Subscribe to Our Newsletter
Simply enter your email to receive the latest news and insights from Pomeas. Stay connected with Pomeas and be the first to discover new innovations in optical excellence.
A practical machine-vision example using a 2/3" 5 MP camera at 80–100 mm working distance for chip surface inspection.
| Item | Value |
|---|---|
| Camera | HIKROBOT MV-CS050-10GC, 5 MP, 2448 × 2048, C-mount |
| Sensor class | 2/3" (active area ≈ 8.45 × 7.07 mm, 3.45 μm pixels, calculated) |
| FOV range | 10 × 10 mm → 4 × 4 mm |
| Working distance | 80–100 mm |
| Minimum feature | 5–10 μm |
| Targets | Scratches, pits, edge chipping, fine surface texture, IC surface defects |
The required magnification is calculated from the sensor and the FOV — before looking at any nominal zoom ratio:
β = sensor short side ÷ required FOV
Both candidate POMEAS zoom bodies cover 0.71–1.77× natively. The TV adapter and auxiliary lens do not create this range — they decide where in the zoom travel each FOV sits, and with it the working resolution and depth of field.
Object-side sampling (calculated): at 4 mm FOV, 4/2448 = 1.63 μm/px (H) and 4/2048 = 1.95 μm/px (V); at 10 mm FOV, 4.09 / 4.88 μm/px.
| Feature | At 4 mm FOV | At 10 mm FOV | Engineering verdict |
|---|---|---|---|
| 5 μm | ≈ 2.6–3.1 px | ≈ 1.0–1.2 px | Borderline at 4 mm (also near the optical-resolution limit); not detectable at 10 mm |
| 10 μm | ≈ 5.1–6.1 px | ≈ 2.0–2.4 px | Robust at 4 mm; marginal at 10 mm |
A properly matched native 1X TV optical path is generally preferred for high-resolution inspection when the required FOV can be achieved through a matched objective or auxiliary lens. Low-quality or small-diameter 0.5X TV adapters can cause sampling-density reduction, edge-image degradation, additional aberrations and lower contrast — but a high-quality, correctly matched TV adapter also works normally in machine vision systems.
A TV adapter acts on the camera side (image side): it scales the image onto the sensor. An auxiliary lens (front lens) acts on the object side: it changes magnification, FOV and working distance of the front group. They are not interchangeable "0.5×" parts: a 0.5X TV adapter halves the image-side NA and therefore costs resolution, while a front auxiliary changes the object-side geometry and the WD. High-resolution inspection should keep the 1X TV path whenever the FOV can be reached with a matched front lens.
Chip and IC surfaces are flat and specular. Coaxial illumination sends light down the optical axis; flat areas reflect back into the lens and appear bright, while scratches, pits and edge chipping scatter light sideways and appear dark — exactly the contrast these defects need. The coaxial versions of both zoom bodies carry an integrated beamsplitter, so no extra beamsplitter cube is required. Ring lights are suitable for textured or non-specular surfaces, not for this case.
For a 2/3" 5 MP camera (2448 × 2048) requiring approximately 10 × 10 mm to 4 × 4 mm FOV at 80–100 mm WD, the required magnification is calculated as 0.71× to 1.77× — a zoom ratio of about 2.5:1, which both the 6.5X and 12.5X zoom bodies cover without reduction optics.
A 5 μm feature spans only ≈ 1 pixel at the 10 mm FOV position, so 5 μm inspection results in a hard requirement: it must be performed at the 4 mm FOV position, where 5 μm spans ≈ 3 pixels and remains near the system's optical limit.
When the 5 μm requirement is confirmed, the native 1X TV optical path of Configuration B should be preferred; the 0.5X TV adapter of Configuration A is suitable when 10 μm-class defects are the target and cost is the deciding factor. Both configurations' WD and depth of field should be verified on the actual bench before release.
Divide the sensor dimension by the required object-side FOV. For square FOVs, use the sensor's short side so the full area fits: β = 7.07 mm ÷ 4 mm = 1.77×. Then check the other direction (8.45 ÷ 1.77 ≈ 4.8 mm) to confirm coverage. Do this calculation before comparing nominal zoom ratios.
It depends on the FOV, not on the megapixels alone. At 4 mm FOV a 5 μm feature spans ≈ 2.6–3 pixels (borderline, needs good contrast); at 10 mm FOV it spans ≈ 1 pixel and cannot be reliably detected. Sampling is necessary but not sufficient — lens resolution, illumination contrast and focus stability must also support it.
A reduction adapter scales the image down behind the lens and halves the image-side NA, so the diffraction-limited object-side resolution roughly doubles compared with a native 1X path at the same FOV. How much this matters depends on the adapter's optical quality — a well-matched 0.5X TV adapter works normally, but the loss should be verified on the bench when 5 μm features are the target.
A TV adapter works on the image side: it scales the already-formed image onto the sensor. An auxiliary lens works on the object side: it changes magnification, FOV and working distance of the front group. The same "0.5×" label describes two physically different operations and they should not be treated as interchangeable components.
Chip surfaces are flat and reflective. Coaxial illumination illuminates along the optical axis so flat areas return light to the lens and appear bright, while scratches, pits and chipped edges scatter light and appear dark. This produces the bright-field contrast needed for surface-defect detection on specular samples.
Simply enter your email to receive the latest news and insights from Pomeas. Stay connected with Pomeas and be the first to discover new innovations in optical excellence.
