نیمسال اوّل افشین همّت یار دانشکده مهندسی کامپیوتر مخابرات سیّار (626-40) روش های دسترسی چندگانه

What is Multiple Access?  Multiple users want to use the same media to transmit and receive information.  In wired systems, the media can be an Ethernet cable.  In wireless systems, the media is free space. Although might have to pay a lot of money to get the right of using it. 2

Multiple Access vs Multiplexing  Multiplexing: One wants to use one media to transmit and receive information of many users.  All of the users in each side of the channel are connected to one node.  Main multiplexing schemes are: 1.FDM: Frequency Division Multiplexing 2.TDM: Time Division Multiplexing 3.CDM: Code Division Multiplexing 3

Multiple Access vs Duplexing  Duplexing: How transmit and receive paths are separated from each other.  Duplexing is still an issue for single user scenarios in which Multiple Access is not required.  Main duplexing schemes are: 1.FDD: Frequency Division Duplexing 2.TDD: Time Division Duplexing 4

Frequency Division Duplexing  Transmit and receive links use different frequency channels, so each duplex channel consists of two simplex channels in two frequencies.  Example: GSM o Tx: MHz o Rx: MHz o Tx and Rx frequencies for each link are separated by 45MHz.  In order to separate Tx and Rx paths, circuits known as duplexer are required at analog front end of each transceiver. 5

Time Division Duplexing  Transmit and receive links use same frequency channel, but occupy different time slots, so each duplex channel consists of two simplex channels at two time slots.  Simpler RF circuits (no duplexer), but delays should be handled properly  not easy to implement for highly moving users.  Examples: o Cordless phones (DECT) o WLAN (802.11) 6

Multiple Access Techniques (1)  Fixed: used for circuit switched application such as voice. o Frequency Division MA (FDMA) o Time Division MA (TDMA) o Code Division MA (CDMA) o Orthogonal Frequency Division MA (OFDMA) FDMA used for analog systems. TDMA, CDMA and OFDMA used for digital Systems. MA and Duplexing are two different issues. A system such as GSM can be TDMA but FDD.  Statistical: used for packet switched applications such as data 7

Multiple Access Techniques (2)  FDMA o Example: old AMPS systems with analog 30KHz channels.  TDMA o One of important issues is synchronization o Special pattern in each frame used to correlate and synchronize named Preamble o Overhead such as guard bits between frames and coding bits should be taken into account to compute throughput. o Example: for GSM each time slot consists of 6 trailing bits, 26 training bits, 116 information bits, and equals to 8.25 bits guard time so: b T = 8*( ) = 1250bits b OH = 8*( ) = 322bits Efficiency = (1-b OH /b T )*100% = 74% 8

Multiple Access Techniques (3)  CDMA o Based on Direct Sequence Spread Spectrum technique o Use more BW o Frequency is the most valuable asset in a wireless system!, but More resistance to multipath and fading effects Multiplexing many users on the same BW, using proper codes o Another Spread Spectrum technique is Frequency Hopping 9

DSSS vs FH 10

DSSS (1) Code for A = Code for B = Code for C = Example 11

DSSS (2) PN code generation 12

DSSS (3) DSSS Tx DSSS Rx 13

Rake Receiver (1) An integral part of DSSS systems 14

Rake Receiver (2)  DSSS removes most of the energy from multipath.  The received signal components typically experience fading. The system normally synchronizes to the strongest multipath component.  A Rake receiver has N branches that synchronize to N different multipath components.  Different multipath components are combined using Scanning Selection Equal Gain Maximal Ratio  Rake is a diversity combining technique, with diversity branches provided by the environment. 15

Rake Receiver (3)  Channel Impulse Response  When the chip time T c is much less than the rms delay spread, each branch has independent fading (assuming uncorrelated scattering), and Rake provides diversity gain.  When chip time T c is greater than the rms delay spread, multipath components can not be resolved, and there is no diversity gain. 16

Rake Receiver (4) Performance in fading channel 17

CDMA (1)  Spread Spectrum originally aimed at single user applications (mainly for military purposes).  But later found to be useful for multiple access schemes where we use different codes for different users.  CDMA  Other users appear as noise  System highly interference limited  Duplexing can be either FDD or TDD 18

CDMA (2) Design Goals 1)Make the interference look as much like Gaussian noise as possible: o Spread each user’s signal using a pseudo-noise random sequence o Tight power control for managing interference within the cell o Averaging interference from outside the cell as well as fluctuating voice activities of users 2) Apply point-to-point design for each link o extract all possible diversity in the channel 19

CDMA (3) Point-to-point link design  Very low SINR per chip: can be less than -15dB  Diversity is very important at such low SINR.  Time Diversity is obtained by interleaving across different coherence times.  Frequency diversity is obtained by Rake combining of the multi-paths.  Transmit diversity in 3G-CDMA systems (multiple base stations and antennas). 20

CDMA (4) Power control  In the reverse path, signals coming from different users will experience wide range of variations.  Since, cross-correlation of codes of different users is not completely zero, we will experience large interference if interferer’s signal is strong.  This will lead to the so-called “Near-far” problem.  The solution to Near-far problem is using proper “Power Control” algorithms, to ensure received power at base station coming from different users is almost the same. 21

CDMA (5) Power control  Maintains equal received power for all users in the cell.  Tough problem since the dynamic range is very wide (users’ attention can be differ by many 10’s of dB).  Consists of both open-loop and closed–loop.  Closed loop is needed since IS-95 is FDD.  Consists of 1-bit up-down feedback at 800Hz.  Not cheap: consumes about 10% of capacity for voice.  Power control is one the most difficult parts of CDMA systems.  For a long time it was believed to be impossible, but “Qualcomm” proved that it works. 22

CDMA (6) Interference averaging  The received signal-to-interference-plus-noise ratio for a user is defined as:  In a large system, each interferer contributes a small fraction of the total out-of-cell interference.  This can be viewed as providing interference diversity.  Same interference-averaging principle applies to voice activity and imperfect power control. 23

CDMA (7) Strengths  Multipath friendly o Using spread spectrum techniques (DS or FH) creates some sort of frequency diversity that will improve system performance against deep fading. o In addition, in time domain, different paths can be resolved and properly added together to improve performance o Since some paths can be in fade, but others not in fade, Rake receiver improves performance by proper combining of paths (some sort of time diversity). 24

CDMA (8) Strengths  Soft Capacity o Unlike FDMA and TDMA that have fixed number of slots in time or frequency domain and therefore put a hard limit on system capacity, in CDMA number of users can be increased without hard limits. o In this case, more users will show up as additional noise and decrease system performance gradually. 25

CDMA (9) Strengths  Soft Handoff o Unlike FDMA and TDMA in which neighbor base stations can not transmit at the same frequency, in CDMA neighbor stations can use same frequencies and talk to a mobile at the same time. o In this way, when a user cross the boundary of two stations, it can simultaneously talk to two stations and even add their signals together in the same way they combine multipath signals. o Therefore, instead of switching one base to another, during handoff multiple signals are used and some sort of macro-diversity is achieved. o This procedure is known as “Soft Handoff” and will improve handoff quality. 26

CDMA (10) Example: IS-95  Most wide-spread 2G CDMA system o Channel Bandwidth: 1.25MHz o Processing Gain: 128 o Bit Rate: 9.6Kbps o Data rate reduced to 1.2Kbps during silence times, so Tx power can be reduced during vacant bits.  FDD used with 45MHz separation o Forward link band: 869 – 894MHz o Reverse link band: 824 – 849MHz 27

CDMA (11) Example: IS-95 Forward link  In forward link 64 Walsh-Hamard codes are used to differentiate up to 63 users.  Each code is multiplied by I and Q PN codes, unique for each base station, to ensure spreading of signals and to reduce interference from neighbor cells.  Orthogonality preserved since users are added synchronously at base.  However, multipath can cause un-orthogonal signals arriving at receiver to be combined by Rake receiver.  Pilot signal used to ensure users can properly use coherent detection and also detect proper base signals during handoffs. 28

CDMA (12) Example: IS-95 Forward link 29

CDMA (13) Example: IS-95 Reverse link  6 Symbols mapped to one of 64 Walsh codes which are the same for all users, used for modulation and spreading.  Then, user specific codes of length used to separate users and base stations from each other.  In this way, the robostness to in-cell interference increases compared with a short code.  Open-loop and fast, closed loop power control used to control transmit power of each user.  Fast closed loop power control essential in fading environments.  A 800bps forward channel used for closed loop signals sent back to the mobile (1dB step changes).  -50 to 23dBm dynamic range, accuracy = 1.5 – 2dB  3 finger Rake used at base station 30

CDMA (14) Example: IS-95 Reverse link 31

CDMA (15) Issues  Main advantages o Allows interference averaging across many users o Soft capacity limit o Allows soft handoff o Simplified frequency planning  Challenges o Very tight power control to solve the near-far problem o Keeping orthogonality only for users in the cell o More sophisticated coding/signal processing to extract the information of each user in a very low SINR environment o Synchronization issues 32

CDMA (16) Synchronization issues  Carrier vs. Chip synchronization  Acquisition vs. Tracking  Matched filter vs. Correlator for acquisition  Early-Late gates for tracking  PLL vs. DLL 33

Multicarrier Modulation (1) Frequency selective channel with no equalizer  Earlier saw that this is possible with DSSS and Rake receiver.  Another option: Multicarrier or OFDM modulation 34

Multicarrier Modulation (2)  Breaks data into N non-overlapping substreams  Substream modulated onto separate carries o Substream bandwidth is B N = B/N for B total bandwidth o B N < B c implies flat fading on each subcarrier (no ISI)  Use BPF of width B N to separate signals at receiver 35

OFDM (1)  It is quite clear how an ideal OFDM system works in frequency domain if we use continuous time non-overlapping subcarriers.  Actually, in this case, these functions (cos(2 π f i t)) are “eigen-functions” of the LTI system.  But, one main problem with this approach is that in reality infinite-length subcarriers can not be used.  Also it is not practical to generate continuous time signals and multiply them with input signals.  So, the main question is how to implement an OFDM system practically in discrete-time domain? 36

OFDM (2)  If we only transmit a finite number of N symbols, d[0] to d[N-1], sinusoids are no longer eigen- functions of the system.  One way to restore this property is by adding a cyclic prefix (CP) to the symbols (of length L-1): x = [ d [N-L+1],... d [N-1], d [0],... d [N-1]]  Now, if only look at channel output for m = L to N+L-1 and define channel vector of length N as h = [ h 0, h 1,... h L-1, 0,... 0], then it can be easily verified that the new output is equal to “cyclic convolution” of d and h. 37

OFDM (3)  Consequently, for cyclic convolution in time domain, the DFT of output is given by multiplication of DFT of d and DFT of channel vector: Y i = D i.H i  So, by using the cyclic prefix and transmitting the DFT of signal through channel, we again get each component separately at the output (eigen- functions are back).  In this way channel will be transformed into a set of independent parallel channels. 38

OFDM (4)  Efficient IFFT structure at transmitter:  Reverse structure (remove CP and use FFT) at receiver. 39

OFDM (5) Challenges  Peak-to-average power ratio o Adding multiple substreams can result in high peak signal values o Impacts amplifier efficiency o Solutions include clipping, coding, and tone reservation.  Inter-carrier Interference o Subcarrier orthogonality compromised by timing jitter, frequency offset, and fading. o Frequency and timing offset causes interference between carriers. 40

OFDM (6) High PAR  For single carrier we have:  In multicarrier, assuming coherent addition of subcarriers, peak power increases linearly with N 2, while average power added over all subcarriers increases linearly with N.  It can be shown that PAR increases approximately linearly with number of subcarriers N. 41

OFDM (7) High PAR 42

OFDM (8) High PAR 43

OFDM (9) Inter-carrier interference  Mismatched oscillators, Doppler shift or errors in timing synchronization cause subchannel interference (loss of subcarrier orthogonality).  Mitigating by minimizing number of subchannels and using pulse shapes robust to timing errors. 44

OFDM (10) Effects of phase/frequency imperfections 45

OFDM (11) Fading across subcarriers  Leads to different BERs for subcarriers o Performance limited to worst subcarrier status  Compensation techniques o Frequency equalization  Noise enhancement o Precoding: compensate channel variations at transmitter  Accurate channel estimate  Power inefficient o Coding across subcarriers  Works well for small B C (large delay spread) in order to use codes over many subchannels and recover data o Adaptive loading (power and rate)  For small subcarrier bandwidths Doppler spread becomes important (higher user mobility): o Should have f D <<B N in order to ignore Doppler spread  Current OFDM-based wireless systems: a, g, a,

OFDM (12) Example: Flash OFDM  Bandwidth = 1.25MHz  Number of data subcarriers=113  OFDM symbol = 128 samples = 100μS  Cyclic prefix = 16 samples = 11μS delay spread 47

Multiuser OFDM (1)  We have seen OFDM as a point-to-point modulation scheme, converting the frequency- selective channel into a parallel channel.  It can also be used as a multiple access technique called “OFDMA”.  By assigning different time/frequency slots to users, they can be kept orthogonal, no matter what the multipath channels are. 48

Multiuser OFDM (2) In-cell Orthogonality  The basic unit of resource is a virtual channel: a hopping sequence  Each hopping sequence spans all the subcarriers to get full frequency-diversity.  Coding is performed across the symbols in a hopping sequence.  Hopping sequences of different virtual channels in a cell are orthogonal.  Each user is assigned a number of virtual channels depending on their data rate requirement. 49

Multiuser OFDM (3) In-cell Orthogonality 50

Multiuser OFDM (4) Out-of-cell interference averaging  The hopping patterns of virtual channels in adjacent cells are designed such that any pair has minimal overlap.  This ensures that a virtual channel sees interference from many users instead of a single strong user.  This is a form of interference diversity. 51

Statistical MA (1)  So far discussed fixed MA, FDMA, TDMA, and CDMA in which a fixed channel is assigned to a user during the whole talk period.  Another MA scheme used for data services is based on packet nature of data transmitted.  Therefore, channel is not assigned to a user and any user only uses the channel while has bursty data to transmit.  Since multiple users can try to use the channel at the same time, “collision” may naturally occur in packet multiple access systems.  In general, in order to avoid collision, some sort of ACK mechanism is used.  Main problems in packet systems are: o Loss of packets (low throughput) o Packet delay and jitter 52

Statistical MA (2)  Main statistical MA techniques used in practice are:  Slotted aloha  Carrier Sense MA (CSMA)  These techniques are also commonly known as MAC layer algorithms in multi-user networks. 53

Statistical MA (3) Slotted ALOHA  Lets assume users send packets of length τ  Users are synchronized to send at intervals of about τ  After sending a packet if collision occurs (no ACK received), the sender waits for a random time and sends at next available time slot.  If we assume packet generation to have Poisson distribution with mean arrival time of λ, then traffic occupancy without collision will be R= λτ.  Probability of no collision is also given by: P(n) = R n e -R /n!  P(n=0) = e -R  Throughput of slotted ALOHA is then given by: T = Re -R  Therefore, maximum throughput will be for R=1, and we will get: T = e -1 =

Statistical MA (4) Medium Access Control  The Medium Access Control (MAC) is responsible for scheduling data of users contesting for the same medium in a suitable way. For a MAC protocol, the figures of interest are data throughput and delay.  Optimum MAC,mechanisms for centralized and decentralized wired and wireless systems are very different. Recent focus is on decentralized wireless relaying protocols, which have been successfully deployed in mobile ad-hoc networks (MANETs)  Due to the absence of central control and synchronization in MANETs. Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) is a highly efficient random access scheme that is widely used in wireless communication systems such as wireless LANs.  The most widespread deployed protocol is the IEEE b MAC which is based on CSMA/CA. 55

Statistical MA (5) Carrier Sense MA  In ALOHA schemes, users do not consider channel status when they transmit over the channel.  In CSMA, users sense the channel status by listening to the channel to reduce chance of collision and re-transmission.  If no carrier is detected on the channel, then this scheme has very good performance.  In wireless systems, propagation delays may be large and so the performance reduces, however, still CSMA is better than ALOHA. 56

Statistical MA (6) CSMA Techniques  Two common CSMA techniques are: 1)CSMA/CD: used in wired Ethernet o Users start transmission and then listen to channel and abort transmission if collision is detected, use exponential backoff for future transmissions of collided packet. 57

Statistical MA (7) CSMA Techniques 2) CSMA/CD: used in wireless LAN o In wireless networks CSMA/CD is not popular for two reasons: Implementation of CD requires a full duplex radio that can sense the channel at the same time as transmission Due to hidden terminals, even if collision happens at receiver, the transmitters can not detect collisions. o One solution for CA is that transmitter before sending any data, broadcasts a signal onto the network (RTS) in order to tell other devices not to broadcast (directly or through Access Point). o If a user wants to send a packet and channel is busy, it will set a counter and decrement it over time to try again. 58

Statistical MA (8) MAC: Center of gravity in wireless networks  Transmit power levels  Error rates, Interference behavior  Frame Lengths  Throughput, Interference behavior  Scheduling timings  Delay, Interference behavior  IP packet buffering  QoS 59