.png)
Serval
Mission-agnostic Data Processing Unit designed for realtime on-board data processing with energy-efficient AI. Built around the AMD Versal AI Edge SoC, Serval Nano DPU combines application and real-time processors, FPGA, DSP resources and AI Engines in a reconfigurable platform. High-speed payload interfaces, NVMe memory and an independent Supervisor enable integration with demanding payloads and spacecraft platforms.
• Supervisor - telemetry, monitoring, boot control and FDIR
• Processing Core - based on the AMD Versal AI Edge Adaptive SoC
• Interface Board - off-the-shelf or tailor-made
• Optional cold-redundant configuration
• AMD Versal AI Edge VE2302
• Dual-core Arm Cortex-A72
• Dual-core Arm Cortex-R5F
• Programmable Logic
• 34 AI Engine-ML tiles
• 464 DSP Engines
• 8 GiB DDR4 with ECC
• 32 GiB pSLC eMMC
• 2 × 64 MiB NOR flash for Versal boot images (redundant, can be updated during the mission)
• 2 × 960 GiB NVMe mass memory (cold-redundant)
• 1 × Gigabit Ethernet
• 4 × 10G GTYP
• 19 × LVDS (XPIO)
• 7 × HDIO
• MIO: SPI, I2C, GPIO
• Supervisor telemetry & control interfaces: 2 × CAN, 2 × UART, PPS, optional SPI/I2C
• SpaceWire
• RS-422
• 4 × 10G Optical Interface
• 10G/40G Ethernet, CoaXPress-over-Fiber, SpaceFibre
Other customisable interfaces available upon request
• Custom 64-bit Linux built with Yocto
• FreeRTOS-based Supervisor firmware
• Vitis AI development workflow
• Qualification for launch and space environments at NASA GEVS levels
• Operating temperature: -25 °C to 75 °C
• Aluminium enclosure with an efficient thermal path and a dedicated ASoC heat-spreading system
• Dimensions: 95.89 mm × 90.17 mm × 39.86 mm (PC/104-compatible)
• Mass: 616 g
• Supply voltage: 6.5 V to 14 V
• Power consumption: 6.5 W to 25 W
Advantages
Compute Power at low SWaP
Adaptive SoC with real-time CPUs, FPGA, AI Engines & high-speed I/O
Algorithm acceleration
Runs classical and AI workloads on board efficiently
Ease of integration
PC/104 form factor with standard or custom Interface Board
Key applications
Earth Observation
Acquires high-rate image data, then applies cloud screening, AI inference and compression onboard, delivering up to 4× less data to downlink from optical and hyperspectral missions.
Autonomous Navigation
Performs real-time image processing and AI inference onboard, supporting autonomous decision-making without waiting for the next ground-station pass.
Scientific and Multi-Instrument Missions
Handles data from several payloads in one unit, with multi-gigabit interfaces, adaptable Interface Boards and redundant mass storage, so a single DPU can serve the whole instrument suite.
Compatibility
Leopard DPU is compatible with the HyperScape100 hyperspectral camera by Simera Sense, enabling real-time onboard data processing, reduced data downlink, and faster access to actionable insights directly in orbit.
Leopard DPU is integrated with Parsimoni's SpaceOS, combining high-performance onboard data processing with secure application deployment, lifecycle management, and in-orbit software updates.
Use cases
John Doe
CEO, XYZ Corp
Jane Smith
CTO, ABC Inc
Mike Johnson
COO, DEF Corp
Sarah Williams
CFO, GHI Inc
Would you like to try it out?
A Smart Mission Lab is a remote testing environment that provides users with access to real satellite hardware, allowing them to evaluate and test it before making a purchase.
Let's talk
Contact us for any inquiries or collaboration opportunities—our team is here to assist you.
