has to be translated to the visible format - there are so many other factors involved (camera internal JPEG engine/RAW conversion software, resultant file format and compression level, display setting on the monitor, capability of the monitor to display properly, color calibration etc.; and if printed the quality of the printer, the paper used, the color calibration of the printer etc.)

3)There are a lot more things to what makes a camera or a camera system than the seneor - and honestly, in the good old days, the sensor wasn't even a factor - the sensor was the 36 exposure roll that you bought separately - you bought cameras for actual camera features. AF system, lens availability, ergonomics, features are more important. All decent sized (read APS-C, APS-H and 35mm) sensors will give results identical for all practical purposes results when shot at the reasonable ISO(100-800) so long as they are placed in a well functional body, has some good glass in front of it, and someone who knows what he is doing when behind the eyepiece.

4)The DXO Mark forgets that the most important part of the camera is the meter, because a good sensor can't do squat if the meter messes up the exposure.....

and even if you are referring to the site for sensor purposes, look at the difference between the a900 and the D3x, for a 10 point difference despite what is essentially the same sensor, abeit with a few minor tweaks and optimisations. These minor tweaks are not alone responsible for this gap, the AA optical filter, the processing pipeline (even RAW data is processed before writing to the file) are big factors, and also how can we know that the DXOMark software can handle the proprietary RAW data formats of different companies equally?

[synn]

They don't deny all that. The first thing they mention is that these benchmarks are not to be taken as a measurement of the performance of the CAMERA, but solely of the sensor. In fact, all your points are answered more or less satisfactorily in their FAQ section.

The D3x vs A900 is interesting, because i remember reading that Sony is keeping the A900 sensor for itself and the D3X uses a different version.

Also, DXO software is used by all the major camera manufacturers for their internal testing, so i guess this is as accurate as it comes.

[Psycho_McCrazy]

The D3x vs A900 is interesting, because i remember reading that Sony is keeping the A900 sensor for itself and the D3X uses a different version.
While there are differences, there is no denying the fact that the base silicon is the same, and Nikon has done some real good work to get the performance out of it.

What I am surprised about is the D300 to D90 gap in sensor score (esp. High ISO). maybe it has to do something with the different AA filter used on the D90, that while reduces some detail might be allowing more light onto the sensor.
however, by looking at the actual images from the cameras, there is not so significant difference in the ISO performance of these two as the score discrepancy suggests.

[Daniel Browning]

I came across this website (That some of you may have already seen) called DXO Mark

I appreciate the good parts of having a resource like DXOMark. I only wish they would publish the raw files used for their data. That would be far more valuable to me. Some of their analysis is flawed (such as failing to account for pattern noise) Thankfully, it is possible to reverse engineer some of their underlying data and some photographers have done that.