When an OEM engineering team evaluates an uncooled thermal camera module, resolution and NETD dominate the first conversation. Those numbers matter. But ask the project manager who just shipped a thermal product what actually burned weeks off the schedule, and the answer rarely involves sensor specs. It involves adding a USB frame grabber because the core only output CVBS. Or designing a dedicated power stage because the module accepted exactly one input voltage. Or spending a sprint on NUC calibration firmware because the core pushed that work to the host processor.

These integration costs are not theoretical. Every extra component on the BOM — a converter IC, a voltage regulator, an interface adapter board — comes with procurement lead time, PCB real estate, firmware overhead, and validation effort. A thermal camera module that handles more of this internally does not just reduce the unit cost. It compresses the development timeline.

This article walks through five thermal camera modules that share a common design approach: each SKU is built around a specific interface combination, and each one pushes integration work onto the module rather than offloading it to the OEM. Across the lineup: four distinct video interface combinations, simultaneous dual-stream output on four of five models, a 4.5V–24V input tolerance that eliminates dedicated power regulation, and on-chip NUC with auto-shutter. The point is not that one model solves everything. The point is that the product line is wide enough that you pick the module that already matches your architecture — you do not add hardware to compensate for what the module should have handled in the first place.

Four Interface Combinations, One Product Line

Most thermal imaging cores ship with a single video output per SKU. You choose USB, or CMOS parallel, or CVBS at purchase time. If your design later needs a second output format, you add external hardware.

These five modules span four distinct interface combinations. Each is a deliberate choice for a different integration profile.

Here is what interface diversity means in practice: you are not locked into one signal path. You do not add a frame grabber because the module only does analog. You do not add a video DAC because the module only does digital. You select the SKU whose interface combination matches what your PCB already expects. And if your product line spans multiple architectures, different SKUs share consistent mechanical footprints.

Integration That Runs on the Module

Interface diversity solves the signal-path problem. On-module integration solves everything else. Three areas move the needle for an OEM schedule.

NUC and Shutter Control — Handled Internally

Microbolometer sensors drift with temperature. Non-uniformity correction recalibrates the pixel array to suppress fixed-pattern noise. Some thermal modules push this to the host: the core streams raw data, and the integrator's firmware team writes the calibration routine, manages shutter timing, and maintains correction tables. That is easily a sprint of firmware effort.

Every module in this lineup runs NUC on its own firmware. The mechanical shutter fires on an automatic schedule. The host can trigger a manual calibration via UART (B640A1, B640A2) or USB host software (B256A1) if needed, but the default auto-shutter mode requires zero host intervention. For a team developing a compact thermal viewer, the module’s automatic NUC handling reduces host-side firmware work.

Power Input — Share the Battery Rail Directly

B640A1 accepts 4.5V to 18V. B640A2 takes 5V to 18V. B256B1 and B960B1 extend that to 5V–24V. B256A1 runs off USB 5V.

A 3S lithium pack at 11.1V nominal, a 12V vehicle bus, a 5V USB rail — the module connects directly to any of these. No dedicated buck converter. No LDO. No filtering network designed around a single-module load. The entire lineup draws 0.8W typical for the A-series and under 2W for the B-series. Passive cooling through the module housing handles the thermal load in most enclosures. The thermal design budget goes to zero.

Mechanical Consistency Across Resolutions

B640A1, B640A2, and B256A1 share the same 21×21mm cross-section, with lengths ranging from 29.8mm to 34.48mm. A 21 × 21 mm cross-section fits within compact gimbal housings, UAV payloads, portable inspection equipment and other space-constrained imaging systems. Because three resolution and interface configurations share this footprint, an OEM can design one enclosure and select different thermal cores for different product variants. Retooling for a resolution upgrade is not necessary.

B256B1 and B960B1 use smaller or different form factors optimized for their target applications — the B256B1 camera head at 19.5×19.5mm, the B960B1 at 22×22mm — but the principle holds within each sub-family.

Where Each Interface Combination Fits

Different products stress different parts of the interface architecture. Here is how the combinations map to common OEM applications:

How the Rest of the Market Handles Interfaces

Comparing interface architectures across manufacturers at comparable price points:

None of these are poor products. The decision turns on interface fit. If a design has PCB space and budget for an adapter board and the firmware team has bandwidth for NUC, a Boson may be the right call. If the requirement is simultaneous analog and digital output from a module that runs directly off a battery rail, the multi-interface approach eliminates components and engineering hours that the single-output alternatives do not.

Seven Things to Verify on Any Thermal Camera Module Datasheet

Before committing to a thermal camera module for an OEM design, here is what to check:

1. Simultaneous output. If the design needs both analog and digital video, confirm the datasheet says "simultaneous" — not "switchable," not "selectable." Verify with the manufacturer. Of the five modules discussed here, four output two streams at the same time.
2. Input voltage range. Wide-range input (4.5–18V or 5–24V) means the module shares a battery rail directly. Narrow input means you add a dedicated regulation stage to the BOM.
3. Where NUC runs. On-module auto-NUC with automatic shutter removes calibration from the firmware schedule. Host-managed NUC adds work. Every module covered here runs NUC internally.
4. SDK platform coverage. A Windows-only SDK creates a problem if the product runs embedded Linux or Android. B256A1 covers all three.
5. Cross-SKU footprint. If the product line may scale to different resolutions later, check whether variants share the same mechanical footprint. B640A1, B640A2, and B256A1 all use 21×21mm cross-section.
6. Connector type and orientation. USB Type-C (B256A1) exits differently than a side-exit SMD connector (B640A1, B640A2). Verify against your enclosure layout before committing.
7. Frame rate at full resolution. Some thermal modules drop frame rate at native resolution. These modules hold 50Hz or 60Hz at full resolution — relevant for tracking moving objects or mounting on a moving platform.