20 May 2007
Evolution of 802.11 (physical layer)
The purpose of this document is to explain the basic ideas laying in the foundation of the technologies adopted by IEEE 802.11 standards for wireless communications at the physical layer. It is designed for audience working with or administrating the devices complying to the named standards, and willing to know their principles of operation believing that such knowledge can help to make educated decisions regarding the related equipment, choose and utilize the available hardware more efficiently.
Using Radio Waves For Data Transmission
Designing a wireless high speed data exchange system is not a trivial task to do. Neither is the development of the standard for wireless local area networks. The major problems at the physical layer here caused by the nature of the chosen media are:
The 802.11 standards had to address them all.
802.11 First Standard For Wireless LANs
The Institute of Electronic and Electrical Engineers (IEEE) has released IEEE 802.11 in June 1997. The standard defined physical and MAC layers of wireless local area networks (WLANs).
The physical layer of the original 802.11 standardized three wireless data exchange techniques:
The 802.11 radio WLANs operate in the 2.4GHz (2.4 to 2.483 GHz) unlicensed Radio Frequency (RF) band. The maximum isotropic transmission power in this band allowed by FCC in US is 1Wt, but 802.11 devices are usually limited to the 100mWt value.
The physical layer in 802.11 is split into Physical Layer Convergence Protocol (PLCP) and the Physical Medium Dependent (PMD) sub layers. The PLCP prepares/parses data units transmitted/received using various 802.11 media access techniques. The PMD performs the data transmission/reception and modulation/demodulation directly accessing air under the guidance of the PLCP. The 802.11 MAC layer to the great extend is affected by the nature of the media. For example, it implements a relatively complex for the second layer fragmentation of PDUs.
The use of infrared for WLAN has not been accepted by public. There was no successful commercial implementations of 802.11 IR technology.
|The frequency hopping was the first step in
the evolution to the DSSS and more complex data transmission techniques. The
idea is to transmit on a given frequency for a very short time and switch to
another frequency according to a pre-defined frequency hopping pattern known
to both transmitter and receiver. This allows to deal with high energy
interference in a narrow band as well mutual interference of two FHSS
transmitters positioned close to each other.
802.11 frequency hopping separates the whole 2.4Ghz ISM band into channels spaced 1MHz. The transmitter has to change channels at least 2.5 times per second (every 400msec or less). The hopping patterns are described by 3 sets containing 26 hopping sequences each. The sets are defined in such way that the sequences in each set, when set up on different access points, provide minimum mutual interference.
The FHSS is quite stable to interference, cost effective and simple data transmission technique but it is not widely used for WLANs nowadays. Mostly because of its growing requirements to bandwidth when data transmission rates are increased.
The idea was patented by Hedy Lamarr, a famous
actress, in 1942. The story is that she overheard and understood the FHSS
idea during one of the meetings her husband (an Austrian businessman selling
weapons to Hitler) had with his customers. She managed to patent
it after leaving her husband and escaping to US right before the WWII.
Direct Spread Spectrum Sequencing
This is one of the most successful data transmission technique for today. The DSSS is used in cellular networks (CDMA systems), Global Positioning Systems (GPS) and of course, Wireless LANs.
The idea is to multiply the data being transmitted to a pseudo random binary sequence of a higher bit rate.
The pseudo random binary sequence (PRN or PN) is called the chipping sequence and the bit rate of the sequence is called the chipping rate. The data is unrecoverable from the result of such multiplication unless the chipping sequence is known.
In order to be used by a DSSS system a PRN sequence should posses following properties:
The mentioned above properties are the properties of the white noise. The spectrum of the white noise is a flat line, and the multiplication of the data to the PRN have the effect of flattening the spectrum of the resulted signal. The Rc=1/Tc in the picture below corresponds to the chipping rate. The conventional modulation of the original data stream (without the chipping sequence) would result in the shoulders of the spectrum being placed at the distance of the data rate 1/Ts (i.e. symbol rate) from the carrier frequency.
The result of the multiplication (the sequence of chips) is transmitted to the receiver using any of the standard modulation techniques. At the receiver's side, the signal is demodulated and the received bit stream is multiplied to the PRN sequence generated by the receiver. When the transmitter's and the receiver's sequences match and are synchronized in time the correlation function will generate a spike of size n if the data bit was 1, and -n if the data bit was 0 (signal -1). If the sequences do not match or not synchronized in time the correlation will stay close to 0. The receiver, when synchronized can use the output of its correlator to determine the quality of the reception and react to the synchronization drift.
The use of different uncorrelated PRN sequences is exploited in CDMA systems to allow multiple simultaneously operating data channels in the same frequency band.
The DSSS transmission is also secure if the PRN is long enough and known to transmitter and receiver only. Basically the DSSS signal can be below the noise level and therefore undetectable.
Besides the mentioned above useful properties DSSS is self synchronizing and immune to the narrowband interference. These are the properties that justified its use in 802.11.
The 802.11 DSSS separates 2.4 ISM band into 11 overlapping channels spaced 5MHz. The 802.11 transmitter always sends symbols (1 or several chips) at the rate of 11Mbps, which requires 22MHz bandwidth. The 802.11 does not use the benefits of the multiple coding sequences. All 802.11 nodes transmit using the same PRN, the 11 bit long Barker code (+1, –1, +1, +1, –1, +1, +1, +1, –1, –1, –1), and therefore only three non-overlapped channels can simultaneously operate in the ISM band without interference.
The data being transmitted should pass several stages and be prepended by PLCP (Physical Layer Convergence Procedure) before transmission. The DSSS transmission starts by sending PLCP preamble and header.
The PLCP Preamble:
- Synchronization (Sync) field consists of 128 bits. It is filled in with predefined numbers and allows receiver to synchronize to the transmission.
- 16-bit Start Frame Delimiter (SFD) field is used to mark the start of every frame.
The PLCP Header:
- 8-bit Signal or Data rate (DR) field indicates how fast the data will be transmitted. The only two possible values for the June 1997 version of 802.11 are 00001010 for 1 Mbps DSSS and 00010100 for 2 Mbps DSSS.
- 8-bit Service field is reserved for future use.
- 16-bit Length field indicates the length of the ensuing MAC PDU.
- 16-bit Cyclic Redundancy Code (CRC) field is used for error detecting.
The preamble and the PLCP header are transmitted at 1Mbps regardless of the current data transmission speed. After the preamble the payload prepared by the MAC layer is sent to the receiver at the rate specified in the services field.
The transmitter uses DBPSK and DQPSK modulation, which results in 1Mbps and 2Mbps data transmission rates correspondingly.
802.11b Need For Speed
In 1998, Lucent Technologies and Harris Semiconductor (later owned by Intersil) proposed a standard called Complementary Code Keying (CCK) to achieve 5.5Mbps and 11Mbps transmit rates. The IEEE adopted the CCK and released the 802.11b in 1999.
The standard included a new option to transmit PLCP header with a short (56 bits) preamble. In the short preamble mode only the Synchronization and Start Frame Delimiter fields are transmitted at 1Mbps. The rest of the PLCP header is transmitted at 2Mbps (using DSSS DQPSK) and the data payload at either the same 2Mbps, or using CCK at 5.5Mbps or 11Mbps.
The 802.11b also introduced the auto rate fallback mechanism missed in the original 802.11, thus standardizing the procedure of adjusting the data transmission rate depending on the link quality.
Complementary Code Keying
The CCK modulation is based on the use of the polyphase complementary codes. The codes posses nearly orthogonal (close to zero autocorrelation if shift is is not 0) properties. The polyphase complementary codes are not binary, they are complex codes. The picture below shows a polyphase code with its real component placed in the vertical plane and the complex component in the horizontal plane. Assuming the data transmission rate is set to 11Mbps, the CCK modulator is fed by bytes of data at the rate of 1.375MBytes/sec. The modulator uses 6 bits of each byte to pick one of 64 unique orthogonal eight chips long polyphase complementary codes (like the one on the picture). The other two bits of the byte are used to rotate the whole code word (0, 90, 180 or 270 degrees). Finally, 11 million times per second, the real and complex parts of the resulted code go to the I(in-phase) and Q(quadrature) channels of the IQ modulator. The resulted symbol rate is 11Mbps, the bandwidth occupied by the channel is 22MHz and consequently the CCK modulation may coexist with original 802.11 DSSS.
802.11a Even Faster
The 802.11a introduced to the WLAN world a new modulation technique called Orthogonal Frequency Division Multiplexing (OFDM). The similar to the OFDM modulation techniques have been used in the modems world, primarily since the approach allows higher data transmission rates in the smaller bandwidth. Besides proposing the new modulation method, 802.11a also switches from the rapidly getting overused 2.4GHz ISM band to 5GHz ISM band. The 5GHz ISM bandwidth is not continuous. There are two areas 5.15GHz - 5.35GHz and 5.725GHz - 5.825Ghz. Both areas are separated by 802.11a into 12 overlapping carriers (similar to 802.11 channels) spaced 20MHz.
Each carrier separated into 52 subcarriers (also called tones) positioned according to the picture below. Four of the subcarriers are called pilot and have to transmit a sequence that can be used by the receiver for synchronization control.
The subcarriers of the OFDM signal are modulated in such a way that adjacent channels even though their shoulders overlap, do not interfere with each other. The chapter "Theory of OFDM Operation" in online article  gives a very good explanation (exactly at the level targeted in this document) of how the desired orthogonality of modulated OFDM tones is reached and at what cost.
Similarly to the DSSS the OFDM transmission prefixes each data packet with PLCP synchronization sequence (PLCP preamble).
The PLCP preamble in OFDM is sometimes called training sequence. It consists of 10 short OFDM symbols used by the receiver to tune AGC (automatic gain control), select antenna and do coarse timing synchronization estimate. The following two long OFDM symbols allow the receiver to fine tune for the data transmission.
The picture below shows the OFDM packet data layout. It starts with training sequence (PLCP preamble), followed by the SIGNAL field and data. The data is followed by 6 tail bits and padding (not shown on the picture). Both the training sequence and the 24 bit SIGNAL field are transmitted at 6Mbps rate. The SIGNAL field tells the receiver at what rate the following data will be transmitted and indirectly defines the subcarriers' modulation technique employed. The BPSK, QPSK, 16-QAM and 64-QAM are the available choices. The SIGNAL field also delivers the length (12 bit) of the following data and includes a zero bit sequence for the data scrambler synchronization. The total training sequence and SIGNAL field transmission times add up to about 20 µs, which is an overhead equivalent to approximately 140 bytes transmission at the maximum transmission rate of 54Mbps defined by the standard.
On the transmitter side the actual data to be sent get padded, scrambled and distributed among the subcarriers according to the modulation technique/rate chosen in the fashion that assures equal among the subcarriers' transmission energy. The signal (in the time domain) is digitally calculated from the combination of all the modulated subcarriers. The resulted numbers are used directly to form the signal on the air. The whole process is reversed at the receiver side.
802.11g Faster But Compatible
Although the 5GHz band is not as crowded as 2.4GHz, one may expect decrease in the operational range when upgrading from 802.11b to 802.11a. Besides, the problem of upgrading the whole network may seem to be scary enough. The new 802.11g standard is called to solve all the problems. The standard is backward compatible with 802.11b and 802.11. The support for the old modulation method is mandatory. For the high transmission rates it introduces the OFDM to the 2.4GHz ISM band.
802.11n Chasing The Ethernet
The emerging new 802.11n standard addresses the need for wireless networks to provide data transfer rates similar to those of the Fast Ethernet. The idea that allows 802.11n compatible equipment to transfer even more data within the same bandwidth is relatively simple. Lets consider a conventional 802.11g adapter radio. One of the major problems that radio receiver has to deal with is the multipath propagation. Essentially the receiver has to filter various echo of the main signal it is tuned to, which is similar to having multiple transmitters of that signal in the environment with no echo. The receivers successfully do the job, so why not to install two transmitters, then install two receivers and tune each of them to the individual transmitter. If that works, data can be sent through two channels and the 54Mbs rate of the standard 802.11g equipment is doubled.
Why would it work in the real environment where the additional transmitters would generate more echo and would make it more difficult to filter the useful signals? The intuitive answer is that another dimension, "space" is now used for the data transmission and therefore increasing the bandwidth use efficiency. In addition, since the receivers work together and each one is synchronized to its own signal, one receiver's reception can be used to couterphase or nullify its component of the signal for the opposite receiver and therefore improve the overall quality of the reception.
The 801.11n is not yet approved standard, however pre-standard
equipment utilizing concepts of MIMO (Multiple Input Multiple Output) is already
available for purchase and is relatively inexpansive. We probably can expect the
standard to be ratified in the nearest future.
In the recent years we have witnessed as the demand and availability of the inexpensive technology made wireless network access real for everyone. The 802.11 standards have been growing like mushrooms after the rain in order to address the needs of the consumers. Since the consumers of the WiFi services are placed on the edge of the network, with the introduction of the transmission rates of 100Mbs and over the immediate attention of the "802.11 community" is likely to shift even more from speed and the physical layer to wireless network structure, reliability, control, manageability, ease of use and security issues. In the future, as the demand for the speed grows, hopefully we will have a "quantum leap" similar to the one that took Internet end users from telephone modems to Cable and DSL.
"High Rate" Wireless Local Area Networks by Kanoksri Sarinnapakorn - March 15, 2001
ANSI/IEEE Std 802.11, 1999 Edition