Color Accuracy and ISO Performance

Early reports speculated that the 5DS cameras were capped at lower ISOs due to an increase in color accuracy over previous Canon cameras that came at a cost of low light, high ISO performance. One of Canon's best high ISO performers was the EOS 6D, but its Sensitivity Metamerism Index (SMI), which describes the ability of a camera to reproduce accurate colors, was relatively low at 69 (click on 'Color Response'). This prompted some to hypothesize that color selectivity of the filters in the 6D's color filter array was sacrificed to let in more overall light to boost low light performance. Increasing color selectivity on the 5DS cameras, then, may have come at the cost of low light performance, while catering to, for example, landscape photographers who care about color.

It turns out that the SMI, while improved over the 6D, remains the same for the 5DS R and 5D Mark III. Again, follow the links and click on 'Color Response' to view SMI. SMI for daylight (CIE-D50) and tungsten-balanced (CIE-A) light are 74 and 75 for both cameras. Oddly enough though, the 5DS scores for daylight and tungsten-balanced light are 82 and 71, respectively. Given that all Canon sensors hover around the 75 range for CIE-D50, 82 for the 5DS does seem like an outlier, especially consider the 71 rating for CIE-A, and we've reached out to DxO for a comment on this.

Regardless, color accuracy seems about on par with what we'd expect from Canon cameras, with the 5DS perhaps breaking into SMI numbers generally reserved for Nikon and Sony cameras.** Hence, we'd venture to guess that a dramatic increase in color accuracy is not the reason for the capped ISO. Furthermore, results for the 5DS R at least still fall behind those from a Nikon D810, which scores 80 and 78 for CIE-D50 and CIE-A, respectively.

Whether or not there are color accuracy tradeoffs, the fact of the matter is that ISO performance of the 5DS cameras, when viewed at common output size, is quite good. It's only a little behind the Nikon D810 according to DxO's low light score, and as verified by our own Raw studio tests, which you can see in our normalized comparison here. Back in February, when we interviewed him at CP+ 2015, Mr. Maeda of Canon suggested it was a performance bar on pixel-level image quality that drove Canon to cap ISO performance on the 5DS cameras, but since high ISO performance is right up there with the 6D when viewed at common size, we think a higher ISO limit might have been appreciated by some of the cameras' users. 

So what does all this mean?

Tulip sunrise at Skagit Valley Tulip Festival, Mt. Vernon WA. Taken with a pre-production Canon 5DS, with a Canon 11-24mm f/4L lens. Click to enlarge

Canon's new 5DS and 5DS R cameras have certainly raised the bar when it comes to resolution, and we look forward to seeing the combined resolving power of 5DS bodies with Canon lenses in DxOMark's 'Perceptual Megapixel' ratings of Canon mount lenses. But while DxOMark scores them as Canon's best sensors to-date, they don't raise the bar in other performance areas, instead falling behind class-leaders in both ISO and dynamic range performance. The sheer light-gathering and low-noise abilities of cameras like the Nikon D810 and Pentax 645Z mean I hardly find myself shooting with graduated ND filters or using HDR bracketing techniques in the field when dealing with high contrast scenes with these cameras.

This opens up creative opportunities that are too numerous to get into here. Dynamic range shortcomings won't matter if you're using the 5DS bodies in the studio, but for the very landscapes and environmental portraits these cameras will otherwise excel at capturing, they're an unfortunate shortcoming. Our own Raw dynamic range analyses along with DxO's indicate Canon has a ways to go before catching up to others but, hey, 50 megapixels!


** Nikon and Sony cameras, most using Sony (or Toshiba) sensors, tend to consistently score at 80 and above for CIE-D50, while maintaining class-leading ISO performance. The D7200, for example, shows the best ISO performance to date, yet scores a SMI of 85. Our point being: color accuracy and ISO performance certainly are not mutually exclusive, but they can be if a manufacturer chooses to boost ISO performance by sacrificing color accuracy.