Measuring Dynamic Range
In a new addition to our future reviews we have now established a method for testing the dynamic range of a digital camera. DPReview in conjunction with Digital Domain Inc. (QImage Pro etc.) have developed a testing method and software application to analyse the results of test shots to establish the dynamic range of the sensor system of a digital camera. We will be updating some of our existing reviews with these test results.
Dynamic range has been something we've always talked about in our reviews but previously have not been able to quantify. Thanks to a collaboration between DPReview and Digital Domain Inc. (QImage Pro - Michael Chaney) we now have a test procedure for measuring dynamic range.
- Put simply we have to establish two images, one which represents the
lowest level of light at which the sensor system* can stil distinguish
detail and one which represents the highest level of light the sensor
system can capture before over exposing.
- This is carried out by shooting a test target with the camera in manual
exposure mode and progressively increasing (or decreasing) the exposure
- The resultant images are then analysed and run through a program which analyses the two images that represent minimum and maximum exposures, measuring both noise and recorded brightness levels on the target and performing a series of calculations to establish the dynamic range between the two extremes.
The result of this are five figures:
- Noise - a calculated figure which represents the average noise level
in the image (using standard deviation from averages)
- Range - - the dynamic range (based on signal versus noise levels)
of the sensor system represented as a magnitude ratio
- Bits - the number of digital bits required to store the resultant
dynamic range (thus a result of 7.8 bits would require 8 bits to store
- Density - a figure more commonly used in film and film scanners, for
reference (log10 - a density of 2 is 10^2 or 100:1)
- dB - the dynamic range in decibels; a measure often used in describing signal to noise ratio in capture devices.
Note: Range, bits, density and dB are simply different ways of characterizing the same dynamic range using different units. Although all four represent the same dynamic range, we provide them as a frame of reference since you may find reference to any of these in other sources.
We repeat the above test for all the ISO sensitivities available on the camera (as dynamic range changes depending on the sensitivity level) and the resultant figures can be used to compare the dynamic range of one sensor system (camera) against another.
We also repeat the test for a "best case scenario"; at the lowest sensitivity with in-camera sharpening disabled. In-camera sharpening defintely has a detrimental effect on dynamic range (as it tends to amplify noise).
Interestingly, because we can also measure noise this gives us a secondary result, that is the relative noise level at certain ISO sensitivities for different cameras.
* "sensor system" description includes the CCD, amplifiers and DSP's and firmware within the camera.
Shot using the above method, output was JPEG (capturing as RAW introduces too many variables). Higher is better for all numbers except noise where lower numbers indicate less noise.
|Nikon Coolpix 990||100*||0.13||446:1||8.8||2.6D||53|
|Fuji S1 Pro||320*||0.10||663:1||9.4||2.8D||56|
* Sharpening Off
(ISO Sensitivity scale on above graph is logarithmic)
DISCLAIMER: ACCURACY AND APPLICABILITY OF THE DATA
Note that although every attempt was made to minimize errors in measurements, there are some variables such as JPEG compression and quantization differences like gamma and contrast curves that can affect measurements. We feel that the results are a fairly accurate indication of the dynamic range of the cameras when used with the most common settings for the particular camera. Although comparability from one camera to the next will be affected somewhat by settings such as sharpening, curves, etc. we feel that the comparison is valid under most "normal" shooting conditions.
Also note again that we are measuring the capability of the camera sensor system and NOT the capability of a given camera to CAPTURE that dynamic range. Although a given camera may return a dynamic range that is higher than another model, it's firmware may not choose to CAPTURE the maximum range since doing so may compromise contrast and overall "pop" in the final images. For this reason, the results on this page are more representative of hardware capability than image capture capability, i.e., just because a camera has the *capability* doesn't mean that this will pass through to final images.