Optics and CFA (Color Filtering Array)
Around the same time, I noted that many pictures coming off Sony mirrorless cameras with adapted Canon optics had much more pleasing results than Sony mirrorless with Sony optics mounted (as I was considering a Sony if I could use a custom color profile on it, IE pull the same trick (custom JPEG engine/picture profiles) as AF on the A6300 was so much better than my then EOS M3, turns out, many folks have tried this too, in vain, it’s really hard to match those colors in post as it turns out), but not quite as good as Canon optics on a Canon body. So obviously optics influence color response too, perhaps that was part of the equation I as missing? But, then I realized it’s the same glass for the M2 / M3 / M5 here too, yet the EOS M3 went cold on the same copy of glass at that. Thus, I was convinced it was still a picture profile issue as the M2 and 5DIII use a DIGIC5 and the M3 uses a DIGIC6 and the M5 uses a DIGIC7 thus Canon must be making tweaks to the default color responses of the picture profiles with each DIGIC revision was my hypothesis. Of the bunch, the EOS M / M2 and 5DIII seem to produce the exact same color response. The EOS M3 produces its own color response and also the coolest of the bunch. The M5 on the other hand appears to hit some kind of middle ground where it’s not as warm as the 5DIII / EOS M2, but not as cool as the EOS M3.
Sony NEX5 - EF 35mm f/2
EOS M3 - EF 35mm f/2 IS USM
5D III - EF 85mm f/1.2L II USM Sony NEX5 - EF 85mm f/1.8
At the same time, I started to harass Canon directly on the subject, as their support is superb, and obviously with how detailed they are with mapping out correctional data for their own DPP4 software and calibrating bodies and lenses via Canon Professional Services, they must have the older picture profile defaults available, but, will they give them up, or not, in a very Canon manner? They tend to hoard features for a cost was my fear, or be unable to get the data that engineering has in the “back office” to a consumer. Well after bugging them twice I got the same answer: they haven’t changed a thing with the picture profiles between camera revisions. I refused to believe it, and did more research.
As I did more digging and comparing, I noted, that my existing EOS M5 at the time and EOS M3 before it, although cooler than older cameras colors, they were more true to life. I mulled over this conclusion and thought perhaps either Canon’s customer service reps had bad data, and that Canon was obviously trying to be sneaky, and adapt their colors to be closer to real life as they progressed their models, which apparently true to life accuracy is overrated as I like the more saturated warmer colors that are less than true to life. Perhaps being accurate to real life is overrated…
5D III - EF 100-400 f/4.5-5.6L IS USM
EOS M5 - EF-M 11-22 f/4-5.6 IS STM
Several folks (non-Canon reps) I corresponded with at the time indicated it was a color filtering array issue (CFA) which influences RAW data, not JPEG data (picture profile) and thus it was a sensor response issue. It was also suggested that it’s human psychology that we prefer warmer responses (I’d certainly agree, give me them back!). By recommendation to confirm that it was a CFA issue, I ran a comparison using Dpreview’s studio tool and sure enough, using RAW comparison (which bypasses jpeg engines unique to each camera) I could see the influence the sensor has on skin tones, and color response between different brands, and, different cameras even from the same brand. What was a surprise is that cameras of the same era, IE that 18MP APS-C era and the 5D III / 6D had the same responses, even though these were different sensors, they were using the same CFA of that era, and furthermore irregardless of what DIGIC as that 18MP APS-C got used over and over and over again for a while, the colors didn’t change between the 60D, T5i, etc etc. So picture profiles were out. CFA was confirmed to be a contributor too, just like optics and the JPEG engine (picture profile). This was the last missing link, or so I thought…
Monday, April 30, 2018