Raw Dynamic Range

Raw Bit Depth doesn't affect a sensor's dynamic range but it can act as a limit that prevents you recording that full range. The GH5S is the first Micro Four Thirds camera to capture 14-bit Raw files, so we'll look to see whether there's any benefit to be gained from the move.

The existing GH5 was measured as capturing a fraction over 12 stops of dynamic range*: the limit of what can be recorded in a linear 12-bit Raw file. With its bigger pixels, we'd expect the GH5S to exceed this, hence the need to move to beyond 12-bit (though we wouldn't expect to see much difference when the two cameras' images are scaled to the same size). Any differences that do exist should lie in the very darkest few tones of the images.

We should see this in our Exposure Latitude test. This is essentially a test of the ability to keep lowering exposure to protect highlights, while still being able to pull shadow detail back into the image. Sure enough, the differences between 12 and 14-bit mode become more apparent the deeper into the shadows you have to delve.

There's very little difference between 12 and 14-bit modes after a 3EV lift but as soon as you dig deeper, the 12-bit file starts to look noisier. The lesson, as theory would lead you to expect, is: the value of shooting 14-bit is in high DR scenes. This is only likely to make a significant difference at ISO 160 and 800, the two points at which minimal post-capture amplification is being applied: at other settings, you're unlikely to be capturing more than 12EV of tonal information, and there's no value to retaining more data than you have information.

Compare it to the GH5 at the same scale and you'll see there's no magic here: while the GH5S's larger pixels mean you should sample them at 14-bit to capture the camera's maximum dynamic range, the GH5 can capture its full DR in 12-bits, and produce a comparable image by doing so.


ISO Invariance

Our ISO Invariance test looks at one of the key properties underpinning the Exposure Latitude test: how much of the noise is contributed by the camera and its electronics. Here we shoot a series of different ISO settings with the same exposure, then brighten the images to match one another: any different in noise level can only be electronic, read noise.

On the GH5S we know that the camera has a dual gain sensor: one with two pixel modes that give different output voltages from the same initial signal levels. Or, to put it another way: it has two base ISOs. The effect of that is apparent when we compare the GH5S to the GH5: ISO 160 pushed by 3.3EV is noisier than ISO 1600 (meaning there's a small amount of electronic noise being added), but comparison with the GH5 shows that this is mainly because the GH5S's ISO 1600 is even cleaner than its sibling's (the GH5S's second mode is as close to noiseless as we've seen, in terms of how much noise comes from the camera).

For stills, this means you'll probably get your best results by shooting at either ISO 160 or, if the scene allows, ISO 800, then brightening later, rather than shooting at higher ISOs. There'll be virtually no noise difference between the two approaches but it'll prevent the loss of highlight data that would otherwise be amplified to clipping.

What does this mean for video?

To an extent, not a lot. The sensor appears to be capturing more than the 12-and-a-bit stops of DR that the GH5 could capture. That means the camera starts with a maximum of something around 13 stops of DR, depending on your tolerance for noise, at the sensor level. However, this figure will only be available in the lower of the camera's two base ISO settings. From a stills perspective this is ISOs 160. However, don't forget that the shape of the tone curve partly defines the ISO rating, so the base ISO modes may differ, depending on your color mode. In Log mode the two base ISO settings are ISO 320 and 1600.


*Dynamic Range is the measure of the range of tones that can be distinguished between, starting at clipping and working down to the last usable tone. This second limit is, to a degree, a matter of personal taste. Here we're quoting figures using the engineering definition, where the lower limit is a signal-to-noise ratio of 1: ie it's no longer possible to distinguish between the captured signal and the sum total of all noise. Most photographers and videographers would consider this to be unbearably noisy, so don't assume you'll get >12EV of usable dynamic range from the GH5S, even if the engineering figures suggest you might.