What Is Dynamic Range? When the Scene Won't Fit on the Sensor

At sunset you turn toward the ridge and see everything at once: orange in the sky, texture in the foreground rocks, every detail in between. You shoot that exact frame. On screen the sky is a white blob, or the rocks are a black mass. Never both together.

That isn’t an exposure mistake. The camera metered correctly and you pressed the shutter correctly. The gap between bright and dark in that scene is simply larger than what the sensor can record in one frame. That gap is what this article is about: it’s called dynamic range, it’s measured in stops, and it’s the most common physical wall in photography.

Let’s set the axis straight first. Reading the histogram is about measuring whether you hit that wall, exposure compensation is about shifting the camera’s decision, and image noise is about the cost of the dark side. This article asks a different question: when the scene doesn’t fit the sensor, what do you actually do?

What dynamic range actually is

Dynamic range is the ratio between the brightest and the darkest signal a system can record at the same time. In photography that ratio is quoted in stops, because every stop doubles or halves the light. The language of the stop scale applies here unchanged: 12 stops of dynamic range means the brightest recordable point is 2¹² = 4096 times brighter than the darkest.

It’s discussed in two different places, and mixing them up is the root of the confusion:

• The scene’s dynamic range: In the landscape in front of you, how many stops brighter is the brightest point than the darkest? This is a fact that has nothing to do with your camera.
• The sensor’s dynamic range: How many stops can your camera record meaningfully in a single frame? That’s a hardware property with nothing to do with the scene.

Scene 9 stops, sensor 12 stops: you’re comfortable, everything stays inside no matter how you frame. Scene 14 stops, sensor 12 stops: you will lose two stops. The only question is which end you lose them from.

The sensor’s limits: a full bucket on one side, noise on the other

Each pixel on the sensor collects light the way a bucket collects water. A photon arrives, an electron is freed, and the number of electrons in the bucket is read out and turned into a brightness value. The limits come from there.

The ceiling — saturation (full well capacity). When the bucket fills, it overflows. After that, no matter how much more light arrives, the result is identical: maximum value. A cloud five times brighter and a sun ten times brighter are both recorded as the same white. The data describing the difference was never created. That’s why a blown highlight doesn’t come back — it isn’t erased information, it’s information that was never written.

The floor — the noise level. There’s always some noise at the bottom of the bucket: read noise from the electronics and photon noise from the statistics of light itself. Below that floor there is still technically a number, but it can’t be told apart from the noise.

So the engineering definition is simple:

Dynamic range (stops) ≈ log₂ (full well capacity ÷ read noise)

The most important consequence of that equation: dynamic range and noise are two faces of the same coin. A sensor with a lower noise floor gains wider dynamic range at the same time. The mechanism is covered in detail in what is image noise; here we only use the result.

Which “12 stops” are we talking about?

Manufacturer and test-site numbers confuse people because two different yardsticks are in circulation:

• Engineering DR: Counts down to the point where signal equals noise (SNR = 1). On a modern full frame body at base ISO that lands at 13-15 stops. Impressive on paper.
• Photographic / usable DR: Counts down to where the result still looks usable, i.e. before the shadows turn to mud. On the same body it’s around 10-11 stops.

The 3-4 stops between them aren’t a lie — they’re the “technically present but you won’t want to look at it” zone. Use the photographic number when you’re making decisions.

Source / target	Approximate dynamic range
Modern full frame, base ISO (engineering)	13-15 stops
Same body, usable shadow limit	10-11 stops
APS-C, base ISO	12-13 stops
MFT / 1 inch	11-12 stops
Phone sensor, single frame	~10 stops
What a JPEG file can carry	~8-9 stops
Visible on a bright SDR screen	~9-10 stops
Print on matte paper	~6-7 stops
Human eye, one instant (no adaptation)	~14 stops
Human eye, scanning and adapting	20+ stops

The bottom two rows explain the disappointment we started with: your eye re-adapts constantly as it scans the scene, while the camera makes one decision and takes one frame. What you saw was really a blend your brain performed.

Raising ISO narrows the range

Every stop above base ISO costs roughly one stop of dynamic range. The reason is mechanical: raising ISO doesn’t enlarge the bucket, it amplifies the signal before readout — so saturation arrives sooner. A body that gives 13 stops at ISO 100 is down around 9 stops at ISO 1600.

In practice: in high-contrast scenes, base ISO is not negotiable. This is where the “stay at base when there’s light” rule from what is ISO bites hardest. One exception: some bodies have a second base ISO (320/400/640) where range jumps back up; the measurement recipe in the noise article tells you whether yours does.

Measuring the scene’s range

No guesswork needed — your camera already has a light meter. Spot metering reads only 1-5% of the frame, so metering two different spots and counting the stops between them gives you the scene’s range directly.

Some typical values worth memorising for a rough read:

Scene	Approximate range
Overcast day, portrait outdoors	4-6 stops
Landscape in shade or at golden hour	6-8 stops
Midday street, hard shadows	9-12 stops
Interior with a window, exterior visible	12-15 stops
Sunset with the sun in frame	15-20 stops
Street lamp at night, dark corners	20+ stops

This table also explains why golden hour and blue hour are so beloved: the scene’s range narrows naturally at those times, so the camera isn’t stretched. Midday sun isn’t disliked by photographers so much as by physics — the difference between hard light and soft light is a dynamic range issue before it’s an aesthetic one.

If it doesn’t fit: what will you sacrifice?

When the scene’s range exceeds the sensor’s, you will lose something. The priority rule is clear and rests on two physical facts:

• A clipped highlight holds no data. The bucket overflowed; software can’t create something from nothing. Try to recover it and you get a flat, grey, plastic patch.
• A shadow isn’t clipped, it’s just buried in noise. There is weak data down there. Lift it and the noise grows too — but the detail arrives.

The one exception is when the loss is deliberate: in a backlit portrait, letting the face go fully black for a silhouette isn’t a mistake, it’s a decision. Choosing a loss and suffering one are different things.

Route 1: Narrow the scene’s range

The most elegant fix isn’t to push the sensor, it’s to make the scene easier. If you can touch the light, the problem is solved before the shutter.

• Fill flash. Lifting the shadows narrows the scene’s range directly. One stop of fill in a midday portrait takes an 11-stop scene down to 8; flash photography basics covers dosing it with TTL and flash compensation.
• Reflector. Same job, no batteries. In natural light portraits, a white surface held opposite the sun opens the shadow under the chin by two stops.
• GND filter. In landscapes it closes the 3-4 stop gap between sky and land in glass, before the light reaches the sensor. The ND and polarizing filter guide explains soft/hard/reverse.
• Polarizer. Cutting glare off leaves, water and rock pulls the scene’s brightest points down; the range narrows by a stop or two.
• Timing. Waiting an hour buys more dynamic range than a several-thousand-dollar body upgrade. Waiting for a cloud to cover the sun does the same.
• Frame it out. Keeping the sun out of the frame removes the brightest point from the scene entirely. It sounds like giving up, but it’s the cheapest fix there is.

Route 2: Get the most out of one frame

When you can’t touch the scene — sports, street, a wedding ceremony where the light changes in seconds — you have to spend that single frame as efficiently as possible.

Shoot RAW. This is the decision most directly tied to this article. Converting to JPEG, the camera applies a tone curve that compresses 12-14 stops of data into about 8 — and throws away the ends while compressing. In a RAW file those ends survive. That’s exactly the “recovery headroom” from RAW vs JPEG.

Stay at base ISO. For the reason above: every ISO stop eats a stop of range.

Expose to the right (ETTR). Pushing the histogram right without clipping highlights puts more light into the shadows — raising the signal-to-noise ratio down there. You pull it back down in editing. The rule: stop before the clipping warning lights up. The details are in reading the histogram.

Watch for channel clipping. Highlights don’t always overflow in all three colour channels at once. At sunset the red channel usually saturates first; in a blue sky, blue does. The luminance histogram hides this — look at the RGB histogram. If only one channel is saturated, colour can partly be rebuilt from the other two.

Route 3: Multiple frames — exposure blending

If the scene is 14 stops and the sensor is 11, one frame is mathematically impossible. The fix: shoot several frames at different exposures and combine the good parts. That’s exposure blending; done automatically it’s called HDR merging.

Shooting the frames properly

The movement problem

The only real enemy of blending is motion. A branch in the wind, a passing cloud, someone walking through — anything that moves between frames leaves a ghost in the merge.

• Most merging software has a “deghosting” option that takes a single frame as reference in moving areas.
• Cleaner still, when blending by hand: mask the entire moving region from one single frame.
• For choppy sea or a crowded square, a single frame plus a GND filter usually causes less trouble than blending.

Merging: automatic or by hand?

Automatic HDR merge (Lightroom “Photo Merge > HDR”, Photoshop, Aurora) aligns the frames, produces a 32-bit file and lets you tone-map it with the normal sliders. Lightroom’s DNG output is good for this — it still behaves like a RAW, so recovery headroom stays wide.

Manual blending (luminosity masks): you place the dark frame over the bright one and open up the sky with a brightness mask. More work, but the result is almost always more natural, because you control the transition yourself.

The “HDR look” is not the HDR technique

That grey-skied, haloed, oversaturated look from the 2010s isn’t HDR itself; it’s the result of bad tone mapping. The technique only says “collect all the data in this scene” — how it looks afterwards is your call.

A well-blended frame doesn’t look HDR at all — it just has detail in both the sky and the foreground. The signs of a bad one are recognisable: a bright halo along the horizon, shadows that read grey and flat, a sky that becomes dead texture. All of them share one cause: flattening contrast too far.

Managing range in editing

Now you have a wide-range file and you need to fit it onto an 8-stop screen or 6-stop paper. What you’re doing is tone mapping, and it helps to know what the sliders actually do.

• The Highlights / Shadows sliders work locally, not globally: they target pixels within a particular brightness band. That’s why cranking both to the extremes flattens the frame; use them in moderation.
• Whites / Blacks set the ends. Starting tone mapping here is a good habit: place the white and black points first, then shape the middle.
• The tone curve is the most honest tool, because it shows you where each band goes. A gentle S curve gives back the micro-contrast that compression takes away.
• Local adjustments (masks, brushes, graduated filters) beat global ones nearly every time: handling sky and foreground separately gathers the range without making the frame look flat.

The full ordering lives in the RAW editing workflow. One warning bears repeating here: the moment you lift a shadow, the noise in it grows too. Lifting shadows four stops carries a noise bill similar to raising ISO four stops. That’s exactly why ETTR and base ISO matter so much.

How much recovery headroom is actually there?

In practice, modern RAW files hold roughly:

• ~1 stop in the highlights, up to 1.5-2 stops if only one channel is saturated. If all three are saturated, zero.
• 4-5 stops in the shadows, at base ISO on an ISO-invariant body. The price is noise and colour blotching.

That asymmetry explains why “I’ll underexpose and lift it later” is such a widespread reflex. But note: the reflex is only right when the highlights are genuinely at risk. Underexposing deliberately when nothing is at risk is buying noise for free — we covered why that’s the most expensive habit there is in low light without flash.

The output side: it doesn’t end at the sensor

How well you recorded the photo is only half of what the viewer sees. The other half is the output medium.

• sRGB / 8-bit screens: 256 levels per channel. An over-compressed frame can show banding here — especially in smooth sky gradients. Working in 16-bit and converting as the last step prevents it.
• Print: Even the most generous black on matte paper stops around 6-7 stops. When preparing for print, leaving shadows slightly more open than they look on screen is nearly always right.
• HDR displays: The HDR mode on phones and newer monitors genuinely shows a wider range, which is why the same photo can look “blown out” on a phone and normal on a computer. The “HDR photos” phones produce usually carry a gain map: bright areas open up on a compatible display and fall back to a normal SDR frame elsewhere.

Common mistakes

• Blaming the camera when the sky goes white. If the scene is 15 stops, no body collects it in one frame; the problem is the scene, not the sensor.
• Underexposing on purpose to save shadows. If the highlights aren’t at risk, all you produce is noise.
• Trusting the rear-screen histogram as if it were the RAW’s. You’re measuring the embedded JPEG preview.
• Reading ETTR as “a little clipping is fine”. The limit of ETTR is exactly where clipping begins, not one pixel past it.
• Raising ISO in a high-contrast scene. The shutter speed you gain costs you precisely the stops you needed.
• Changing aperture while bracketing. Depth of field shifts frame to frame and the merge fails at the edges.
• Merging three copies of one RAW as HDR. No new information is added; contrast just flattens.
• Pushing Highlights and Shadows to the extremes together. The frame goes grey and lifeless — the classic “bad HDR” look.
• Grading a print by screen brightness. Paper’s range is about half the screen’s.

Dynamic range is the invisible fourth side of the exposure triangle: aperture, shutter speed and ISO decide how much light you collect; dynamic range decides how much of that light gets recorded. Once you can read a scene’s range, you know what you’ll lose before you press the shutter — and losing something knowingly is, most of the time, the only real control photography offers.