“A camera is just a sensor with an opinion.”
We have spent five modules talking about universal standards (ACES, CIE XYZ). Now, let’s talk about the proprietary “Secret Sauce” that manufacturers bake into their cameras.
Why does an ARRI Alexa look different from a RED V-Raptor? It’s not just the sensor hardware; it’s the math they wrap around it.
6.1 The ARRI Gold Standard
For over a decade, the ALEV III sensor (Alexa Classic through Mini LF) has been the benchmark for digital cinema. Why? Because ARRI prioritized Highlights over everything else.
The LogC Curve (LogC3 vs. LogC4)
ARRI’s encoding is designed to mimic film density.
- LogC3: The classic curve. It allocates a massive amount of data to the highlights (shoulder), ensuring that roll-off is gentle and desaturated.
- LogC4 (Alexa 35): With the new ALEV IV sensor (17 stops), LogC3 ran out of room. LogC4 is a new curve that pulls Middle Grey down to 32% (from 39%) to make room for 2.5 extra stops of highlight information.
LogC3 Formula (ARRI Wide Gamut 3):
\[y = \begin{cases} 5.555556 \times x + 0.052272 & \text{if } x < 0.010591 \\ 0.247190 \times \log_{10}(5.555556 \times x + 0.047996) + 0.385537 & \text{otherwise} \end{cases}\]Where $x$ is scene linear exposure (normalized) and $y$ is the LogC3 code value (0-1). Middle Grey (18% reflectance, $x = 0.18$) maps to $y = 0.39$.
LogC4 Formula (Alexa 35):
\[y = \begin{cases} (x - 0.0011361) / 0.068512 & \text{if } x < 0.0059569 \\ 0.27861 \times \log_{10}(x \times 10.6723 + 1) + 0.33122 & \text{otherwise} \end{cases}\]Where Middle Grey ($x = 0.18$) maps to $y = 0.32$. This shift from 39% to 32% creates headroom for 2.5 additional stops in highlights while maintaining 10-bit precision.
Why Middle Grey Matters: The position of Middle Grey determines how the 10-bit code values (0-1023) are distributed across the dynamic range. LogC3’s 39% was optimal for 14 stops. LogC4’s 32% accommodates 17 stops by “borrowing” code values from midtones and shadows—acceptable because modern sensors have lower noise.
The “K1S1” Magic
The “Alexa Look” isn’t just the Log curve; it’s the Display Rendering Transform (DRT).
The famous K1S1 (and later the ARRI Rec.709 LUT) applies a very specific, non-linear tone map.
- Shoulder Desaturation: As colors get brighter, ARRI’s LUT aggressively desaturates them. A bright red taillight turns white/orange, not “digital red.” This mimics how film emulsion runs out of dye.
- Matrix Crosstalk: The ARRI matrix allows colors to “bleed” into each other slightly, creating a dense, organic palette rather than a sterile, separated one.
Interactive figure in development. It will be published only when it has a cited source or an original, documented demonstration.
6.2 The Color Wars
For years, camera manufacturers took different philosophical approaches to color science. ARRI favored Perceptual Pleasing, RED favored Mathematical Flexibility, and Sony/Canon positioned themselves somewhere in between.
RED IPP2 (Image Processing Pipeline 2)
For years, RED was the “Wild West.” You had DragonColor, DragonColor2, RedGamma3, RedGamma4… it was a mess.
Then came IPP2, which unified everything into a workflow remarkably similar to ACES.
The Separation of Church and State
IPP2 separates the Technical from the Creative.
- REDWideGamutRGB: The container. It encompasses every color the sensor can see.
- Log3G10: The curve. It defines Middle Grey and provides 10 stops of highlight headroom (hence “3G10” = 3 stops below, 10 above middle grey).
Log3G10 Formula (RED IPP2):
\[y = \begin{cases} 15.1927 \times x & \text{if } x < -0.01 \\ 0.224282 \times \log_{10}(x + 0.01) + 0.444666 & \text{otherwise} \end{cases}\]Where $x$ is scene linear exposure and $y$ is the Log3G10 code value (0-1). The name “3G10” refers to the dynamic range allocation: 3 stops below middle grey, 10 stops above = 13 stops total.
Technical Detail: Unlike LogC which uses $\log_{10}(ax + b)$, Log3G10 uses $\log_{10}(x + c)$ with a simpler offset. This makes the math cleaner for ACES transforms since both RED and ACES use similar logarithmic structures.
The Output Transform
In the old days, you baked the look in the camera. In IPP2, the R3D file is always Log3G10. The “Look” (Contrast, Rolloff) is just metadata applied at the very end of the chain.
Why This Matters: This makes RED footage incredibly flexible in an ACES workflow because Log3G10 → ACES AP0 is a clean, mathematically defined path.
The Philosophical Divide
- ARRI Philosophy: “Give cinematographers a beautiful starting point that matches film.” (Perceptually optimized from the start)
- RED Philosophy: “Give colorists maximum data and let them decide.” (Technically pure, creatively flexible)
- Sony Venice: A hybrid approach—X-OCN (eXtended Original Camera Negative) with S-Gamut3.Cine attempts to balance both worlds by providing ACES-friendly primaries with perceptually tuned shoulder rolloff.
Interactive figure in development. It will be published only when it has a cited source or an original, documented demonstration.
6.3 RAW Mechanics: The Bayer Pattern
Underneath the Log curves and LUTs, all these cameras (mostly) share the same anatomy: The Bayer Filter Array.
The Mosaic
Sensors are monochromatic. They only count photons. To see color, we place a mosaic of colored filters over the pixels.
The pattern is BGGR (Blue, Green, Green, Red).
- 50% Green: Because the human eye (M-Cone) is most sensitive to green luminance.
- 25% Red / 25% Blue: We need less resolution for these colors.
Demosaicing (Debayering)
The process of turning this mosaic into an RGB image is called Demosaicing.
The algorithm looks at a “Red” pixel (which has no Blue or Green data) and guesses the missing values by looking at its neighbors.
- Debayer Quality: High-quality debayering (like ARRI ADA-7) looks for edges and textures to avoid “zippering” artifacts. Low-quality debayering just averages the neighbors, resulting in soft images.
Interactive figure in development. It will be published only when it has a cited source or an original, documented demonstration.
The Hacker’s Way: Dcraw & Libraw
If you want to see what the sensor actually saw—without ARRI’s matrix or RED’s IPP2 sauce—you need a command-line tool called dcraw (or its modern successor Libraw).
This is for the DIT who wants to verify sensor noise or check for true clipping without any LUTs hiding the truth.
The “Truth” Command
Run this in your terminal to convert a RAW file to a linear TIFF with no white balance and no color interpolation:
dcraw -D -4 -T image.raw-D: Document mode. No debayering. You see the raw greyscale Bayer pattern.-4: Linear 16-bit output. No gamma curve.-T: Save as TIFF.
Converting to ACES
If you want to debayer straight to ACES AP0 (bypassing the manufacturer’s SDK):
dcraw -o 5 -4 -w -T image.raw-o 5: Output color space = XYZ (for ACES conversion).-w: Use camera white balance.
Why do this? Sometimes manufacturer SDKs apply hidden noise reduction or sharpening. Dcraw gives you the raw, mathematical pixel data. It is the biologist’s microscope for digital images.
Next Module: We finish the course with the “Final Boss”—Computational Color. Algorithms, Inverse LUTs, and Security.
References & Further Reading
- ARRI (2020). ALEXA 35 - System and Workflow Overview. ARRI Technical Documentation.
- ARRI (2017). ALEXA LogC Curve - Usage in VFX. ARRI White Paper WP-5.
- RED Digital Cinema (2021). IPP2 White Paper - Color Science and Image Pipeline 2.
- Sony (2018). Venice Technical Manual - X-OCN and S-Gamut3.Cine. Sony Professional Solutions.
- Coffin, D. (2023). dcraw - Decoding RAW Digital Photos in Linux. Available at: cybercom.net/~dcoffin/dcraw
- OpenImageIO Project. RAW File Processing and Color Management. Available at: openimageio.org