Every integrator remembers their first vision station, usually because it ran over budget in places nobody predicted and under-delivered in places everyone did. Here is a realistic build list for a first automated inspection cell - a single camera checking parts on a conveyor or indexing table - with honest guidance on where to spend and where to save.
The five components
Camera: $270-$550. Start mid-resolution, not high-resolution. For most first stations a 5 MP global shutter camera is the sweet spot - the MV-CS050-10GM (Sony IMX264, 2448 x 2048, 24 fps, GigE) handles the majority of presence, dimension and surface checks, its 3.45 um pixels tolerate affordable lenses, and global shutter forgives motion. Tight on budget? The 1.6 MP 65 fps color MV-CS016-10GC covers simple presence checks at around $200 less. High-resolution needs (tiny defects over large fields) push toward 12-20 MP - the MV-CU120-10GM or MV-CS200-10GM - but understand the sampling math first, because the lens cost climbs with sensor format.
Lens: $60-$300. A 5 MP-class C-mount FA lens in the right focal length. Two calculations, thirty seconds each: focal length from field of view and working distance, image circle from sensor format. This is not the place to save aggressively - but it is also not the place for a 25 MP lens on a 5 MP camera.
Lighting: $50-$250. The best money-per-image-quality on the whole list. A diffuse bar or ring in the right wavelength, plus a shroud to kill ambient. If parts move fast, add a strobe controller. Underspending here is the number one reason first stations disappoint.
Mounting: $30-$150. Rigid extrusion brackets, a breadboard or a machined plate. Vibration is the silent killer of vision stability; wobbly mounts show up as intermittent "bad images" nobody can reproduce. Overbuild this - it is cheap insurance.
Compute and I/O: $150-$400. An industrial fanless PC or even a respectable mini-PC runs classic vision software comfortably; a GigE NIC with decent drivers costs little. Add a trigger line from your PLC or photo-eye. Avoid the temptation to run the station on the same PC that streams cat videos - deterministic matters.
Total: roughly $600-$1,600 for a serious first station
The spread is dominated by lighting complexity and compute choices, not the camera. That surprises buyers who budget from the camera outward.
Where first builds go wrong
Overbuying resolution. Sampling math first: if your smallest defect is 0.3 mm and your field of view is 150 mm, you need 0.1 mm/px, which is 1500 pixels across - a 2 MP camera suffices. The extra megapixels buy nothing but lens cost and bandwidth.
Underbuying stability. The $40 bracket that flexes costs a week of debugging later.
Skipping the golden part. Keep one known-good and one known-bad part at the station. Every lighting change, every threshold, gets validated against them. It is the cheapest test lab in existence.
Forgetting refurbished stock. Cameras from major lines - the MV-CS050-10GM, MV-CU060-10GM and friends - are available refurbished at 25-40% off with real warranties. For a pilot station that may be rebuilt twice, refurbished is a rational choice, not a compromise.
The growth path
Design station one so camera and light are on separate rigid mounts with spare travel - station two, three and four always want bigger fields, faster lines and different wavelengths. GigE infrastructure scales: the switch that carries one camera today carries four tomorrow, and the same PC drives them all. Build the first station as the seed of the fleet, not a one-off.
Our catalog lists stock for every camera in new and refurbished condition with full specifications, so each line of this budget can be priced against real availability before you commit.