Camera selection for reading panel displays at a distance

You pointed a consumer webcam at a bench of instruments — voltmeters, a high-voltage supply module — and the readings came back soft and unreadable. That failure is predictable, and so is the fix. Reading markings on instrument front panels from one to five meters is a pixel-budget problem first and a camera-choice problem second. This guide works the numbers: how many pixels a marking needs, what field of view that implies, which focal length delivers it at your working distance, and which cameras and lenses do the job. It ends with mounting and aiming, because at these distances the bracket matters as much as the optic.

Start from the smallest marking you must read

Two different jobs hide inside “monitor the front panel”. Detecting that an indicator is lit, or that a switch moved, needs only 3 to 5 pixels across the feature — the standard detection rule in machine vision. Reading what the panel says is harder: OCR tooling asks for about 30 pixels over a character's height and 10 across its width, and license-plate systems target roughly 100 pixels across the whole plate, the hardest case. A static bench panel sits between those cases, so commit to two numbers: 30 pixels across the smallest character height when software must read it, and 15 to 20 pixels when a person will zoom the saved image and read seven-segment digits by eye.

Call the smallest marking height h. The pixel budget is h ÷ 30, in millimeters per pixel. On the high-voltage module in this application the smallest features run about 5 mm, so 5 ÷ 30 = 0.17 mm per pixel is the design number, and every choice below is arithmetic on it. To turn a field width and that pixel size into a pixel count, the camera resolution calculator does the division.

Two formulas size the whole build

Field of view and focal length follow from the sensor and the working distance:

At panel-viewing magnification m sits near 0.03, so the sizing shortcut f ≈ sensor width × working distance ÷ field of view lands within one lens step of the thin-lens value: order by the shortcut, fine-focus with the ring. The same two formulas worked at close range are in Blister inspection working distance and field of view formula.

Sensor widths below are nominal type figures; a given sensor's exact active area moves them a few percent, never a lens step.

Worked example: 5 mm markings on a 200 mm panel at 2 m

Inputs, each with its source:

• Smallest marking to read: the module's smallest feature, sized in the section above; its 10 mm labels are the easy case.
• Reading threshold: 30 pixels over character height, the OCR figure above.
• Panel width: 200 mm plus a quarter for margin — a 250 mm target field of view. Set this from your own panel; the arithmetic is identical.
• Working distance: 2000 mm, the closest a bracket can sit.
• Sensor: 1/1.7-inch type, 7.5 mm nominal width, about 2000 pixels across — a 2.5 MP class camera.

In words: divide marking height by the pixel target to get pixel size; divide field of view by pixel size to get the pixel count; multiply sensor width by working distance and divide by field of view to get focal length.

• Pixel size: 5 ÷ 30 px = 0.17 mm per pixel.
• Pixels across the field: 250 mm ÷ 0.17 mm per pixel = 1500 px, which a 2000-pixel sensor clears.
• Pixel size achieved: 250 mm ÷ 2000 px = 0.125, call it 0.13 mm per pixel.
• Pixels on the marking: 5 ÷ 0.13 mm per pixel ≈ 38 px, a quarter above the bar.
• Focal length: 7.5 mm × 2000 mm ÷ 250 mm = 60 mm.

Result: at 2 m, a 1/1.7-inch 2.5 MP camera behind a 60 mm lens puts 38 pixels on each marking. A 50 mm lens widens the field to 7.5 × 2000 ÷ 50 = 300 mm and still leaves 5 ÷ 0.15 = 33 pixels on it — take 50 mm for context around the panel, 60 mm for margin on the digits. The decision line: at 30 pixels, software reads; at 15 to 20, a person reads; below that you are detecting, not reading. Turning the frames into numbers is its own task, covered in Digital meter OCR: capture and read numeric displays.

From five meters: the trade

Distance does not change the pixel budget — pixel size is set by field of view and pixel count — it changes the focal length and punishes loose mounts. The same 250 mm field from 5000 mm needs f = 7.5 × 5000 ÷ 250 = 150 mm, and one degree of yaw swings the frame 5000 × tan 1° = 87 mm, a third of the panel lost to a nudge. The better buy at 5 m is pixels, not glass: a 12 MP camera (4000 × 3000) behind the same 60 mm lens covers 7.5 × 5000 ÷ 60 = 625 mm at 625 ÷ 4000 = 0.16 mm per pixel — 32 pixels on the smallest marking, with a whole rack section in frame instead of one panel.

Decision: if you can move the camera to 2 m, do it, and buy the 2.5 MP camera with a 50 or 60 mm lens. If 5 m is fixed, buy the 12 MP camera and keep the 60 mm lens. Buy the 150 mm lens only when a single panel must fill the frame and the mount is bolted down.

Why the webcam came up short

Run the same arithmetic on the camera that disappointed you. A 1080p webcam with a 60-degree horizontal field frames 2 × 2000 × tan 30° ≈ 2300 mm of bench at 2 m. Spread over 1920 pixels that is 1.2 mm per pixel, so the smallest marking gets 4 pixels — the detection floor, never a reading. Its focus is factory-set for a face at half a meter, its exposure hunts as displays and bench lighting change, and its rolling shutter smears anything that moves. None of that is a defect; it is a half-meter camera asked to work at four times its design distance. What an industrial camera adds is a lens you choose and exposure you control, not just megapixels — Webcam vs industrial camera for simple vision tasks itemizes the differences.

Choosing the camera

Three buys cover this job, all of them area scan cameras:

The MV-CU020-90GC is a 2.5 MP GigE camera with a 1/1.7-inch sensor, a global shutter, color and a C-mount. That is the single-panel, 2 m answer. The LS-CMR-0166 is a 12 MP USB3 camera with a 1/1.7-inch sensor, 4000 × 3000 resolution and 30 fps. That is the 5 m and multi-panel answer.

The MV-CS200-10UM is a 20 MP USB3 camera with a 1-inch sensor, an IMX183, mono and a C-mount. That is the wide-group answer: a 20 MP 1-inch sensor runs about 13 mm wide with roughly 5400 pixels across it, so behind a 35 mm lens at 2 m it frames 13 × 2000 ÷ 35 = 740 mm — three or four panels — at 0.14 mm per pixel, 36 pixels on the smallest marking.

Rolling shutter is acceptable while panels sit still; move to global shutter if the camera will pan or if a display's scan interacts with the sensor's. Color earns its keep when you must tell wire jackets or LED tints apart; mono renders cleaner digit edges for OCR. USB3 suits one or two cameras beside the PC; GigE suits a camera per bench with 100 m runs on standard cable. Two deeper reads: Small pixel camera selection: resolution versus usable detail and High resolution machine vision cameras: how many pixels?. The camera finder filters by shutter, format and interface.

Choosing the lens

Order the lens on three lines: the focal length from the example, an image circle at or above the sensor format, and a resolution rating at or above the sensor's pixel count. Two lenses carry the two mounting styles:

The LS-FFL-0679 is a 35 mm C-mount machine vision lens rated 6 MP on 2/3-inch sensors. The LS-MBL-0031 is a 25 mm M12 board lens rated 5 MP on 1/1.8-inch sensors. Where they land: on the 2.5 MP camera at 2 m, the 35 mm frames 7.5 × 2000 ÷ 35 = 430 mm — two adjacent panels — and puts 5 ÷ 0.22 = 23 pixels on the smallest marking, person-readable but short of the 30 pixel software bar. The 25 mm M12 is the hidden-install answer: on a board camera at 1000 mm it frames 7.2 × 1000 ÷ 25 = 290 mm, and a 2000-pixel board sensor delivers 0.15 mm per pixel, about 34 pixels on it.

Board cameras have their own lens guide — M12 lens selection for board cameras: format, focal length and IR cut — and the general walk-through is How to choose a machine vision lens focal length. The lens finder filters the 50 to 100 mm C-mount list, where the 60 mm class lives.

Chain the camera, lens and bracket in one pass, starting at the camera step:

Mounting, aiming and light

Asking for a camera that turns left-right and top-down is the right instinct aimed at the wrong hardware. At 2 m, one degree of yaw moves the frame 2000 × tan 1° = 35 mm, a seventh of the panel, so you aim once and lock. Put each camera on a ball head or an articulating bracket, tighten it and leave it; a motorized pan-tilt spends money the pixel math has already spent, and a PTZ webcam pays for its zoom range with the small sensor that failed you. Two fixed cameras read more than one movable one.

On high-voltage benches, mount and aim with the supply off and discharged, and keep bracket and cables outside the module's clearance envelope.

Instrument windows are glossy, so keep the camera a few degrees off the mirror angle of the room lights. If digits bloom, shorten the exposure and light the bezel from the side — Automatic iris, exposure and gain control in machine vision covers the control loop. Multiplexed LED and VFD readouts can band at short exposures: if you see stripes across the digits, lengthen the exposure past the display's refresh period.

Checks before you order

• Measure the smallest marking with calipers — the whole build divides it by 30.
• Measure the working distance from lens front to panel with the bracket you will actually mount.
• Run the arithmetic: field of view ÷ horizontal pixels, then marking ÷ that number; at least 30.
• Match the lens's rated image format to the sensor's format or larger.
• Match resolution rating to sensor pixels: a 6 MP lens belongs in front of the 2.5 MP camera, not the 12 MP one.
• Match mounts, and know what an adapter changes — C mount vs CS mount: adapters, focus and clearance.
• Test one exposure on the real display for banding before you cable the room.
• USB3 for one or two cameras within a few meters; GigE for long runs and multi-camera benches.

When the usual answer is wrong

One camera for the whole wall: a 3000 mm wall at 2 m on a 1/1.7-inch sensor calls for f = 7.5 × 2000 ÷ 3000 = 5 mm — fisheye territory — and 3000 ÷ 2000 = 1.5 mm per pixel, three pixels on the smallest marking. That detects lit indicators and nothing more; split the wall between cameras.

More megapixels as a cure-all: pixels only count when the lens resolves them. A 6 MP lens in front of a 12 MP sensor delivers soft pixels that no software sharpens. Lens contrast and resolution: evaluating fine detail shows how to read the MTF curves before you buy.

Recessed readouts: a display 100 mm behind the bezel stresses depth of field. At m = 0.03 with the circle of confusion set to two pixels of the 3.75 µm pitch (7.5 µm), depth of field runs 2 × N × C × (1 + m) / m²: about 70 mm at f/4, about 135 mm at f/8. Stopping down costs two stops of light, so add light or lengthen the exposure rather than closing the iris further.