SDAA410 June   2026 AM2611 , AM2612 , AM2612-Q1 , AM2631 , AM2631-Q1 , AM2632 , AM2632-Q1 , AM2634 , AM2634-Q1 , AM263P2 , AM263P2-Q1 , AM263P4 , AM263P4-Q1

 

  1.   1
  2.   Abstract
  3.   Trademarks
  4. 1Introduction
    1. 1.1 The Challenge of Determinism in Robotics
    2. 1.2 Why standard ethernet fails for real-time communication
      1. 1.2.1 Head-of-Line Blocking
      2. 1.2.2 Lack of Time Synchronization
      3. 1.2.3 No Traffic Scheduling
    3. 1.3 Time-Sensitive Networking (TSN) based proposed solution
      1. 1.3.1 What is TSN?
      2. 1.3.2 IEEE 1588 (802.1AS gPTP - generalized Precision Time Protocol)
      3. 1.3.3 IEEE 802.1Q (VLAN)
      4. 1.3.4 IEEE 802.1Qbu/Qbr (IET - Interspersing Express Traffic / Frame Preemption)
      5. 1.3.5 IEEE 802.1Qbv (EST - Enhancements for Scheduled Traffic)
      6. 1.3.6 CPSW Specific hardware features
  5. 2Sample Use Cases: Distributed Motion Control in Robotics
    1. 2.1 Representative scenario
    2. 2.2 Network Topology Requirements
      1. 2.2.1 Why Daisy-Chain?
      2. 2.2.2 Real world applications of daisy chain ethernet solutions
    3. 2.3 Communication Requirements
    4. 2.4 Test Implementation
  6. 3System Overview and Architecture
    1. 3.1 Hardware Architecture
      1. 3.1.1 AM261x LaunchPad
      2. 3.1.2 CPSW Sub-System overview:
    2. 3.2 Software architecture
  7. 4Sample Implementation
    1. 4.1 Standard Ethernet + CPSW InterVLAN routing
      1. 4.1.1 What is Inter-VLAN Routing
      2. 4.1.2 How This Implementation leverages Inter-VLAN Routing:
      3. 4.1.3 Test-1 Benchmarks
    2. 4.2 Integrating gPTP Time Synchronization (IEEE802.1AS)
      1. 4.2.1 What is PTP time synchronization?
      2. 4.2.2 How this implementation uses GPTP time synchronization
      3. 4.2.3 Test-2 Benchmarks
    3. 4.3 Integrating VLAN (IEEE802.1Q)
      1. 4.3.1 What is VLAN?
      2. 4.3.2 How this implementation leverages VLAN
      3. 4.3.3 Test-3 benchmarks
    4. 4.4 Integrating IET Frame Preemption (IEEE802.1Qbu/Qbr)
      1. 4.4.1 What is IET (Interspersed Express Traffic)?
      2. 4.4.2 How this implementation leverages IET
      3. 4.4.3 Test-4 Benchmarks
    5. 4.5 Integrating EST scheduling (IEEE802.1Qbv)
      1. 4.5.1 What is EST?
  8. 5Conclusion
  9. 6Challenges and Debug considerations
    1. 6.1 Network Topology Verification
    2. 6.2 Traffic Flow Analysis
    3. 6.3 Host Port Traffic Monitoring
    4. 6.4 PHY Link Management
    5. 6.5 Packets not forwarded to next node
    6. 6.6 Error Handling and Retries
    7. 6.7 High latency or Jitter for high priority packets
    8. 6.8 gPTP not synchronizing
  10. 7References

2.4 Test Implementation

The test system uses five AM261x LaunchPads in a daisy chain, representing:

  • Spirent/TI AM64x processor: Central motion controller (or gateway to controller)
  • TI AM261x LaunchPad Nodes 1 – 5: Joints in the robot arm
  • The setup uses a 100M RGMII full-duplex link.

The external PC/ Spirent/ TI AM64x processor, acting as the central motion controller, generates packets to each node in the network every cycle. So, at the start of every 1ms cycle, the packets are generated for and sent to the 5th Node (farthest node), then to the 4th node and the last packet to the 1st Node (nearest node).

The reason for sending the first packet to the last node is to avoid network congestion between external PC and Node-1. Traffic in the network:

    1. High priority traffic (priority = 7): 64B Unicast packet sent to every node at start of a 1ms cycle, (burst of 5 packets for 5 AM261x-LPs). As soon as the 5 packets are transmitted, background traffic starts.

    2. Background traffic (priority = 0): 512B background traffic is sent from packet generator to nodes for remaining time in the 1ms cycle (addressed to farthest node first, nearest node last).

      Standard IPG gap (12B) between High priority and Background traffic.

    3. AM261x generated traffic: Every 1ms, each node sends a Unicast packet to packet generator. (516B)
Table 2-1 Packet priorities
Packet priority (VLAN PCP based) Packet size Packet source Packet destination
P-0 (highest) – High priority control traffic 64B Packet generator AM261x Node
P-5 – Node generated traffic 512B AM261x Node Packet generator
P-7 (lowest) – Background traffic 64B or 512B Packet generator AM261x Node

To get a highly accurate timestamp, special timestamping hardware like ProfiShark or other hardware network taps can also be used. One thing to note here is that when Wireshark is used with Windows, the timestamps observed are not accurate to the scale of microseconds and the results seen might differ. The AM261x can support up to 8 different priorities based on application requirements.

The flow of packets into the daisy chain network is shown below:

Ethernet traffic in a cycleFigure 2-1 Ethernet traffic in a cycle
Sample packet propagation in the daisy chain networkFigure 2-2 Sample packet propagation in the daisy chain network

The cycle starts with sending five high priority P0 frames to the Daisy Chain network, followed by low priority background traffic to the network until the end of the cycle.

The round-trip time is calculated for each node for the high priority traffic. This round-trip time includes the wire propagation time, CPSW processing and switching time for both, packet transmit from external packet generator to the daisy chain network and back to the packet generator. This setup helps in validating:

  • Whether TSN can maintain sub-150 microsecond latency to the farthest node (Node 5)
  • How latency scales with the number of hops
  • Whether frame preemption effectively isolates high-priority traffic from background load
  • Whether gPTP maintains sufficient synchronization accuracy.

While the actual robot arm can have more than five joints, the five-node test chain provides sufficient validation of the scaling behavior and establishes a reference for engineers to extrapolate performance for longer chains.