The individual light-collecting element on the sensor. Preferred over "pixel" when discussing capture — pixel technically describes the output unit displayed on screen, not the collecting element itself.
The semiconductor element within each photosite that converts incoming photons into electrical charge. The physical mechanism behind all digital capture.
Technically the output unit — a value in the final image. Often used interchangeably with photosite in casual usage, but in the context of these diagrams photosite is more precise when referring to the sensor element itself.
The region within a photosite that holds accumulated electrical charge. "Well capacity" refers to how many electrons it can hold before clipping. In these diagrams, the well is shown on a 0–16 383 scale matching a 14-bit sensor.
The maximum number of electrons a photosite can hold before it saturates. Exceeding FWC destroys information — any additional photons produce no additional signal. Typical values range from 10 000 to 100 000 electrons depending on sensor design.
The actual charge carriers being counted. Electron count is the most physically precise way to describe signal level — it is proportional to the number of photons that arrived, before any digital conversion is applied.
Generic term for the accumulated electrons in a well. "Charge level", "charge accumulation", and "charge integration" all refer to the same process of collecting electrons over the exposure period.
The circuit that measures the charge in each well and converts it to an integer. The ADC is where the analogue electron count becomes a discrete number the computer can store. All information above or below its range is permanently lost at this stage.
The integer output of the ADC — the raw value stored in a raw file. Also called ADU (analog-to-digital unit). A 14-bit ADC produces DNs in the range 0–16 383. DN is proportional to electron count within the sensor's linear range.
The unprocessed DN straight from the sensor, before any demosaicing, white balance, gamma encoding, or colour correction. Raw files preserve these values — every other format has already applied at least one irreversible transform.
The number of discrete levels the ADC can resolve. A 14-bit ADC produces 2¹⁴ = 16 384 levels (0–16 383). More bits mean finer gradations between steps, reducing quantisation error — particularly important in shadow regions where the signal is small.
Electronic noise introduced during the readout process — the act of measuring the well's charge and passing it to the ADC. Measured in electrons RMS. Read noise sets the absolute noise floor: signals smaller than read noise cannot be distinguished from it.
Statistical noise inherent in photon counting — because photons arrive randomly, the actual count varies from the expected value by approximately √N, where N is the electron count. Shot noise is unavoidable and sets the theoretical upper limit on SNR for any given signal level.
Charge that accumulates in a well even without light, due to thermal energy in the semiconductor. Relevant mainly in long exposures. Increases with temperature and exposure duration. Less significant for film scanning than for astrophotography.
The combined minimum signal level below which information cannot be recovered. Primarily determined by read noise at low signal levels, and by shot noise at higher levels. Signal must exceed the noise floor to carry usable information.
The key figure of merit for any captured signal. Higher electron count = better SNR — which is why maximising the number of electrons captured before the ADC (exposing to the right, using a bright light source) improves image quality independently of any post-processing.
The condition where a photosite's well is full. Additional photons arriving after saturation produce no additional charge — the well cannot hold more electrons. Saturation is the upper hard limit of the sensor's dynamic range.
Recording a value at or above full well capacity. Clipped data is permanently lost — no software process can recover detail from a clipped channel. In film scanning, clipping the raw blue channel by over-exposing to compensate for the orange mask is a common and irreversible error.
The remaining well capacity between the current signal level and full well. Preserving headroom means the sensor has room to record brighter values without clipping. In the series diagrams, headroom is shown as the unused upper portion of the well bar.
Charge spilling from a saturated photosite into its neighbours, corrupting adjacent pixels. Largely eliminated in modern sensors by anti-blooming structures, but historically a significant problem in CCD-based scanners with bright highlights.
The fraction of incident photons that actually produce an electron in the well. A QE of 0.6 means 60 out of every 100 photons are converted — the other 40 are reflected or absorbed without producing charge. Modern sensors typically achieve 50–80% QE at their peak wavelength.
The fraction of the photosite's total area that is actually photosensitive. The remainder is occupied by wiring, transistors, and other circuitry. Microlenses placed over each photosite focus light onto the active area, effectively increasing the fill factor.
Loosely used to describe how efficiently a sensor converts light to signal. Formally expressed as QE × fill factor. Often conflated with ISO in camera usage, but ISO is a gain setting applied after capture — it amplifies signal and noise equally and does not change the physical sensitivity of the photosite.
photosite refers specifically to the light-collecting element on the sensor, and pixel refers to the output value. The distinction matters when discussing what the sensor captures versus what the image file contains.