16, 20 or 24 MP camera: pick by aspect ratio and format

Three cameras from one sensor generation offer 16, 20 and 24 megapixels on the same 2.74 µm pixel, and a spec table invites you to read that column top-to-bottom as a quality ranking. It is not one. The three arrays differ in shape — wide, square and tall — and your part has a shape too. The right choice falls out of two numbers (field of view and smallest defect) plus one constraint (lens image circle). This guide gives you the arithmetic, a worked example on a 200 × 115 mm part, the bandwidth cost of each option and the checks to run before ordering.

One pixel, three shapes

The trio belongs to one stacked-CMOS global-shutter generation with the same 2.74 µm pixel: the IMX532, IMX531 and IMX530. Per-pixel detail and lens MTF demands therefore match across all three, and the arrays differ only in geometry. Geometry decides coverage, lens selection and bandwidth — which is why the decision below is about shape, not size.

Active areas are array counts times the 2.74 µm pitch, and the 16.8 and 17.5 mm diagonals follow from √(width² + height²); 19.3 mm is the sensor data-sheet figure. Note the trap: the 20 MP array is 816 columns narrower than the 16 MP array (4512 vs 5328). Megapixels measure area, not width — and your field of view cares about width.

The arithmetic that picks the array

Two short formulas settle the pixel budget:

Budget 3 pixels across the smallest defect for reliable detection, and 4–5 pixels when the defect must be measured rather than flagged. The calculator below runs the same arithmetic from your field of view and feature size, and its button opens the camera finder filtered to the pixel count:

Worked example: a 200 × 115 mm bracket

Flat bracket, one overhead camera, field of view 200 × 115 mm — an aspect of 1.74:1, nearly the wide array's own 1.75:1. Smallest defect that matters: a 0.12 mm scratch. Budget: 3 pixels across it.

Required columns: (200 mm / 0.12 mm) × 3 = 5000 px. Required rows: (115 mm / 0.12 mm) × 3 = 2875 px.

• IMX532: 5328 ≥ 5000 (6.6% margin) and 3040 ≥ 2875 (5.7% margin) — fits.
• IMX531: 4512 < 5000 — 11% short on width despite the bigger megapixel figure.
• IMX530: fits, but you would run a central 5328 × 3040 ROI and stream 24.6 MB frames to use 16.2 MB of them.

Pixel scale on the part: 200 mm / 5328 = 37.5 µm per pixel, and the scratch lands at 0.12 mm / 0.0375 mm = 3.2 pixels. Decision: the 16 MP array matches this part in shape and budget. Setting that pixel scale honestly is where most aspect-ratio mistakes start — the simulated camera resolution and pixel scale guide shows the workflow.

Image circle: the constraint that eliminates cameras

Inch-type format names go back to vidicon tubes: a 1-inch type covers a circle of about 16 mm diagonal, so 1.1-inch means roughly 17.6 mm and 1.2-inch roughly 19.2 mm. The image sensor format convention records that 16 mm rule.

Against the trio:

• 16.8 mm diagonal: the corners sit 0.4 mm outside a 1-inch circle, so a 1-inch lens vignettes all four corners. Use a 1.1-inch lens.
• 17.5 mm diagonal: clears a nominal 1.1-inch circle by 0.1 mm. Choose by the lens's printed image circle rather than the format name on the box; margins this thin deserve the millimetre figure.
• 19.3 mm diagonal: 0.1 mm beyond a nominal 1.2-inch circle. Specify a lens with a printed circle of 19.3 mm or more. No M12 (S-mount) lens reaches that, which rules out compact board-lens builds at 24 MP — and cameras this size move to M58, M72 or F-mounts, because the mount's clear opening must pass the whole circle.

Not every 20 MP camera needs the 1.1-inch class: the MV-CS200-10UM puts 20 MP on a 1-inch sensor over USB3, small enough for standard 1-inch optics.

What the extra pixels cost in bandwidth

Frame size is columns × rows × bytes per pixel. At 8 bits (mono or Bayer colour) a full frame weighs 16.2, 20.4 or 24.6 MB across the trio. Practical payload rates: GigE streams near 110 MB/s against a 125 MB/s line rate, USB3 near 360–400 MB/s, and 10GigE or CXP-12 links near 1.2 GB/s.

The fps columns are frame MB divided by throughput. A data-sheet anchor: the LS-CMR-0166 streams 4000 × 3000 at 30 fps over USB3. That works out to 12 MB × 30 = 360 MB/s — the top of USB3's practical range — and the same stream would carry a 24.6 MB frame at about 14.6 fps.

So the fast end of this decision is an interface decision: the MV-CH250-90YC-M58S-NN pairs a 25 MP global-shutter sensor with a CXP-12 link and an M58 mount sized for the big circle.

Motion adds a second axis: global shutter freezes the whole array at once while rolling shutter scans row by row — the global vs rolling shutter comparison covers when the difference bites, and RGB camera frame rate limits continues the bandwidth story. Readout mode moves these numbers too, as 4x4 binning vs 2x2 shows.

Which resolution wins, case by case

• Wide parts, 1.6:1 aspect or wider: the 16 MP array — the most columns per megapixel of the trio, the smallest diagonal (16.8 mm), the friendliest lens and bandwidth budget.
• Square parts, round parts, or parts you re-orient in the fixture: the 20 MP square, which keeps a rotating field inside the lens circle. Quick check on a 150 × 150 mm field with a 0.15 mm defect at 3 px: (150 / 0.15) × 3 = 3000 px per side. The square array gives 4512 (a 50% margin); the wide array's 3040 rows clear 3000 by only 1.3% — too little room for alignment drift.
• Detail in both dimensions: the 24 MP array at 5328 × 4608, provided the lens circle reaches 19.3 mm and the interface moves full frames at your line speed.
• Conveyor speed over GigE: none of the three at full frame; run 16 MP with a central ROI, or step up to USB3 and beyond.
• Anything that moves: a global-shutter variant of whichever array wins.

When more megapixels is the wrong answer

Pixel count is a budget, not a goal. If the arithmetic lands near 4000 columns or below, a 12 MP camera on a 1.1-inch format keeps the coverage with lighter bandwidth. The MV-CH120-11UC is exactly that class — and its global shutter covers anything that moves.

Oversampling is the second way the megapixel number misleads: when the array delivers twice the pixels your defect budget needs, you stream twice the data for no extra information — bin or crop back to frame rate. And on colour sensors the megapixel figure overstates resolved detail anyway, because demosaicing interpolates: color camera spatial resolution puts numbers on that.

Checks before you order

1. Write down field of view and smallest feature, then compute columns and rows at 3 px (detection) or 4–5 px (measurement); keep 10% margin on each axis.
2. Add the sensor diagonal and match it to a lens's printed image circle at your working distance — zoom and close-focus positions shrink usable circles. The calibration target field of view guide shows how coverage margins behave in practice.
3. Match mounts, not just lenses: C-mount handles the 16.8–17.6 mm diagonals; beyond that expect M58, M72 or F-mounts on camera and lens alike.
4. Multiply frame MB by required fps and compare with the interface ceilings above; keep 20% headroom for bursts.
5. Choose shutter type separately from resolution: global for anything that moves.
6. Light the whole field, not just its centre — balance across large fields drifts fast, as the large field of view lighting guide quantifies.

Where to go next

Filter the area scan cameras by the resolution class your arithmetic lands in, then match the sensor format to a lens circle on the C-mount lenses page. The camera finder filters both axes at once, and the camera step of the configurator assembles camera, lens, light and controller around your part. Questions that survive the arithmetic go to the contact page; more sensor-side reading sits in the cameras topic.