This article provides information about the communications aspects of Universal Serial Bus (USB): Signaling, Protocols, Transactions. USB is an industry-standard used to specify cables, connectors, and protocols that are used for communication between electronic devices. USB ports and cables are used to connect hardware such as printers, scanners, keyboards, mice, flash drives, external hard drives, joysticks, cameras, monitors, and more to computers of all kinds. USB also supports signaling rates from 1.5 Mbit/s (Low speed) to 80 Gbit/s (USB4 2.0) depending on the version of the standard. The article explains how USB devices transmit and receive data using electrical signals over the physical layer, how they identify themselves and negotiate parameters such as speed and power with the host or other devices using standard protocols such as USB Device Framework and USB Power Delivery, and how they exchange data using packets of different types and formats such as token, data, handshake, and special packets.
Mode | Abbrev. | Signaling rate | Typical data rates | Introduced in | |
---|---|---|---|---|---|
Low speed | LS | 1.5 Mbit/s (187.5 KB/s) | USB 1.0 | ||
Full speed | FS | 12 Mbit/s (1.5 MB/s) | USB 1.0 | ||
High speed; also, Hi-speed | HS | 480 Mbit/s (60 MB/s) | w: 25–30 MB/s r: 30–42 MB/s | USB 2.0 | |
SuperSpeed | SS | 5 Gbit/s (625 MB/s) | w: 70–90 MB/s r: 90–110 MB/s | USB 3.0 | |
SuperSpeed+ | SS+ | 10 Gbit/s (1.25 GB/s) | USB 3.1 | ||
SuperSpeed+ | SS+ | 20 Gbit/s (2.5 GB/s) | USBc3.2 | ||
USB4 Gen 3 | 40 Gbit/s (5 GB/s) | USB4 | |||
USB4 Gen 4 | 80 Gbit/s (10 GB/s) | USB4 2.0 |
The maximum signaling rate in USB 2.0 is 480 Mbit/s (60 MB/s) per controller and is shared amongst all attached devices. Some personal computer chipset manufacturers overcome this bottleneck by providing multiple USB 2.0 controllers within the southbridge.
In practice and including USB protocol overhead, data rates of 320 Mbit/s (38 MB/s) are sustainable over a high-speed bulk endpoint.[1] Throughput can be affected by additional bottlenecks, such as a hard disk drive as seen a in routine testing performed by CNet, where write operations to typical high-speed hard drives sustain rates of 25–30 MB/s, and read operations at 30–42 MB/s;[2] this is 70% of the total available bus bandwidth. For USB 3.0, typical write speed is 70–90 MB/s, while read speed is 90–110 MB/s. Mask tests, also known as eye diagram tests, are used to determine the quality of a signal in the time domain. They are defined in the referenced document as part of the electrical test description for the high speed (HS) mode at 480 Mbit/s.
According to a USB-IF chairman, "at least 10 to 15 percent of the stated peak 60 MB/s (480 Mbit/s) of Hi-speed USB goes to overhead—the communication protocol between the card and the peripheral. Overhead is a component of all connectivity standards".[3] Tables illustrating the transfer limits are shown in Chapter 5 of the USB spec.
For isochronous devices like audio streams, the bandwidth is constant and reserved exclusively for a given device. The bus bandwidth therefore only has an effect on the number of channels that can be sent at a time, not the speed or latency of the transmission.
The host controller divides bus time into 1 ms frames when using low speed (1.5 Mbit/s) and full speed (12 Mbit/s), or 125 μs microframes when using high speed (480 Mbit/s), during which several transactions may take place.
USB signals are transmitted using differential signaling on a twisted-pair data cable with characteristic impedance.[6]
A USB connection is always between a host or hub at the A connector end, and a device or hub's upstream port at the other end.
The host includes 15 kΩ pull-down resistors on each data line. When no device is connected, this pulls both data lines low into the so-called single-ended zero state (SE0 in the USB documentation), and indicates a reset or disconnected connection.
The following terminology is used to assist in the technical discussion regarding USB PHY signaling.
Signal | Line transition state | Description | Low speed (D− pull-up) | Full speed (D+ pull-up) | |||
---|---|---|---|---|---|---|---|
D+ | D− | D+ | D− | ||||
J | Same as idle line state | This is present during a transmission line transition. Alternatively, it is waiting for a new packet. | low | high | high | low | |
K | Inverse of J state | This is present during a transmission line transition. | high | low | low | high | |
SE0 | Single-ended zero | Both D+ and D− is low. This may indicate an end of packet signal or a detached USB device. | low | low | low | low | |
SE1 | Single-ended one | This is an illegal state and should never occur. This is seen as an error. | high | high | high | high |
Line state/signal | Description | Low speed | Full speed | High speed |
---|---|---|---|---|
Detached | No device detected. Both lines are pulled down by 15 kΩ pull-down resistors on the host side. | SE0 ≥ 2 μs | ||
Connect | USB device pulls up either D+ or D− to wake the host from the detached line state. This starts the USB enumeration process. This sets the idle state. | D− pull-up | D+ pull-up | As full speed, then chirp in reset |
Idle / J | Host and device transmitter at Hi-Z. Sensing line state in case of detached state. | As detached or connect state. | ||
Sync | Start of a packet line transition pattern. | KJKJKJKK | 15 × KJ, then KK, for 32 symbols total. | |
EOP | End of packet line transition pattern. | SE0, SE0, J | ||
Reset | Reset USB device to a known initial state. | SE0 ≥ 2.5 ms | ||
Suspend | Power down the device, such that it would only consume 0.5 mA from VBUS. Exits this state only after a resume or reset signal is received. To avoid this state a SOF packet (high speed) or a keep alive (low speed) signal is given. | J ≥ 3 ms | ||
Resume (host) | Host wants to wake device up. | K ≥ 20 ms then EOP pattern | ||
Resume (device) | Device wants to wake up. (Must be in idle for at least 5 ms.) | Device drives K ≥ 1 ms Host then sends a resume signal. | ||
Keep alive | Host wants to tell low speed device to stay awake. | EOP pattern once every millisecond. |
USB data is transmitted by toggling the data lines between the J state and the opposite K state. USB encodes data using the NRZI line coding:
To ensure that there are enough signal transitions for clock recovery to occur in the bitstream, a bit stuffing technique is applied to the data stream: an extra 0 bit is inserted into the data stream after any occurrence of six consecutive 1 bits. (Thus ensuring that there is a 0 bit to cause a transmission state transition.) Seven consecutively received 1 bits are always an error. For USB 3.0, additional data transmission encoding is used to handle the higher signaling rates required.
A special protocol during reset, called chirping, is used to negotiate the high speed mode with a host or hub. A device that is high speed capable first connects as a full speed device (D+ pulled high), but upon receiving a USB RESET (both D+ and D− driven LOW by host for 10 to 20 ms) it pulls the D− line high, known as chirp K. This indicates to the host that the device is high bandwidth. If the host/hub is also HS capable, it chirps (returns alternating J and K states on D− and D+ lines) letting the device know that the hub operates at high bandwidth. The device has to receive at least three sets of KJ chirps before it changes to high speed terminations and begins high speed signaling. Because SuperSpeed and beyond uses wiring that is separate and additional to that used by earlier modes, such bandwidth negotiation is not required.
Clock tolerance is 480.00±0.24 Mbit/s, 12.00±0.03 Mbit/s, and 1.50±0.18 Mbit/s.
USB 3 uses tinned copper stranded AWG-28 cables[7] with impedance for its high-speed differential pairs. Electrical signalling uses a linear feedback shift register and 8b/10b encoding with spread spectrum clocking, sent at a nominal 1 Volt with a 100 mV receiver threshold; the receiver uses equalization training. Packet headers are protected with CRC-16, while data payload is protected with CRC-32. Power up to 3.6 W may be used. One unit load in Super Speed mode is equal to 150 mA.[8]
After the sync field, all packets are made of 8-bit bytes, transmitted least-significant bit first. The first byte is a packet identifier (PID) byte. The PID is actually 4 bits; the byte consists of the 4-bit PID followed by its bitwise complement. This redundancy helps detect errors. (A PID byte contains at most four consecutive 1 bits, and thus never needs bit-stuffing, even when combined with the final 1 bit in the sync field. However, trailing 1 bits in the PID may require bit-stuffing within the first few bits of the payload.)
Type | PID value (msb-first) | Transmitted byte (lsb-first) | Name | Description |
---|---|---|---|---|
0000 | 0000 1111 | |||
Token | 1000 | 0001 1110 | SPLIT | High-bandwidth (USB 2.0) split transaction |
0100 | 0010 1101 | PING | Check if endpoint can accept data (USB 2.0) | |
Special | 1100 | 0011 1100 | PRE | Low-bandwidth USB preamble |
Handshake | ERR | Split transaction error (USB 2.0) | ||
0010 | 0100 1011 | ACK | Data packet accepted | |
1010 | 0101 1010 | NAK | Data packet not accepted; please retransmit | |
0110 | 0110 1001 | NYET | Data not ready yet (USB 2.0) | |
1110 | 0111 1000 | STALL | Transfer impossible; do error recovery | |
Token | 0001 | 1000 0111 | OUT | Address for host-to-device transfer |
1001 | 1001 0110 | IN | Address for device-to-host transfer | |
0101 | 1010 0101 | SOF | Start of frame marker (sent each ms) | |
1101 | 1011 0100 | SETUP | Address for host-to-device control transfer | |
Data | 0011 | 1100 0011 | DATA0 | Even-numbered data packet |
1011 | 1101 0010 | DATA1 | Odd-numbered data packet | |
0111 | 1110 0001 | DATA2 | Data packet for high-bandwidth isochronous transfer (USB 2.0) | |
1111 | 1111 0000 | MDATA | Data packet for high-bandwidth isochronous transfer (USB 2.0) |
Packets come in three basic types, each with a different format and CRC (cyclic redundancy check):
Field | Sync | PID | EOP | |
---|---|---|---|---|
Bits | 8 | |||
Signal | KJ KJ KJ KK | XXXX XXXX | SE0 SE0 J |
Handshake packets consist of only a single PID byte, and are generally sent in response to data packets. Error detection is provided by transmitting four bits, which represent the packet type twice, in a single PID byte using complemented form. The three basic types are ACK, indicating that data was successfully received; NAK, indicating that the data cannot be received and should be retried; and STALL, indicating that the device has an error condition and cannot transfer data until some corrective action (such as device initialization) occurs.[9] [10]
USB 2.0 added two additional handshake packets: NYET and ERR. NYET indicates that a split transaction is not yet complete, while ERR handshake indicates that a split transaction failed. A second use for a NYET packet is to tell the host that the device has accepted a data packet, but cannot accept any more due to full buffers. This allows a host to switch to sending small PING tokens to inquire about the device's readiness, rather than sending an entire unwanted DATA packet just to elicit a NAK.
The only handshake packet the USB host may generate is ACK. If it is not ready to receive data, it should not instruct a device to send.
Token packets consist of a PID byte followed by two payload bytes: 11 bits of address and a five-bit CRC. Tokens are only sent by the host, never a device. Below are tokens present from USB 1.0:
USB 2.0 also added a PING Token and a larger three-byte SPLIT Token:
Field | Sync | PID | ADDR | ENDP | CRC5 | EOP | |
---|---|---|---|---|---|---|---|
Bits | 8 | 7 | 4 | 5 | |||
Signal | KJ KJ KJ KK | XXXX XXXX | XXXX XXX | XXXX | XXXXX | SE0 SE0 J |
Field | Sync | PID | Frame number | CRC5 | EOP | |
---|---|---|---|---|---|---|
Bits | 8 | 11 | 5 | |||
Signal | KJ KJ KJ KK | XXXX XXXX | XXXX XXXX XXX | XXXXX | SE0 SE0 J |
Use: The first transaction in each (micro)frame. An SOF allows endpoints to identify the start of the (micro)frame and synchronize internal endpoint clocks to the host.
S/C mode | Field | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
0, SSPLIT | Sync | PID | Hub address | S/C | Port number | S | E | EP | CRC5 | EOP | |
1, CSPLIT | U | ||||||||||
Bits | 8 | 7 | 1 | 7 | 1 | 1 | 2 | 5 | |||
Signal | KJ KJ KJ KK | XXXX XXXX | XXXX XXX | X | XXXX XXX | X | X | XX | XXXXX | SE0 SE0 J |
Field | Sync | PID | DATA | CRC16 | EOP | |
---|---|---|---|---|---|---|
Bits | 8 | 0–8192 | 16 | |||
Signal | KJ KJ KJ KK | XXXX XXXX | (XXXX XXXX) × byteCount | XXXX XXXX XXXX XXXX | SE0 SE0 J |
A data packet consists of the PID followed by 0–1,024 bytes of data payload (up to 1,024 bytes for high-speed devices, up to 64 bytes for full-speed devices, and at most eight bytes for low-speed devices),[12] and a 16-bit CRC.
There are two basic forms of data packet, DATA0 and DATA1. A data packet must always be preceded by an address token, and is usually followed by a handshake token from the receiver back to the transmitter. The two packet types provide the 1-bit sequence number required by stop-and-wait ARQ. If a USB host does not receive a response (such as an ACK) for data it has transmitted, it does not know if the data was received or not; the data might have been lost in transit or it might have been received but the handshake response was lost.
To solve this problem, the device keeps track of the type of DATAx packet it last accepted. If it receives another DATAx packet of the same type, it is acknowledged but ignored as a duplicate. Only a DATAx packet of the opposite type is actually received.
If the data is corrupted while transmitted or received, the CRC check fails. When this happens, the receiver does not generate an ACK, which makes the sender resend the packet.[13]
When a device is reset with a SETUP packet, it expects an 8-byte DATA0 packet next.
USB 2.0 added DATA2 and MDATA packet types as well. They are used only by high-bandwidth devices doing high-bandwidth isochronous transfers that must transfer more than 1024 bytes per 125 μs micro frame (8,192 kB/s).
A hub is able to support low bandwidth devices mixed with other speed device via a special PID value, PRE. This is required as a USB hub functions as a very simple repeater, broadcasting the host message to all connected devices regardless if the packet was for it or not. This means in a mixed speed environment, there is a potential danger that a low speed could misinterpret a high or full speed signal from the host.
To eliminate this danger, if a USB hub detects a mix of high speed or full speed and low speed devices, it, by default, disables communication to the low speed device unless it receives a request to switch to low speed mode. On reception of a PRE packet however, it temporarily re-enables the output port to all low speed devices, to allow the host to send a single low speed packet to low speed devices. After the low speed packet is sent, an end of packet (EOP) signal tells the hub to disable all outputs to low speed devices again.
Since all PID bytes include four 0 bits, they leave the bus in the full-bandwidth K state, which is the same as the low-bandwidth J state. It is followed by a brief pause, during which hubs enable their low-bandwidth outputs, already idling in the J state. Then a low-bandwidth packet follows, beginning with a sync sequence and PID byte, and ending with a brief period of SE0. Full-bandwidth devices other than hubs can simply ignore the PRE packet and its low-bandwidth contents, until the final SE0 indicates that a new packet follows.
Full speed preamble, PRE | Hub setup enable output to low speed devices. | Low speed packet example | Hub disable output to low speed devices. | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Field | Sync | PID (PRE) | Sync | PID | ADDR | ENDP | CRC5 | EOP | |||
Bits | 8 | 8 | 7 | 4 | 5 | ||||||
Signal | KJ KJ KJ KK | XXXX XXXX | KJ KJ KJ KK | XXXX XXXX | XXXX XXX | XXXX | XXXXX | SE0 SE0 J |
OUT transaction (3 packets total) | ||||
---|---|---|---|---|
Host | Host | Device | ||
Packet PID | OUT | DATAx | ACK | |
Packet type | Token | Data | Handshake | |
Description | Tell device on ADDRx to start listening for incoming data packet on endpoint EPx. | Tell USB device the data that you want to send to it. | Device tells the host that it has successfully received and loaded the data payload to buffer EPx. |
IN transaction (3 packets total) | ||||
---|---|---|---|---|
Host | Device | Host | ||
Packet PID | IN | DATAx | ACK | |
Packet type | Token | Data | Handshake | |
Description | Tell device on ADDRx to send any data that it has on its endpoint buffer EPx. | Device checks its EPx endpoint buffer and sends the requested data to host. | Host lets device know that it has successfully received the payload and has loaded the payload into its EPx buffer. |
This is used for device enumeration and connection management and informs the device that the host would like to start a control transfer exchange.
SETUP transaction (3 packets total) | ||||
---|---|---|---|---|
Host | Host | Device | ||
Packet PID | SETUP | DATA0 | ACK | |
Packet type | Token | Data | Handshake | |
Description | Tell device on ADDRx to start setup mode and be ready for a data packet. | Send to device the 8 bytes long setup packet. | Device acknowledge reception of SETUP data and updates its setup state machine. |
A setup transaction transfers an 8-byte setup packet to the device. The setup packet encodes the direction and length of any following data packets.
Field | Offset | Bytes | Bits | Description | |
---|---|---|---|---|---|
bmRequestType | 0 | 1 | 0–4 | Recipient: USB software component being addressed 0 = Device 1 = Interface 2 = Endpoint 3 = Other 4–31 (reserved) | |
5–6 | Type: Used with bRequest byte 0 = Standard (supported by all USB devices) 1 = Class (Depends on USB device class) 2 = Vendor 3 (reserved) | ||||
7 | Direction: 0 = Host to device, or no data transfer (wLength 0)1 = Device to host (wLength > 0 bytes of status returned) | ||||
bRequest | 1 | 1 | Setup command: When Recipient = 0 (Device) and Type = 0 (Standard), defined requests are: 0 = GET_STATUS (2-byte read) 1 = CLEAR_FEATURE (0 bytes; feature selected by wValue) 3 = SET_FEATURE (0 bytes; feature selected by wValue) 5 = SET_ADDRESS (0 bytes; address in wValue) 6 = GET_DESCRIPTOR (wLength-byte read; descriptor type & index in wValue) 7 = SET_DESCRIPTOR (wLength-byte write; descriptor type & index in wValue) 8 = GET_CONFIGURATION (1-byte read) 9 = SET_CONFIGURATION (0 bytes; configuration selected by wValue) | ||
wValue | 2 | 2 | Parameter value: Interpretation depends on bRequest | ||
wIndex | 4 | 2 | Secondary parameter: Specifies the Interface or Endpoint this request is addressed to. For string descriptors (Recipient = Device), this is the language code. | ||
wLength | 6 | 2 | Data transfer length: The number of bytes to be transferred after the setup packet. |
The control transfer exchange consist of three distinct stages, each consisting of their own transactions:
This allows the host to perform bus management action like enumerating new USB devices via retrieving the descriptors of the new devices. Retrieval of the descriptors would especially allow for determining the USB Class, VID, and PID, which are often used for determining the correct USB driver for the device.
Also, after the descriptors is retrieved, the host performs another control transfer exchange, but instead to set the address of the USB device to a new ADDRx.