· VisionSelect Team · Guides · 4 min read
TDI Line Scan Selection Guide: When You Need It and How to Specify It
TDI line scan cameras dominate wafer, panel and high-speed web inspection. This guide covers when TDI is required, the five specs that matter, and the verification points that catch wrong selections.
If you inspect flat, moving surfaces — wafers, display panels, PCB, battery electrode, film or paper — you will eventually meet TDI. This is the guide we use internally before recommending a TDI camera, and it is where the most expensive selection mistakes happen.
When TDI is the right answer
TDI (Time Delay Integration) exposes the same object point multiple times as it passes the sensor, summing charge synchronously. Sensitivity scales with the number of integration stages while noise grows far more slowly. You need TDI when three conditions coincide:
- Continuous web or stage motion — the object never stops, so exposure time is set by line rate and magnification, not by a strobe.
- Very high resolution across the web — thousands to tens of thousands of pixels across the scan direction.
- A tight light budget — short exposure per line, or illumination that must stay within safe intensity limits (UV wafer inspection, sensitive displays).
If your object is stationary during exposure, an area scan camera is simpler and cheaper. If sensitivity is not a constraint, a conventional line scan camera will do. TDI earns its price in the overlap.
The five specs that decide
| Spec | Why it matters | Typical questions to answer |
|---|---|---|
| Sensor & lines | 2K–16K+ across the web; pixel size sets the light budget and diffraction limit | How many pixels across your web width? What is the smallest defect across-scan? |
| Line rate | Lines per second; sets the maximum web speed at your magnification | What is your web speed in mm/s, and your magnification (mm per pixel)? |
| TDI stages | 32/64/96/128 stages; more stages = more sensitivity, more charge-transfer constraints | What is your per-line exposure time? Does it fit your light budget? |
| Bit depth & noise | 8/10/12-bit; subtle defects live in the low bits | What grey-level difference defines your smallest defect class? |
| Interface & optics | CoaXPress or Camera Link; F-mount or custom; telecentric or conventional | Cable length to host? Lens mount and working distance available? |
The three specs nobody puts on the datasheet
- Charge transfer efficiency (CTE). TDI sums charge over many stages; imperfect transfer smears signal along the scan direction. It does not appear on most datasheets — it appears in your image as a ghost trail behind bright features. Ask for it, or test for it.
- Spectral response. UV and SWIR variants exist for wafer and panel inspection, but quantum efficiency at your wavelength is a curve, not a number. Get the QE curve at your wavelength.
- Illumination reality. A TDI camera never works alone: the light budget assumes a specific illuminator at a specific distance. If your light cannot deliver the required irradiance safely, no camera choice rescues the application.
Verification points before the PO
A TDI selection is only as good as its measured behaviour. Our verification plan for TDI candidates covers:
- Resolution across-scan with representative optics (not the sensor’s pixel count) on a standard target.
- Line-rate stability at full rate — sustained, with no dropped lines or frame anomalies over a soak test.
- Noise and shading uniformity at your working line rate and gain, not at the datasheet’s reference conditions.
- CTE check using a bright point or edge target: look for trailing smear along the scan direction at high contrast.
- Spectral response verification for UV/SWIR variants at your actual wavelength.
- Interface behaviour on your host: bandwidth, jitter, and SDK stability under sustained streaming.
Where TDI selections go wrong
- Buying stages you cannot feed. A 128-stage camera with an under-powered illuminator produces a noisy image and a false conclusion that “TDI doesn’t work here”.
- Ignoring the lens. TDI pixels are often small; a lens that cannot resolve them wastes the sensor. Lens and camera must be specified together.
- Confusing pixel size with defect size. Your across-scan sampling must satisfy the same 3–5 pixel rule as any other camera — with margin for alignment drift.
- Trusting the demo image. Demo images are shot with the best light, the best lens and the best part. The verification plan exists to replace the demo image with your data.
The short version
TDI is a physics-driven choice for high-resolution, motion, low-light applications — and it is unforgiving of approximations. Specify sensor, line rate, stages, bit depth and optics as one system, then verify the system before the purchase order.
That is exactly what our Selection Sprint and Sample Verification do for TDI projects: the shortlist is built around your web speed and light budget, and every candidate is verified against your acceptance criteria before you commit.
