Multicarrier Modulation

Chapter 1 of MODULE -2
 Multicarrier Modulation
 1st Part of Module -2
 3.2 OFDM basic
 3.3 OFDM in LTE
 3.4 Timing and Frequency Synchronization
 3.5 Peal-to-average ratio
 3.6 SC-FDE
 3.1 - Multicarrier Concept
 ISI free channel – symbol time Ts >> delay spread ζ
 Wideband channels provides high data rates, but Ts<< ζ – ISI
is severe.
 To overcome this problem - multicarrier modulation divides
high-rate transmit bitstream into L low-rate substreams.
 L is chosen so that each of the subcarriers has effective symbol
time LTs >> ζ – ISI free.
 Individual substreams are sent over L parallel subcarriers
maintaining the total desired data rate- Orthogonal subcarriers.
 Each subcarrier data rate << total data rate and BWsubcar <<
BWsystem.
 No. of subcarrier ensures - each BWsubcar << BWc –
subcarrier experience flat fading–small ISI on each subcarrier.
 Exp – 3.1
 A wideband wireless channel has a delay spread of 1µs. In
order to overcome ISI, assume a requirement that 10ζ ≤ Ts.
1. What is the maximum BW allowable is this system if the ISI
constraint is to be met without using multicarrier modulation?
2. If multicarrier modulation is used, and we desire a 10MHz BW,
what is the required no. of subcarriers?
Ans- 1 - If it is assumed that Ts=10ζ, to satisfy the ISI free
condition the maximum BW would be 1/Ts = 100KHz.
Ans-2 - If multicarrier modulation is used symbol time T=LTs. The
delay spread mandates that the new symbol time is still
bounded to 10% of the delay spread, 1/LTs = 100KHz. But the
10MHz BW requirement gives 1/Ts = 10MHz. Hence L> 100 so
L might be 128 subcarriers to allow the full 10MHz BW to be
used with negligible ISI.
 Basic Multicarrier Tx & Rx

 Transmitter - A high rate stream of R bps is broken

into L parallel streams each with rate R/L and then
multiplied by a different carrier frequency.
 Receiver – each subcarrier is decoded separately
requiring L independent Rx.
 Elegant Approach to ISI

 Rbps – high data rate signal, B – passband bandwidth

 L – no. of parallel substreams each with rate R/L passband

bandwidth B/L. After passing through the channel H(f) the

received signal is shown. B/L<<Bc – each subcarrier
experience flat fading.
 Symbol duration on subcarrier is increased to T=LTs.

 If L is large then symbol duration

Exceeds the channel delay spread

T>>ζ – ISI free communication
 Disadvantages – large BW,

high quality & expensive LPF-

maintaining orthogonality, L
independent mod & demod units.
 3.2 - OFDM Basics
 To overcome the requirement for L RF units in both Tx & Rx –
OFDM uses DFT – discrete fourier transform.
 Block transmission with Guard Intervals – L data symbols
are grouped into blocks – OFDM symbol.
 OFDM symbol lasts for T seconds, where T=LTs.
 To separate individual OFDM symbol guard band in introduced
between each OFDM symbol.

 As long as guard time Tg >> ζ channel delay spread – each

OFDM symbol will not interfere with itself- no ISI
 Circular Convolution and DFT

 To remove ISI within each OFDM symbol – when an input data

stream x[n] is sent through a LTI FIR channel h[n] the output
is the linear convolution of input and the channel.
 y[n] = x[n] * h[n]
 For circular convolution

 DFT of channel output

 In the frequency domain
 Unique property of DFT – duality between circular
convolution in the time domain and multiplication in the
frequency domain.
 OFDM Cyclic Prefix

 L point DFT is defined as

 IDFT

 If channel frequency response H[m] is known, input can be

found
 Cyclic Prefix

 To realize OFDM in practice – FFT algorithm is utilized for

computing DFT and IFFT algorithm for computing IDFT –
reduces no. of multiplication & addition.
 IFFT operation at the transmitter allows all the subcarriers to be
created in the digital domain and requires a single radio link.
 Channel must provide circular convolution for IFFT/FFT to
create ISI free channel –
 If maximum delay spread has a duration – v+1 samples, then
by adding a guard band of v samples between OFDM
symbols, each OFDM symbol is made independent of those
coming before and after it.
 Such OFDM symbol is represented in time domain as a length
L vector X=[x1, x2, …xL]
 After applying a cyclic prefix of length v – transmitted signal

 Channel output is
 Where h – is a length v+1 vector describing the impulse
response of the channel during the OFDM symbol.

Output ycp has (L+v) + (v+1) -1 = L+2v samples – input (L+v)
impulse response (v+1)

1st v samples of ycp has interference from the preceding
OFDM symbol - discarded. Last v samples disperse into the
subsequent OFDM symbol – discarded.
 This leaves exactly L samples for the desired output Y.
 Cyclic Prefix


Consider only the 1st element of y – yo. Due to cyclic prefix yo
depends on xo and the circularly wrapped values
 That is

 Due to circular convolution – a cyclic prefix that is at least as

long as the channel duration allows the channel output y to be
decomposed into a simple multiplication of the channel
frequency response H=DFT{h} and the channel frequency
domain input X=DFT{x}.
 Rate Loss = Power Loss = L/(L+v). If L>>v – inefficiency can
be reduced.
OFDM Cyclic Prefix creates Circular
Convolution
 OFDM Block Diagram

 1st step in OFDM is to break a wideband signal of BW B into L

narrowband signals (subcarriers)each of BW B/L.
 Symbol rate is maintained but each subcarrier experiences
flat fading or ISI free communication – as long as cyclic prefix
>> delay spread is used.
 The L subcarriers for a given OFDM symbol are represented by
a vector X, which contains the L current symbols.
 To use a single wideband radio instead of L independent
narrow band radios, the subcarriers are created digitally using
an IFFT operation.
 In order for the IFFT/FFT to decompose the ISI channel into
orthogonal subcarriers, a cyclic prefix of length v must be
appended after the IFFT operation.
 OFDM Block Diagram

 L+v symbols are sent serially through the wideband channel.

 At the RX the cyclic prefix is discarded and the L received
symbols are demodulated using FFT operation - giving L data
symbols, each of the form Yl =Hl Xl + Nl for subcarrier l
 Each subcarrier is equalized by FEQ – by dividing by Hl
 Result is Xl=Xl + Nl / Hl.
OFDM in LTE
 3.3 - OFDM in LTE

 Inputs are L independent QAM symbols (vector X), these L

symbols are treated as separate subcarriers.
 These L data-bearing symbols can be created from a bit stream
by a symbol mapper and serial-to-parallel converter.
 L point IFFT then creates a time domain L-vector X that is cyclic
extended to have a length L(1+G), G- fractional overhead. G =
0.07 - normal cyclic prefix, G = 0.25 -extended cyclic prefix.
 This vector is then parallel-to-serial converted into a wideband
digital signal, that can be amplitude modulated at a carrier
frequency wc/2π.
 Data rate of LTE system
with 16QAM
OFDM Parameters in LTE
 Frequency Equalization

 Equalization is the processes of reversing the distortion

occurred during signal transmission.
 To estimate the received signal, complex channel gain must
be known – knowledge of amplitude & phase.
 For simple modulation QPSK, phase information is sufficient as
amplitude doesn’t transmit information.
 After FFT is performed data symbols are estimated using one-
tap frequency domain equalizer of FEQ as
 Hl = complex response of the channel at the
frequency fc + (l-1)Δf
 It both corrects the phase and equalizes the amplitude before
the decision device.
3.4 - Timing and Frequency Synchronization

 Synchronization is the most challenging problem in OFDM.

 To demodulate OFDM signal – 2 synchronization is used :
 Timing synchronization – determining symbol offset timing
and optimal timing instant.
 Frequency synchronization – Rx carrier frequency must be
as closely as possible tuned to Tx carrier frequency.
 Frequency synchronization of OFDM is stringent due to
orthogonality of subcarriers.
 In time domain – IFFT modulates each data symbol onto a
unique carrier frequency.
 Due to time window of T = 1 μsec – frequency response of
each subcarrier is a sinc function with zero crossing every 1/T
= 1MHz.
OFDM Synchronization in Time
 OFDM Synchronization in Time

 Shown with a Fourier Transform

 If carrier frequency synchronization is perfect – desired

subcarrier amplitude will be maximum and ICI – inter carrier
interference is zero.
 If carrier frequency is misaligned to some amount δ – some of
the desired energy is lost and ICI – inter carrier interference is
introduced.
OFDM Synchronization in Frequency
 1. Timing Synchronization

 Effect of timing error in symbol synchronization is not severe

due to the presence of cyclic prefix.
 If perfect time synchronization is not maintained – still a
timing offset of δ sec can be tolerated without any degradation
in performance, as long as
 Tg – guard time / cyclic prefix duration and Tm – maximum
channel delay spread.
 δ < 0 – sampling earlier the ideal instant of time and δ > 0 –
sampling at a later instant of time.
 As long as is maintained timing offset results in
a phase shift per subcarrier of which is fixed for all
subcarriers. If δ remains constant fixed phase offset occurs
that can be removed by FEQ without any loss in performance.
 1. Timing Synchronization
 Acceptable range of δ - timing synchronization margin
 If the timing offset δ is not within this range
ISI occurs.
 For δ > 0 - Rx losses some of the desired energy and
incorporate undesired energy from the subsequent symbol.
 For - desired energy is lost and interference from
the preceding symbol is included in the Rx window.
 1. Timing Synchronization

 Observations
 SNR loss for both these case :
 SNR decreases quadratically with the timing offset.
 Longer OFDM symbols are immune to timing offsets – more
no. of subcarriers helps.
 Since in many δ << LTs – timing synchronization errors are
not so critical as long as the induced phase change is
correct.
 2. Frequency Synchronization

 OFDM gives a high degree of BW efficiency compared to other

wideband systems – subcarrier packing is extremely high.
 The zero crossing frequency offset δ=0 – no interference
between subcarriers.
 Frequency offset is not always 0 due to mismatched
oscillators at the Tx & Rx and Doppler frequency shifts due
to mobility.
 Since precise crystal oscillators are expensive some degree of
frequency offset are present in the LTE system with OFDM.
 E.g. – if oscillator is accurate to 0.1ppm – foffset = fc 0.1ppm
 If fc = 3GHz & Doppler = 100Hz then foffset = 300+100Hz – will
degrade the orthogonality of the received signal – received
samples of the FFT will contain interference from the adjacent
 Analyzing ICI on OFDM Performance

 Matched filter Rx for subcarrier l 1/LTs = ∆f

LTs = duration of the data portion of the OFDM symbol,
T=Tg+LTs
 Interfering subcarrier m is given by
 If the signal is demodulated with a fractional frequency offset δ,
where I δ I ≤ ½

 ICI between subcarriers l and l+m using a matched filter (FFT)

is the inner product between them

 As δ =0 Im =0 and m=0 Im =0
 Analyzing ICI on OFDM Performance

 Total average ICI energy per symbol on subcarrier l

 C0 = constant that depends on various assumptions

 εx = average symbol energy
 Approximation indicates that this expression assumes there are
infinite no. of interfering subcarriers.
 SNR loss induced by frequency offset:
 Analyzing ICI on OFDM Performance

 SNR loss vs δ, relative to subcarrier spacing. Solid lines- fading

channel & dotted lines – AWGN channels.
 Observations:
 SNR decreases quadratically with frequency offset.
 SNR decreases quadratically with no. of subcarriers.
 For low loss < 0.1dB δ= 1-2% of subcarrier spacing
 3.5 - Peak-to-Average Ratio

 OFDM signals have high PAR – peak-to-average power ratio

(PAPR).

 Because – in the time domain a multicarrier signal is the sum

of many narrowband signals – this value can be large and
small – larger than average value.

 High PAR is an important challenge as it reduces efficiency and

increases cost of RF power amplifier.

 Same power amplifier can be used with reduced input power –

input backoff IBO – causes lower average SNR at the Rx.
 PAR Problem

 When high-peak signal is transmitted through nonlinear device

– HPA / DAC – generates in-band distortion – severely
degrading the system performances.
 Typical Power amplifier response with input & output backoff
regions
 To avoid distortion -
 High peak power signals
are transmitted in the
Linear region – by
decreasing average power
of input signal.
 Input backoff IBO resulting
in proportional output
backoff OBO.
 PAR Problem
 High backoff - reduces power efficiency of HPA, limit mobile
battery life, reduces coverage range & increases cost of HPA.
 Input backoff is defined as:
 Pinsat = saturation power, above which is nonlinear region,
 Pin = average input power
 Amount of backoff is usually ≥ PAR of the signal.
 Power efficiency of the HPA can be increased by reducing
the PAR of the transmitted signal – efficiency of class A
amplifier is halved when the input PAR is doubled.
 High PAR needs – high resolution for both Tx DAC & Rx
ADC as dynamic range of the signal is proportional to PAR.
 High resolution of DAC & ADC – increases complexity, cost,
power.
 Theoretical efficiency limits of 2 classes of HPA – typical
OFDM PAR range – 10dB – PAR efficiency is 50-75% lower
than single carrier system
 Quantifying the PAR

 Since multicarrier systems transmit data over a no. of parallel

frequency channels, resulting waveform is superposition of L
narrowband signals.
 Each of the output samples from an L-point IFFT operation
involves sum of L complex numbers.
 Due to Central Limit Theorem resulting output values{x1, x2,
…xL} can be accurately modelled as complex Gaussian
random variables with zero mean & variance
 Amplitude of the output signal :
 As is complex Gaussian like narrowband fading the
envelope is Rayleigh distributed.
 Output power is which
is exponentially distributed with mean
 Quantifying the PAR

 PAR of the transmitted signal is defined as :

 Usually PAR is considered for a single OFDM symbol,
consisting of L+Ng samples or a time duration of T.
 Discrete PAR for IFFT output :

 Complementary cumulative distribution function CCDF= 1-CDF

of the PAR is the most commonly used measure.
 Van Nee and de Wild introduced a simple and accurate
approximation of the CCDF for large L(≥64)


CCDF of PAR for QPSK OFDM: solid line –
simulation result & dotted line – approximation
 Clipping and Other PAR Reduction
Techniques
 To avoid HPA operating in the nonlinear region, input power is
reduced to an amount equal to PAR – this will reduce SNR by
the same amount.
 PAR reduction –
 Some highest PAR values can be clipped off – reducing some
amount of distortion – still some amount of distortion remains.
 Clipping – also called soft limiting – truncates the amplitude of
the signal that exceed the clipping level as
 Original signal is output after clipping
 A = clipping level.
 Clipping ratio is defined as
 Clipping reduces PAR at the expense of distorting the desired
signal.
 Clipping and Other PAR Reduction
Techniques
 Drawbacks of clipping :
 1. Spectral regrowth – frequency domain leakage – causes
interference to users in neighboring RF channels.
 Clipping noise can be expressed in frequency domain by DFT.
 Clipped frequency domain signal is
 Ck = clipped of signal in the frequency domain.
 Ck increases as clipping ratio is reduced
from 7-3dB.
 As clipping ratio decreases distortion
increases.
PSD of the original and clipped OFDM
signals with clipping ratio γ = 3, 5, 7dB
 Clipping and Other PAR Reduction
Techniques
 2. Distortion of the desired signal – in-band distortion
 Attenuated signal is given by
 d[n] = uncorrelated additive noise and α = attenuation factor

 Attenuation factor α is plotted

as a function of clipping ratio γ.
 Attenuation factor is negligible
when clipping ratio > 8dB
 Variance is given by
 Clipping and Other PAR Reduction
Techniques
 Signal-to-noise-plus-distortion Ratio
 Bit error probability can be evaluated

 BER of OFDM system with

L=2048 subcarriers and 64QAM
 Other PAR reduction methods
 Peak cancellation
 Iterative clipping & filtering
 Tone reservation
 Active constellation extension
 LTE’s Approach to PAR in the Uplink

 In downlink – PAR is less important as base stations are fewer

in no. and higher in cost.

 If PAR is very high it can be reduced at the cost of complexity

and performance trade-offs.

 Input power backoff is done to keep spectral regrowth and in-

band distortion within the acceptable level.

 In uplink – mobiles are many in no. and sensitive to cost.

Single-carrier frequency division multiple access SC-FDMA is
used.
 3.6 - Single-Carrier Frequency Domain
Equalization (SC-FDE)
 SC-FDE maintain 3 most important benefits of OFDM:
 Low complexity even for severe multipath channel
 Excellent BER performance
 Decoupling of ISI from other type of interference.
 Single carrier transmission reduces PAR relative to multicarrier
modulation.
 SC-FDE System Description
 In SC-FDE, IFFT is moved to the end of Rx instead of the Tx.
 SC-FDE uses a cyclic prefix as long as the channel delay
spread.
 Transmitted signal is QAM symbols with low PAR 4-5dB, as an
unmodulated sin signal has PAR 3dB, so PAR can’t be reduced
below that of SC-FDE system.
 Comparison between OFDM & SC-FDE
system

 Due to cyclic prefix the received signal appears to be

circularly convolved:

 Input can be estimated and the resulting signal is then

converted back into time domain using IFFT to give
 SC-FDE Performance vs OFDM

 Primary difference in performance – noise treatment.

 In both the SC-FDE & OFDM Rx FEQ inverts each frequency
by 1/Hl. – SNR remains unchanged as both signal & noise are
inverted by same amount.
 SC-FDE- FEQ operates on the frequency domain dual of the
data symbols - noise effects all the data symbols instead of
one.
 OFDM performs better when coding is strong or constellation is
large.
 Since higher order constellations are more sensitive to
noise, this affects SC-FDE more as all the symbols are
affected by amplified noise.
 Design Considerations for SC-FDE & OFDM

 SC-FDE has a lower complexity Tx but a higher complexity Rx

compared to OFDM.
 In LTE system – uplink can use SC-FDE and downlink can use
OFDM – to reduce PAR, cost and power.
 The preamble & pilot tone is inserted in the frequency domain
for OFDM but these are inserted in time domain for SC-FDE.
 Disadvantage of SC-FDE – more dispersive spectrum
compared to OFDM.
 OFDM has more sharper spectrum resulting in less co-channel
interference – but due to high PAR OFDM is more subject to
clipping.
 OFDM is more popular than SC-FDE.