SuperSpeed / USB 3.0: More than a 10X increase in throughput
By Dan Harmon, Consumer & Computing Interface Product Marketing Manager, Texas Instruments
SuperSpeed USB (USB 3.0), has gotten much attention as products utilizing the standard become available. The primary change is the 10-fold increase in speed compared to USB 2.0 high-speed (480 Mbps to 5 Gbps); however, there are several other benefits. This article discusses what is new and improved with the USB 3.0 protocol and also looks at power management versus USB 2.0.
Faster, better, and slightly bigger
You might know by now that USB 3.0 is the latest evolution of the Universal Serial Bus specification. Brian O’Rourke of In-Stat has called USB “the most successful PC interface of all time” (Reference 1). The greatest difference in the new version of USB is that maximum throughput increases from 480 Mbps (USB 2.0 high–speed) to almost 5 Gbps (SuperSpeed). The USB Implementer’s Forum (USB-IF) has focused on creating consumer value in 3 main areas: 1.) Improve power efficiency over the bus; 2.) Continue the tradition of backward compatibility to prior versions of USB; and 3.) Improve the efficiency of data transfer.
Improve Efficiency: Use less power to move even more data
Most portable devices require batteries, regardless of whether they are a host or a peripheral, so improving power efficiency to increase operating time was of primary concern. Several aspects of the new specification are designed to reduce the overall power profile of SuperSpeed-capable devices including:
- Removing device polling
- Provision of more, intermediate-level power states
- Removing broadcast packets
- A ten-fold increase in transfer rate
SuperSpeed improves power consumption, replacing the continuous device polling requirement of USB 2.0 with asynchronous notification. With USB 2.0 technology, the host controller polls each device on the bus to check if any data needs to be transferred to the host system. Polling devices increase traffic and consume power because devices are sending NAKs (Not Acknowledged) to the host system even when they do not have any data to send. All of this checking at full power consumes energy and that means there is a tremendous power savings when polling is eliminated.
The next change was to take host packet transfer from broadcast to a targeted, device-directed method. A USB 2.0 host sends data to a device via broadcasting the data to all downstream ports. Each downstream hub then broadcasts the packets again on all of its downstream ports until the last device on the farthest branch of the tree receives the packets, which may or may not be relevant to that particular device. Each of the devices on the bus must process the data (which consumes power) to determine whether they are the intended target of the transfer. With a directed data transfer, SuperSpeed devices that are not needed can remain in lower power states.
In SuperSpeed USB, packets get sent only to the intended device. The host has to specifically know where each device is located in the USB network and the hub that feeds that device. This targeted communication significantly reduces power usage. The transmissions only target the intended device and of course, all the host and hub ports along its path down the tree.
The third area of power reduction was to define two intermediate idle states. In USB 2.0, there are two device states: ACTIVE and SUSPEND. In SuperSpeed USB, FAST EXIT IDLE (U1) and SLOW EXIT IDLE (U2) were added so devices can go into ever deeper sleep states with prolonged inactivity. Devices operating at SuperSpeed can now lower their power consumption in idle modes when they are not transmitting or receiving data. USB 3.0 allows for further power savings by requiring a link to put itself into a low power state when its linked neighbors are idle. This concept, called Link Power Management (LPM), is a localized power management that was introduced later on in the life of USB 2.0 and carried over into USB 3.0 (thus not all USB 2.0 devices have it).
In the FAST EXIT IDLE state, the link is idled but the clock for the device stays on. In the SLOW EXIT IDLE state, both link and clocking are switched off, which causes a greater signal latency (response time) before data can be acted upon. ACTIVE and SUSPEND states are the same in both USB 2.0 and USB 3.0.
|Link State Designation
||Link is Active
||FAST EXIT IDLE
||Link is idle (in a low power state) with transmit & receive temporarily inactive. Can exit fast so signal latency is low. The term applies to SuperSpeed only.
||SLOW EXIT IDLE
||Link is idle, state U1 is in effect and clock is inactive as well. Deeper power savings at a cost of increased signal latency. Applies to SuperSpeed only.
||Suspended state. Clock is off. Full latency is reached.
The 10X speed increase enables a greener overall footprint. A 5-Gbps transceiver doesn’t necessarily require less power to transmit data than a 480-Mbps transceiver. A 5-Gbps SuperSpeed transceiver requires a higher peak current than a USB 2.0 480-Mbps transceiver.
However, there are a couple of factors at play here. The overall power footprint is lower, not just the peak current that is consumed in a time slice when the transceiver is active. And with a 10 times decrease in actual transceiver active-time, the total power required to transmit a known amount of data can be far less. For example, to move a file from a PC to a flash drive is from 20% (two times peak in 1/10 the time) to 50% (5 times peak in 1/10 the time) the total power required for the same action with USB 2.0.
Combine the efficiencies of SuperSpeed USB and you find that it consumes approximately one-third (or less) the power of USB 2.0 via:
How does “backwards compatibility” work?
- Elimination of polling
- Elimination of broadcast communication
- Addition of 2 IDLE power states
- Lowering the average power required to transmit
Another key attribute that became a focus during USB 3.0 development was to maintain compatibility with USB 2.0. In the effort to accommodate backwards compatibility, it was determined that the existing connectors and cable assemblies were not going to be good enough to reliably move data at 5 Gbps. The developers also found that the signaling needed to be done over individual conductors. A full duplex differential signaling method was chosen based on the similar-speed PCI Express specification.
To maintain compatibility, no changes were made to USB 2.0 signaling. Two new differential pairs were added to accompany the existing USB 2.0 differential pair (D+ and D-), VBUS and GND. Including the ground shield for the two new SuperSpeed differential pairs (i.e., 4 more conductors), the total number of conductors in a USB 3.0 cable is nine. So nine contacts must be made in the connector, too.
Well then, what does backwards compatibility actually mean? For the end-user, all existing USB-compliant products will connect and work seamlessly with all SuperSpeed certified products! So the existing cables (i.e., plugs) must fit into the new receptacles. The opposite is also true. The new cables for SuperSpeed products must have plugs that fit into the old receptacles.
Of course, any cable and connector that supports the new USB specification and SuperSpeed data transfers will be new, or of the new technology. So there are both new conductors in the cable portion and new contacts in the plug on a newer technology cable. Indeed, new receptacles must also have new (and more) contacts as well, in order to accept the new connections that are required.
This can get confusing, but in the USB world there are two basic types of connectors (“connectors” meaning receptacles and/or plugs). The A-receptacle pretty much ships with every PC today. The A-receptacle (female in form factor) accepts A-plugs (plugs being of a male form factor, if you will). These are very familiar as the cable we see on USB mice, keyboards and most flash drives, for instance. The B-type receptacle and plug are seen connecting on the peripheral side. In USB 2.0, there are three sizes of B connectors, (ordered by size from largest to smallest): standard, mini, and micro. For the A-side, there is only standard and micro. In addition, there is a micro-AB that is used for USB On-the-go products.
Backwards compatibility requires compatibility in both protocol and physical connection, so the new SuperSpeed USB A-type receptacle must accept both the new (USB 3.0) and old (USB 2.0) type A plugs. In turn, the older USB 2.0 A-type receptacles must also accept the new USB 3.0 A-plugs. And of course, if either the plug or receptacle is physically only USB 2.0 compliant, then the data transfer will be limited to USB 2.0 speeds. The solution for requiring nine conductors in USB 3.0 was to add five new conductors on the insert side of the existing form factor of plugs and receptacles. This allows USB 3.0 to have the same connector form factor as USB 2.0 and preserves full backwards compatibility to USB 2.0 (see Figure 1).
Figure 1: SuperSpeed USB A–type receptacle and plug
For peripherals, the challenge is more difficult in one way, as there are several type-B connector sizes. Remember that type-B is for the peripheral side of the USB equation.
Of note, especially for end-users, is the fact that the new USB 3.0 type-B cable assemblies will not insert into old USB 2.0 type-B receptacles. All USB 3.0 type-B connectors are physically larger than USB 2.0 type-B connectors. A USB 3.0 device connected to a USB 2.0-only hub or PC must use a USB 2.0 cable and connector. The older, smaller profile plugs will insert into the lower section of the larger profile USB 3.0 receptacles. Recall that there are only 4 conductors in the older USB technology, so it is common sense that the old cables cannot conduct SuperSpeed even if it wanted to. So a 4 conductor, old-style cable forces any device you have to work at USB 2.0 speeds. You can see that a USB 3.0 cable is critical in operating at SuperSpeed even if everything else is USB 3.0-compliant.
So you need the new, bigger USB 3.0 B-type plug (with nine conductors) for operating a USB 3.0 device at SuperSpeed. Old USB 2.0 cables are capable of being inserted into USB 3.0-supported machines. The new B-receptacle (regardless of size) must accept both the old and new B-type plugs for backwards compatibility. Therefore, change to the form factor does not preclude insertion of an existing USB 2.0 cable.
Now look at the options for the three USB 2.0-compliant peripheral-side B-type connectors. The standard-sized, B-type connector is the most straightforward. This larger receptacle is typically found on larger form-factor peripheral devices like printers, scanners, and the like. The USB-IF accommodated the additional USB 3.0 requirements by adding a “bump” with extra contacts to the top of the B-connector. As you can see in Figure 2, the USB 3.0-compliant standard sized B-type receptacle will accept Type- B plugs in either the new USB 3.0 form factor or the old USB 2.0 form factor.
Figure 2: SuperSpeed USB B-side plug and receptacle
A decision was made by the USB standards body to not update the Mini-B connector. Primarily due to governmental regulations around the world that mandate charging over the micro-B connection. The concern was that there was a proliferation of chargers with custom connectors for every new device, causing a disposal problem.
Moving forward, the Micro-B receptacle is the new standard interface for charging small products. Handsets have migrated to micro-B USB for charging, as well as other portable devices. There is a USB Implementer’s Forum group, called the Device Working Group, specifically focused on battery charging (Reference 2.)
The existing USB 2.0 Micro B receptacle (Figure 3) is too small to add five more conductors. A major change had to be made. A dual connector solution was settled upon that can accept either a single USB 2.0 Micro B-plug or the new SuperSpeed dual Micro B-plug, as shown below.
Figure 3: SuperSpeed USB Micro B-side receptacle and plug
Waste not, want not
The next key improvement is that the efficiencies of bus usage were improved. The elimination of polling has already been discussed. Additionally, the full duplex structure of SuperSpeed allows concurrent bi-directional data flow. Recall that USB 2.0 architecture is half duplex, i.e., data travels either upstream or downstream, but not at the same time. Think of USB 2.0 (half-duplex) as a two-lane highway that allows only one car to travel in one direction at a time. Two way communication is possible, but in only one direction at a time. SuperSpeed is full duplex, which can be likened to a two-lane highway where cars can travel in both directions at the same time.
What does elimination of polling really mean? Polling is a clear waste of bus usage, since the host must ping each device to determine if data is waiting for permission to travel on the bus. In honor of my mom who was a school teacher for many years, I will make a classroom analogy for you. If the classroom were USB 2.0, the teacher would poll the class, that is, go around the classroom and ask each student if they had a question, one by one. When a student had a question, the teacher would answer the question. Then she would continue around the room until every student had been queried, and then she would start all over again.
If we liken the classroom to a SuperSpeed USB 3.0 system, a student would raise his or her hand when they had a question and the teacher would acknowledge the pupil and answer the question. The latter method is commonly referred to as “asynchronous.” Specifically, it is an asynchronous notification of data transfers going in to the teacher. In SuperSpeed, a peripheral device will send an Endpoint Ready (ERDY) signal to the host when it has new data for the host. The host then sends an ACK signal (ACKnowledge) to the peripheral when it is ready to process the transfer.
In USB 2.0, the half-duplex bus has a single differential pair to transfer data, which causes two problems for bus efficiency. Remember the highway analogy? In half-duplex, the flow of traffic has to be managed for each data flow change in direction. This “turn around” of data flow takes some time. Each lane, one for each direction, has to be flagged for use. Maybe there’s a single car bridge in the middle that allows only one car at a time through. This is horribly inefficient use of a two-lane road, and the same is true for USB 2.0.
Another consequence of half-duplex communication is that a transfer must complete its transaction before the next one can start. The receiving device ACKnowledges (ACK) that it received the data, and the transmitting device then needs to wait for the ACK before another data packet can be sent across the bus.
With SuperSpeed there are two differential pairs for every device, one pair for transmitting and one pair for receiving, and there is no time required to set up for a “turn around” of data direction. Full duplex also allows the transmitting device to continue to send data packets before it receives acknowledgement that the previous packet did arrive. This extra wait time is eliminated.
All of this change requires more intelligence on both devices in case of errors. What does this mean? If an ACKnowledge comes back with an error, a device has to be equipped with the intelligence and memory to go back, identify, and re-transmit a previous transmission. However, the protocol governing SuperSpeed limits a device to how many data “payloads” it can have active and unacknowledged at one time.
Speeding to a better future
Data transmission, in general, inherently experiences bottlenecks from time to time. A bottleneck is a spot in the signal chain that limits performance for the entire chain. Whereas USB 2.0 high-speed is perfectly sufficient for some applications, for many PC applications USB 2.0 has become a bottleneck. Storage media has become capable of transferring data at a rate that exceed what USB 2.0 can do. Larger storage devices obviously take longer to transmit more data. The hiccups that you see with streaming video using USB 2.0 are likely due to slow transfer speeds, although 480 Mbps seemed really fast at the time. But this was the whole point of developing a new specification: to eliminate bottlenecks.
Today there are media capable of reading and writing data near 5 Gbps. We have moved the bottleneck elsewhere. Will SuperSpeed USB be a bottleneck again someday? With time, yes. However, with the exception of some elite performance devices, SuperSpeed USB should have enough headroom to put the pinch point somewhere else for at least the next five years.
To conclude, USB 3.0 provides:
- A decrease the power required to transfer data
- Backwards compatibility with USB 2.0 products
- ncreased bandwidth efficiency
- A greater than 10x increase in raw bit rate.
These features are a marked benefit, both in the sync-and-go consumer experience as well as extended battery life. USB 3.0 should be more than adequate for some time to come.
 USB: The Universal Connection, IN020007MI, Multimedia Interfaces, March 2002, http://www.instat.com/abstract.asp?id=161&SKU=IN020007MI
 USB-IF Device Working Group, http://www.usb.org/about/dwg_charter/
To read more from Dan on SuperSpeed USB, please see Dan's Consumer & Computing Interface blog (http://e2e.ti.com/blogs_/b/interface/default.aspx on TI’s E2 Community at http://e2e.ti.com/.
About the Author
To learn more about SuperSpeed USB interface solutions, visit: http://www.ti.com/superspeedusb-ca.
To see our video that demonstrates the speed of SuperSpeed USB3 versus High-Speed USB 2.0, visit http://www.ti.com/usb3video-ca.
Dan Harmon is Product Marketing Manager for Consumer & Computing Interfaces at Texas Instruments as well as serving as TI’s USB-IF Representative and TI’s USB 3.0 Promoter’s Group Chair. He earned a BSEE from the University of Dayton and a MSEE from the University of Texas in Arlington.