| ||
Taking a Walk Inside Bluetooth EDR
David McCall, Cambridge Silicon Radio (CSR)
Dec 21, 2004 (10:02 AM)
URL: http://www.commsdesign.com/showArticle.jhtml?articleID=55800768
There's no doubt that Bluetooth technology has started to gain hold in the communication sector. The technology, which was once bashed for lack of adoption, is now shipping in 10s of millions in applications like cellular phones, wireless headsets, and automobiles.
But, like all wireless protocols, Bluetooth also needs to evolve to meet the changing demands of consumers. In particular, there is a clear call in the consumer electronics space to support the wireless transfer of higher-bandwidth/higher-quality audio signals, a request that will stretch the throughput capabilities delivered by existing 1-Mbit/s Bluetooth designs.
To address the need for next-generation audio and other short-range, high-bandwidth wireless applications, the Bluetooth Special Interest Group has developed an Enhanced Data Rate (EDR) protocol, which pushes the bandwidth delivered over Bluetooth connections up to a peak rate of 3 Mbit/s. In this article, we'll look at the reasons why the bandwidth of Bluetooth connections needed to be increased. We'll then take a detailed look at the modulation changes made in version 2.0 of the Bluetooth spec to support EDR operation.
Why is EDR Needed?
When the Bluetooth specification was initially drawn up back in the late 90s, great care was taken to ensure that the specification allowed flexibility and adaptability in addition to ensuring that that this wireless communication medium remained robust and secure. Since its inception, Bluetooth has seen a number of adaptations that have led to major improvements in performance and interoperability as well as co-existence with other wireless standards such as 802.11. EDR Bluetooth is the latest such adaptation.
EDR operation is defined under the Bluetooth 2.0 specification. Specifically, the Bluetooth 2.0 spec defines modulation changes and additional packet types that allow designers to deliver a peak rate of 3 Mbit/s (2.1 Mbit/s real throughput) over a Bluetooth connection. In comparison, the Bluetooth 1.2 spec defined a maximum data rate of 1 Mbit/s (723 kbit/s real throughput) over a Bluetooth connection.
For any communications technology, faster is almost always better, and Bluetooth is no exception. However, the reasoning behind EDR, and the selection of the 3-Mbit/s data rate in particular, goes beyond a simple desire for more speed.
At present, there is no single Bluetooth application that demands more than the current 1-Mbit data rate. Even a high-quality stereo audio stream (using the mandatory SBC codec) tops out at 345 kbit/s.
Although future individual applications are expected to maintain data rate demands at this level, it seems almost certain that as Bluetooth grows in popularity, users will increasingly run multiple Bluetooth links at the same time. This is particularly true for PCs, where it is easy to imagine using a Bluetooth mouse and keyboard and listening to streaming stereo audio over a pair of Bluetooth headphones, all at the same time. EDR gives Bluetooth the extra capacity it needs to maintain all these links at data rates that users find acceptable.
Taking the example of stereo audio plus a mouse and keyboard and standard rate Bluetooth: A mouse and keyboard can each typically take 11% of the maximum theoretically available bandwidth and a medium quality stereo audio stream, using the default SBC codec, takes 35%. This leaves a margin of 43%, which would be plenty for a wired network.
However, for a wireless network the theoretical maximum is rarely reached due to interference and non-ideal packet scheduling. In practice the 43% margin is just enough to maintain acceptable performance in most situations — assuming the packet scheduling is just right. High levels of interference will, however, cause glitches in the audio or sluggish mouse and keyboard response.
The situation changes if the user wants to listen to a high-quality audio stream, which requires 345 kbit/s of data rather than the 237 kbit/s for medium quality. The stereo audio stream now takes up 53% of the available bandwidth, which, when combined with the 22% required for a mouse and keyboard, leaves a 25% margin. This simply isn't enough to allow for the retransmissions necessary to maintain acceptable performance under even the lightest interference.
Switching to EDR solves the problem. The mouse and keyboard still take 11% of the maximum available bandwidth, but a high-quality A/V audio stream now only takes up 18%. This leaves a margin of 60%, which is easily enough to maintain acceptable performance under even heavy interference and still leave room for additional applications, like printing a file or synchronising data.
Reducing Power; Providing Compatibility
Backwards compatibility with v1.2 was also a high priority during the development of EDR. The final specification is 100% backwards compatible and allows networks containing a mixture of EDR and standard rate devices. The new modulation schemes are also very compatible with standard rate, so a single transmit and receive chain can do both. This means that designing a product using EDR is no more complicated than designing a Bluetooth 1.2 product.
How EDR Works
All Bluetooth data is transmitted as part of a packet. Standard rate packets are made up of four sections:
In the Bluetooth 1.2 specification, all three transmit sections use Gaussian frequency-shift keying (GFSK) to modulate the over-air RF signal. In GFSK, the carrier frequency deviates by +/-160 kHz to indicate a one or a zero thus encoding one bit per symbol. The symbol rate is 1 MSymbol/s, leading to a peak data rate of 1 Mbit/s. But, when designers account for access codes, headers, and guard bands, Bluetooth systems actually deliver a maximum payload data rate of 723 kbit/s.
In the Bluetooth 2.0 specification, EDR packets still use GFSK modulation for the access code and header. However, EDR uses one of two different modulation schemes for the payload: one mandatory and one optional (more below). The change of modulation scheme also requires the insertion of a small guard band and synchronization sequence between the header and the payload (Figure 1).
The EDR version of Bluetooth was ratified in November of 2004 and although it appears to be a mere addendum to the latest version 2.0 of the specification, its effect is far reaching.
In addition to supporting higher-quality audio, EDR is important because it helps lower power consumption in a Bluetooth design. The amount of power drawn by a Bluetooth radio depends on the length of time it is active. Since EDR allows data to be transmitted 3 times faster, the radio only needs to be active for a third of the time and draws a third of the power as a result.
So, how does EDR work? Essentially, the EDR protocol defines additional packet types that employ new modulation schemes for payload data.
The 2-Mbit Improvement
The mandatory 2X (2-Mbit) data rate uses π/4 differential quaternary phase-shift keying (π/4-DQPSK). As the name implies, this modulation scheme varies the phase of the carrier rather than the frequency. Quaternary refers to the fact that there are four possible phase positions for each symbol, allowing two bits of data to be encoded per symbol. The symbol rate remains the same; hence the 2X data rate increase.
π /4-DQPSK is differential because it is the phase position of each symbol relative to the previous position (i.e. the differential move) that determines the 2-bit value that is encoded. The π/4 designator means that the differential move is +3π/4, +π/4, π/4 or 3π/4. Translating from radians to the more familiar degrees gives +135, +45, -45, or -135 degrees.
The four moves described above result in eight possible phase positions in total, each separated by 45 degrees. However, in an EDR design, only four positions separated by 90 deg. are possibilities for any given symbol (Figure 2). The avoidance of +/-π (+/-180 deg.) phase jumps of guards against large amplitude variations and enables demodulation without knowledge of the carrier phase. A raised cosine pulse shaping filter is used to reduce sideband emissions.
The 3-Mbit Improvement
The 3X (3-Mbit) data rate uses eight-phase differential phase-shift keying (8DPSK), which is similar to π/4-DQPSK but allows differential moves to any of the eight possible phase positions. In 8DPSK, the separation between the possible phase positions for each symbol is reduced to 45 deg. (see Figure 2 above).
A PSK receiver needs to decide what phase the incoming transmission is in to decode the digital data. Noise masks the true phase of the transmission. The larger the separation between adjacent possible phase positions for each symbol, the more noise can be tolerated before the incoming signal is decoded incorrectly. The smaller phase difference between adjacent positions, along with use of +/-π phase jumps, means that 8DPSK is more vulnerable to interference. It does, however, allow the encoding of 3 bits of data per symbol giving the EDR peak data rate of 3 Mbit/s.
New Packets
The Bluetooth 2.0 specification defines ten new EDR packets: 5 each for the 2X and 3X data rates. Two of the five are 3- and 5-slot extended synchronous connection oriented (eSCO) packets that use reserved bandwidth and are typically used for voice communications. The remaining three are 1-, 3-, and 5-slot asynchronous connectionless (ACL) packets that are used for data transfer.
None of the new EDR packets employ forward error correction (FEC). Instead the existing channel quality driven data rate (CQDDR) algorithm is extended to automatically switch back to standard rate packets with FEC when necessary.
One of the interesting aspects of the Bluetooth 2.0 spec is how the new EDR packets are identified in the packet header. This identification is important since it tells the receiving radio whether it needs to switch modulation schemes between the header and the payload. A Bluetooth packet-header contains 4 bits for packet identification, which was ample for the 15 packet types defined in the Bluetooth 1.1 spec. However, adding ten new EDR packets, plus the three additional standard rate eSCO packets that were part of v1.2, means that the header has run out of room. Changing the header format would jeopardise backwards compatibility, so some clever re-use was required.
The solution is to define different modes of operation for the link, and create a new message that EDR compatible radios exchange to switch between modes. One mode of operation is 1 Mbit/s, which is no different from standard rate. When two EDR radios talk to each other they can exchange messages agreeing to switch to 2- to 3-Mbit/s mode where the same packet-header codes mean different things. For example, in 1-Mbit/s mode, the code 1011 refers to a standard rate, 3-slot long packet without error correction, but in 2- to 3-Mbit/s mode, the same code refers to a 3X EDR, 3-slot long packet.
No Upper Layer Impact
All of the above information relates to the lower Bluetooth stack layers that run on-chip — the layers below the HCI interface. That is because EDR really does not have much, if any, impact on the upper layers. An EDR capable chip can stream data faster, run at lower power or carry out more simultaneous operations at one time than a standard rate device, but other than asking the upper layers to supply data more quickly it makes no other demands.
The only other difference between standard rate and EDR radios worth mentioning is that a 3X speed increase requires 3X on-chip buffering. Thus, some additional RAM will be common on EDR capable devices.
Silicon Avaialbility
Fully qualified EDR capable silicon is now available. The first iterations are flash-based chips where the firmware can easily be upgraded to cope with any changes in the specification during the initial shipment to market. These will be joined in early 2005 by ROM-based products that lack the flexibility of flash, but are substantially less expensive.
The interesting thing about EDR Bluetooth is the subtlety of change from version 1.2 of the Bluetooth specification and how radically that can affect several fundamental qualities of the standard: the data rate; the power consumption; the user experience. That is not to say that the changes to make EDR possible are minor ones. EDR does, after all, require a change to new hardware and the associated costs that go along with that. However, for a Bluetooth device engineer, these changes are, at most, a diversion, in that much of EDR specification appears to be the same as standard rate Bluetooth 1.2 spec, and therefore remains just as easy to integrate into a design as it was before.
Editor's Note: Additional information on the EDR spec can be found at www.bluetooth.org.
About the Author
David McCall is a senior product consultant for CSR. David received an MSEE from Edinburgh University, with a focus on IC engineering. He can be reached at david.mccall@csr.com.