Raw Dynamic Range: Real World

By Rishi Sanyal

The previous two pages demonstrated that while the a7S fares well next to its competitors with respect to dynamic range, especially when compared to Canon DSLRs, it isn't quite class-leading, falling behind what we've come to expect from Sony Exmor sensors.

It can sometimes be a bit hard to read about 'Dynamic Range' or see test data and translate it into real-world impact. Here we demonstrate the processing latitude offered by the Raw files of the Sony a7S and a7R, which illustrates the two cameras' respective base ISO dynamic ranges. We do this by exposing each camera for the highlights, and then assessing the noise levels of shadows when they're brought up to visible levels.

The Raw dynamic range of a camera is defined by the difference between the brightest tone a camera can capture and the darkest usable tone that is not swamped in noise. While DxO provides rigorous measurements that put a number on this attribute (they place the a7S and a7R about 1 EV apart in this regard), we visually show you what the difference in base ISO Raw dynamic range between these cameras looks like.

Our methodology

We shot a moderately high dynamic range sunset scene with both the a7S and the a7R, without filters of any kind. Our choice of exposures for this test parallel how a landscape shooter may have shot this scene, and requires a bit of elaboration.

The best way to capture a high dynamic range scene in a single exposure is to 'expose to the right' (ETTR). Using this method, you give the camera as much exposure as possible until the brightest tones in the scene you wish to capture are just short of clipping in the Raw file. Not only does this ensure that highlights are preserved, it also ensures that other tones - mainly shadows and midtones - have a fighting chance despite the shorter exposure required to protect highlights. The more light you collect, the cleaner your tones are because you (1) minimize the risk of darker tones running into the noise floor of the sensor from too short an exposure, and (2) minimize noise inherent in the light itself ('shot noise', due to the random arrival of photons), which decreases or 'averages out', in a sense, the more light you collect. Hence ETTR philosophy champions giving the sensor as much exposure as you can, but not so much that you clip bright portions of the scene you wish to preserve.

Typically, when you then pull an exposed-for-highlights Raw file into your favorite Raw processing software, the shadows and midtones will be dark. This is largely due to the low dynamic range and brightness of most of our monitors these days, requiring us to push shadows in Raws exposed for the highlights just to make them visible. A comparison of the noise levels in these pushed shadow regions then gives one an idea of the dynamic range capabilities of the camera.

Above, we've captured Raw files using exposures that map some of the brightest tones in our scene - the clouds above the sun - to Raw values just short of clipping*. That is, 1/3 EV longer exposures on either camera clip the green channels in these clouds. We decided not to try and retain detail in the sun because that is probably photographically irrelevant. We're essentially trying to 'fit' the dynamic range of the scene into the dynamic range of the camera by mapping the brightest tones we wish to preserve as close to clipping on the sensor as possible, but not further below it, for the reasons we gave above.

Who wins?

In a nutshell? a7R, hands down. As you can see in the normalized comparison above, where we've downscaled the a7R image to the a7S' 12MP resolution, the a7R offers more detail and cleaner shadow/midtone imagery compared to the a7S. Downscaling the a7R image also appears to have the added benefit of making any noise present look more fine grained; the a7S' noise looks coarse in comparison.

So much for more pixels = more noise, a fallacy we've now debunked a number of times in this review.

If Sony should've been bragging about the dynamic range of any of its cameras, it should've been the a7R, not the a7S. Sony marketing claimed 15.3 EV dynamic range for the a7S, which appears to be approximately exaggerated by 2 EV according to DXO tests. The a7R appears to be 1 EV ahead of the a7S in those tests. To put this in perspective, though, both cameras are well ahead of offerings from Canon, but if you're looking for dynamic range for landscapes, or exposure latitude for flexibility in post-processing (or because you've simply mis-exposed a shot), the a7R is the winner here. Even in non-normalized, full-resolution comparisons, where the a7R continues to offer both more resolution and more pixel-level dynamic range than the a7S.

An aside: Sony Raw compression

One thing we're going to call Sony out for is its continued use of compression in Raw - which can affect your ability to perform some of the edits we performed above. Because of the compression scheme Sony uses even in Raw files (no, you can't turn this off), high contrast edges in a scene are sometimes prone to posterization. While it's not uncommon to see manufacturers offer compressed Raw modes that allow for smaller file sizes with minimal cost to image quality, some implementations are better than others. Typically, these modes will take advantage of the fact that most recorded tones inherently have some fluctuations associated with them due to photon/shot noise (random fluctuations in incoherent light itself), and it's only important to accurately record the magnitude of these deviations. Recording finer variations in signals doesn't hurt, but just over-quantizes shot noise, and so is typically redundant. This is how Nikon offers 12-bit Raw files that encode larger ranges of dynamic range, for example.

Sony's compression scheme is two-step, however, and while the first step of compression (that takes advantage of the shot noise fluctuations described above) seems perfectly reasonable, it's in further spatial compression of that data where things go wrong, and gaps in retained information become visible. What results are posterization artifacts in transitions between light and dark tones. The problem extends to many Sony cameras, including all of the a7-series bodies. Have a look at the borders between the bright Space Needle buildings and the dark blue sky (which has been lifted only 1.5 EV in post) in the rollover below:

100% crop #1
100% crop #2
Full image

You can find the original Raw file here. It's important to note that - in thise case - we didn't even have to make a drastic edit to make the posterization visible; in fact, posterization was evident even in out-of-camera JPEGs that had 'DRO' set to 'Auto' (as this mode performs some shadow boosting to even out tones). We feel that Raw shooters looking for the best image quality could do without the potential pitfalls of Sony's compression methods, especially considering the competition offers significantly more 'lossless' Raw. Not to mention the price point of some of these bodies. We hope to see this fixed in future Sony cameras.

If you want to learn more about Sony's compression scheme, Iliah Borg and Alex Tutubalin of LibRaw/RawDigger fame have done a fairly in-depth study.


*To be precise: the brightest pixels have a 14-bit Raw value of 15,860 in the green channel, just short of the maximum 'clipped' signal of 15,988, as analyzed by RawDigger. Raw files are available for download for your own analysis.