The actual sensitivity of each indicated ISO is measured 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 a calibrated Sekonic L-358), middle gray matched. We estimate the accuracy of these results to be +/- 1/6 EV (the margin of error given in the ISO specifications). Note that these tests are based on the sRGB JPEG output of the cameras, in accordance with ISO 12232:2006, the standard used by camera manufacturers.
Barring minor fluctuations caused by white balance drifts and noise, measured ISO from all three of the cameras on test here is within +/- 1/6 EV of indicated ISO across their entire ISO range.
Noise and Noise Reduction (JPEG)
This is our standard studio scene comparison shot taken from exactly the same tripod position. Lighting: daylight simulation, >98% CRI. Crops are 100%. Ambient temperature was approximately 22°C (~72°F).
Note: this page features our new interactive noise comparison widget. By default, we show you the default noise reduction settings of the camera tested, and three other models of the same class. You can select from all available NR options, and from other cameras. The 'tricolor' patches beneath the familiar gray/black/portrait images are taken from the same test chart, and show how noise impacts upon blue, green and red areas of a scene.
As you can see from the images in this comparison, at their respective base ISO settings and up to ISO 200, images from all three of these cameras are essentially noise free at their defaut NR settings. By ISO 400, it is clear that the P7000's approach to luminance noise reduction is somewhat conservative, but detal is still very high from all three cameras. At ISO 800, slight grittiness creeps into all of the images, and it is clear that the LX5's NR+2 setting takes the edge off fine detail reproduction for the sake of slghtly smoother output.
At ISO 1600 and above, images from all three of the cameras in this grouptest start to show increased noise levels, but the best of the bunch in terms of detail reproduction is the P7000 with NR set to 'low'. We consider that the S95 strikes the best balance between noise reduction and detail retention, but there isn't a great deal between the three models at this point, and all are capable of producing high-quality small prints at these settings.
At ISO 3200, the differences between the three models are clear - the S95's noise reduction has smeared away a lot of fine detail, whilst the P7000 (which uses the same sensor) manages to retain it, but at the expense of much higher overall luminance noise levels. The LX5 is a midpoint between the two, and whilst high-contrast detail is relatively well described, areas of plain tone are distractingly blotchy. Only the Nikon P7000 can manage ISO 6400 at its full resolution of 10MP and image quality is fairly poor at this setting, as we'd expect. Useable results (just) can be coaxed out of its .NRW RAW files though (see below).
RAW noise (ACR 6.3 - noise reduction set to zero)
With ACR's noise reduction set to zero (which isn't the same thing as 'off', but provides as near to a level playing field as practical), it is clear that the differences visible in the JPEG files are largely due to in-camera processing. We'd expect the Canon S95 and Nikon P7000 to give very similar results here, but even the LX5, which uses a totally different sensor, gives very similar results right up to ISO 3200.
Although the lines on our noise graph show some separation between the cameras in this comparison, the 'real world' differences are minimal. In terms of detail reproduction, the distinctions between these cameras are almost certainly due to Adobe Camera RAW's approach to demosaicing, more than the output of the sensors themselves. What is interesting is that despite sharing a sensor with the S95 and P7000, the Samsung EX1's RAW output appears sharper and more detailed when converted in the same way. If this changes when Adobe releases the full public version of ACR 6.3 we will update this review accordingly.