I have an FS700 + Nanoflash (undecided re:O7Q) and have been working with 5D3 RAW in Resolve and ACR. I've been reading and watching FS700+07Q videos and see that there are issues with color calibration, purple fringing, etc.
From working with 5D3 RAW, my understanding is that Resolve takes the linear data and converts to a colorspace and gamma as a first step. I was able to get the most accurate results by using Rec 709 colorspace and gamma and manually setting white balance and tint by eye to match known lights and scene colors. Even when matched somewhat closely this way, adjusting the luma curve, for example causes incorrect color shifts. Adjusting luma should only affect luma- colors shouldn't shift (much), right? I wrote a custom interactive 3D color corrector a few years ago, allowing direct manipulation of an XYZ RGB colorcube*. What this issue reminds me of is instead of the cube being properly aligned/rotated with respect to RGB colors, it's off, so that when going in the direction of increased color intensity (luma/brightness), the direction is off, and instead of all colors going up as expected the colors shift incorrectly. This is worse if the effective color cube is warped. So minimally it looks like a matrix (linear transform, not lossy) is needed and possibly a 3D LUT (non-linear transform, possibly lossy; don't need both).
Reading the docs here: http://www.convergent-design.com/Portals/0/Odyssey/docs/FS700Raw_PostWorkflow.pdf I see BMD Film is recommended as the colorspace. My understanding is BMD Film takes the linear raw data and converts to a log like, but not actual log color space ( http://www.bmcuser.com/showthread.php?2306-Cineform-Raw-BMD-Film-Curve ).
The next step is to use one of the supplied 3D LUTs:
* similar idea as this app, which looks very cool: https://www.facebook.com/groups/3dlutcreator/
From working with 5D3 RAW, my understanding is that Resolve takes the linear data and converts to a colorspace and gamma as a first step. I was able to get the most accurate results by using Rec 709 colorspace and gamma and manually setting white balance and tint by eye to match known lights and scene colors. Even when matched somewhat closely this way, adjusting the luma curve, for example causes incorrect color shifts. Adjusting luma should only affect luma- colors shouldn't shift (much), right? I wrote a custom interactive 3D color corrector a few years ago, allowing direct manipulation of an XYZ RGB colorcube*. What this issue reminds me of is instead of the cube being properly aligned/rotated with respect to RGB colors, it's off, so that when going in the direction of increased color intensity (luma/brightness), the direction is off, and instead of all colors going up as expected the colors shift incorrectly. This is worse if the effective color cube is warped. So minimally it looks like a matrix (linear transform, not lossy) is needed and possibly a 3D LUT (non-linear transform, possibly lossy; don't need both).
Reading the docs here: http://www.convergent-design.com/Portals/0/Odyssey/docs/FS700Raw_PostWorkflow.pdf I see BMD Film is recommended as the colorspace. My understanding is BMD Film takes the linear raw data and converts to a log like, but not actual log color space ( http://www.bmcuser.com/showthread.php?2306-Cineform-Raw-BMD-Film-Curve ).
The next step is to use one of the supplied 3D LUTs:
FS700_Linear_SLOG_2
FS700_Linear_709
The names seem to imply a linear starting point, however BMD Film isn't linear- it's 'almost log'. I understand this is a new product, and things are still being refined- is that why these are called "simulated log/Rec709"? It would appear you'd want to start with actual linear data before applying a 3D LUT to a calibrated colorspace? Canon raw looks really good processed in ACR- if there's a way to get FS700 footage looking like that with little effort (especially skintones), that would be a win (one way would be with a calibrated 3D LUT).FS700_Linear_709
* similar idea as this app, which looks very cool: https://www.facebook.com/groups/3dlutcreator/
Last edited: