Similar to other high-speed serial interfaces operating at gigabits-per-second, USB SuperSpeed (SS) defines port-to-port protocols to ensure the reliable and efficient transfer of protocol packets across each link.

Key characteristics of SS port-to-port operation:

• Packet error checks are performed at every link.
• A packet must be successfully transferred on the current link before it is forwarded to the next link.
• These protocols are fundamentally different from the USB 2.0 Token/Data/Handshake model and are not directly comparable.

Core Features of Port-to-Port Protocols

The SS port-to-port protocol is built around three primary features:

• Link Flow Control
• Link Error Detection and Recovery
• Link Power Management

Link Training

Before any data transfer can occur, both link partners must establish a reliable connection through link training. This process uses Low-Frequency Periodic Signaling (LFPS) and Training Sequences (TS1/TS2).

Link Training Stages

LFPS (Low-Frequency Periodic Signaling)

• A low-frequency out-of-band signaling mechanism used during:
  • Connection detection (Rx.Detect)
  • Link state transitions (e.g., waking from U1/U2/U3)
  • Speed negotiation
• LFPS bursts are sent as simple square waves at ~20–40 MHz (much lower than the 5 Gbps data rate)
• Does not require the receiver to be trained or synchronized

Training Sequences (TS1 and TS2)

• TS1 — Sent to train the receiver: achieves bit alignment, symbol lock, and polarity detection
• TS2 — Sent after TS1 is received successfully; both partners exchange TS2 to confirm the link is ready
• Once both sides send and receive TS2, the link transitions to U0 (Active)

Link Flow Control

• Uses a credit-based flow control mechanism.
• Tracks buffer availability at the link partner.
• Operates bidirectionally, with each link partner advertising available credits to the other.
• Prevents buffer overflow and ensures smooth packet delivery.

How Credits Work

1. Each link partner has a receive buffer capable of holding a fixed number of header packets.
2. On link initialization, the receiver advertises its available buffer capacity as header credits.
3. The transmitter decrements a credit for each header packet sent.
4. The receiver sends a Link Credit Advertisement after processing a packet to return the credit.
5. If credits reach zero, the transmitter must stop sending header packets until credits are replenished.

Link Error Detection and Recovery

• Verifies that each header packet is transferred successfully across the link.
• Ensures in-order delivery of header packets.
• Error handling behaviour depends on the error type:
  • Automatic retries until successful transfer, or
  • Link retraining after a specified timeout period.

Link Commands

Link-level communication uses Link Commands (LC), which are short 4-symbol packets used for:

Header Sequence Numbers

• Each header packet carries a 3-bit sequence number (0–7).
• The receiver tracks expected sequence numbers to detect:
  • Missing packets
  • Duplicate packets
  • Out-of-order delivery
• On error, the receiver sends LBAD, and the transmitter replays all unacknowledged header packets.

Recovery Escalation

1. Retry — Transmitter replays unacknowledged packets (up to a retry limit)
2. Link Retraining — If retries fail, the link returns to Recovery state to retrain
3. Link Reset — If retraining fails, a full link reset may occur

Link Power Management

• SuperSpeed links support fine-grained power states: U0, U1, U2, U3 (see Power Management (SuperSpeed))
• Port-to-port protocols manage transitions between these states using link commands (LGO_U, LAU, LXU)
• Either link partner (upstream or downstream port) can initiate a transition to a lower-power state
• The partner can accept (LAU) or reject (LXU) the request

Summary