A BSI sensor, short for backside-illuminated sensor, is a CMOS image sensor in which the silicon wafer is flipped and thinned so that light reaches the photodiodes from the back, without first passing through the metal wiring layers. This layout lets more of the incoming light reach the light-sensitive area of each pixel, which improves sensitivity, noise performance and angular response.
In photography, BSI design appears in cameras and sensors of every size, from smartphone modules with pixels around one micrometer to full-frame mirrorless bodies. It is also the structural basis for stacked sensors, which add a separate logic layer beneath the pixel layer.
How backside illumination works
In a conventional front-side illuminated (FSI) CMOS sensor, the photodiodes are built into the silicon, and several layers of metal interconnects and transistors are stacked on top of them. Light must travel through gaps in this wiring before it reaches the photodiode, so part of it is blocked, reflected or scattered along the way.
A BSI sensor reverses this arrangement. After the photodiodes and wiring are fabricated, the wafer is bonded to a support wafer, turned over and thinned from the back until only a few micrometers of silicon remain above the photodiodes. The color filter array and microlenses are then placed on this thinned back surface, so light enters directly into the silicon while the wiring sits behind it.
The result is a much larger effective light-gathering area per pixel and a shorter optical path between the microlens and the photodiode. Designers also gain freedom to route wiring without worrying about shading the pixel, which supports faster readout circuits and more complex pixel designs.
BSI compared with other sensor architectures
The main difference between sensor architectures lies in where the circuitry sits relative to the light path. The table compares BSI with its neighbors in the Topic.
| Architecture | Light path | Main advantage | Main drawback |
|---|---|---|---|
| FSI CMOS | Through metal wiring layers | Simpler and cheaper to manufacture | Lower fill factor and angular response |
| BSI CMOS | Directly into thinned silicon | Higher quantum efficiency | More complex process, crosstalk control needed |
| Stacked BSI CMOS | As BSI, with logic on a separate layer | Very fast readout and processing | Highest cost and complexity |
| CCD | Through gate electrodes | Uniform charge transfer | Slower, more power-hungry, largely obsolete in cameras |
A stacked sensor is always backside-illuminated, because the logic wafer must be bonded to the side of the pixel layer that carries the wiring. A BSI sensor, however, does not have to be stacked; many cameras use non-stacked BSI sensors as a balance between image quality and cost.
Where the BSI advantage matters most
The benefit of backside illumination depends strongly on pixel size. When pixels are small, the wiring of an FSI design occupies a large share of each pixel’s area and creates a deep tunnel that light must pass through. When pixels are large, the wiring blocks a smaller proportion of the light and the gain from BSI is more modest.
| Pixel pitch | Typical device | Relative BSI benefit |
|---|---|---|
| Below 1.5 micrometers | Smartphone cameras | Large; BSI is effectively mandatory |
| 2 to 4 micrometers | Compact cameras, high-resolution APS-C | Clear gain in sensitivity and corner performance |
| 4 to 6 micrometers | High-resolution full frame | Moderate gain, notably at the frame edges |
| Above 6 micrometers | Low-resolution full frame, video-oriented bodies | Small gain in light collection |
Beyond raw sensitivity, BSI improves angular response: light arriving at steep angles is collected more efficiently. This matters for wide-angle lenses on mirrorless cameras with short flange distances, where rays strike the sensor corners obliquely and FSI designs suffer from corner shading and color shifts.
Practical benefits for image quality
For photographers, the improvements of a BSI sensor appear in several measurable properties. The size of each gain varies with sensor generation, pixel size and processing, so the table describes the direction and mechanism rather than fixed numbers.
| Property | Effect of BSI | Mechanism |
|---|---|---|
| Quantum efficiency | Higher | Fewer photons lost in the wiring layers |
| Low-light noise | Lower at a given exposure | More signal relative to read noise and shot noise |
| Corner performance | Better with wide-angle lenses | Improved acceptance of oblique light |
| Readout speed | Often faster | Wiring can be optimized without shading pixels |
| Dynamic range | Equal or slightly better | Depends more on read noise and full-well capacity |
BSI does not change the fundamentals of sensor size. A full-frame sensor still collects more total light than an APS-C or Micro Four Thirds sensor at the same exposure, regardless of whether either uses backside illumination.
Manufacturing challenges and misconceptions
BSI manufacturing requires wafer bonding and precise thinning, which add steps and cost compared with FSI production. Thin silicon also raises optical and electrical crosstalk between neighboring pixels, which manufacturers counter with deep trench isolation walls between photodiodes, and the newly exposed back surface must be treated to keep dark current low.
- BSI does not mean stacked: many BSI sensors have no separate logic layer and read out at conventional speeds.
- BSI does not guarantee lower noise in every comparison: processing, read noise and sensor generation matter as much.
- BSI does not remove rolling shutter: readout speed depends on the readout circuits, not on illumination direction.
- The label alone says little: two BSI sensors of the same size can differ markedly in dynamic range and noise.
Stripped to the essentials
A BSI sensor is a camera sensor built so that light reaches the pixels more directly. In older designs, tiny wires sat in front of each pixel. Part of the light was blocked by them.
BSI sensors move these wires behind the pixels. It is like moving power cables from the front of a window to the wall behind it, so more daylight comes in. The result is cleaner images in dim light, especially with small sensors such as those in phones.
Where it is applied
A smartphone manufacturer designs a main camera module with a sensor that has pixels of about one micrometer across. With a front-side illuminated layout, the metal wiring would shade a large share of each tiny pixel, and evening photos would show heavy noise and weak corners with the wide-angle lens.
The team chooses a BSI sensor with deep trench isolation between pixels. In low-light tests at a restaurant table, the module produces images with visibly less noise and more even brightness toward the corners than a comparable FSI prototype at the same exposure settings. The same layout later allows the manufacturer to move to a stacked BSI version with faster readout for video, without redesigning the optical stack.