Overview
The quantum efficiency (QE) spectrum is a key parameter of CMOS image sensors. It describes the sensor's photon-to-electron conversion efficiency as a function of wavelength, reflecting how sensitive the sensor is across different parts of the spectrum and influencing image quality, sensitivity, and color reproduction. QE spectra typically show peaks and valleys across the visible band, and those features directly affect imaging performance.
What the QE Spectrum Reveals
The QE spectrum indicates the sensor's response efficiency at different wavelengths. Since photon energy is inversely proportional to wavelength, photons at different wavelengths produce different responses in a CMOS image sensor. By examining the QE spectrum, engineers can assess sensitivity and color fidelity across wavelengths and identify process- or design-related issues that degrade performance.
Common Defects Revealed by QE Spectrum
The QE spectrum can help diagnose several internal issues in CMOS image sensors. Common categories include:
- Back-side illumination (BSI) processing and design
- Optical crosstalk
- Color filter quality and performance
- Silicon wafer or backside thinning (THK) conditions in BSI processing
A. BSI Processing and Design
How BSI Works
Back-side illumination (BSI) places the photodiode at the back of the silicon substrate so light enters the pixel from the substrate side. Compared with front-side illumination (FSI), where metal interconnects and circuitry sit above the photodiode and obstruct some incoming light, BSI reduces front-layer obstruction and improves light collection efficiency and diffraction behavior. This improves resolution and sensitivity, especially at various wavelengths.
Figure 1. BSI operation diagram
Front-Side Illumination (FSI)
FSI sensors receive light on the front side, where wiring and circuit layers sit above the photodiode, reducing effective light collection and potentially lowering image quality. BSI relocates light entry to the wafer backside, enabling higher light utilization, higher pixel density, and faster readout without additional front-side blocking structures.
Why BSI Is Important
BSI is an important manufacturing technique that significantly increases CMOS image sensor sensitivity and quantum efficiency, which is beneficial for low-light imaging. BSI also improves resolution, dynamic range, and signal-to-noise ratio (SNR). For modern imaging applications that demand high quality and performance, BSI has become a mainstream technology for high-end sensors.
Evaluating BSI Using QE Spectrum
The QE spectrum provides a wavelength-resolved measure of a sensor's photodetection ability and can be used to assess the effectiveness of BSI processing.
Case 1
Comparative QE measurements show that BSI can substantially increase pixel sensitivity for RGB channels. For example, wafer-level QE analysis comparing FSI and BSI processes in a 1.75 um pixel showed QE improvement from about 40% for FSI to nearly 60% for BSI, demonstrating BSI's advantage in light collection.
Case 2
Another example compares BSI-1 and BSI-2 processes on a 1.4 um pixel using a 65 nm process node. The QE spectrum shows BSI-2 delivers about a 10% improvement in absolute sensitivity compared with BSI-1, indicating BSI-2 raises the sensor's intrinsic photodetection capability. When applied to 1.1 um pixels, BSI-2 also shows higher blue-channel QE while green and red channels remain similar to the 1.4 um case.
Figure 2. Wafer-level QE analysis comparing different BSI process variants on a 65 nm production flow.
QE spectra are therefore an important tool for optimizing image sensor manufacturing and selecting BSI variants that improve sensitivity without changing pixel size.
B. Optical Crosstalk Inspection
What Is Optical Crosstalk?
Optical crosstalk occurs when light intended for one pixel diffuses, reflects, or refracts into neighboring pixels, causing interference between adjacent pixel signals. This effect reduces inter-pixel distinction and contrast, degrading image quality and SNR.
Figure 3. Optical crosstalk illustration
Why Detecting Optical Crosstalk Matters
Optical crosstalk is a critical issue for CMOS sensors because it degrades image contrast and SNR. Reducing optical crosstalk is a key design and process objective to maintain pixel-level signal integrity.
Detecting Crosstalk with the QE Spectrum
Abnormal features in a QE spectrum at specific wavelengths can indicate crosstalk. When crosstalk is present, QE measurements may show unexpected responses in one channel caused by photons entering from neighboring pixels. QE-based diagnostics can guide design or process changes to mitigate crosstalk.
Shrinking pixel size is necessary for high-resolution imaging but increases the importance of addressing optical crosstalk and related issues.
For example, a 0.9 um stacked pixel made using a 45 nm advanced CMOS process shows that crosstalk significantly impacts SNR and image quality. To address this, a pixel process with deep trench isolation (DTI) was implemented to suppress crosstalk while preserving dark performance.
Figure 4. Pixel cross-section schematic: (a) control pixel; (b) crosstalk-improved pixel with DTI.
The control pixel layout shows the optical path through the microlens (ML), color filter (CF), and photodiode (PD). Photons from neighboring pixels can refract through multilayer structures and reach the central PD, causing crosstalk. The DTI structure forms trenches that block photons from neighboring pixels, reducing crosstalk and improving SNR.
Figure 5. QE spectrum of 0.9 um pixels: dashed line is the control pixel, solid line is the crosstalk-improved pixel. The reduction in optical aperture area slightly lowers blue and red channel responses, but improved color filter materials increase the green channel QE. The three marked points indicate evidence of crosstalk suppression in R, G, and B channels. QE measurements provide direct evidence of crosstalk reduction.
Wafer-level QE measurements can directly demonstrate crosstalk suppression and allow comparison of sensor variants across wavelengths to verify mitigation strategies.
