Dynamic Range in video

Like the GH5 and 4, the GH5S can shoot using Panasonic's V-Log L gamma curve, to increase the dynamic range you capture and to make it more usable. This is how it works:

Since you can't shoot Raw video on these cameras, the next best thing is to ensure you make the best use of the limited bit depth of the 8 or 10-bit footage the camera can capture. To achieve this, a Log curve crams lots of highlights into the top ~20% of the file's values, then devotes around ~60% of the remaining file to capturing as much of the tonal information about the midtones as possible, then leaves the remainder for the shadows.

The idea is to devote as much of the file's capacity for tonal information to the most important tones in the image. The result is footage that initially looks grey and low-contrast because so much of the middle of the tonal range has been stretched out across the file, but in a way that should allow you to give a gentle roll-off to the highlights and allow significant manipulation of the midtones without the risk of posterization.

All of this makes it crucial to expose Log footage accurately, so that the tonal information you're trying to capture isn't crushed into the highlight or shadow portion, where only limited information is retained. It also means it's important for the gamma curve to be a good match for the sensor's characteristics.

An 'L' of a price to pay?

Where it gets interesting is that the GH5S doesn't use the same V-Log curve as the company's Varicam line of professional camcorders. Instead it uses a truncated version called V-Log L, designed to offer compatibility with fully fledged V-Log. This use of a less ambitious curve makes sense: the Varicam cameras tend to be capable of capturing greater dynamic range than the GH5S, so it would be detrimental to use the same gamma curve on the Lumix (it would encourage the use of lower exposure to capture more highlights, leading to increased shot noise in all the most valuable tones).

The rather flat looking VLog-L curve, showing clipping recorded at a value of 768 and nothing recorded below 128 (80 and 7.5 IRE, respectively). While this maintains compatibility with the full VLog curve, it wastes 384 of the possible 1024 values (or 96 of the 256 values in 8-bit mode).

Instead, V-LogL clips the top of the VLog curve: encouraging you to apply more exposure, to get cleaner mid-tones at the cost of losing some very bright highlights. It's a perfectly sensible approach. However, it's the next decision that is likely to divide GH shooters: rather than mapping the camera's brightest captured tone to the brightest value in the file, Panasonic has mapped it to 80% brightness. This doesn't throw away any captured data but means that some of the capacity for tone retention (the entire point of shooting Log) is left unused. The footage also never reaches black, so there are wasted values at the bottom of the files, too.

The V-Log L curve (blue) is essentially the lower section of the full V-Log curve, which includes the red section to capture additional highlights. The GH5S's clipping point is mapped to an output level of 768, to provide cross-compatibility.

The result is a file that can be easily integrated into a mixed V-Log/V-Log L workflow but one where there are fewer values available to record midtones and shadows than there could/should be, risking posterization and noise. The problem is most likely to be visible in 8-bit mode, where leaving around 64 values blank mean there are fewer than 192 left for capturing everything else. The drawback is less detrimental in 10-bit mode, since there are still 768 available values, even with the highest values left blank. It's still not optimal for users who aren't using Varicam cameras alongside their GH5S but a partially empty 10-bit space still allows a vast improvement over using a full 8-bit container.

TLDR:

Our recommendation would be to only ever shoot V-Log L in 10-bit. This will mean using an external recorder if you plan to shoot 4K at 59.94 or 50p with Log.

HLG as Log capture mode

Is it better to use HLG mode, rather than V-Log L, for greater flexibility in the edit?

Hybrid Log Gamma (HLG) is primarily designed as a no-processing-necessary way of shooting footage for HDR playback (ie: showing greater dynamic range on high-dynamic-range screens, not to be confused with garish attempts to compress large amounts of dynamic range into a disconcerting psychedelic nonsense). However, at its heart HLG is a Log curve that, like V-LogL, tries to squeeze lots of highlights and shadows into a file but while devoting most values to the tonal information of the midtones.

HLG mode makes much more sensible use of the available space, using all 1024 values. It appears the linear region is smaller, meaning more tonal information is crammed into the lower values however, so it'll be interesting to see what impact that has in terms of flexibility for grading.

Because it isn't trying to provide compatibility with a different camera's workflow, HLG uses a much wider range of the available values in the file, which should mean more data values are available to retain tonal information. And, although this means you can't simply use an off-the-shelf V-Log to... LUT when you first import the footage, the HLG curve looks like it's going to be a widely-enough accepted standard that we'd expect to see HLG to standard DR LUTs become increasingly common.

Of course there's more to optimizing video than just the gamma curve, and HLG's means of retaining color information isn't necessarily as sophisticated as the one Panasonic's engineers have developed for VLog L (HLG is, after all, intended as a ready-to-watch medium). But we'll have to do more tests and wait until there are cameras and footage in the hands of more experienced colorists, before the full picture will be known.

So how much Dynamic Range does the GH5S have?

Having talked so much about the ways of capturing and preserving the most possible dynamic range, just how much DR can the GH5S capture?

Manufacturers love to throw around DR figures for their cameras, especially on the video side of things. We've shot our dynamic range brightness wedge to see how many of the 1/3EV increments we could distinguish between, before the footage clipped to black or to an unacceptable noise level.

Interestingly, when shot side-by-side with the a7R III in S-Log2 mode, we got pretty similar results, but for different reasons. The GH5S's darkest tones become progressively noisier until they become indistinguishable after around 11.3EV. The big sensor Sony shows a little bit of noise and then suddenly hits a wall at a similar point. This makes sense: the GH5S will have noisier deep shadows than a camera getting much more light. However, the 8-bit Sony runs out of values to keep distinguishing between tones, so the full benefit of that extra DR gets lost.

What will be really interesting will be when we step out from the studio and shoot some side-by-side footage. The Sony should capture cleaner midtone information than the Panasonic but again will have fewer values to record this subtlety. The Panasonic will capture a slightly less clean signal but will preserve that information with its greater bit-depth. We'll be shooting more side-by-side footage soon.