USB Split Control Transfers

When a host and device share the same USB speed (Full‑Speed at 12 Mbps or High‑Speed at 480 Mbps), transactions proceed directly: the host sends data packets, the device responds, and the protocol remains uniform. However, when a High‑Speed host needs to communicate with a Full‑Speed device, the disparity in signaling and timing requires a special mechanism. Without it, the High‑Speed host’s 480 Mbps signaling would overwhelm a Full‑Speed device limited to 12 Mbps.

This situation commonly arises when a host must support both:

  • A USB‑Net gadget operating at High‑Speed
  • A USB‑Mouse operating at Full‑Speed

Direct transfers to the Full‑Speed device will fail. The USB specification solves this via Split Control Transfers, which break a single control transfer into several phases, allowing the downstream Full‑Speed device to operate at its native speed while isolating it from the High‑Speed link timing.

1. Overview of Split Transactions

A split transaction consists of three phases:

  1. Start-Split (SSPLIT)
    • Initiated by the host controller on the High‑Speed bus
    • Contains the original SETUP packet (bmRequestType, bRequest, wValue, wIndex, wLength)
    • Routed downstream toward the hub’s Transaction Translator (TT)
  2. Micro-Frame Delay
    • The hub buffers the SSPLIT request
    • After one or more 125 µs micro‑frames, the hub issues an internal Full‑Speed SETUP to the attached device
  3. Complete-Split (CSPLIT)
    • Issued by the host to collect the response (DATA and STATUS stages)
    • Commands the hub’s TT to deliver the Full‑Speed device’s reply back over the High‑Speed link

2. Detailed Phases

2.1 SETUP Stage

  • Host → Hub (High‑Speed SSPLIT)
    • PID: SPLIT_START
    • Payload: Standard SETUP packet fields
  • Hub → Device (Full‑Speed SETUP)
    • After buffering, the hub’s TT generates a Full‑Speed SETUP with correct timing

2.2 DATA Stage (if applicable)

  • For control transfers with a DATA stage, the hub performs two split transfers:
    1. SSPLIT (Host → Hub)
      • Indicates DATA–IN or DATA–OUT direction
    2. CSPLIT (Host → Hub)
      • Retrieves or sends data to the Full‑Speed device

2.3 STATUS Stage

  • Always a zero‑length packet
  • Host issues a final SSPLIT to signal the hub to send the Full‑Speed zero‑length status packet, followed by a CSPLIT to acknowledge completion

3. Role of the Transaction Translator

  • Located inside USB hubs capable of handling dual‑speed ports
  • Converts High‑Speed packets into Full‑Speed timing and vice versa
  • Buffers split transactions and ensures micro‑frame alignment

4. When to Use Split Transfers

  • Supporting mixed‑speed devices on the same root hub
  • Ensuring legacy Full‑Speed peripherals function on a High‑Speed bus
  • Required for class‑compliant devices like USB Networking (RNDIS, CDC‑ECM) coexisting with Full‑Speed HID

5. Example Control Transfer Sequence

Phase Host Packet Hub Action Device Packet
SETUP SSPLIT (SETUP) Buffer & translate Full‑Speed SETUP
DATA‑IN SSPLIT (IN) Trigger Full‑Speed IN Full‑Speed DATA
DATA‑IN CSPLIT (CSPLIT) Forward data to host —
STATUS SSPLIT (OUT ZLP) Trigger Full‑Speed ZLP Full‑Speed STATUS
STATUS CSPLIT Acknowledge completion —

6. Key Benefits

  • Compatibility: Allows High‑Speed hosts to support older Full‑Speed devices
  • Efficiency: Minimizes host intervention by batching downstream transactions
  • Transparency: USB class drivers remain unchanged—split handling is managed by the host controller and hub firmware

By leveraging Split Control Transfers and the hub’s Transaction Translator, USB stacks can seamlessly communicate across speed boundaries, ensuring reliable operation of mixed‑speed USB topologies.