​ Principle of Line-Scan TDI Industrial Cameras and Their Applications in Semiconductor Metrology

1. What Is TDI?

A TDI camera (Time Delay Integration Camera) is a high-performance line-scan camera specifically designed for imaging high-speed moving objects. By means of a unique synchronized charge-transfer mechanism, it can deliver high signal-to-noise ratio (SNR) and high-resolution images even under extremely low-light conditions. TDI cameras are widely used in industrial inspection, remote sensing, semiconductor manufacturing, biomedical imaging, and defense applications.

TDI imaging is well suited for applications that involve time-varying linear processes or targets with a highly asymmetric aspect ratio. In particular, TDI is ideal for low-light scanning applications, where conventional line-scan cameras are unable to generate usable images.

2. TDI Operating Principle

Time Delay Integration is a scanning technique that generates continuous images of moving objects by using a stack of linear sensor arrays aligned with and synchronized to the object’s motion direction. As the image moves from one line to the next, the accumulated charge is synchronously transferred, enabling higher sensitivity and resolution under lower illumination compared with conventional line-scan cameras.

In FFT-CCD sensors, the signal charge of each line is vertically transferred during readout. In TDI mode, this vertical transfer timing is synchronized with the motion of the target object, so that the number of charge integrations equals the number of vertical stages of the CCD pixel. In TDI operation, the signal charge must be transferred in the same direction and at the same speed as the moving object. The relationship can be expressed as:

v=f×d

Where:

v = object motion speed / charge transfer speed

f = vertical transfer frequency

d = pixel size

As illustrated conceptually, when the charge from the first stage is transferred to the second stage, the second stage generates and accumulates additional photoelectrons through photoelectric conversion. This process continues until the charge reaches the final stage M (the number of vertical stages), resulting in a total accumulated signal that is M times the initial charge.

Since the signal charge from each line is output via the CCD horizontal shift register, continuous two-dimensional images can be obtained. In this way, TDI achieves M-times higher sensitivity than a conventional linear image sensor (with SNR improved by √M). Compared with frame-based operation, TDI mode also provides improved sensitivity uniformity.

The following illustration intuitively demonstrates the charge accumulation process and the resulting image formation.

Typical TDI Applications in Semiconductor Manufacturing

3.1 Wafer Defect Inspection

In wafer fabrication, even microscopic defects can cause device failure. TDI cameras are used to scan wafer surfaces inline, detecting and localizing defects through high-resolution imaging. Thanks to their high SNR under low illumination, TDI cameras are particularly effective for identifying fine scratches, particle contamination, and material non-uniformities.

3.2 Photomask Inspection

Photomasks serve as templates for transferring circuit patterns onto wafers. Any defect on a photomask will be replicated on the final product, leading to yield loss. TDI cameras enable extremely precise inspection of photomasks, ensuring defect-free patterns and maintaining the high quality standards required in semiconductor production.

3.3 Wafer Alignment and Metrology

Precise alignment is a critical step in semiconductor manufacturing. TDI cameras are used for accurate wafer alignment and dimensional measurement, supporting advanced packaging technologies such as 3D stacked chips. This improves both device performance and long-term reliability.

3.4 Fine Structure Analysis

As semiconductor feature sizes continue to shrink, the analysis of fine and nano-scale structures has become increasingly important. With their superior spatial resolution and sensitivity, TDI cameras can effectively capture nanometer-level details, supporting the development of next-generation semiconductor materials and processes.

Thanks to their ability to operate at extremely high line rates, TDI cameras are particularly well suited for semiconductor metrology applications.

4. Why Can TDI Cameras Achieve Such High Speed?

The exceptionally high line rates of TDI cameras result from their specialized sensor architecture, charge-domain processing, and hardware-level synchronization. Key factors include:

4.1 No Per-Line Readout → In-Pixel Charge Transfer

Conventional line-scan cameras: After each line exposure, all charges must be read out through on-chip or off-chip ADCs and then reset before the next line can start.

TDI cameras:

Multiple pixel rows are vertically stacked.

Charges are transferred directly from one row to the next within on-pixel vertical shift registers (CCD or CMOS global-shutter structures).

Only the final row is read out.

Result: The line period is approximately equal to a single charge-transfer time (nanoseconds to microseconds), rather than the sum of exposure plus readout time.

4.2 Parallel Processing and Pipeline Operation

TDI imaging is inherently pipeline-based:

Row N is collecting new photons,

Row N–1 is transferring its charge into row N,

Row M (the final row) is being read out.

All rows operate simultaneously without waiting, similar to a CPU pipeline.
Throughput = 1 / transfer period, which is far higher than serial readout architectures.

4.3 Dedicated High-Speed CCD / CMOS Architectures

High-end TDI sensors (e.g., from Teledyne e2v or Sony IMX series) employ:

Buried-channel CCDs with charge transfer efficiency > 99.999%, supporting MHz-level transfer clocks;

Low-noise output amplifiers with single- or dual-channel high-speed readout;

Multi-tap outputs that divide a line into multiple segments for parallel readout.

For example, a 128-stage TDI sensor can achieve a 200 kHz line rate, corresponding to only 5 μs per line.

4.4 No Mechanical Shutter + Global Shutter Operation

TDI uses global exposure, with all rows integrating simultaneously.

Exposure and charge transfer are fully controlled electronically.

No mechanical delay is involved; response speed is limited only by electronic charge transport.

4.5 External Synchronization (Encoder / FPGA)

In industrial systems, TDI line rates are driven by high-precision external clocks or encoders.

FPGA-based timing generators can achieve sub-nanosecond jitter (jitter < 0.1%),

Ensuring strict synchronization even at line rates above 100 kHz.

5. Future Development Directions of TDI Technology

Looking ahead, TDI technology is evolving along four main dimensions: performance scaling, architectural integration, intelligence, and application expansion.

5.1 Higher Line Rates and Larger Resolutions

Targets: >1 MHz line rates and >16k-pixel line lengths.

Drivers: Advanced semiconductor nodes below 3 nm demand nanometer-scale precision combined with high throughput.

Technical approaches:

Multi-tap parallel readout (e.g., 8/16-tap CMOS TDI),

Backside-illuminated (BSI) and stacked CMOS processes,

On-chip ADCs and digital preprocessing.

Manufacturers such as Sony and Gpixel have already introduced CMOS TDI sensors with 8k–16k pixels and line rates around 300 kHz.

5.2 Full Transition from CCD to CMOS TDI

Trend: CCD TDI (high sensitivity but high power and limited integration) → CMOS TDI (low power, high integration, programmable).

Advantages:

ROI (Region of Interest) TDI support,

Embedded on-chip ISP,

HDR and multi-gain operation.

Challenge: CMOS charge transfer efficiency remains slightly lower than CCD, though the gap has narrowed significantly with pinned photodiodes and global shutter designs.

5.3 Intelligent TDI: AI and Edge Computing Integration

Direction: Integrating “TDI + real-time AI inspection” within FPGA/SoC platforms (e.g., ZYNQ, Jetson).

Typical applications:

Inline wafer defect classification (YOLO + TDI data streams),

Adaptive exposure and gain control based on preceding lines,

Triggered high-resolution re-inspection of abnormal regions.

Value: Reduced data bandwidth and increased inspection efficiency.

5.4 Multispectral, Polarization, and 3D TDI

Extensions:

Multispectral TDI (visible, NIR, SWIR) for material identification,

Polarization-sensitive TDI for detecting surface stress and metallic scratches,

TDI combined with structured light or laser triangulation for high-speed 3D profiling.

Example: Multispectral TDI in aerospace remote sensing to differentiate vegetation, minerals, and water bodies.

5.5 Quantum Imaging and Single-Photon TDI

Frontier research:

Combining TDI with SPAD (Single-Photon Avalanche Diode) arrays,

Enabling photon-counting-level TDI imaging.

Applications: Quantum communication, fluorescence lifetime imaging (FLIM), deep-space exploration, and ultra-low-light scenarios.

5.6 Software-Defined (Programmable) TDI

Concept: Dynamically configurable via software:

Switchable TDI stages (e.g., 32 / 64 / 128),

Programmable integration direction (forward/reverse) for bidirectional motion,

Seamless switching between TDI and frame modes.

Platform support: Programmable CMOS sensors combined with high-speed interfaces such as CoaXPress 2.0 and 10GigE Vision.

You may contact us at chenguo@mindvision.com.cn to gain more in-depth technical insights and practical applications in the fields of machine vision and optical imaging.