An image sensor is a grid of tiny wells that catch light and turn it into numbers. That is the whole job. Everything else on a Camemaker sensor page (megapixels, format, shutter, HDR, interface) is a way of saying how those wells are built and how fast you can read them.
Each well is a photodiode. Photons arrive, charge collects, an analog-to-digital converter counts it. Fill the well and the highlight clips. Leave it almost empty and the dark is noise. That is already the HDR story, only smaller.
In one sentence. A sensor is a chessboard of buckets. Resolution is how many buckets. Pixel size is how big each bucket is. Format is how large the whole board is.
What the page is really listing
When you open a sensor in the image-sensor library, read the fields in this order. Not in the order marketing likes.
- Optical format. The size of the board. It decides which lenses can cover it.
- Pixel size. How much light each well can hold, and how hard the lens must work.
- Resolution. How many wells. Useful only after format and pixel size are honest.
- Shutter. Whether the grid is read all at once, or row by row.
- CFA. Whether each well sees colour through a filter, or sees all of the light in mono.
- Interface and FPS. Whether you can actually get those wells off the chip in time.
What a sensor is not
It is not a module. It is not a lens. It is not an ISP. Two shop pages can share a part number and still be different products once the holder, the PCB colour, the glass, and the firmware change. The library lists the die so you can compare dies. The camera you ship is still a stack.
It is also not a promise of image quality. A 12 MP sensor with a weak M12 and a smeared CRA match will lose to a quieter 2 MP stack that was built as a system.
A real-world picture
Think of a rain gauge. A wide gauge collects more rain before it overflows. A field of many tiny gauges draws a finer map, but each one overflows sooner and reads noisier when the rain is light. That is pixel size versus megapixels. The size of the whole field is optical format. The speed at which you walk the field and write the numbers down is interface and frame rate.
Your eye does something similar, except the retina adapts locally and the brain lies helpfully. Silicon does not. If the well is full, the street is chalk. If the well is empty, the garage is noise.
How to use this on Camemaker
Start from the scene, not the part number. Door to daylight, night road, fast object, colour barcode, mono OCR. Then open the library and filter format, shutter, pixel size, and CFA. Leave brand for last. Brand is a supply question. The wells are the physics question.
The later posts in this series take each field apart. This one is only the map: the sensor is the board of wells. Everything else is how you fill them and how you read them.