The span from dimmest to brightest
The ratio between the largest and smallest signals a camera can accurately capture in a single image. High dynamic range allows imaging of samples with both bright and dim features without saturating bright regions or losing dim ones to noise.
Dynamic range is the ratio between the largest and smallest signals a pixel can distinguish in one exposure: full well divided by read noise. A sensor holding 15,000 electrons with 1.5 electrons of read noise has a dynamic range of 10,000 to 1.
Expressed in decibels that is 20 log₁₀(10,000) = 80 dB. Expressed in bits it is log₂(10,000) = 13.3 — which is the number worth carrying, because it is almost never the bit depth on the specification sheet.
Dynamic range is the gap between the faintest and the brightest a single exposure can record: how much the bucket holds, divided by how much the electronics wobble.
A typical scientific sensor manages about ten thousand to one. That sounds enormous until you compare it with the number of levels the camera digitizes into, which is usually larger — and that mismatch is the interesting part.
A 16-bit camera digitizes into 65,536 levels. The sensor above distinguishes 10,000. The extra levels are not information; they are the converter describing noise with great precision.
This is the difference between bit depth, which is a property of the converter, and dynamic range, which is a property of the sensor. The effective number of bits — ENOB — is set by the smaller of the two, and it is essentially always the sensor. Matching a converter to a sensor's real range is the whole of the sizing question: a 14-bit converter at 16,384 levels comfortably covers a 10,000:1 sensor, and the 16-bit option is buying file size rather than fidelity.
The headline number is not fraudulent — the levels are genuinely there — but reading it as range is a category error, and a common one.
A 16-bit camera records over sixty-five thousand distinct levels. Its sensor can only tell about ten thousand apart. The remaining levels are describing the wobble very precisely.
Bit depth belongs to the electronics; dynamic range belongs to the sensor; and the smaller one is what you actually get. A larger number on the box mostly buys larger files.
Dynamic range is about simultaneity. It constrains you only when a bright structure and a dim one must be captured in a single exposure.
If you can take two exposures — a short one for the nuclei, a long one for the marker — the sensor's dynamic range stops being the limit and your patience becomes it instead. Sequential acquisition or exposure bracketing sidesteps the constraint entirely, at the cost of time and of assuming nothing moved between frames.
So the question is never 'is this dynamic range enough' in the abstract. It is whether your bright and dim targets are in the same channel, in the same frame, and moving. For fixed tissue with sequential channels, dynamic range is close to irrelevant. For live imaging of a fast process spanning a large intensity range, it is the specification that decides the experiment.
Dynamic range only bites when bright and faint must appear in the same photograph. If you can take two — a brief one for the bright thing and a long one for the faint — the limit goes away.
Which means it barely matters for fixed samples imaged channel by channel, and matters enormously for something fast and live that spans a wide range of brightness.
Nuclei stained with DAPI alongside a weakly expressed marker, 100 times dimmer. To keep the nuclei below a 15,000-electron ceiling, the marker lands near 150 electrons. Its shot noise is √150 = 12.2, so its SNR is about 12 — workable.
Now suppose the marker is 1,000 times dimmer. The nuclei still cap the exposure, the marker lands at 15 electrons, and its SNR falls to about 3.9 — below reliable detection. No bit depth fixes this, because the limit is the sensor's 10,000:1 range against a 1,000:1 intensity span that must share it with the noise floor.
The fix is not a camera. Image them sequentially with different exposures, or separate the channels, and the constraint dissolves.
Bright nuclei beside a marker a hundred times fainter works: set the exposure so the nuclei just avoid overflowing and the marker still lands comfortably above the noise.
Make the marker a thousand times fainter and it does not. The nuclei still dictate the exposure, and the marker drops to where it cannot be reliably picked out. The answer is not a better camera — it is two exposures instead of one.
