ISO / Sensitivity accuracyIn a new addition to our reviews we are now measuring the actual sensitivity of each indicated ISO sensitivity. This is achieved using the same shots as are used to measure ISO noise levels, we simply compare the exposure for each shot to the metered light level (using Sekonic L-358), middle gray matched. We estimate the accuracy of these results to be +/- 1/6 EV. Both the Sony DSC-R1 and Olympus E-500 used exposures which matched our metered light level (returning middle gray on our GretagMacBeth ColorChecker), as we have noted previously the Canon EOS 350D (Digital Rebel XT) proved to be about a third of a stop more sensitive than indicated.
ISO Sensitivity / Noise levelsISO equivalence on a digital camera is the ability to increase the sensitivity of the sensor. The works by turning up the "volume" (gain) on the sensor's signal amplifiers (remember the sensor is an analogue device). By amplifying the signal you also amplify the noise which becomes more visible at higher ISO's. Many modern cameras also employ noise reduction and / or sharpness reduction at higher sensitivities.To measure noise levels we take a sequence of images of a GretagMacBeth ColorChecker chart (controlled artificial daylight lighting). The exposure is matched to the ISO (ie. ISO 200, 1/200 sec for consistency of exposure between cameras). The image sequence is run through our own proprietary noise measurement tool (version 1.4 in this review). Click here for more information. (Note that noise values indicated on the graphs here can not be compared to those in other reviews). Room temperature is approximately 22°C (~72°F), simulated daylight lighting. Sony DSC-R1 vs. Canon EOS 350D (Digital Rebel XT) vs Olympus E-500Note that we have chosen to compare to the EOS 350D for these tests because price-wise (including a similar lens) it's closer to the DSC-R1 than the EOS 20D. Also the noise profile of the EOS 350D and EOS 20D are as good as identical.
The DSC-R1's noise reduction strategy appears to be no noise reduction from ISO 100 to 400, then at ISO 800 noise reduction begins and there is a decrease in sharpening, at ISO 1600 the noise levels become too much for the noise reduction to contain. It's interesting to observe the difference between the DSC-R1 and Olympus E-500 at ISO 1600, the Olympus image appearing much softer due to noticeably heavier noise reduction. The DSC-R1's higher ISO noise is characterized mostly as chroma blotches, most notably from the red channel which cause cyan colored patches in gray areas and red patches in shadows. Areas with a lower 'signal' (shadows) will suffer more. Compared to the two other cameras here the DSC-R1 performs well up to ISO 800 (with some softening), at ISO 1600 and 3200 images are still usable but there's a clear increase in blotchy noise which is especially noticeable in shadow areas (seen on our black patches above). Luminance noise graphOur new luminance noise graph shows noise levels in both the middle gray and black patches. As you can see the R1's primary noise 'problem' is in shadows (the dotted lines), you can also see the effect of noise reduction / sharpening reduction at ISO 800 with a flattening of the gray noise line.
Indicated ISO sensitivity is on the horizontal axis of this graph, standard deviation of luminosity (normalized image) on the vertical axis. Chroma (color) noise graphIn addition to measuring luminance noise we now also measure chroma (color) noise. Note the sudden drop at ISO 800, this is the noise reduction system taking effect, then at ISO 1600 noise levels are back up again. Overall not a bad performance, about the same as the other two cameras
Indicated ISO sensitivity is on the horizontal axis of this graph, standard deviation of each of the red, green and blue channels (normalized image) are on the vertical axis. Effect of heat build-up from live view (over time) on noiseOne thing we were concerned about was the build-up of heat in the sensor due to the live view. To test this we took a sequence of noise tests at two minute intervals. Between sequences the live view was constantly maintained. Below ISO 1600 there is almost no difference, at ISO 1600 we saw a small spike near the end of the test. At ISO 3200 we can see a very slight increase in noise but this fell away at the end of the test (camera modifying noise reduction level to compensate for sensor temperature?).
Indicated ISO sensitivity is on the horizontal axis of this graph, standard deviation of each of the red, green and blue channels (normalized image) are on the vertical axis. Black patch / Shadow noiseAs noted above the DSC-R1's noise is primarily characterized by color blotches in shadows (simulated by the black patch on our test chart). Here we're comparing the breakdown of noise in shadow areas by color channel. The crops below show the full color crop from our black patch and directly below this the contents of Red, Green and Blue color channels. As you can see noise from the red and blue channels produce the mottled patchy appearance of noise in shadows, green will always perform better because there are twice as many green photosites and sensors tend to be more sensitive to it. For this comparison we chose to use the Canon EOS 20D (which has essentially identical response to the EOS 350D but supports sensitivity up to ISO 3200).
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
