ISO Sensitivity / Noise levels
Standard Test Scene
ISO equivalence on a digital camera is the ability to increase the sensitivity of the sensor to enable faster shutter speeds and/or better performance in low light. The way this works in a digital camera is by "turning up the volume" (gain) on the CCD's signal amplifiers. Nothing is without its price however and doing so also typically increases visible noise (random speckles visible all over the image).
The F700 responded very oddly to our new noise measurement test, it actually produced lower noise readings at higher sensitivities. After examining higher ISO images in detail we discovered that this is because the camera drops its sharpening considerably to keep noise down. Thus instead of posting crops of gray patches I thought it more appropriate to provide examples of what effect this has on image detail.
As you can see from the 100% crops below the F700's noise reduction system is either softening the images or works in conjunction with the rest of the image processing system to turn down sharpening.
|ISO 200, 1/34 sec, F5.6|
|ISO 400, 1/70 sec, F5.6|
|ISO 800, 1/125 sec, F5.6|
|ISO 1600, 1/250 sec, F5.6 (Forced 1MP: 1280 x 960)|
One of the big selling points (and a major point of interest) behind the F700 is that it is the first Fujifilm digital camera to use Fujifilm's extremely interesting 'double photodiode' SuperCCD SR sensor. The concept behind this design was described at the beginning of this review, essentially the primary photodiode has normal sensitivity and thus captures the 'normal' image, the secondary photodiode is much less sensitive (for example two stops less - pure conjecture) which captures bright detail which was overexposed on the primary photodiode. Some clever image processing then combines these 'two images' into one high dynamic range image.
When I first used the F700 I have to be honest and say I didn't see what I was expecting. However after more use and more examination of F700 images it's clear that there is better reproduction of bright detail, however it's not as impressive as I had hoped nor does it do away with the harsh 'clipping' which is such a signature of most digital camera images.
The problem appears to be that the secondary photodiode while contributing to the final image takes effect on only a very small portion of image data at the very bright end, and even then can easily be pushed into clipping. The improvement is there, but in my opinion not as much as anyone had hoped.
Secondarily the F700 appears to have a fairly contrasty tone curve, I would have expected the opposite of this camera, that images would be more neutral with a flatter look which has plenty of shadow and highlight detail. And of course there is no control of contrast from the camera menus.
The first sample image below shows that the F700 does manage to maintain more detail near to clipping (more of the sky detail is maintained) however it's nowhere near as obvious as we had expected. The second sample is a little more convincing, skin tone on the forehead is maintained by the SuperCCD SR sensor, that said it's not much and I would have been very interested to see how close the SuperCCD HR image would have been with a flatter tone curve.
|FinePix F700 (SuperCCD SR)||FinePix S5000 Zoom (SuperCCD HR)|
The graph below was produced by measuring the luminosity value of a gray chart shot at a range (about 40) of exposures with each camera. This graph represents the tone response curve of the camera, I wouldn't strictly say it accurately reports dynamic range. However, it does show that the F700 is still reproducing detail about 1.5 stops further than the S5000 (although the F700's shadow detail doesn't extend as far as the S5000). Interestingly the Canon PowerShot S50 with a low contrast setting got pretty close to the F700. The FinePix F700 and S5000 measurements were taken at ISO 200, the PowerShot S50 at ISO 50 (which in theory should have less dynamic range).
Horizontal (X) axis represents stops (EV's), the vertical axis the measured average luminosity value (where 128 is mid-gray, 0 is black and 255 is white).