You think the log curve is a stylistic choice. It isn't. It comes from a mismatch between how light behaves and how your eye reads it.
A 10% increase in brightness at 5% luminance is obvious. The same 10% increase at 90% luminance is invisible. Record light linearly, and you hand the same numerical precision to the peak white as to the deepest shadow. That's wasteful: the highlight data eats bandwidth and bit depth the shadow needed to avoid banding and noise.
Log encoding compresses the upper range instead. The response curve flattens as light intensity climbs, which shifts numerical weight toward the lower end of the scale — more steps, more precision, more usable range in the darks.
How the Curve Maps Light to Numbers
In a linear system, double the light means double the value: 50% light is 0.5, 100% light is 1.0. A 12-bit linear file spreads 4,096 steps evenly across that range. A scene with a 10:1 contrast ratio fits fine. A 16:1 ratio starts clipping highlights or crushing shadows — there's no headroom left for error.
A log curve remaps that relationship. The initial rise is steep; as input light increases, the output value climbs more slowly. The first half of the digital range can end up representing only the bottom 10% of the scene's dynamic range, while the remaining half covers the top 90%. That's not an exaggeration — it's what a log scale does by definition. Flattening the highlight response means the dimmest parts of the frame get the most distinct values, which is exactly where a slight coding error would otherwise show up as a visible band rather than nothing at all.
Why Linear Fails in High Contrast
Linear recording assumes uniform response across the sensor's range, which is often true of the sensor itself. It's the wrong assumption for how you'll judge the result. Log encoding tracks the way brightness differences become harder to see as light increases — the same reason a photographic characteristic curve has always rolled off at both ends instead of running straight.
Shoot log and the highlights are intentionally underexposed in the digital file — flat, gray, not a defect but the cost of buying shadow detail. You can brighten shadows in post without dragging up noise, because the data is there. You can pull highlights down without crushing them to black, because there's headroom.
Linear doesn't give you that choice twice. Expose for the highlights and the shadows go noisy and banded. Expose for the shadows and the highlights clip. Log defers that decision instead of forcing it at capture.
The Trade-off Is Bit Depth
None of this is free. An 8-bit log file bands in the shadows anyway, because the compressed range doesn't leave enough steps to work with. You need at least 10-bit, ideally 12-bit, before the curve pays for itself — which is why professional cameras and film stocks lean on high bit-depth formats. Those extra bits aren't spread evenly. They're concentrated exactly where a viewer would notice their absence.
Film has run on a log-like curve for a century before digital borrowed the idea. A negative's characteristic curve is S-shaped: toe flat, shoulder flat, midtone steep. That shape already preserved dynamic range better than a straight linear response would. Digital log profiles are a translation of that physical behavior, not a new invention.
Light is energy, and eyes read brightness differences unevenly across the range — coarser at the top, finer at the bottom. Linear recording spends precision on differences you can't tell apart. Log spends it where you can.
Try it in the Cineon studio to see how a log curve interacts with color primaries, which is less obvious than its effect on luminance alone — our piece on ACES covers how a color space carries that data through without clipping chroma in the shadows. The two are linked. You can't separate how you encode light from how you encode the color riding on top of it.
A flat, gray log image isn't a lack of contrast. It's a reservoir of detail sitting compressed, waiting for a grade to unpack it.