The following graphs show how much noise there is in the images. It is performed by photographing a gray scale step chart, which has gray rectangular patches going from white to black, with various gray levels in between. I used an ISO of 100 for all the tests.

First, the Sigma SD14. The top graph shows the average density of the grayscale patches (black line) and the first order (blue dotted line) and second order (green dotted line) density fits. In this case, the average aligns with the second order fit, which means that the camera rolls off its response to bright regions in a scene so as not to blow out the highlights. The other two cameras here (Canon and Samsung) also do this, as do consumer digital cameras in general. You can see that the Samsung rolls off its response just a bit more than the Sigma or Canon.

In the bottom graph the highest noise is in the blue channel, ranging from 0.6% to 0.93%. (% noise is shown on the Y axis, gray level zones are on the X axis, with zone 1 being the lightest and higher zones being darker.)

For the Canon EOS-5D, noise was highest in the blue channel for most of the gray levels, and part of the time, it was highest in the red channel. Noise ranged from a low of 0.4% to a high of 0.87%. This is slightly better than the Sigma SD14, and predictable based on the fact that the Canon sensor is much larger (noise goes down as the sensor size goes up, and goes up as the number of pixels increases on the same sized sensor).

So, we would predict that noise on the Samsung NV10, which has a smaller sensor than either the Samsung or Canon, would be higher, and it is. The red channel has the highest noise, and ranges from a low of 0.94% to a high of 1.82%.

Because the noise at ISO 100 is lowest for the Canon, and highest for the Samsung, with the Sigma in between, we would expect the same order for performance at higher ISOs, such as ISO 800. Thus, the Sigma would be the middle candidate for use in low light where you needed to set the ISO at a high number.

Below are Sigma, Canon, and Samsung tests using a Gretag Macbeth ColorChecker SG color chart, which has 140 colored squares. The test results are a combination of what the colored squares should look like (the Gretag Macbeth standard) and how the camera photographed them (JPEG output). The photos are split diagonally to show the standard vs. the camera results. For each photo below, the Gretag Macbeth standard is the top left half of each square, and the bottom right half is the way the camera photographed that square.

None of the three cameras reproduced all the squares ideally, but each one tended to alter different colors. The Sigma did better on flesh tones than the Canon, and about the same as the Samsung. The photos were taken in full sun, middle of the day.

Conclusions

The Sigma SD14 DSLR is a fine camera. It has the easiest-to-use interface I have yet experienced, has a wide variety of lenses to fit it, and takes great photographs. Its infrared capability makes it unique among DSLRs, and its 14.1 megapixels will allow enlargements of just about any size.

 

- John E. Johnson, Jr. -

Additional References:

Infrared Photography

http://msp.rmit.edu.au/Article_03/04b.html

http://www.rit.edu/~andpph/text-ir-uv-book.pdf

http://dpfwiw.com/ir.htm

MTF

http://bobatkins.com/photography/technical/mtf/mtf1.html

http://www.normankoren.com/Tutorials/MTF5.html#whytest

Vignetting

http://www.vanwalree.com/optics/vignetting.html