Simulated camera resolution and pixel scale: set the FOV right

Simulating a vision cell before the cameras arrive is the right order of work, and it breaks in a predictable place: the view window opens at the buffer size. A 4024 × 3036 simulated camera floods a 1920 × 1080 screen, and every API grab re-opens that oversized window until the tool hangs and must be restarted. The fix is to separate optics from display. This guide derives the field of view and pixel scale from focal length, sensor size and working distance, keeps the preview at screen size, and captures full-resolution images headlessly — so the station you validate is the station you order, cameras included.

Derive the field of view from the optics

The simulation needs four inputs, not a format name: focal length f, working distance WD, buffer resolution and pixel pitch. A 1/1.7" label is a size class, not an active area — compute the width from the buffer: 4024 pixels at 1.85 µm span 7.44 mm, 3036 pixels span 5.62 mm. With f and WD in the same units, the exact thin-lens relations are:

Use the exact form whenever the magnification m exceeds about 0.05. The pinhole shortcut FOV ≈ width × WD/f overstates the field by m itself: 5% out at m = 0.05, 10% out at 0.10. The same geometry drives real lens picks — it is the working distance and field of view formula, and object-space pixel size from magnification tabulates the scale relation.

Worked example: 50 mm at 523 mm on a 12 MP sensor

Inputs: f = 50 mm, WD = 523 mm, buffer 4024 × 3036, pitch 1.85 µm.

• Active area: 4024 × 1.85 µm = 7444 µm = 7.44 mm wide; 3036 × 1.85 µm = 5617 µm = 5.62 mm high; diagonal sqrt(7.44² + 5.62²) = 9.33 mm — the 1/1.7" class in real numbers.
• Magnification: m = 50 / (523 - 50) = 50 / 473 = 0.1057.
• Field of view: 7.44 / 0.1057 = 70.4 mm wide; 5.62 / 0.1057 = 53.1 mm high.
• Pixel scale: 70.4 mm / 4024 px = 17.5 µm per object pixel. Cross-check: 1.85 µm / 0.1057 = 17.5 µm — 57 pixels per millimetre on the part.
• Shortcut check: 7.44 × 523 / 50 = 77.9 mm, 10.6% too wide — the m-sized error predicted above.

Enter FOV 70.4 × 53.1 mm at 523 mm and the 4024 × 3036 buffer; every length the simulation then reports reads in true scale.

The calculator runs this focal-length and FOV relation, and its button filters machine vision lenses to the focal length you need.

Keep the preview window at screen size

A 4024 × 3036 view holds 12.2 million pixels against 2.07 million on a 1920 × 1080 screen — 5.9 times more than the display has, which is the whole lag. Set each view window to a fixed screen-sized value and let the tool scale the buffer into it: at 1920 × 1080 the frame fits at 36% (1080 / 3036); at 960 × 540 it fits at 24%, leaving room for a camera grid. The buffer stays 4024 × 3036; scaling touches only what is drawn. If the camera properties offer an 'Allow Resize' checkbox, that governs dragging the window after it opens — it does not shrink the initial size, so set the window dimensions explicitly.

Capture full-resolution images without a live window

1. Create the camera with the optics: f = 50 mm, WD = 523 mm, active area 7.44 × 5.62 mm, buffer 4024 × 3036.
2. Frame a 70 × 53 mm rectangle at 523 mm. It should fill the view edge to edge; if it does not, the FOV field is wrong, not the rectangle.
3. Fix each view window explicitly at 1920 × 1080, or 960 × 540 in a multi-camera grid, so no buffer size ever resizes the display.
4. Grab in snapshot mode — the SNAPSHOT option of the simulator's camera-add call — which writes the image to file or buffer without opening a live view.
5. Where your tool insists on a view, open it minimized (the MINIMIZED option) and read the image from the buffer.
6. For a seven-camera station, keep one handle per camera, snapshot in sequence, and reopen any closed view with setParam('Open', 1) before its next grab. Confirm each returned image is 4024 × 3036; full resolution is the point.

Verify the model before ordering hardware

• Field: the 70.4 × 53.1 mm rectangle fills the frame at 523 mm, and a measured bar reads 70.4 mm exactly. The geometry is closed-form, so any mismatch is a set-up typo, not model error.
• Scale: a 1 mm feature spans 57 pixels. A feature needs 3-5 pixels of sampling, so this set-up resolves 53-88 µm details; a 20 µm defect is out of reach. Close the gap with a shorter working distance, tighter framing or more pixels — the trade-offs are in small-pixel camera selection: resolution versus usable detail.
• Coverage: the lens you buy must cover the 9.33 mm diagonal; apply the margin logic from calibration target field of view and lens coverage margins.
• Motion: rolling shutter skews parts that move during exposure — if parts travel under the camera, plan a global-shutter model or a strobed flash.

From model to hardware

The catalog part behind the simulation: the LS-CMR-0053 is a 12 MP GigE area scan camera on a 1/1.7" rolling-shutter IMX226 sensor, and the LS-CMR-0166 carries the same 1/1.7" 12 MP sensor on USB3 at 30 fps:

For the lens, set the lens finder to 50-100 mm focal length at 12 MP resolution so the purchased optic reproduces the simulated field; the focal-length chooser walks that decision. Where the cell wants a compact short focal length instead, the LS-FFL-0567 is a 16 mm, C-mount lens for a 1.1" format, rated 20 MP with F2.8-22 and <0.3% distortion on a 16-22 mm image circle. That circle clears the simulated 9.33 mm diagonal with margin.

Carry the three results to the counter — FOV 70.4 × 53.1 mm, buffer 4024 × 3036, scale 17.5 µm per pixel. The camera step of the configurator filters area scan cameras to models that match those numbers. If the field of view changes, rerun the three formulas before you shop — two minutes of arithmetic against days of rework; for anything unusual, talk to the applications team.