Analysis of DxO 5DM3 low-light ISO score (vs D800)

Started Apr 19, 2012 | Discussions thread
Shop cameras & lenses ▾
Flat view
Horshack Veteran Member • Posts: 5,803
Analysis of DxO 5DM3 low-light ISO score (vs D800)

I haven't examined the full SNR chart yet but in looking at the 18% SNR, which is what I believe DxO is using to generate the low-light ISO sports score, the 5DM3 is much closer to the D800 than what the score would indicate, depending on the importance you place on color separation/depth. DxO's site says the score is generated as follows:

"Sports Score is based on Low-Light ISO performance (values in ISO index). Low-Light ISO indicates the highest ISO sensitivity to which your camera can be set while maintaining a high quality, low-noise image (based on a Signal-to-Noise-Ratio [SNR] of 30dB, a dynamic range of 9EVs and a color depth of 18bits). As cameras improve, the highest ISO setting to produce 30dB, 9EVs, 18-bit images will continuously increase, making this scale open. Low-Light ISO performance is of primary importance in photojournalism, sports and action photography."

The 18% gray 30db cross-over point for the both the 5DM3 and D800 is around nominal ISO 3200. The D800's ISO 3200 SNR is 31.1db, the 5DM3's is 30.3, for a 1/3 stop difference. However, the D800's measured ISO at 3200 is ISO 2211 and the 5DM3's is 2518, so the 5DM3 will expose 19.1% brighter at ISO 3200. If you factor in that difference than the SNR delta between the two is only 1/16 stop.

For the DR component of the low ISO score, both cameras are nearly identical at ISO 3200, the D800 at 10.09EV and 5DM3 at 10.0.

Where the difference comes in is the color depth. At nominal ISO 3200 the D800 has a depth of 18.9 bits and the 5DM3 at 17.8 bits (5DM3's 17.8 bits is below DxO's 18-bit threshold for the low-light ISO score). Even after backing in the 19.1% measured ISO difference the D800 still leads here. It's not clear how much weighting DxO gives the color depth but based on their description they give extra "points" for performance that measures above their threshold minimums, which for color depth is 18-bits.

As you move further down the SNR/DR slope beyond the ISO 3200 level from which the sports-ISO is calculated in these cases, the 5DM3 and D800 are virtually identical for SNR on a measured ISO basis, and the 5DM3 takes a small lead starting at nominal ISO 6400 (less than or equal to 1/4 stop).

On a purely sensor-measurement scale, in terms of QE adjusted for color-selectivity (CFA), I think the DxO score accurately represents the difference between the two sensors, since the 5DM3 has a less-selective CFA which lets more light into the sensor at the expense of color separation and thus more noise when matching saturation between the two sensors. That's probably not too critical in actual use though since most have come to expect DR/color compromises at higher ISOs.

In summary, if you discount the color depth performance difference, which has implications for color separation/noise esp. if you post-process, the bodies are much closer to each other than the composite score would indicate for general usage. This is especially true beyond DxO's arbitrary 30db SNR, 18-bit color depth, and 9EV DR thresholds. I don't mean "arbitrary" in a pejorative sense because any objective measurement must establish well-defined metrics when producing a composite score, but in doing so they discount High ISO performance before/after the chosen metrics and do not account for non-linear performance differences between sensors, which occur between the D800 and 5DM3 (5DM3 closes the gap with the D800 after ISO 3200, although it lags in color depth the entire scale).

Canon EOS 30D Canon EOS 5D Nikon D800
If you believe there are incorrect tags, please send us this post using our feedback form.
Flat view
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow