Mobile and Satellite Communications Lecture 6 Multiple Access Techniques Multiple Access How can many uncoordinated users share the same radio spectrum? Shannon s theory factors that determine the capacity of a channel. For the AWGN channel: S C = B log 2 1 + 1
Communications Resources Three basic resources that we can between users: Time (TDMA) Frequency (FDMA) Power (CDMA) Each implies a Multiple Access strategy we might adopt. Multiple Access Strategies Time Division Multiple Access (TDMA) Frequency Division Multiple Access (FDMA) Code Division Multiple Access (CDMA) 2
Duplexing Allowing communication in both directions simultaneously Problem: Transmitted power: 30dBm Received power: -100dBm Leakage from the Tx side would swamp the Rx signal Duplexing - Solutions Have separate, electrically isolated Tx and Rx antennae Expensive, and impractical in a portable device Frequency Division Duplexing (FDD) transmit and receive at different frequencies Time Division Duplexing (TDD) rapidly switch between Tx and Rx operations, so it appears seamless to the user 3
Frequency Division Duplexing Txand Rx at different frequencies A filter can be used to separate Txand Rx signals Frequency separation Frequency separation enough so can filter out Tx signal But, not too great so is out of antenna bandwidth Only choice in analogue systems (AMPS and ETACS) FDD - Practical Isolation often require in excess of 100dB Requires a good filter, and require accurate impedance matching Notch filters are a common choice 4
FDD - Practical Typically standardise the frequency separation to optimise the electronics Example: Base Station pass-reject duplexer for GSM Time Division Duplexing Only possible in digital cellular systems Audio codecs mean transmission can be bursty A user is allocated a regular repeating timeslot Tx and Rx timeslots can be offset Allows time for transceiver to switch between Tx and Rx 5
TDD No duplexer filter is required Example GSM 8 timeslots per frame, 3 timeslot offset between Tx and Rx Issues: TDD - Practical Switching time large power differential in Tx and Rx Synchronisation across entire network, with varied MS distances and multi-pathing Removes Duplexer from MS, but not the Base Station (transmits to multiple users) 6
Multiple Access Techniques Allowing multiple users to share available communication resources. Two broad types (both employed in cellular networks): 1.Random Access uncoordinated users access channel resources (computer networks, cellular control channels -> packet based) 2.Fixed Access user assigned a channel for entire duration of session (circuit switched) Multiple Access in Cellular Systems Example: Mobile initiated call 7
Multiple Access in Cellular Systems Mobile attempts to access the network on a Random Access Channel Network can t assign when the MS will require access/network resources Channel Assignment performed on an assigned Control Channel Traffic exchanged on a fixed, assigned channel Cellular Systems types of channels Cellular Systems employ two different types of channels: 1. Random Access Channels for network registration, paging, location updates (RACH, PCH) 2. Fixed, Assigned Channels for the exchange of control or traffic data (voice channels, dedicated control channels) 8
Multiple Access - types Random Access MA ALOHA Slotted ALOHA Carrier Sense MA (CSMA) Allocated MA FDMA TDMA CDMA Frequency Division Multiple Access Users are each assigned disjoint frequency bands (as channels) Essentially divide the allocated spectrum into distinct channels Common for Radio and Television systems The only possibility in analogues systems -> used for AMPS and ETACS (1G systems) 9
FDMA Capacity Modulated signal would require a bandwidth, B chan This would also include a guard band protect against adjacent channel interference Guard bands at end of spectrum, to protect against other systems/operators, B guard Number of channels available is then: = B sys 2B B chan guard FDMA Guard bands Size of required Guard band determined by: performance of channel filters to reject out of band signal range of signal amplitude variation between channels out of band spectral occupancy of transmitted signal 10
FDMA Example - AMPS Analogue data -> requires FDMA Did not have high performance filters available in the 1980s Recall 7 cell cluster assignment FDMA Example - AMPS Channel Assignment in seven maximally separated groups: A set: Channel numbers: 1, 8, 15,... B set: Channel numbers: 2, 9, 16,... etc... spacing within cell 7x30kHz = 210kHz Later employed a 120 sectored arrangement increased frequency spacing to 630kHz improved SIR S = I ( 3 ) 2 2 11
FDMA Example - AMPS No guard bands used in AMPS between adjacent channels Channel assignment to cells done to minimise adjacent channel interference FDMA Advantages Narrow bandwidth channels -> flat-fading, so no complex receivers (equalisers) required No synchronisation overheads (suits uncoordinated users) Simple, analogue circuitry and signal processing 12
FDMA Disadvantages Dedicated channel per user for voice conversation is inefficient more 50% of average conversation is silence Guard bands waste spectrum typical 10 25% of channel bandwidth Require complex band-pass filters and RF filters to reject adjacent channel interference Require duplexers (must be FDD) FDMA in Satellite Systems Initial satellite systems carried analogue data (TV and telephone signals) -> hence required FDMA Filter technology well-understood Later, allowed demand assignment of channels -> DAMA (Demand Assigned MA) Challenge in satellite FDMA is intermodulation products (IM) 13
Intermodulation (IM) Satellites require high-powered amplifiers Common was Travelling Wave Tube Amplifier (TWTA) before the development of Solid- State High Powered Amplifiers (SSHPA) TWTA tend to be non-linear Operated at back-off limit to linear region TWTA RF signal feed along a delay-line Induces current in electron beam (velocity modulation) Electron beam induces current in the line, amplifying the signal Gains of 70dB are possible! 14
IM Products Due to non-linear characteristics in amplifier Model amplifier characteristic as: This is simplest polynomial approximation Input signal contains a range of FDMA signals: V in = V 1 cos( 2π f 1 t) + V 2 cos( 2πf 2 t) V = AV + b A little bit of trig-identities shows that the output signal is: out in ( V ) 3 in IM Products Desired signal output: ( 2π ft) + AV cos( πf t) V out = AV 1 cos 1 2 2 2 Intermodulation product is the mixed signal at near-by frequencies: 2 2 = bv V cos( 4πft 2πf t) + bv V cos( 4πf t πft) V IM 1 2 1 2 2 1 2 2 1 Other components are in different frequency ranges (triple input frequencies) 15
IM Products Components at frequencies 2f1 f2 and 2f2 f1 can lie within band. Example, C-band transponder at frequencies 3718MHz and 3728MHz IM products at 3708MHz and 3738MHz If signal bandwidths were B1 and B2, then IM signal will be spread over bandwidths 2 1 B2 and Increases the noise floor: C = 1 B + 2B 2+ B1 + 1 ( C ) ( C ) in IM 1 TWTA Back-off TWTA saturates at input power of -100dBW Input back-off of -2.2dB Output power at saturation is 17dBW (or 50W) Output back-off is 1dB, giving output power of 16dBW 16
Time Division Multiple Access Users are allocated a repeating time-slot Tx and Rx in a round-robin fashion Naively, same capacity as FDMA system: Analogue Information signal bandwidth Sample at rate 2B chan B chan Spectral allocation of B sys implies 2B sys signalling basis functions to communicate with per second Hence, number of simultaneous user in roundrobin is: 2Bsys users 2B chan TDMA - Advantages Not including guard bands this is the same as an FDMA system In practise, TDMA systems have higher capacity: Speech compression Digital modulation schemes Advanced digital signal processing techniques 17
TDMA Scheme General repeating frame structure shown below Each user will be assigned a timeslot in the frame time slot #n TDMA Features Initial preamble mark the beginning of the frame (say for cellular downlink or satellite comms) Guard interval allows users to switch and power-up. Also allows for synchronisation slips 18
TDMA Features Synchronisation three types 1. Carrier Synchronisation If use coherent demodulation Use a PLL to track the carrier Send a string of 1111...1 -> unmodulated carrier, for receiver to lock-on to Output reference carrier enables demodulation TDMA Features 2. Bit clock synchronisation Alternating sequence of 1 s and 0 s Allows receiver to determine the beginning and end of bits 3. Word synchronisation Generally by transmission of unique word Correlate received sequence against known unique word -> align with frame structure 19
Word Synchronisation - Issues This pattern appears in the data? the longer the word, the less likely window the detection based on previous sync Bit errors in the received unique word shorter the word pattern, the less likely lower the detection threshold (say 25 out of 28 bits) Allows corrections of signal inversion due to PLL phase ambiguity TDMA Efficiency Preamble, guard, and synchronisation bits waste signal space The proportion of total frame size used to communicate actual data is the frame efficiency Determined from number of overhead bits and total bits sent per frame: η f b = 1 b OH TOT 20
TDMA Capacity Most systems are a combination of FDMA and TDMA MF-TDMA Like GSM, 200kHz channels with 8 time-slots per channel If TDMA system supports mtimeslots per channel, number of simultaneous users is: mb = ( 2B ) sys B chan guard TDMA Example - USDC US Digital Cellular was the 2G US equivalent to GSM Was a TDMA scheme, evolved closely from AMPS Adopted the same 30kHz channels, FDD with 45MHz spacing 6 timeslots per frame, at full-rate speech a user would get 2 timeslots/frame 21
TDMA Example - USDC Frame duration was 40ms, with 1944 bits/frame Modulation technique was π/4-oqpsk Symbol rate of 24.3kbaud. Root-raised cosine (Nyquist) shaping pulses, with α=0.35. Fits into the required 30kHz channels Hence, air data rate was 48.6kbps 324 bits per timeslot TDMA Example - USDC Frame structure different for Forward and Reverse channels 22
USDC Forward Link Timeslot 28 bits for synchronisation (carrier and bitclock) SACCH Slow Associated Control Channel. Used for power control information CDVCC Coded Digital Verification Colour Code. Effectively the MS address, and used for word sync 260 bits for data USDC Reverse Link timeslot Guard interval MS are uncoordinated transmitters Ramp-up allow transmitter circuitry on MS to power on 260 data bits output of channel codec, (260, 159) 159 raw speech samples Frame efficiency is USDC η f 159 = 49.1% 324 Voice codec rate was 7.95kbps, so produced 2x159 every 40ms 23
TDMA Advantages Discontinuous transmission natural idleness of speech Reduces battery power Allows things like MAHO mobile can perform other functions while idle No duplexer Bandwidth on demand multiple timeslots assigned if needed TDMA Disadvantages Synchronisation is required across all users, to prevent over-lapping timeslots Overheads needed can account for a large segment of traffic (USDC 50%) High instantaneous data rates implies wide bandwidth signals -> signal bandwidth is often larger than channel coherence bandwidth. This requires equalisation (GSM) 24
TDMA in Satellite Systems Some advantages of TDMA: Adaptability to voice, video, or data traffic One signal in transponder at one time, so intermodulation issues Challenges in Satellite applications: Larger bandwidth implies greater noise power -> impractical in VSAT systems Issues of equalisation TDMA - Satellites Synchronisation more complex in satellite TDMA systems Doppler effects due to motion Local oscillator drift 25
Spread Spectrum Communications Evolved out of military applications Aim to spread signal over a wide bandwidth so it is undetectable to eaves-droppers Ratio of spread bandwidth to original signal bandwidth is called the spreading gain B Gains of 20dB to 60dB are common G= B trans signal SS - Applications Military high tolerance against jamming (intentional interference). Also communication to be un-detectable Accurate position and velocity estimation -> used in GPS Multiple access technique allowing a large number of uncoordinated users to share radio spectrum (3G cellular networks, WLANs,...) 26
SS -Types 1. Frequency Hopping (FHSS) transmitter pseudo-randomly alters transmission frequency, and receiver can follow 2. Time-Hopped (THSS) signal is transmitted in short bursts pseudorandomly 3. Direct-sequence (DSSS -> CDMA) a highrate code signal is used to spread the data signal to a fixed, wide bandwidth Frequency Hopped Spread Spectrum Used in Bluetooth (IEEE802.15), some WLAN (802.11), and in GSM Instantaneous transmission is narrowband Carrier frequency is changed over some available, wideband spectrum 27
FHSS Instantaneous bandwidth of signal, Spectrum available, Number of available channels -> same as the spreading factor (or processing gain) = G= B sys signal B signal Generally, the greater the number of frequencies the better the FHSS system B B sys FHSS The other parameter is the hopping period the time between hops Relative to symbol period, T hop Defines fast or slow frequency hopping T s 28
Fast FHSS Hop period less than symbol period A single symbol is thus transmitted at several different frequencies Good diversity technique unlike the different frequencies are all experiencing fading at the same time The level of diversity depends on signal bandwidth, coherence bandwidth, and available hopping spectrum Hard to implement in practise, due to difficulty in switching oscillator frequency rapidly Slow FHSS Hopping rate is less than symbol rate Most common in practise When combined with inter-leaving and channel coding, can still offer significant diversity gain 29
FHSS - Implementation FHSS can be done with any modulation scheme (but usually simple FSK) FHSS Example - Bluetooth IEEE802.15 standard Intended for short-range (up to 10m) unlicensed use 2.4GHz band. 79 channels with 1MHz bandwidth each Gaussian-filtered FSK (GFSK) modulation Data rate is 720kbps, so symbol period is 1.4µs Performs 1600 hops per second, so hoping period is 625µs. Slow FHSS system 30
FHSS Example - Bluetooth Hopping sequence established on connection between the two devices (two cell phones, or an MP3 player and sound system, etc.) Performance metric is the collision probability the likelihood two different devices will hop to the same frequency. Both transmissions would then be lost For Bluetooth, this is 1/79 (since there are 79 available frequencies) FHSS System Performance Collision probability, When they collide, BER becomes 0.5 BER of FHSS system: PFH = 0.5Pcoll + ( 1 Pcoll) BER mod Here, BER mod is the bit error rate of whatever modulation scheme is used P coll This is a function in turn of the signal to noise ratio at the receiver, E b 0 31
Soft Capacity As more users come onto system, the probability of a collision will increase The performance of the system will degrade -> BER increases System capacity is reached when the degradation reaches its limit Capacity is not determined by the number of available channels, like in FDMA & TDMA FHSS Collision Probability Assumptions: Slow FHSS, sends nbbits at each hop N frequencies are available Users are uncorrelated Each user can hop to anyone frequency For simplicity, begin by considering just two users on the system User A is using frequency foat this instant 32
FHSS Collision Probability Consider user B Probability it s using the same frequency, fo, at this instant is: 1/N Probability not using fo: 1 1/N Probability it will hop in this cycle: 1/nb If user B hops, then probability it hops to fo: 1/(N-1) Hence, we calculate the probability of a collision between these two users: 1 1 1 1 1 1 ( ) P = + = 2 1 1 + nb 1 n b FHSS Collision Probability If there are Kactive users on the system: K 1 P coll = 1 ( 1 P2) Improve system performance, increase number of available frequencies Have neglected channel coding, and issues of fast-frequency hopping Need to consider demodulator structure 33
Direct Sequence Spread Spectrum Will refer to as Code Division Multiple Access (CDMA) Used in 2.5G cellular networks (IS-95); 3G cellular standards (W-CDMA and cdma2000); and WLAN (802.11a-g family) Basic idea the narrowband data signal is mixed with a high-rate code signal, with a noise-like spectrum CDMA The high-rate rate (or spread ) signal is then converted to RF and transmitted The receiver converts the received signal back to the baseband. It then correlates the received signal with the known spreading sequence (or code sequence), and is able to de-spread and recover the data 34
CDMA CDMA signal processing: CDMA - Spreading Spreading can be thought of as multiplying the signals, if they re represented as polar line codes Alternative is XOR of bit stream mathematically isomorphic groups. 35
CDMA - Spreading Bandwidth of a signal is proportional to the data rate Hence, spreading increases the data rate of signal Spreading factor: T T b c G = = = c R R data B B code data CDMA Spreading The spreading factor represents the ratio by which the SNR is increased at the output relative to that on the channel hence is often called the processing gain Spread signal: r( t) = Amt ( ) c( t) cos( 2πfc t) + n( t) SNR over the channel: S R spread A = B Receiver multiplies spread signal by the code: (1x1 = 1; -1x-1 = 1) -> recovers the signal r A 2 A 2 0 2 2 code 2 ( t) = mt ( ) c ( t) + n ( t) c( t) = mt ( ) n ( t) c( t) I + I 36
CDMA - Spreading Hence, SNR at the output of the detector: S R out A = B 0 2 2 data = G S R spread Implies reasonable detection of the signal, even when the signal is well below the noise floor on the channel CDMA MATLAB Simulation 1 37
CDMA Multiple Users Multiple users can be accommodated if they are given orthogonal spreading sequences By orthogonal, we mean they re uncorrelated signals: R CC τ = 2 c1 tc2 t If the signals of two users are present at the detector: rt Am1 tc1 t cos 2πf t + Am2 t τ c2 t τ cos 2πfc t Recover the signal of User 1: ( ) ( ) ( τ) dt 0 1 ( ) = ( ) ( ) ( ) ( ) ( ) ( ( τ) ) n( t) c + A A r 1 2 2 2 ( t) = m ( t) + cos( πf τ) c ( λ) c ( λ τ) dλ+ n( t) If the code sequences were uncorrelated, then there ll be no contribution from User 2 s signal at the output T b c 1 2 0 CDMA Spreading Sequences Properties sought for individual code sequences: The main issue is acquisition Seek signals with autocorrelation function as close to a delta function as possible: R c = ( τ) c( t) c( t τ) dt δ( τ) Sliding window correlator receiver 38
CDMA - Autocorrelation Only get an output when the code is aligned to itself Receiver code acquisition CDMA Code Family For a family codes, we need a set of codes with strong auto-correlation properties as well as no cross-correlation R C 1 C = 2 1 2 ( τ) c ( t) c ( t τ) dt 0 For the down-link (BTS transmits), we can synchronise the code sequences and our cross-correlation can be restricted to: T b c 0 ( t) c ( t) dt 0 1 2 = 39
CDMA Walsh codes A family of such codes can be formed using the Walsh-Hadamard matrices Can be constructed to any power of 2 H 2 1 = 1 1 1 H 2 H = H H The rows have the desired auto-correlation and cross correlation properties Used in IS-95 Downlink Not suitable for the up-link, as orthogonality is not preserved on a relative shift of the codes H CDMA MATLAB Simulation 2 40
CDMA PN Sequences For the Uplink, when we can t synchronise MS transmissions, we require more general orthogonal sequences Seek to form Pseudo-Noise Random sequences (or PN sequence) appear random but really have hidden structure Properties of a PN sequence: Equal number of 1 s and 0 s Run lengths of 1 s or 0 s follow coinflipping statistics, ie., probability of N straight is 1/2^N Uncorrelated with shifted versions of itself CDMA PN sequences A common method of forming them is to use linear feedback shift registers (LFSR) Binary registers with feedback Understood as state machines Generally after Maximal length LFSR (MLSR) use m registers, and produce a sequence of m length 2 1 41
CDMA PN sequences The output sequences from MLSR are called m-sequences Can generate a family by off-setting these sequences m-sequences have properties desired in spreading codes good auto-correlation and cross-correlation properties IS-95 used an m-sequence of length 2 42 1 with various off-sets CDMA Gold Sequences Can obtain even better properties by summing together two different m- sequences (XOR) These are called Gold Codes Used in W-CDMA 42
CDMA - Advantages 1. Universal Frequency Re-use: No frequency planning. Adjacent cells can use the same spectrum Cells can be distinguished by unique spreading codes 2. RAKE Receivers: good diversity technique Wide-bandwidth, B spread >> B c Distinguish individual multipath components T chip <<σ τ RAKE Receiver Auto-correlation property of spreading code A multipath component won t affect detector output R c τ = ctct τ dt Can use multiple fingers to recover the later multipath components ( ) ( ) ( ) δ( τ) 43
CDMA - Advantages 3. Soft Handoff a MS can be handled by two or more base stations simultaneously -> they just appear as different multi-path components on the RAKE receiver input CDMA - Advantages 4. Source variability when no data to transmit, then don t send anything. Saves power on MS, and reduces noise level on system. Voice Activity Detection (VAD) 5. Bandwidth flexibility doesn t have the same tight bandwidth restrictions as FDMA/TDMA. Can add things like FEC at cost of spreading gain alone 44
CDMA - Issues There are a few issues/challenges associated with CDMA systems: Soft Capacity Limit The near-far effect Power control CDMA - Capacity Like FHSS, there s no fixed number of channels in a CDMA system (as long as we can find new orthogonal codes) The signals of other users appear as noise to a specific user The more users added, the noise level increases -> eventually performance degrades for all users. This is soft capacity 45
CDMA - Capacity It is easy to show that the SNR with K users Ps S R1 is: SI R= = B P G s ( K 1) 1+ ( K ) 0 code + 1 S R G 1 Other users signals appear as noise Performance improves with increasing spreading factor, G SNR reduction increases BER, system performance degrades CDMA Near-far Effect Recall this is a factor for all cellular networks It is a particular problem in CDMA, since another users signal appears as noise 46
CDMA Power Control The near-far effect makes power control a premium in CDMA networks Must happen rapidly, to account for shadowing events Fast: Open-loop Power Control adjust Txbased on Rx signal power Slow: Closed-loop Power Control send power measurements back to BTS and receive instruction Contention-based Multiple Access Techniques Well suited for packet-based data systems Widely used in Computer WLANs Used in random access channels (RACH) in cellular networks How uncoordinated MS get access to network resources, and have FDMA/TDMA/CDMA traffic channels assigned 47
ALOHA The very simplest Contention-based MA scheme Whenever a user has a data packet to transmit on the channel it does so It then waits for an ACK If no ACK is received, then it waits a random amount of time and transmits again ALOHA - Performance Measure performance of Contention-based schemes as throughput: the amount of traffic that can successfully make it through the channel Analyse ALOHA Assumptions: Fixed packet duration, τ Packets arrive following a Poisson n distribution: ( λτ) λτ Pr { n packets in τ} = e n! 48
ALOHA - Throughput Throughput is the probability a packet is successful multiplied by the traffic load offered Probability a packet is successful == probability there is no collision == probability no packet is generated within the vulnerable period ALOHA - Throughput Vulnerable period is 2τ Probability no packets are created in this time: 2λτ Pr{ no collision} =e Rate of traffic generation: R=λτ Therefore, throughput is: T = R Pr Max. throughput is 0.184 Erlang, when offered traffic is 0.5 Erlang 2R { success} = Re 49
Slotted ALOHA The choice for the RACH in AMPS, ETACS, GSM, USDC, IS-95, W-CDMA, and cdma2000 Same as ALOHA, but user must wait until the beginning of the next timeslot to transmit data Slotted ALOHA - Throughput This reduces the vulnerable period to τ R The throughput is then: T = R Pr{ success} = Re Maximum throughput is then 0.37Erlang It is effectively twice as efficient as pure ALOHA 50
Carrier-Sense MA (CSMA) User listens to channel to sense if it is vacant before attempting transmission This further reduces collision probability, increasing throughput Different ways it can be done: 1-persistent CSMA if channel is idle, it transmits. Otherwise, waits until it is idle and transmits immediately p-persistent CSMA when channel becomes idle, transmits with probability p non-persistent CSMA if channel is idle, transmits. If not, waits a random amount of time and then has another look CSMA - Practical WiFi (IEEE802.11) uses a version of nonpersistent CSMA Ethernet 1-persistent CSMA with Collision Detection (CD) 51
Demand Access Multiple Access (DAMA) Assign channels on demand in a satellite system (transponder) Particularly for telephony and data traffic, fixed FDMA or TDMA assignment is inefficient Requires the use of a Common Signalling Channel (CSC) the role of RACH in satellite system Users send short packets on a CSC, requesting a fixed data channel for a certain period. Assignment is performed by a controlling, master Earth station (minimise processing on-board) DAMA Uplink is usually FDMA Figure shown has 45kHz channels For QPSK modulation with RRC pulse-shaping (α= 0.5), can carry 64kbps of data A user may be assign one or more of these FDMA uplink channels, as required 52
DAMA Downlink is usually a single wide-band, high data rate TDMA signal TDM is performed on the satellite A continuous downlink signal, to allow serviced Earth stations to maintain synchronisation VSAT limit bandwidth that can be used (noise bandwidth at receiver) 53