The actual sensitivity of each indicated ISO is measured using the same shots as are used to measure 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 the 'Standard Output Sensitivity' method described in ISO 12232:2006, the standard used by camera manufacturers.
Like previous Panasonic G-series models, we found that the measured ISO from the GF2 is roughly 1/3 EV higher than indicated and that this holds true throughout the entire ISO range, so ISO 100 indicated = ISO 125 measured (approx) and so on. This is similar to the Sony NEX-3 and Nikon D3100 in the comparison below, while the Olympus E-PL2's indicated and measured ISOs match (making it about 1/3 EV less sensitive than the other cameras at any given ISO setting). It's important to note that a discrepancy this small has little practical impact upon everyday photography.
Noise and Noise Reduction (JPEG)
The GF2's noise reduction is pretty strong by default, with somewhat more emphasis on attacking both chroma and luminance noise than we've seen on recent models such as the DMC-G2. Indeed our graphs suggest that the lowest (-2) setting is roughly equivalent to the G2's default setting. The result is a reasonably good impression of detail retention, but a gradual loss of color saturation and accuracy.
The outcome is very good up to ISO 400, but by ISO 1600 the processing starts to have rather more trouble. ISO 3200 surrenders the game to the effects of noise and has to smear away all the fine detail in order keep the output clean. ISO 6400 looks to be distinctly a step too far, and for emergency use only.
Real-world usage shows the ISO settings up to about 1600 to be very usable in natural light, but increasingly troublesome under artificial light (particularly tungsten, which contains a very low proportion of blue light). If you don't like the GF2's default approach to noise reduction you can of course tone it down a bit; the lowest setting lets through a little more noise but gives a bit more detail in return.
What's quite noticeable in this comparison is how much better Olympus's JPEG processing deals with noise from essentially the same sensor in the E-PL2. It does a remarkable job of detecting and emphasizing detail while effectively removing noise in tonally-smooth areas of the image. The end result is a much more agreeable output than the GF2's.
RAW noise (ACR 6.3 noise reduction set to zero)
The amount of NR applied 'under the hood' is not high, but it does vary by camera (Adobe is attempting to normalize output across different sensors), so inevitably we are still looking at a balance of noise and noise reduction, rather than pure noise levels. However, the use of the most popular third-party RAW converter is intended to give a photographically relevant result, rather than simply comparing sensor performance in an abstract manner.
As usual, the RAW data pretty much echoes the JPEG output, with noise remaining relatively low until ISO 400. Between ISO 800 and 1600 the noise starts to encroach on the image and it's clear why the JPEG engine is starting to produce less detailed images to counter-act this.
Even a glance at the ISO 3200 and 6400 images shows exactly why the JPEG image quality takes such a dramatic plunge at these settings, with noise starting to overwhelm the image. It should also be remembered that these samples are shot under daylight-balanced lighting, giving the best possible results - under artificial lighting that often provides little blue light, you can expect the results to be less positive.