Alvium VSWIR series

Select the sensor by balancing object-space resolution, field of view, pixel size and permitted data rate. The IMX991 and IMX990 use 5 µm pixels and provide 0.30 MP and 1.30 MP respectively, while the IMX993 and IMX992 use 3.45 µm pixels and provide 3.20 MP and 5.30 MP. The smaller pixels support finer spatial sampling when the lens and working distance can resolve the additional detail. The larger pixels reduce the number of pixels required for a given sensor area, which can simplify data handling, although practical sensitivity still depends on illumination, exposure and optics. Engineers should calculate the required object-space micrometres per pixel before choosing the camera model.

USB3 Vision suits short, direct connections to a PC or embedded host, while GigE Vision and 5GigE Vision provide Ethernet-based networking and PoE options. MIPI CSI-2 is intended for close integration with compatible embedded processors and carrier boards, where the host design controls the data path. GMSL2 and FPD-Link III extend CSI-2-based cameras over rugged coaxial or shielded twisted-pair connections through a serializer and deserializer. Coaxial versions support cable runs up to 15 m, while documented STP limits are up to 8 m for GM2 and 10 m for FP3. The final choice should account for bandwidth, cable movement, connector type, power delivery and host-driver availability.

Yes, the sensors cover a specified spectral range from 400 to 1,700 nm, encompassing visible wavelengths and the SWIR band. This enables a single monochrome camera to observe features that may respond differently under visible, near-infrared and SWIR illumination. The camera does not create color-separated spectral channels by itself, so wavelength discrimination requires controlled illumination, filters or a hyperspectral optical arrangement. Lenses, windows and protective materials must also transmit the wavelengths used by the application. In practice, the optical stack and lighting design are as important as the camera when combining visible and SWIR measurements.

Throughput depends on resolution, frame rate and selected pixel format. At Mono8, the 3.20 MP G5 model is specified at 131 fps and approximately 525 MByte/s, while the 5.30 MP G5 model reaches 84 fps at the same stated transport rate. The USB models provide 123 fps at 3.20 MP and 77 fps at 5.30 MP, reflecting the available USB transport bandwidth. Higher bit depths increase the data volume, so 10-bit or 12-bit acquisition may reduce the achievable frame rate or require packed formats. Regions of interest can reduce transmitted pixels and may permit higher acquisition rates when only part of the scene is required.

The lens must cover both the sensor format and the wavelength range used by the inspection. Sensor formats extend from Type 1/4 to Type 1/1.4, so the required image circle grows with the selected resolution and sensor model. C-Mount and CS-Mount options are available across the range, with S-Mount offered on selected models and configurations. A lens designed only for visible imaging may lose transmission, focus consistency or contrast in the SWIR band. Engineers should therefore verify spectral transmission, chromatic focus shift, working distance, aperture and image-circle coverage before finalizing the optical design.

These cameras do not use thermoelectric sensor cooling, so heat transfer into the surrounding mechanical structure needs to be considered during installation. Published operating limits refer to housing temperature and should not be treated automatically as unrestricted ambient-temperature limits. Depending on model and interface, the upper housing-temperature limit is +55 °C, +60 °C or +65 °C, with a typical lower limit of −20 °C. Board-level and open-housing installations also require the system designer to provide a suitable thermal path and EMC treatment. Temperature monitoring can be used during prototype testing to confirm that the intended enclosure, mounting plate and airflow keep the camera within its permitted range.

Available functions include defect-pixel correction, adaptive noise correction, black-level adjustment, contrast, gamma, lookup tables, convolution, lens-shading correction and configurable regions of interest. These operations can prepare or reduce image data before it reaches the host, although exact availability depends on the interface, model and firmware. Multiple ROI and binning can be useful when only defined inspection zones or reduced spatial data are required. Sequencers, counters and timers allow acquisition settings to change in a controlled series without continuous host intervention. The feature set should be checked against the intended software access method, particularly when moving between Ethernet, USB and CSI-2 implementations.

The series provides hardware triggering and programmable GPIOs, with the exact line arrangement determined by the interface family. GigE and 5GigE versions also support Ethernet-oriented functions such as Action Commands and Precision Time Protocol on applicable models. CSI-2, GMSL2 and FPD-Link III variants provide trigger and peripheral control through their camera I/O arrangements and compatible host hardware. Sequencers, counters and timers can coordinate exposure settings with repetitive machine cycles. During system design, trigger latency, cable propagation, host buffering and illumination timing should be validated together rather than assessed as independent components.