From OFDM to LTE

Fabrizio Tomatis (ST-E) – Based on slides from Andrea Ancora (ST-E)

 Agenda
Introduction
– OFDM History
– Principles of Wireless Communications
OFDM principles
– Serial-to-parallel conversion
– Cyclic prefix
– Frequency domain orthogonality
– Fourier transform
– OFDM advantages and disadvantages
Dimensioning of OFDM system parameters
– Principles
– Example of LTE
OFDMA
– Orthogonal Frequency Domain Multiple Access
– Combination of OFDMA and TDMA
OFDM implementation issues
– PAPR and non-linearities
– Frequency offset and time varying channel sensitivity
– Insufficient cyclic prefix
– OFDM pilot structure and channel estimation
Summary and conclusions

CONFIDENTIAL 2
Andrea Ancora, Matthew Baker
Introduction:
OFDM History
Principles of Wireless Communications

CONFIDENTIAL 3
Andrea Ancora, Matthew Baker
 Introduction: OFDM History

The first OFDM patent was filed in US by Chang at Bell Labs in 1966 and the patent was
issued in 1970.
“R.. W. Chang, Synthesis of band-limited orthogonal signals for multichannel data transmission, Bell Systems Technical Journal,
vol. 46, pp. 17751796, December 1966.”
“R.W. Chang - US Patent 3,488,445, 1970 ”
A first analysis on this parallel system was done in 1967, at that time only analog design
was proposed.
“B. R. Saltzberg, Performance of an Efficient Parallel Data Transmission System, IEEE Trans. On Communications, vol. 15, no. 6,
pp. 805811, Dec. 1967.”

The employment of the discrete Fourier transform (DFT) to replace the banks of
sinusoidal generators and the demodulators was suggested by Weinstein and Ebert in
1971, which made the implementation OFDM cost effective.
“S. B. Weinstein and P. M. Ebert, Data Transmission by Frequency-Division Multiplexing using the Discrete Fourier Transform,
IEEE Trans. on Communications, vol. 19, no. 5, pp. 628634, Oct. 1971.”

Furthermore, the complexity could drastically be reduced by the application of reduced

computational complexity algorithms such as the Winograd Fourier Transform (WFT) or
the fast Fourier transform (WFT).
“A. Peled, A. Ruiz, Frequency domain data transmission using reduced computational complexity algorithms,
Proc. IEEE Int. Conf. Acoust., Speech, Signal Processing, pages: 964-967, Denver, CO, 1980.”
Since then, the processing power of modern digital signal processors has
increased allowing OFDM to find its way into commercial use

CONFIDENTIAL 4
Andrea Ancora, Matthew Baker
 Introduction: OFDM in telecommunication systems
OFDM has developed into a popular scheme for digital communication, whether wireless or
over copper wires, used in applications such as digital television and audio broadcasting,
wireless networking and broadband internet access.
Cable
ADSL and VDSL broadband access via Plain Old Telephone Service (POTS) copper wiring.
Power Line Communication (PLC) for Broadband over Power Lines (BPL) .
Multimedia over Coax Alliance (MoCA) home networking between televisions, set top boxes and other
entertainment devices.
Wireless
The Wireless LAN (WLAN) radio interfaces IEEE 802.11a, g, n and HIPERLAN/2.
The digital radio systems DAB/EUREKA 147, DAB+, Digital Radio Mondiale, HD Radio, T-DMB and
ISDB-TSB.
The terrestrial digital TV system DVB-T.
The terrestrial mobile TV DVB-H, T-DMB, ISDB-T and MediaFLO downlink.
The cellular communication systems Flash-OFDM and the 3GPP Long Term Evolution (LTE)
downlink.
The wireless MAN/ Fixed broadband wireless access (BWA) standard IEEE 802.16 (or WiMAX).
The Mobile Broadband Wireless Access (MBWA) standards IEEE 802.20, IEEE 802.16e (Mobile
WiMAX) and WiBro.
The wireless Personal Area Network (PAN) Ultra wideband (UWB) IEEE 802.15.3a implementation
suggested by WiMedia Alliance.

CONFIDENTIAL 5
Andrea Ancora, Matthew Baker
Introduction:
OFDM History
Principles of Wireless Communications

CONFIDENTIAL 6
Andrea Ancora, Matthew Baker
 Introduction: Wireless communications principles (1)
Any wireless telecommunication system can be modeled as an equivalent base-band
cascade of:
- Low-pass transmitted signal generator, characterized by a transmission bandwidth W;
- Time-varying low-pass channel filter, characterized by
- Coherence bandwidth B (frequency selectivity)
- Doppler spread fd (time variation);
- Additive interference source, typically modeled as white (un-limited and frequency flat) or colored
(band limited and spectrally shaped) noise.
- Adaptive receive filter, whose filter coefficients (or, equivalently, spectral shape) are varied
according to some received signal criterion to reconstruct at best the transmitted signal, e.g.:
- Match transmitted signal spectrum (MF)
- Undo channel response (ZF)
- Minimize received signal mean squared error (MMSE)
- Minimize symbol error rate (ML)
- Minimize bit error rate (MAP) Interference

Adaptive
+ filter
Transmitter Channel Receiver

CONFIDENTIAL 7
Andrea Ancora, Matthew Baker
 Introduction: Wireless communications principles (2)

The choice of the transmission scheme in conjunction with the channel characteristics
mainly determines:
– The maximum achievable transmission rate
– The receiver complexity
In single Tx & Rx systems, the maximum achievable transmission rate (i.e. in absence of
coding, channel and interference impairment, control overhead) is given by
- R=W·M b/s
- where W is the transmitted symbol bandwidth and M is the modulation order (bits per symbol)
In the presence of noise, this rate is practically unfeasible and is instead bounded by the
well-known Shannon limit
R=W·log2(1+SINR) b/s
- where SINR is the signal to interference+noise power ratio

CONFIDENTIAL 8
Andrea Ancora, Matthew Baker
 Introduction: Narrow-band vs. broad-band transmissions (1)

At receiver side, the received signal is the result of the convolution operation between the
time domain transmitted signal waveform and channel impulse response, i.e. low-pass
filtering, plus a noise term

Frequency selectivity occurs whenever the transmitted signal x(t) occupies an interval
bandwidth [-W/2,W/2] greater then the coherence bandwidth B of the channel c(t) (defined
as the inverse of the channel delay spread Td)
Narrow-band (e.g. GSM) transmissions are those where the symbol duration (defined as
T=1/W) is greater than the channel delay spread
T>Td or, equivalently, W<B
Broad-band (e.g. HSDPA/LTE) transmissions are instead those where the symbol
duration is small compared to the channel delay spread
T<Td or, equivalently, W>B

CONFIDENTIAL 9
Andrea Ancora, Matthew Baker
 Introduction: Narrow-band vs. broad-band transmissions (2)

Signal (Symbol period)

T Td<T (Narrow band)

convolution Inter-symbol interference

Channel

Td Td>T (Broad-band)
(Delay spread)

In general, in case of frequency selectivity, the received signal requires an equalization

procedure (filtering) to eliminate the effect of inter-symbol interference.
The complexity (i.e. cost) is proportional to the square of the channel memory and could
be prohibitively high.
C= k·(Td/T)2 or, equivalently, C= k·(W·Td)2

CONFIDENTIAL 10
Andrea Ancora, Matthew Baker
 Introduction: Narrow-band vs. broad-band transmissions (3)
Exploiting the well known time-frequency duality (and then the equivalence of time-domain
convolution to spectral multiplication), the frequency selectivity effect can be seen as the
frequency components of x(t) with separation exceeding B subject to different frequency
channel gains.
The receiver aims at compensating independently each channel gain, each compensation
requiring consequent complexity. Equalization necessary
Broad-band Non-uniform channel gain to compensate independently
Tx signal each channel gain
spectrum
P P P

x
f f f
Channel
P
W/2 W/2 W/2
spectrum

B
Channel gain constant across
f bandwidth of transmitted signal
Narrow-band P P
Tx signal
spectrum
Therefore no equalization necessary
x (only one channel phase
f f compensation operation)
W/2 W/2

CONFIDENTIAL 11
Andrea Ancora, Matthew Baker
OFDM principles:
Serial-to-parallel conversion
Cyclic prefix
Frequency domain orthogonality
Fourier transform
OFDM advantages and disadvantages

CONFIDENTIAL 12
Andrea Ancora, Matthew Baker
 OFDM principles: Serial-to-Parallel Conversion
The Orthogonal Frequency Division Multiplexing consists in mixing the advantages of
broad-band transmissions, namely the high data rate proportional to the total system
bandwidth W, and the low required receiver complexity inherent to narrow-band
transmissions.
As seen before, in case of broad-band transmissions, the channel delay extends beyond
one symbol period.
To avoid this, the high rate data symbols can be serially-to-parallel converted and
modulated around N different carrier frequencies.
The high rate symbols are then converted into N parallel longer duration symbols.
By varying N, the modulated symbol duration can be made sufficiently low compared to
the channel delay spread and N equivalent narrow-band transmissions are obtained.
exp(-j2π·t·f1)

High symbol rate

x
Low symbol
S/P rate exp(-j2π·t·fn)
Σ
x

CONFIDENTIAL 13
Andrea Ancora, Matthew Baker
OFDM principles
Serial-to-parallel conversion
Cyclic prefix
Frequency domain orthogonality
Fourier transform
OFDM advantages and disadvantages

CONFIDENTIAL 14
Andrea Ancora, Matthew Baker
 OFDM principles: Cyclic prefix
Nevertheless, in the time domain, the modulated symbol so obtained would still suffer
from inter-symbol interference due to the memory of the channel.

symbol inter-symbol interference

duration
By inserting a guard-interval between the modulated symbols, the inter-symbol
interference can be avoided and the memory of the channel emptied.
In OFDM, the guard interval is obtained via a cyclic prefix that consists of inserting a
replica of last G samples at the beginning of each symbol.
While implying some wastage of spectrum, it has the positive effect of reducing the
sensitivity to time synchronization.

CONFIDENTIAL 15
Andrea Ancora, Matthew Baker
OFDM principles
Serial-to-parallel conversion
Cyclic prefix
Frequency domain orthogonality
Fourier transform
OFDM advantages and disadvantages

CONFIDENTIAL 16
Andrea Ancora, Matthew Baker
 OFDM principles: Frequency-domain orthogonality (1)

Equivalently, in the frequency domain, modulating each incoming data symbol around N
equispaced carriers might generate inter-carrier interference due to spectrum leakage.
Again, a guard band could be used to put sufficiently apart the N narrow-band signals.
Similarly to the cyclic-prefix, this would imply spectrum wastage.

Guard band

frequency frequency

CONFIDENTIAL 17
Andrea Ancora, Matthew Baker
 OFDM principles: Frequency-domain orthogonality (2)
Instead, in Orthogonal Frequency Division Multiplexing the inter-carrier interference is
avoided more efficiently by exploiting the DFT operation.
This is equivalent to modulating N symbols using overlapping orthogonal waveforms
(Fourier basis) without cross-talk between the N sub-channels.

Hence, no guard bands are required and the spectrum wastage is limited to the use of the
cyclic-prefix guard interval in the time domain.

CONFIDENTIAL 18
Andrea Ancora, Matthew Baker
OFDM principles
Serial-to-parallel conversion
Cyclic prefix
Frequency domain orthogonality
Fourier transform
OFDM advantages and disadvantages

CONFIDENTIAL 19
Andrea Ancora, Matthew Baker
 OFDM principles: Fourier transform (1)
Further complexity reduction is achieved by choosing N to be radix-2 and benefiting from
the efficiency of the fast Fourier transform (FFT):
– complexity is proportional to only N·log2(N).

At transmitter side, the IFFT accomplishes the task of modulating around multiples of the
sub-carrier frequency ∆f and multiplexing each independent channel at the same time.

Cyclic Prefix

CONFIDENTIAL 20
Andrea Ancora, Matthew Baker
 OFDM principles: Fourier transform (2)
At receiver side, the FFT demodulates around multiples of the sub-carrier frequency
∆f and demultiplexes each independent channel at the same time.
The main advantage of OFDM transmissions is then to turn the convolutive channel
into a multiplicative one in order to simplify the equalization task.

Cyclic Prefix removed

CONFIDENTIAL 21
Andrea Ancora, Matthew Baker
OFDM principles
Serial-to-parallel conversion
Cyclic prefix
Frequency domain orthogonality
Fourier transform
OFDM advantages and disadvantages

CONFIDENTIAL 22
Andrea Ancora, Matthew Baker
 OFDM principles: Advantages and disadvantages (1)
Advantages
Low complexity equalization, O(N·log2N), compared to CDMA case, O(N2), with same
performance.
Transmitter and receiver architecture easily scale with system bandwidth, i.e. by increase of
FFT order.
Robust against narrow-band co-channel interference, i.e. suppressing only some sub-channels.
Robust against inter-symbol interference (ISI) and channel selectivity due by multi-path
propagation.
High spectral efficiency, as almost the whole available frequency band can be utilized .
Efficient implementation using FFT, i.e. numerically stable and supporting digital processing.
Low sensitivity to time synchronization errors.
Sub-carrier (de)modulation and sub-channel (de)multiplexing happening at same time.
Summary of disadvantages
Sensitive to Doppler.
Sensitive to frequency synchronization problems.
High peak-to-average-power ratio (PAPR), requiring high dynamic linear transmitter circuitry
suffering from poor power efficiency.

CONFIDENTIAL 23
Andrea Ancora, Matthew Baker
 OFDM principles: Advantages and disadvantages (2)
OFDM requires very accurate frequency synchronization between the receiver and the
transmitter; with frequency deviation the sub-carriers will no longer be orthogonal, causing
inter-carrier interference (ICI), i.e. cross-talk between the sub-carriers.
Frequency offsets are typically caused by mismatched transmitter and receiver oscillators,
or by Doppler shift due to movement in LOS case.
Whilst Doppler shift alone may be compensated for by the receiver, the situation is
worsened when combined with multi-path channels, as reflections will appear at various
frequency offsets, engendering a Doppler spread which is much harder to correct.
This effect typically worsens as speed increases, and is an important factor limiting the
use of OFDM in high-speed vehicles. Several techniques for ICI suppression are
suggested, but they may increase the receiver complexity.

CONFIDENTIAL 24
Andrea Ancora, Matthew Baker
Dimensioning of OFDM system parameters:
Principles
Example of LTE

CONFIDENTIAL 25
Andrea Ancora, Matthew Baker
 Dimensioning of OFDM system parameters: Principles
An OFDM system needs proper dimensioning with respect to deployment scenario
Given a system bandwidth W, two main parameters must be defined:
The sub-carrier spacing, equivalent to define the number N of parallel sub-channels, such
that:
∆f=W/N < Bmin
(i.e. the sub-carrier spacing is lower than the minimum channel coherence bandwidth)
– Or equivalently
T’=N·T > Td,max
(i.e. the OFDM symbol duration is greater than the maximum channel delay spread)
– The OFDM symbol duration must be chosen such that
T’=N·T << 1/fd,max
(i.e. the channel must be constant over one OFDM symbol duration)
– N must be radix-2 for efficient implementation
The cyclic prefix length G, must be chosen such that:
G> Td,max
(i.e. the guard interval is greater than the worst-case channel length)
– Or equivalently
G·Bmin > 1
– G must be as small as possible to avoid spectral efficiency wastage.
Hence, the channel characteristics, delay spread and maximum Doppler frequency, are
uniquely considered in the system dimensioning.
In case of wireless channel, worst-case assumptions must be taken.

CONFIDENTIAL 26
Andrea Ancora, Matthew Baker
Dimensioning of OFDM system parameters
Principles
Example of LTE

CONFIDENTIAL 27
Andrea Ancora, Matthew Baker
Dimensioning of OFDM system parameters: Example of LTE (1)
In the LTE downlink case, propagation scenarios cover:
– Indoor, urban and sub-urban propagation conditions determining different
delay spread
– Highly mobile user equipments within a cell that can have relative speed of
350 km/h (and more)
– As MBMS makes use of Macro diversity (combining from multiple cells),
a long cyclic prefix is required to cover the full range of delay spread.
As a result, 3 OFDM modes are envisaged for efficient use of system
resources depending on the deployment.
– They are as follows:
• 15kHz sub-carrier spacing (less sensitive to Doppler, so better suited for high-
mobility applications). Two alternative cyclic prefix lengths:
– Normal CP (5us): to suit small cell deployments such as indoor and urban.
– Extended CP (17s): allowing for larger cell deployments such as in suburban / rural.

• 7.5kHz sub-carrier spacing with 33us cyclic prefix length:

– intended for Multimedia broadcasting, allowing for longer symbol duration and CP:
better for large cell and for multi-cell combining although more prone to Doppler.

CONFIDENTIAL 28
Andrea Ancora, Matthew Baker
Dimensioning of OFDM system parameters: Example of LTE (3)

CONFIDENTIAL 29
Andrea Ancora, Matthew Baker
Dimensioning of OFDM system parameters: Example of LTE (2)

System band-width is imposed by regulatory organizations and cover a wide variety of

bandwidths ranging in [1.25 MHz .. 20 MHz] to provide operators deployment flexibility.
Backward compatibility against R99 and HSPA favours sampling frequencies being a
multiple of 3.84 MHz (the UMTS chip rate).
– For efficient implementation of multi-mode terminals

The maximum sampling frequency is indicated to be fs=30.72 MHz and the FFT order is
N=2048 for 20 MHz bandwidth.
Given the sampling frequency and the system bandwidth, the ratio between M used sub-
carriers and the N available at FFT output is lower compared to other systems (DVB,
WLAN, etc) slightly reducing FFT efficiency.
The current assumption is M=1200 to give M/N≈0.7 as ratio between used vs. processed
sub-carriers.
Lower sampling frequencies (and proportionally lower FFT orders) are possible for
narrower bandwidth deployments, to reduce complexity:
e.g.: W=5 MHz
N=512/M=300/fs=7.68 MHz and so on..

CONFIDENTIAL 30
Andrea Ancora, Matthew Baker
Dimensioning of OFDM system parameters: Example of LTE (4)

Channel Bandwidth [MHz]

Transmission Bandwidth Configuration [RB]

Transmission
Bandwidth [RB]

Channel edge
Channel edge

Resource block

Active Resource Blocks DC carrier (downlink only)

CONFIDENTIAL 31
Andrea Ancora, Matthew Baker
OFDMA:
Orthogonal Frequency-Division Multiple Access
Combination of OFDMA and TDMA

CONFIDENTIAL 32
Andrea Ancora, Matthew Baker
 OFDMA
Orthogonal Frequency-Division Multiple Access (OFDMA) is a multi-user version of
OFDM digital modulation scheme. Multiple Access is achieved in OFDMA by assigning
groups of sub-carriers to individual users.
Based on feedback information about the channel conditions from each user, adaptive
user-to-RB assignment can be performed.
If the assignment is done sufficiently fast, this further improves the OFDM robustness to
channel selectivity/fast fading of each user and further increases spectral efficiency.
Moreover, OFDMA allows the support of differentiated Quality of Service (QoS), i.e. to
control the data rate and error probability individually for each user.

LTE spectrum allocation

granularity is 12 contiguous
sub-carriers: Resource Blocks.

CONFIDENTIAL 33
Andrea Ancora, Matthew Baker
 Combination of OFDMA and TDMA

OFDMA can also be used in combination with Time Domain Multiple Access (TDMA),
where the resources are partitioned in the time-frequency plane and chunks are assigned
along the OFDM symbol index as well as OFDM sub-carrier index.
As in LTE:

CONFIDENTIAL 34
Andrea Ancora, Matthew Baker
OFDM implementation issues
PAPR and non-linearities
Frequency offset and Doppler sensitivity
Insufficient cyclic prefix
OFDM pilot structure and channel estimation

CONFIDENTIAL 35
Andrea Ancora, Matthew Baker
 OFDM related issues: PAPR and non linearities (1)
OFDM can be seen as a linear operation performed over a large block of frequency-
domain QAM modulated complex symbols with (almost) constant modulus property.
As a result, due to the central limit theorem, the time-domain OFDM symbol can be very
well approximated as a Gaussian waveform and therefore losing the interesting constant
modulus property.
The amplitude of the OFDM modulated signal can then have very large values
– These amplitude variations are a big disadvantage as they reduce the power efficiency (and
increase cost) of the Power Amplifier (PA) in the RF transmitter as the linear operational
amplification region is much increased.
LTE does not specify any PAPR reduction technique and leaves to eNodeB
manufacturers the burden of handling the increased required costs.

(I)FFT

Limited signal dynamic High signal dynamic

(constant modulus signal) (Gaussian signal)
PAPR = 1 PAPR >> 1

CONFIDENTIAL 36
Andrea Ancora, Matthew Baker
 OFDM related issues: PAPR and non linearities (2)
OFDM suffers from distortions of time-domain signal as errors spread over all sub-carriers.
Hence, non-idealities of real RF transceiver, as limited dynamic amplifiers (i.e. clipping
effect) have a direct impact on the quality of received signal and result in an additional
source of impairment.
Inherent OFDM property such as high PAPR largely contributes to this phenomenon.
In case of OFDM, non-linearities can easily be accounted knowing that any non-linear
device present along the transmission chain is equivalent to the cascade of a complex gain
factor and an additional white noise (distortion) source applied to the received signal.
This model is used in setting the performance for LTE and is commonly referred as EVM
source.
As a result, despite SIR growing as high as possible, OFDM signal would be affected by an
irreducible error floor limiting the overall system spectral efficiency.

CONFIDENTIAL 37
Andrea Ancora, Matthew Baker
OFDM implementation issues
PAPR and non-linearities
Frequency offset and time Dopplersensitivity
Insufficient cyclic prefix
OFDM pilot structure and channel estimation

CONFIDENTIAL 38
Andrea Ancora, Matthew Baker
OFDM related issues: Frequency offset and Doppler sensitivity
The orthogonality principle of OFDM relies on the fact that the channel stays constant over
the transmission of each OFDM symbol corresponding to the DFT block.
In wireless environments the multi-path channel is time varying because of the user’s
mobility (Doppler effect). Rapid channel variations over a symbol period introduce a
frequency error.
A frequency error may be also the result of a mismatch between the transmitter and
receiver local oscillators.
The frequency error destroys sub-carriers orthogonality and results in Inter-Carrier
Interference (ICI).

CONFIDENTIAL 39
Andrea Ancora, Matthew Baker
OFDM implementation issues
PAPR and non-linearities
Frequency offset and time Dopplersensitivity
Insufficient cyclic prefix
OFDM pilot structure and channel estimation

CONFIDENTIAL 40
Andrea Ancora, Matthew Baker
 OFDM implementation issues: Insufficient cyclic prefix
OFDM systems are designed to support a cyclic prefix length longer than the channel
impulse response in order to benefit from cyclic periodicity of Fourier transform and result
in an orthogonal multi-carrier system allowing for extremely low complexity equalization.
The condition of a sufficient guard interval is therefore strictly related to the orthogonality
property of OFDM.
In the unfortunate situation when the channel might be longer than the system designed
cyclic prefix length, the orthogonality is destroyed resulting in the introduction of Inter-
Carrier Interference (ICI) and Inter Symbol Interference (ISI) from previous OFDM symbol

CONFIDENTIAL 41
Andrea Ancora, Matthew Baker
OFDM implementation issues
PAPR and non-linearities
Frequency offset and time Dopplersensitivity
Insufficient cyclic prefix
OFDM pilot structure and channel estimation

CONFIDENTIAL 42
Andrea Ancora, Matthew Baker
OFDM implementation issues:
LTE OFDM pilot structure and channel estimation (1)

Transmission schemes can in general be classified as:

– Non-coherent, where useful signal is modulated using differential schemes (low rate) and requiring
energy detection based receiver (e.g. UWB).
– Coherent, where transmitted signal is modulated using QAM schemes (high rate) and requiring
receivers able to recover both amplitude and phase (e.g. LTE downlink).
Coherent systems require accurate estimation of the channel
Channel estimation can be performed in different ways: by using a parametric model of the
channel and exploiting the a-priori knowledge of the model, by exploiting the correlation
properties of the channel, by using blind estimation, or by exploiting some known signals.
To facilitate and limit the complexity of the channel estimation process, LTE uses known
reference signals in the transmitted OFDM symbols allowing for pilot based channel
estimation (at the expense of some overhead).

data Equalized data

x
Demux

CP FFT
removal
pilot Channel
estimation

CONFIDENTIAL 43
Andrea Ancora, Matthew Baker
OFDM implementation issues:
LTE OFDM pilot structure and channel estimation (2)
LTE reference signal (pilot symbols) are available in the time-frequency plane dispersed on a
regular grid.
The RS grid is diamond shaped to allow exploiting diversity and for efficient interpolation in
the frequency-time 2-dimensional plane.
Within one cell, the RS are designed to support multi-antenna transmission: RS are
interleaved such that they do not interfere and each antenna avoid transmission on those
resource elements where RS of the other antennas are present.
Among neighboring cells, the RS resource elements pattern is (mostly) unique to allow for
interference randomization. There are 6 available patterns obtained by frequency shifts
determined by the Cell ID.

Antenna port 0 Antenna port 1 Antenna port 2 Antenna port 3

frequen

R0 R0 R1 R1 R2 R3
cy

R0 R0 R1 R1 R2 R3

R0 R0 R1 R1 R2 R3

R0 R0 R1 R1 R2 R3

time

CONFIDENTIAL 44
Andrea Ancora, Matthew Baker
 OFDM implementation issues:
LTE OFDM pilot structure and channel estimation (3)
The OFDM channel must be estimated over the time-frequency grid and in spatial direction
(NTx x NRx spatial channels):
– The channel grid is estimated performing (2D) interpolation and filtering.
– Each grid for each spatial channel must be estimated separately: no interpolation is possible.

The interpolation and filtering technique is critical for channel estimation performance (and
complexity). It is based on the exploitation the coherence bandwidth and the time-varying
characteristics.
In the practical case, the 2-D problem is split into 2 1-D separated interpolation/filtering in
frequency and time directions to lower memory complexity as data to be decoded would need
to be buffered while the channel estimation is being derived.
Channel

frquency
frquency

f-interpolation
channel

at RS positions

2D
interpolation t-interpolation

time/frequency
Interpolated time time
channel

CONFIDENTIAL 45
Andrea Ancora, Matthew Baker
Summary and conclusions

CONFIDENTIAL 46
Andrea Ancora, Matthew Baker
 Summary and conclusions

OFDM is a mature technology.

It achieves the high transmission rate of broad-band but with the low receiver
complexity of narrow-band transmission.
Makes use of cyclic-prefix to null Inter-Symbol Interference, enabling block-wise
processing.
Orthogonal sub-carriers avoid guard-band wastage.
Benefits from efficient Fourier transform implementation – FFT.
System design allows to trade off between Doppler and delay spread depending on
the deployment scenarios.
Widely deployed:
– especially suited for broadcast or downlink because of low receiver complexity while
requiring high transmitter complexity (expensive PA).
Straightforward multi-user extension into OFDMA.

CONFIDENTIAL 47
Andrea Ancora, Matthew Baker
 CONFIDENTIAL 48
Andrea Ancora, Matthew Baker