Blog

Precision Illumination & Detection for Machine Vision Systems

  • 06 August 2026
Female industrial engineer in manufacturing facility

Machine vision inspection failures rarely start in software — they start upstream, at the moment a photon either reaches the detector cleanly or does not. Opto Diode, a division of ITW based in Camarillo, California, manufactures both halves of that pair — the LEDs and the silicon photodiodes — and how well they are matched is a large part of what separates an inspection line that catches a defect from one that misses it. This guide covers wavelength selection, environmental hardening, and the catalog parts that cover most visible-to-NIR machine vision designs.

400–740 nm Si responsivity band850 & 940 nm NIR illumination-65°C to 150°C hermetic ratingCamarillo, California manufacturing

Why the optical subsystem decides inspection outcomes

The camera and the algorithm get most of the credit, but the matched emitter-detector pair does most of the work. Three failure categories recur across industrial inspection systems, and all trace back to component selection:

  • False rejects under lighting drift. An LED whose output drifts over a shift pulls a thresholded system out of calibration.
  • Missed defects on low-contrast targets. When illumination wavelength sits in the wrong part of the target’s reflectance curve, defects produce less signal than the noise floor.
  • Ambient interference. Overhead fluorescents, sunlight, and adjacent stations leaking infrared into the field of view corrupt signal the detector cannot reject.

Each is addressable at the point where the LED and photodiode part numbers are chosen — before a line of code is written.

Matching an LED to a photodiode

Every photodiode has a spectral responsivity curve — amps generated per watt of light at each wavelength — and every LED has an emission spectrum tens of nanometres wide — 40 nm at half maximum on the OD-850LHT, OD-850WHT and OD-110W, 55 nm on the OD-110WISOLHT. The signal a system produces is the product of those two curves, which is why pairing matters as much as raw output power: a lower-power LED matched to a peak-aligned detector can outperform a mismatched higher-power one.

Opto Diode’s Blue & Red Enhanced silicon photodiodes are specified for responsivity between 400 nm and 740 nm. Within that band the naming refers to which side of the spectrum has been enhanced beyond a standard silicon photodiode’s flatter response: blue-enhanced parts favor the shorter-wavelength end (0.28 A/W typical at 450 nm), red-enhanced parts the longer end (0.35 A/W typical at 632 nm). Silicon keeps responding into the near infrared, but Opto Diode does not publish a specified responsivity for these parts above 740 nm — for 850 nm or 940 nm work, confirm the operating point with applications engineering.

Choosing the wavelength for the inspection task

Different wavelengths reveal different defects. The table below maps common inspection tasks to a starting wavelength and matched detector type.

Wavelength Best for Why it works Matched detector
450 nm (blue) Surface defects, scratches on plastics, cosmetic inspection Scatters strongly off micron-scale features, giving high contrast on light substrates Blue-enhanced Si photodiode
630 nm (red) Color discrimination, label inspection, pharma blister packs Strong contrast on printed marks; good penetration through thin clear films Red-enhanced Si photodiode
685 nm (deep red) Vision-guided robotics, fluorescent ambient suppression Sits beyond peak fluorescent emission, less affected by overhead lighting flicker Red-enhanced Si photodiode
850 nm (NIR) Conveyor vision, part counting, edge detection Outside the visible band operators work in; strong silicon response; high-power emitters available Red-enhanced Si photodiode (confirm NIR operating point)
940 nm (NIR) Outdoor or sun-exposed lines, agricultural sorting Solar irradiance dips near 940 nm (water absorption), improving daylight rejection Red-enhanced Si photodiode (confirm NIR operating point)

Opto Diode’s catalog GaAlAs emitters sit at 810, 850 and 880 nm; the family spans 700–1300 nm, and other center wavelengths — 940 nm among them — are a custom configuration rather than a stock part. Specifying a wavelength is only half the job — for high-speed inspection also confirm the detector’s rise time and junction capacitance against your bandwidth budget.

Designing for the industrial environment

Lab benchtop performance does not predict the factory floor. Four factors do most of the damage and should be specified explicitly.

Thermal stability. LED peak wavelength moves with junction temperature. Opto Diode’s emitter datasheets do not publish a temperature coefficient; for GaAlAs the figure commonly used for planning is 0.2–0.3 nm per °C, which puts a line running 25–60°C at roughly 7–10 nm of movement — enough to walk off a narrowband detector or filter. Specify LEDs in hermetic packaging rated for the full range.

Mechanical durability. Vibration, shock, and washdown cycles stress packages; plastic-encapsulated emitters can crack and admit moisture. Hermetic metal-can packages (TO-5, TO-8, TO-39, TO-46) avoid this.

Optical alignment tolerance. Wide-angle emitters illuminate broadly but waste power; narrower emitters concentrate flux but demand tighter tolerances. Match beam pattern to working distance early in the design.

Electrical noise. Drives, contactors, and solenoids inject noise that can masquerade as optical signal. Low-capacitance photodiodes paired with a right-sized transimpedance amplifier recover SNR without exotic shielding.

Common failure modes and how to engineer them out

Failure mode Root cause Design response
Saturation on bright targets Detector signal exceeds the linear range before the brightest pixel resolves Specify broad-linear-range detectors; reduce optical gain or LED drive rather than sacrificing dynamic range in software
Drift across temperature Dark current rises and LED output decays, eroding thresholds Choose low-dark-current detectors — 1 nA typical / 3 nA max at 10 V reverse on the 5 mm² parts, 3 nA / 7 nA on the 12 mm² parts — and emitters rated across the full operating temperature range
Ambient and strobe interference Overhead lighting, sunlight, or adjacent stations contribute uncontrolled signal Modulate illumination at a known frequency, lock-in detect, and add a bandpass optical filter
Wavelength-mismatch SNR loss LED bin shift across lots moves the emission peak off the detector’s responsivity peak Specify tight LED wavelength binning, or use a detector broad enough to tolerate normal bin variation
Beam non-uniformity across the field of view A single-emitter point source produces a hot center and dark edges Use diffusers or line-source geometries; calibrate per-pixel rather than globally

Opto Diode building blocks for machine vision

Opto Diode manufactures both sides of the pair in Camarillo, California.

Photodiode Active area Spectral optimization Typical inspection use
ODD-5WB 5 mm² Blue-enhanced silicon, TO-5 hermetic package Surface inspection, blue-LED systems
ODD-12WB 12 mm² Blue-enhanced silicon, TO-8 hermetic package Larger field-of-view blue inspection, edge detection
ODD-5WISOL 5 mm² Red-enhanced silicon, TO-5 hermetic package Red and NIR inspection, label and barcode reading
ODD-12W 12 mm² Red-enhanced silicon, TO-8 hermetic package Conveyor vision, larger-area NIR detection
LED / IR emitter Peak wavelength Power output (typ.) Beam Typical use
OD-850LHT 850 nm 22 mW Medium angle (narrower), high-temperature TO-46 Focused NIR illumination in high-temperature environments
OD-850WHT 850 nm 26 mW Wide angle, high-temperature TO-46 Broad-area NIR illumination on conveyor lines
OD-110W 850 nm 140 mW Wide angle, TO-39 High-power NIR for longer working distances
OD-110WISOLHT 880 nm 120 mW Wide angle, high-temperature TO-39 High-temperature, wide-area illumination

Power figures are typical values at the datasheet drive current: 100 mA for the TO-46 parts, 500 mA for the TO-39 parts. Minimum-rated output is lower in each case — check the datasheet before setting a link budget.

When standard parts don’t fit

A minority of designs fall outside standard catalog parts. Opto Diode’s Camarillo manufacturing supports custom active-area sizes and shapes — circular, square, rectangular, quadrant and multi-element layouts — detectors optimized for a specific spectral band including integrated thin-film filters and optical filter windows, photodiode-preamplifier hybrids with custom gains such as the ODA-6WB series, hermetic TO-style packaging and complete optoelectronic assemblies, and emitters configured to a required wavelength, output and beam pattern.

Talk to applications engineering

Contact Opto Diode’s Camarillo, California team to discuss a design, request samples, or scope a custom program. Quote requests get a response within 1–2 business days.

Contact Opto Diode

Download the Product Catalog

FAQ

Should I choose a blue-enhanced or red-enhanced photodiode for machine vision?

+

It depends on the illumination wavelength, though both types share one specified band: Opto Diode publishes responsivity for its Blue & Red Enhanced silicon photodiodes between 400 nm and 740 nm. Blue-enhanced parts such as the ODD-5WB and ODD-12WB favor the shorter-wavelength end — 0.28 A/W typical at 450 nm — and pair with blue illumination for surface-defect and cosmetic inspection. Red-enhanced parts such as the ODD-5WISOL and ODD-12W favor the longer end — 0.35 A/W typical at 632 nm — and pair with red or deep-red illumination for label inspection and bar code reading.

Why does an LED’s output power alone not determine inspection performance?

+

The signal a vision system produces is the product of the LED’s emission spectrum and the photodiode’s spectral responsivity. A lower-power LED matched to a peak-aligned detector can outperform a higher-power LED feeding a poorly matched one.

Why does peak wavelength drift with temperature, and how much should I plan for?

+

LED peak wavelength moves with junction temperature. Opto Diode’s emitter datasheets do not publish a temperature coefficient; the figure commonly used for planning with GaAlAs is 0.2–0.3 nm per °C, which puts a line running from a 25°C cold start to 60°C steady state at roughly 7–10 nm of movement — enough to walk off a narrowband detector or filter. High-temperature-rated emitters buy margin on the package side: the hermetically sealed TO-46 OD-850LHT and OD-850WHT need no derating or heat sink to 80°C, and they and the isolated-case TO-39 OD-110WISOLHT are each rated for a -65°C to 150°C storage and operating range. A package rating does not remove the wavelength shift itself, so budget for it in the optical design.

Why use 850 nm or 940 nm illumination instead of visible light for conveyor and outdoor vision?

+

850 nm is invisible to operators yet still produces a strong silicon photodiode response, and high-power emitters such as the OD-110W are available at that wavelength. 940 nm helps on sun-exposed lines because solar irradiance dips near that wavelength (a water-absorption band).

Reviewed and updated August 2026.

Get a Quotation Now

For any queries, please complete the form and we
will contact you within 48 hours


    Please note: Direct orders must meet a $3,000 minimum.