Board-level vs housed cameras: what you give up

A board-level camera is the sensor, the image engine and the interface of its housed sibling, minus the metal around them. That single subtraction removes EMC shielding, mechanical protection, on-camera frame buffers and, on many families, interface options and firmware features. If you build a camera into your own machine, instrument or embedded carrier, those are exactly the things you must now design in yourself. This comparison lays out what changes between the board and housed versions of one camera family, puts numbers on the interface decision with a bandwidth worked example, and ends with a recommendation per use case, so you can tell which losses matter for your build and which you will never notice.

Two cameras, one sensor

Industrial camera makers usually run one sensor across both a housed line and a board line. The board is a printed circuit carrying the sensor, the interface bridge and a connector set; some versions arrive with a lens mount fitted, others ship as a bare PCB.

The purest case is a board with no lens mount at all:

A bare board means you design the lens holder to the sensor's flange distance and own its alignment. Other board models arrive with the mount already fitted:

And some split the difference with a short CS mount:

The housed equivalents live in the area scan cameras category, with a machined housing, a locked lens mount and industrial connectors. The like-for-like housed part here is a 6 MP USB3 camera on the same 1/1.8-inch sensor format:

Side by side, the bare board and the housed 6 MP USB3 camera share sensor format, interface and shutter type; the difference is packaging:

All 2 share: Sensor type CMOS, rolling shutter; Pixel size 2.4 μm × 2.4 μm; Sensor size 1/1.8"; Resolution 3072 × 2048; Max. frame rate 60.9 fps @ 3072 × 2048 Mono 8; Gain 0 dB to 24 dB; Exposure time 8 μs to 1 sec; Mono/color Mono. Values as printed in each product's data sheet; open a part for its drawing and full specifications.

Because the sensor is shared, image quality is not the trade. Sensitivity, dynamic range and shutter behavior carry over unchanged. What you give up lives entirely around the sensor, and it falls into four groups.

What the housing buys you

Electromagnetic shielding

A metal housing works as a Faraday cage around the sensor and its electronics. Without it, the electromagnetic behavior of a board camera differs from its cased twin: the board radiates more and picks up more. On a clean bench you will never see the difference. Put the same board next to a variable-frequency drive, a servo cabinet, a welder or a radio transmitter, and noise lands in your images as streaks, sparkles or dropped frames that no software filter removes cleanly.

Your enclosure inherits that job: a conductive shield around the board, a real ground path, and filtered or shielded pass-throughs where power, trigger and data leave the enclosure. Our board-level cameras category spans both interface types and every mount flavor:

Mechanical protection and a defined mount

The housing gives a camera its structure: a defined mounting face, a lens mount that carries the optic without loading the PCB, and protection from handling damage during integration and service. On a bare board you rebuild all of it — standoffs at defined points, strain relief on every cable, a lens holder machined to the sensor's flange distance. A board with an M12 or CS mount fitted removes the lens-holder work but not the structural work, because the mount still threads onto the board.

On-camera frame buffers

Housed models in a family usually include frame buffers on the camera; board versions of the same sensor often drop them. The buffer absorbs jitter on the host side — a scheduler, a GUI thread, a shared bus — so a busy host receives complete frames instead of losing them mid-transfer. Continuous streaming into RAM rarely suffers. The loss appears in demanding applications: high trigger rates, burst transfers, or hosts that also run inference, logging or a display. Your design then leans on generous host-side buffering, and that memory must be sized for the high-resolution case — the frame pool out of memory article shows what an undersized pool costs you.

Interface and firmware features

Housed lines normally span the family's widest interface range — GigE and USB3 at minimum, 10GigE or CoaXPress at the top. A board line may carry only one of them, and USB3-only board versions of sensors that also ship in GigE housed models are common. That matters for reach: passive USB3 copper tops out around 3 m, while GigE runs 100 m on standard cable. Feature lists drift too: chunk data, event channels, on-camera lookup tables or compression can exist on the housed SKU and not on the board one. Read the board model's feature list before you commit the design — parity with the housed twin is not guaranteed, in either direction. For how the interfaces themselves compare, see machine vision camera interface comparison: USB to CoaXPress.

Board vs housed at a glance

Worked example: will the interface carry your frames?

Take the 6.3 MP GigE board camera from the previous section, rated at 18 fps. The number that decides whether GigE is a real option for it is the pixel data rate.

At 8-bit mono output:

• Per frame: 6.3 Mpx × 8 bit = 50.4 Mbit
• Per second: 50.4 Mbit × 18 fps = 907.2 Mbit/s = 113.4 MB/s (907.2 ÷ 8)
• Practical GigE Vision payload with jumbo packets: about 118 MB/s
• Headroom: (118 − 113.4) ÷ 118 ≈ 3.9%

It fits, barely. At 10-bit output the same frames need 6.3 × 10 × 18 = 1134 Mbit/s = 141.8 MB/s, and GigE cannot carry them: 118 MB/s is 944 Mbit/s, which divides into 63 Mbit frames about 15 times per second. Your 18 fps sensor becomes a 15 fps camera on GigE. On USB3, where practical payloads land around 350-400 MB/s, 141.8 MB/s runs with 60-65% headroom ((400 − 141.8) ÷ 400 ≈ 65%).

The decision, for this sensor class: at 8-bit, GigE works with jumbo frames but under 5% headroom, so treat any added stream — chunk data, a second region of interest — as overflow. At 10-bit and the full 18 fps, USB3 is the working interface. If you need GigE's 100 m reach at 10-bit, budget roughly 15 fps from this sensor class, or pick a sensor that ships on a faster link. The camera finder filters both categories by interface and resolution, so the check takes a minute.

Which one to choose, case by case

• Your product is the enclosure. Embedded machines, instruments, kiosks, mobile robots: you already provide the shield, the structure and the service path, so a housed camera inside would pay for protection twice. Choose board — after the feature-list check — and start the mechanical design from board level camera housing design from maker STEP files.
• Exposed factory floor. Drives, welders, forklift traffic, technicians under time pressure: choose housed. The housing does the shielding, carries the locking connectors and makes replacement a two-minute swap. The board-side version of the replacement problem is covered in board level camera replacement when a sensor goes end of line.
• Multi-camera embedded rig. Several boards on one carrier keep cabling short and power simple; stability then rests on the match between camera and bridge — board-level camera and USB3 bridge board: match before you stream walks that check.
• Long runs or many cameras on one network. GigE wins, and if the board line of your sensor is USB-only, the housed GigE model is not a luxury — it is the way to get the interface. The host side needs the same discipline; board level camera sensor compatibility: match the host before you buy covers it.
• Harsh environments. Washdown, outdoor, dust: housed, and the rating comes from the whole assembly — housing, lens, connectors — never from the camera alone.

Checks before you order

1. Match the exact model numbers: confirm the board SKU's sensor, shutter and resolution line up with the housed unit you benchmarked.
2. Read the board model's feature list for every camera-side setting you scripted: chunk data, events, on-camera processing, pixel formats.
3. Check the board SKU's interface against the cable plan — reach, bandwidth and multi-camera wiring all follow from it.
4. Decide the mount early: a bare board makes the lens holder and its flange distance your work; board camera lens mount thread sizes: M12 or M6 covers the options.
5. Plan the shield: conductive enclosure, ground path, filtered pass-throughs for power, trigger and data.
6. Add strain relief and an ESD-safe assembly procedure; board connectors are not made for repeated handling.
7. Measure board temperature at maximum duty cycle inside the final enclosure, with the final airflow.
8. Decide the service story: a housed camera swaps in minutes; a failed board can mean enclosure rework.

When the usual answer is wrong

• A clean bench result is not a production result. Electromagnetic behavior changes with the final enclosure, wiring and machine neighbors, so retest with the machine built, not the evaluation kit.
• If the host also runs a GUI, logging or inference, treat a missing frame buffer as a constraint, not a footnote: the failure mode is dropped frames exactly when the machine is busiest. The board camera evaluation checklist builds this into acceptance testing.
• Choosing board purely to save unit cost in a factory installation can backfire: one EMC retrofit can erase the savings several times over.

For the wider selection flow, board level camera selection for embedded vision machines takes the decision from interface to mount; the opposite boundary — consumer against industrial cameras — is covered in consumer camera vs machine vision camera: what you give up. More guides sit on the cameras topic page. Build the shortlist in the camera configurator, and if the trade-off is still borderline for your build, write to us through the contact page.