MARS400-A10-04-C
The NEXCOM MARS400-A10-04-C is a Programmable Automation Controller (PAC) built on the NVIDIA Jetson AGX Orin, delivering 275 TOPS of AI inference and a certified EtherCAT Class A master with 250 μs cycle time in one 172.6 × 145 × 109.5 mm controller. The 12-core ARM Cortex-A78AE CPU and 2048-core Ampere GPU run deterministic motion control and real-time AI vision on the same module — eliminating the two-box PLC-plus-vision-server architecture.
NEXCOM's NexECM EtherCAT stack supports up to 64 slaves, CiA 402 drive profiles and cable redundancy. I/O includes 5× LAN, 4× COM (2× isolated RS-232/422/485), 8-in/8-out isolated GPIO, 64 GB eMMC + 1 TB NVMe SSD, and a 24–48 V DC input rated for −20 °C to +50 °C.
Key Features
- NVIDIA Jetson AGX Orin — 275 TOPS: 12-core ARM Cortex-A78AE + 2048-core Ampere GPU with 64 Tensor Cores; 64 GB LPDDR5 at 204.8 GB/s keeps inference throughput memory-fed.
- EtherCAT Class A master (NexECM): 250 μs cycle time, up to 64 slaves, CiA 402 servo drive profiles, cable redundancy in ring topology — on a dedicated EtherCAT LAN port.
- Tiered storage: 64 GB eMMC 5.1 for OS and runtime, 1 TB PCIe Gen4 ×4 NVMe SSD for models, logs and camera archives, plus Micro SD.
- 5× LAN + 4× COM: Physical separation of EtherCAT, vision, MES and management networks; 2× RS-232 + 2× RS-232/422/485 with 2.5 kV isolation for serial PLCs and drives.
- 8+8 isolated GPIO: Drives actuators, lamps and interlocks directly — no external I/O module for basic on/off logic.
- AI-to-motion loop on one SOM: Vision inference results feed the EtherCAT motion loop without inter-system network round-trips, cutting cycle time versus a separate vision server.
- Wireless-ready: M.2 Key-B (4G/5G + SIM), M.2 Key-E (Wi-Fi/BT), HDMI 2.1, 4× USB 3.2; CE/FCC Class A.
Specification Overview
| Parameter | MARS400-A10-04-C |
|---|---|
| AI module | Jetson AGX Orin, 275 TOPS, 64 GB LPDDR5 |
| EtherCAT | NexECM Class A, 250 μs, 64 slaves, CiA 402, cable redundancy |
| LAN / COM | 5× RJ45 / 4× COM (2× isolated RS-232/422/485) |
| GPIO | 8× isolated DI + 8× isolated DO |
| Storage | 64 GB eMMC + 1 TB M.2 NVMe PCIe Gen4 ×4 |
| Power / Temp. | 24–48 V DC / −20 °C to +50 °C |
| Dimensions | 172.6 × 145 × 109.5 mm |
Application Use Cases
Vision-Guided Robotic Assembly
The GPU localises parts from 3D camera frames whilst the EtherCAT master drives the servo axes at 250 μs — the AI result passes to the motion loop inside one controller, removing the 5–20 ms round-trip of a two-box architecture.
AI-Augmented CNC Quality Control
CNC servo drives connect via EtherCAT (CiA 402) whilst in-line dimensional inspection AI runs on the Ampere GPU from a USB 3.2 camera; quality classifications go to the MES via the isolated COM port.
AMR Fleet Controller
SLAM, obstacle-avoidance AI and EtherCAT wheel-drive control run on one SOM; the 4G/5G M.2 slot handles fleet backhaul and the −20 °C rating covers cold-storage operations.
Regulated Production with AI Inspection
In pharmaceutical and food lines, the EtherCAT master drives conveyors whilst per-product inspection results are logged in real time to the 1 TB NVMe SSD alongside motion data for batch-level traceability audits.
Why Choose the MARS400-A10-04-C
AI and motion on one controller. Replacing the PLC + vision-server pair removes hardware cost and the inter-system latency that limits vision-guided cycle times — sub-millisecond AI-to-motion latency versus 5–20 ms across a network.
Production-tested EtherCAT stack. NexECM is NEXCOM's own master implementation, proven across its controller lines — deterministic 250 μs cycles, out-of-the-box CiA 402 drive compatibility and stack-level cable redundancy, not a generic open-source port.
Full Jetson ecosystem. TensorRT, CUDA, DeepStream, ROS 2 and Isaac ROS run natively, cutting AI application development time versus custom-SDK embedded platforms.
Frequently Asked Questions
What EtherCAT performance does the MARS400-A10-04-C achieve?
The NexECM EtherCAT Class A master runs a 250 μs minimum cycle time with up to 64 slaves in line or ring topology, CiA 402 drive profiles and cable redundancy, on a dedicated LAN port.
What AI throughput does it deliver?
275 TOPS (INT8) from the Jetson AGX Orin's 2048-core Ampere GPU and 64 Tensor Cores, with 64 GB LPDDR5 at 204.8 GB/s bandwidth.
How many COM ports are included?
Four: 2× RS-232 and 2× RS-232/422/485 with 2.5 kV isolation for electrically noisy environments.
What storage is included?
64 GB eMMC 5.1 (OS and runtime) plus a 1 TB PCIe Gen4 ×4 M.2 NVMe SSD (models, logs, archives) and a Micro SD slot.
What power input does it require?
24–48 V DC, covering standard 24 V PLC panel supplies and 48 V industrial bus systems, at −20 °C to +50 °C ambient.
The NEXCOM MARS400-A10-04-C is a Programmable Automation Controller (PAC) built on the NVIDIA Jetson AGX Orin, delivering 275 TOPS of AI inference and a certified EtherCAT Class A master with 250 μs cycle time in one 172.6 × 145 × 109.5 mm controller. The 12-core ARM Cortex-A78AE CPU and 2048-core Ampere GPU run deterministic motion control and real-time AI vision on the same module — eliminating the two-box PLC-plus-vision-server architecture.
NEXCOM's NexECM EtherCAT stack supports up to 64 slaves, CiA 402 drive profiles and cable redundancy. I/O includes 5× LAN, 4× COM (2× isolated RS-232/422/485), 8-in/8-out isolated GPIO, 64 GB eMMC + 1 TB NVMe SSD, and a 24–48 V DC input rated for −20 °C to +50 °C.
Key Features
- NVIDIA Jetson AGX Orin — 275 TOPS: 12-core ARM Cortex-A78AE + 2048-core Ampere GPU with 64 Tensor Cores; 64 GB LPDDR5 at 204.8 GB/s keeps inference throughput memory-fed.
- EtherCAT Class A master (NexECM): 250 μs cycle time, up to 64 slaves, CiA 402 servo drive profiles, cable redundancy in ring topology — on a dedicated EtherCAT LAN port.
- Tiered storage: 64 GB eMMC 5.1 for OS and runtime, 1 TB PCIe Gen4 ×4 NVMe SSD for models, logs and camera archives, plus Micro SD.
- 5× LAN + 4× COM: Physical separation of EtherCAT, vision, MES and management networks; 2× RS-232 + 2× RS-232/422/485 with 2.5 kV isolation for serial PLCs and drives.
- 8+8 isolated GPIO: Drives actuators, lamps and interlocks directly — no external I/O module for basic on/off logic.
- AI-to-motion loop on one SOM: Vision inference results feed the EtherCAT motion loop without inter-system network round-trips, cutting cycle time versus a separate vision server.
- Wireless-ready: M.2 Key-B (4G/5G + SIM), M.2 Key-E (Wi-Fi/BT), HDMI 2.1, 4× USB 3.2; CE/FCC Class A.
Specification Overview
| Parameter | MARS400-A10-04-C |
|---|---|
| AI module | Jetson AGX Orin, 275 TOPS, 64 GB LPDDR5 |
| EtherCAT | NexECM Class A, 250 μs, 64 slaves, CiA 402, cable redundancy |
| LAN / COM | 5× RJ45 / 4× COM (2× isolated RS-232/422/485) |
| GPIO | 8× isolated DI + 8× isolated DO |
| Storage | 64 GB eMMC + 1 TB M.2 NVMe PCIe Gen4 ×4 |
| Power / Temp. | 24–48 V DC / −20 °C to +50 °C |
| Dimensions | 172.6 × 145 × 109.5 mm |
Application Use Cases
Vision-Guided Robotic Assembly
The GPU localises parts from 3D camera frames whilst the EtherCAT master drives the servo axes at 250 μs — the AI result passes to the motion loop inside one controller, removing the 5–20 ms round-trip of a two-box architecture.
AI-Augmented CNC Quality Control
CNC servo drives connect via EtherCAT (CiA 402) whilst in-line dimensional inspection AI runs on the Ampere GPU from a USB 3.2 camera; quality classifications go to the MES via the isolated COM port.
AMR Fleet Controller
SLAM, obstacle-avoidance AI and EtherCAT wheel-drive control run on one SOM; the 4G/5G M.2 slot handles fleet backhaul and the −20 °C rating covers cold-storage operations.
Regulated Production with AI Inspection
In pharmaceutical and food lines, the EtherCAT master drives conveyors whilst per-product inspection results are logged in real time to the 1 TB NVMe SSD alongside motion data for batch-level traceability audits.
Why Choose the MARS400-A10-04-C
AI and motion on one controller. Replacing the PLC + vision-server pair removes hardware cost and the inter-system latency that limits vision-guided cycle times — sub-millisecond AI-to-motion latency versus 5–20 ms across a network.
Production-tested EtherCAT stack. NexECM is NEXCOM's own master implementation, proven across its controller lines — deterministic 250 μs cycles, out-of-the-box CiA 402 drive compatibility and stack-level cable redundancy, not a generic open-source port.
Full Jetson ecosystem. TensorRT, CUDA, DeepStream, ROS 2 and Isaac ROS run natively, cutting AI application development time versus custom-SDK embedded platforms.
Frequently Asked Questions
What EtherCAT performance does the MARS400-A10-04-C achieve?
The NexECM EtherCAT Class A master runs a 250 μs minimum cycle time with up to 64 slaves in line or ring topology, CiA 402 drive profiles and cable redundancy, on a dedicated LAN port.
What AI throughput does it deliver?
275 TOPS (INT8) from the Jetson AGX Orin's 2048-core Ampere GPU and 64 Tensor Cores, with 64 GB LPDDR5 at 204.8 GB/s bandwidth.
How many COM ports are included?
Four: 2× RS-232 and 2× RS-232/422/485 with 2.5 kV isolation for electrically noisy environments.
What storage is included?
64 GB eMMC 5.1 (OS and runtime) plus a 1 TB PCIe Gen4 ×4 M.2 NVMe SSD (models, logs, archives) and a Micro SD slot.
What power input does it require?
24–48 V DC, covering standard 24 V PLC panel supplies and 48 V industrial bus systems, at −20 °C to +50 °C ambient.
Get a Quote
