Notes on Wireless Technologies
I found various pieces of interesting information in the "Handbook of Wireless Networks and Mobile Computing" (ed. Ivan Stojmenovi'c, Wiley 2002), and in "Mobile Radio Networks" (Bernhard H. Walke, Wiley 2002). The below is a summary of some of the useful information that I felt that notes on would be beneficial. I thoroughly recommend the book to anyone with an in depth interest in wireless networks.
WCDMA and TD-CDMA
Code Division Multiple Access (CDMA) is based on the idea of multiple nodes utilising one transmission frequency simultaneously. This is done by modulating a high frequency carrier wave with the signal to be transmitted. The carrier wave is derived from a stream of bits known as the chip code. The number of chips transmitted per bit of information in the original data stream is known as the Spreading Factor. For UMTS, the chip code rate is 3.84 Mchip/s.
Wideband CDMA (W-CDMA, or FD-CDMA) allows all users to transmit simultaneously provided that each uses a different chip code. To allow large numbers of users the spreading factor (SF) is large, normally 512 chips/bit, hence the name wideband. Two separate frequencies are allocated (frequency division multiplexing), one for upload, and one for download. Hence this system is also known as UTRA-FDD. Different uplink channels can be further differentiated by the phase of the carrier (0 or pi).
Time Division CDMA (TD-CDMA) is a hybrid approach. There are slots of a particular time length, within each of which multiple users may transmit simultaneously using much smaller SFs, e.g. 16 chips/bit. This then allows a single frequency to be used for both upload and download, by utilising alternating time slots. Slots are 10 ms long, which therefore allows 15 data frames per slot to be transmitted. This radio interface is known as the UMTS Terrestrial Radio Access Time Division Duplex (UTRA-TDD).
TD-SCDMA, or Time Division Synchronous CDMA, is a Chinese standard, with 1.6 MHz of bandwidth per carrier. EDGE, or Enhanced Data Rates for GSM Evolution, utilises 8-PSK modulation rather than GSM's GMSK to enhance data rates on traditional GSM networks, up 384 kbits/s (though normally 144).
Variable transmission rates in the FDD mode of UMTS (W-CDMA) can be achieved by changing the spreading factor used (e.g. a greater spreading factor implies a greater bandwidth used, hence a greater data rate), or using multicode transmission (i.e. simultaneously using several CDMA spreading codes, to have multiple data streams). In TD-CDMA, each 10 ms frame is subdivided into 15 time slots, each of which can have 16 CDMA channels in it. These time slots can be allocated in order to assign to each user as much or as little bandwidth as they require.
IEEE 802.11 DCF and PCF
The basic IEEE 802.11 MAC layer is the Distributed Co-ordination Function (DCF). In this protocol, nodes listen to the channel, and if it is idle for a time greater than the Distributed Interframe Space (DIFS) then the node transmits its packet. If the channel is busy, the node initialises a random backoff timer. The value of this timer is decreased with the time that the channel is seen to be idle for. When the timer reaches zero the node transmits its data. The time at which a node may begin transmission after a DIFS period expires is slotted, such that the interval between slots is sufficient for a node to detect that another node has begun a transmission in a previous slot.
The acknowledgements mechanism of the protocol involves the receiving node transmitting an ACK after a short period of time known as the Short InterFrame Space (SIFS), which is shorter than the DIFS. If no ACK is received, the sender retransmits.
A Request To Send/Clear To Send (RTS/CTS) protocol is added above the basic protocol to solve the hidden node problem, and also to allow channel conditions to be evaluated using short packets, rather than risk a collision when transmitting longer packets.
The standard also defines a Point Control Function (PCF), where an access point (AP) orchestrates when nodes may transmit, to allow guaranteed QoS. This functions by the AP transmitting a beacon packet after an idle period of length PIFS (which is shorter than the DIFS). This ensures that the beacon packet is higher priority than other data transmissions. The point co-ordinator transmits the expected length of the non-random access period, so that other stations are aware of how long CSMA/CD style transmission will be unavailable. The point co-ordinator will then send polling messages to all nodes requiring contention-free service in turn. After a SIFS, the node is allowed to transmit a frame to another node. After another SIFS has passed from the end of the transmission, the destination node sends an ACK. The sending node (after another SIFS) then sends an ACK to the point controller. The data message may be to the point controller, in which case the ACK for the beacon may be combined with the data packet.
Phase-Shift Keying and Constellation Diagrams
A constellation diagram is a representation of the Argand plane, with one axis representing the amplitude of the complex component of a wave, and the other the imaginary. This is analogous to representing the amplitude of the cosine component, and the amplitude of the sine component. These two components when added together give the transmitted wave.
Symbols in the transmitter's alphabet can be mapped onto the constellation diagram by means of their real and imaginary components. The greater the separation between the symbols on the Argand diagram, the less likely it is that a received signal will be misinterpreted. This is because the receiver uses maximum likelihood detection, i.e. finds the nearest symbol to the received data in the Argand diagram.
The difference in phase of two symbols is simply the angle between the lines connecting each of them to the origin of the Argand diagram. Hence a pure sine wave is 90 degrees out of phase with a pure cosine wave.
For Binary Phase Shift Keying (BPSK), the two symbols are separated by 90 degrees of phase. For Quaternary Phase Shift Keying (QPSK), the symbols are located in a square centred upon the origin of the Argand diagram. Each symbol has a different phase.
Quadrature Amplitude Modulation (QAM) is so named because the two amplitude modulated carrier waves are 90 degrees out of phase, i.e. sine and cosine are quadrature carriers. 2-QAM is simply BPSK, 4-QAM is QPSK.
Higher order QAM is more interesting. 16 QAM has 4 symbols arranged in a square in each quadrant of the Argand diagram. This now means that there are two symbols with the same phase. Hence, the receiver must also measure the amplitude of each carrier rather than only their phase. This makes the modulation scheme more prone to error due to interference.
Phase shift keying is distinct from phase modulation in that the phase of the carrier is changed by an amount related to the input signal, rather than set by it.
Gaussian Minimum Phase Shift Keying (GMSK) is the modulation scheme used in GSM networks. It is a special case of Minimum Phase Shift Keying (MSK), which involves changing the phase of the carrier wave within each bit interval (i.e. whilst the bit is being transmitted). The direction of the phase shift relative to the reference signal indicates the symbol transmitted, e.g. clockwise in the Argand diagram might mean a zero.
With MSK, the phase transitions between each bit can be large. This then causes discontinuities in the outputted signal, which are high frequency components that can cause unwanted interference. To avoid this, in GMSK we filter the signal using a Gaussian filter over each bit period. This ensures that the phase transitions over each bit interval are slow at the beginning and end, and fast in the middle, thus avoiding discontinuities due to instantaneous phase changes.
IEEE 802.11n Overview
For an excellent overview of the possible ways that the IEEE 802.11n (MIMO for WiFi, among other things) could have gone (now out of date), see chapter 15 of O'Reilly 802.11 Wireless Networks.
Essentially, 802.11n has the goal of achieving 100+ Mbps at protocol level (i.e. over and above the throughput required for the transmission of MAC headers, preambles, etc.). This is done by using Multiple Input Multiple Output techniques, where there is more than one antenna at each station. This allows spatial diversity (space-time coding), and hence better reception in the case of multipath. The MAC layer is also enhanced to use frame bursting (one ACK for multiple frames, rather than per frame), MAC header compression, and frame aggregation (where many small frames are packaged up into one super frame).
The PHY will make use of the same spectrum as 802.11b/g, and will use OFDM. It has the potential to use 40 MHz channels (compared to the normal 20 MHz), though these are not normally allowed by regulators (except in Japan). One of the proposals allowed for one 20 MHz channel to be the primary (used for contention and transmission), with a secondary channel to be used for "spillover" (i.e. if a station has the primary channel, it also has exclusive right to the second channel).
The proposal also allowed for 802.11n equipment to modify the Network Allocation Vector to a very long value where 802.11g equipment is present, ensuring that legacy stations do not attempt to transmit, to ensure that the two do not conflict. This long NAV allows the 802.11n stations to then use a modulation scheme/PLCP that legacy stations do not understand, without fear of interference.
Modulation in 802.11n will be 16- or 64-QAM, but with a 7/8 code rate. In advanced beamforming mode, the various OFDM sub-channels' quality is evaluated, and the coding and modulation rates adjusted accordingly in order to use the optimal ones for each sub-channel. One proposal, which was not accepted, was for ABM to also allow for 256-QAM modulation.
Multipath & Fading
Multipath effects occur when a transmitted signal travels along a direct path to the receiver, and in addition via a reflected path, such that multiple copies of the signal are received at slightly different times. If these signals are received close together in time such that they cannot be distinguished from each other, then constructive/destructive interference will occur. This will cause the received signal's amplitude to potentially decrease, possibly making the SNR too low for the signal to be decoded correctly. This is fast fading.
If the multipath components are spaced enough in time that they can be distinguished from each other, then inter-symbol interference is likely to occur. Here, the two (or more) signals have a time difference of arrival greater than the symbol period of the channel. Whilst fast fading cannot be eliminated (it is common to all narrowband channels), inter-symbol interference is decreased by using longer symbol periods and methods of detecting the effects of multipath on the channel.
In order to take advantage of multipath effects (effectively providing multiple copies of the signal at the receiver for "free"), the components must be resolvable in time from each other. Using training sequences, the channel can be characterised to ascertain where in time each multipath component is received. An equaliser can be used to amplify those components that have been subject to fading. However, this step also amplifies any noise that has been added. Therefore care must be taken to not amplify those components that are too low to start with (!).
Fast fading vs Slow Fading and Attenuation
Attenuation or path loss is the reduction in signal power due to absorption of the signal by air or other materials. In a vacuum attenuation causes the inverse square fall-off of power with distance.
Fast fading was described above, and is due to multipath effects causing constructive/destructive interference, which in turn reduces SNR levels.
In contrast, slow fading results from shadowing by buildings or similar objects. Unless simulations of a given environment are carried out, the attenuation due to slow fading is assumed to follow a log-normal distribution, i.e. the distribution of attenuation values in dB is Normal. The variance of the distribution varies with different environments, e.g. an urban canyon and a village street would be different.
Fading can be characterised by comparing the coherence time (the period of time over which a signal's phase can be predicted from its preceding values) to the delay of the channel. If the coherence time is much greater than the delay, then the phase of the signal is unlikely to change unpredictably over the time period a symbol takes to traverse the channel. This is slow fading. Fast fading, meanwhile, occurs when the coherence time is shorter than the delay, and the signal's phase changes unpredictably within a symbol. Because the coherence time is inversely proportional to the Doppler spread (the spread of phase differences of all the transmission paths due to travelling away/towards the transmitter at speed), cluttered environments traversed at speed will result in large Doppler spreads, low coherence times, and hence a greater degree of fast fading.
What the probability distribution of fading is depends on various factors, including whether there is a direct path between transmitter and receiver, and how long that path is. If there is no direct path, then fading is approximated using a Rayleigh distribution. In contrast, if there is a direct path, a better approximation is a Rician distribution. Nakagami fading, meanwhile, is used for longer-range signals, where there is a larger spread of path length differences.
Narrowband & Wideband Signals
Fading (or signal attenuation) occurs to differing degrees depending on what frequency a signal is transmitted on. If the magnitude of the fading is equal across the frequency band of a given channel, this is said to exhibit flat fading. Whether this is the case depends on the bandwidth of the channel as compared to its coherence bandwidth. The latter is inversely proportional to the multipath spread, i.e. the greater the difference in the various transmission paths between the sender and receiver, the smaller the coherence bandwidth, and hence the narrower the range of frequencies that can be considered to exhibit flat fading. A narrowband signal is one where the bandwidth of the transmitted signal is approximately equal to the coherence bandwidth, and hence if part of the channel's frequencies are attenuated, all of them are, and the message is unlikely to be received.
In contrast, wideband signals have channel widths that ensure that even if some frequencies are subject to fading, there will exist other frequencies that are in use on that channel that are not within the coherence bandwidth. These other frequencies will not have their fading correlated with the first set, and hence the message is more likely to arrive at the transmitter.
OFDM functions by using multiple narrowband carriers in a wideband portion of spectrum. Each carrier is modulated by a low bit-rate data stream, and then all the carriers are superposed and transmitted. The receiver decides the stream on each carrier, thus resulting in a high combined data rate. To decode each carrier, the received signal is subjected to a Fast Fourier Transform. This detects the frequency components of the signal. Because an FFT assumes that the input signal is periodic (and hence infinite), the receiver uses a window with width of one symbol period as input to the FFT, with this input effectively being copied infinitely to produce the FFT input signal. This works well if the window is exactly synchronised with the oscillations making up the symbol. If it is not, then the duplicated signal passed into the FFT will not be smooth, and hence will have frequency components that are at frequencies other than those of the carrier that generated them. These "noise" frequencies may well be those of other sub-carriers used in the transmission, and hence may cause decoding errors.
To counter this, each symbol has a cyclic prefix of say 25% of the symbol period. This effectively means that the window can be wrong by up to 25%, and yet the duplicated input to the FFT will still be smooth. These guard periods that contain the cyclic prefixes are a constant overhead to using OFDM. In 802.16 WiMax, the length of these guard periods can be adjusted, i.e. if it very little multipath is expected, then the error in where the window is placed can be expected to be low, and hence the guard period can be short. This will increase the overall data throughput.
When a modulated signal is demodulated, i.e. the carrier wave is "subtracted", we are left with a signal that has frequencies from zero up to some limit. These are known as baseband signals.