-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
7
1. Downlink Physical Layer
1.1 OFDMA Principles OFDMA (Orthogonal Frequency division
Multiple Access) is a multicarrier scheme. Multicarrier schemes
subdivide bandwidth into parallel subchannels, ideally each
non-frequency-selective (spectrally-flat gain), overlapping but
orthogonal. This avoids need of guard-bands, makeing OFDM highly
spectrally efficient, as subchannels can be perfectly separated at
the receiver. This makes receiver less complex, attractive for
high-rate mobiles. Robustness has to be built up against
time-variant channels by employing channel coding. LTE downlink
combines OFDM with channel coding and Hybrid Automatic Repeat
reQuest (HARQ). OFDM is ideal for broadcast/DL applications for low
receiver complexity. OFDM has efficient implementation by means of
the FFT. It uses Cyclic Prefix to avoid ISI, enabling block-wise
processing. Orthogonal subcarriers avoid spectrum wastage in
intersubcarrier guard-bands. Parameters flexibility allows balance
the tolerance of Doppler and delay spread.
Key OFDMA points (a) Orthogonal subcarriers with very small
inter-subcarrier guard-bands. (b) It makes use of a CP to avoid
ISI, enabling block-wise processing. (c) Efficient implementation
by means of the FFT. (d) Achieves high transmission rates of
broadband transmission, with low receiver complexity. (e) Balanced
tolerance of Doppler and delay spread depending on the deployment
scenario. (f) It can be extended to a multiple-access scheme,
OFDMA, in a straightforward manner. (g) Suited for broadcast or
downlink applications because of low receiver complexity while
requiring a high transmitter complexity (expensive PA).
First OFDM patent filed at Bell Labs in 1966, initially only as
analog. In 1971, Discrete Fourier Transform (DFT) was proposed.
Later in 1980, application of the Winograd Fourier Transform (WFT)
or Fast Fourier Transform (FFT) was employed. OFDM then became
modulation of choice for ADSL and wireless systems. OFDM tended to
focus broadcast systems such as - Digital Video Broadcasting (DVB)
and Digital Audio Broadcasting (DAB), and WLANs. Main thing to
control in OFDM was PAPR and thats why in low power WLAN it was
good. First cellular mobile based on OFDM was proposed in 1985 by
IEEE to LTE downlink. Other benefits of OFDM was to operate in
different bandwidth according to spectrum availability.
1.1.1 OFDM - Orthogonal Multiplexing Principle Challenge is
always in having a symbol period Ts < channel delay spread Td.
This generates Intersymbol Interference (ISI), needing complex
equalization procedure. Equalization complexity usually is in
proportion to square of (channel impulse response length). Data
symbols are first serial-to-parallel converted for modulation on M
parallel subcarriers.This increases symbol duration by a factor of
approx M, > channel delay spread.
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
8
Fig 2.1.1.1 OFDM Signal Processing This operation makes
time-varying channel impulse response substantially constant during
each modulated OFDM symbol. Resulting long symbol duration is
virtually unaffected by ISI compared to the short symbol duration.
A Serial to Parallel (S/P) converter collects serial data symbols
into a data block Sk = [Sk [0] , Sk [1] , . . . , Sk [M 1]]T of
dimension M, where the subscript k is the index of an OFDM symbol
(spanning the M sub-carriers). The M parallel data streams are
first independently modulated resulting in the complex vector Xk =
[Xk [0] , Xk [1] , . . . , Xk [M 1]]T . In principle it is possible
to use different modulations (e.g. QPSK or 16QAM) on each
sub-carrier, the channel gain may differ between sub-carriers, and
thus some sub-carriers can carry higher data-rates than others. The
vector of data symbols Xk then passes through an Inverse FFT (IFFT)
resulting in a set of N complex time-domain samples xk = [xk[0], .
. . , xk[N 1]]T . In a practical OFDM system, the number of
processed subcarriers is greater than the number of modulated
sub-carriers (i.e. N M), with the unmodulated sub-carriers being
padded with zeros.
Fig 2.1.1.2 OFDMA tramsmission and reception
A guard period is created at the beginning of each OFDM symbol,
to eliminate the remaining impact of ISI. A Cyclic Prefix (CP) is
added at the beginning of each symbol xk. The CP is generated by
duplicating the last G samples of the IFFT output and appending
them at the beginning of xk. This yields the time domain OFDM
symbol [xk[N G], . . . , xk[N 1], xk[0], . . . , xk[N 1]]T . CP
length G should be longer than the longest channel impulse response
to be supported. The CP converts the linear (i.e. aperiodic)
convolution of the channel into a circular (i.e. periodic) one
which is suitable for DFT processing. The IFFT output is then
Parallel-to-Serial (P/S) converted for transmission through
frequency-selective channel. Here is an example of OFDM LTE signal.
At the receiver, the reverse operations are performed to demodulate
the OFDM signal, CP are removed and ISI-free block of samples is
passed to the DFT. If number of subcarriers N is designed to be a
power of 2, a highly efficient FFT implementation may be used to
transform the signal back to the frequency domain. Among the N
parallel streams output from the FFT, the modulated subset of M
subcarriers are selected and further processed by the receiver. Let
x(t) be the signal symbol transmitted at time instant t . The
received signal in a multipath environment is then given by r(t) =
x(t) * h(t) + z(t), where h(t) is the continuous-time impulse
response of the channel, represents the convolution operation and
z(t) is the additive noise. Assuming that x(t) is band-limited to
[1/2Ts ,1/2Ts], the continuous-time signal x(t) can be sampled at
sampling rate Ts such that the Nyquist criterion is satisfied. Due
to multipath, several replicas of the transmitted signals arrive at
the receiver at different delays. The received discrete-time OFDM
symbol k including CP, under the assumption that the channel
impulse response has a length smaller than or equal to G, Receiver
has to process equalization to recover xk[n] signals. CP of OFDM
changes the linear convolution into a circular one. The circular
convolution is very efficiently transformed by an FFT into a
multiplicative operation in frequency domain. Hence, the
transmitted signal over a frequency-selective (multipath) channel
is converted into a transmission over N parallel flat-fading
channels in
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
9
the frequency domain: Rk[m] = Xk[m] H[m] + Zk[m]. As a result
the equalization is much simpler than for single-carrier systems
and consists of just one complex multiplication per subcarrier.
1.1.2 Peak-to-Average Power Ratio and Sensitivity Major
drawbacks of OFDM is that it has a high Peak-to-Average Power Ratio
(PAPR). The amplitude variations of OFDM signal can be very high,
however PAs of RF transmitters are linear only within a limited
dynamic range. Hence, OFDM signal is likely to suffer from
non-linear distortion caused by clipping, giving out-of-band
spurious emissions and in-band corruption of the signal. To avoid
such distortion, PAs should have large power back-offs, leading to
inefficient amplification. Let x[n] be the signal after IFFT. PAPR
of an OFDM symbol is defined as the square of the peak
amplitude divided by the mean power, i.e. PAPR = Max,n{|x[n]|2}
/ E{|x[n]|
2}
It is observed that a high PAPR does not occur very often.
However, when it does occur, degradation due to PA non-linearities
may be expected. PAPR Reduction Techniques Many techniques are
studied for reducing the PAPR, but not specified for downlink. An
overview of possibilities is provided below.
1. Clipping and filtering . Signal may be clipped, but it causes
spectral leakage into adjacent channels, resulting in reduced
spectral efficiency, in-band noise, degrading BER. To avoid this
problem, oversample the original signal by padding with zeros and
processing it using a longer IFFT. Oversampled signal is clipped
and then filtered to reduce the out-of-band radiation. This may be
is used in LTE.
2. Selected mapping. Whichever phase vector gives Least PAPR,
that is used. To recover phase information, separate control
signalling is used to tell which phase vector was used. It is not
used.
3. Coding techniques. Use code words with lowest PAPR.
Complementary codes have good properties to combine both PAPR and
forward error correction. It is not used.
Sensitivity to Carrier Frequency Offset and Time-Varying
Channels OFDM orthogonality relies that transmitter and receiver
operate with exactly same frequency reference, else perfect
orthogonality of subcarriers is lost, causing subcarrier leakage
(Inter-Carrier Interference (ICI). UE local oscillator frequency
drifts are usually greater than in the eNodeB and are typically due
to temperature and voltage variation and phase noise. This
difference between the reference frequencies is referred as Carrier
Frequency Offset (CFO). The CFO can be larger than subcarrier
spacing - divided into integer part and fractional part. Frequency
error fo = (T+e)df. Where, df is subcarrier spacing,, T is an
integer and 0.5
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
10
where d is the timing offset in samples corresponding to a
duration equal to To.
This phase shift can be recovered as part of the channel
estimation operation, with cyclic prefix but not zero-padding.In
the general case of a channel with delay spread, for a given CP
length the maximum tolerated timing offset without degrading the
OFDM reception is reduced by an amount equal to the length of the
channel impulse response: To TCP Td. For greater timing errors, ISI
and ICI occur. Timing synchronization becomes more critical in
long-delay spread channels. Initial timing is achieved by the
cell-search and synchronization procedures. Thereafter, for
continuous tracking of timing-offset, either CP correlation or
Reference Signals (RSs) is used. If an OFDM system, CP is
sufficiently designed of lengthG samples such that Channel impulse
Response L
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
11
further overhead, sub-carrier spacing is kept at 7.5kHz and an
extended CP of approx 33 s is used.
Fig 2.1.5.1 -FDD Frame Structure
Fig 2.1.5.2 TDD Frame Structure
Note that, with normal CP, the CP for the first symbol in each
0.5 ms slot is slightly longer than the next six symbols, to
accommodate an integer (7) number of symbols in each slot, with
assumed FFT block-lengths of 2048. For 20 MHz, FFT order of 2048 is
assumed for efficient implementation. However, in practice the
implementer is free to use other Discrete Fourier Transform sizes.
These parameterizations are designed to be compatible with a
sampling frequency of 30.72 MHz, which is 8*3.84Mhz(UMTS sampling
rate), for backward compatibility. Thus, the basic unit of time in
LTE, is defined as Ts = 1/30.72 s. Lower sampling frequencies (and
proportionally lower FFT orders) are always possible to reduce RF
and baseband processing complexity for narrower BW: Example, for 5
MHz, FFT order and sampling frequency could be 512 and fs = 7.68
MHz respectively, while only 300 subcarriers are actually modulated
with data. For simple implementation, direct current (d.c.)
subcarrier is left unused, to avoid d.c. offset errors.
1.1.6 Transmission Resource Structure LTE downlink, consist of
user-plane and control-plane data from higher protocol stack layers
multiplexed with physical layer signalling. A DL resources possess
dimensions of time(slot),
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
12
frequency(multiple of 180khz) and space(layer). Layer is defined
by multiple antenna transmission and reception. The largest unit of
time is the 10 ms radio frame, further subdivided into ten 1 ms
subframes, each of which is split into two 0.5 ms slots. Each slot
has seven OFDM symbols in normal CP (six if extended CP). In
frequency domain, resources are grouped in units of 12 subcarriers
(15*12kHz=180 kHz), such that one unit of 12 subcarriers for a
duration of one slot is termed a Resource Block (RB).
Fig 2.1.6.1 FDD Downlink Frame sample
Smallest unit of resource is the Resource Element (RE) - one
subcarrier for a duration of one OFDM symbol. A RB comprised of 84
REs in normal CP (72 RE in extended CP). Within certain RBs, some
REs are reserved for synchronization signals (PSS/SSS), reference
signals (RS), control signalling and critical broadcast system
information (CFICH,PHICH,PDCCH). Remaining REs are used for data
transmission(PDSCH), and are usually allocated in pairs (in time
domain) of RBs.
Fig 2.1.6.2 TDD Downlink Frame sample
Two types of frame structure are defined:
1. Frame Structure Type 1(Frequency Division Duplex,FDD) assumes
all subframes are available for DL, in paired radio spectrum, or
standalone downlink carrier.
2. Frame Structure Type 2(Time Division Duplexing , TDD) in
unpaired spectrum, basic structure of RBs and REs remains same, but
only a subset of subframes are available for
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
13
downlink; remaining subframes are used for uplink or for special
subframes which allow switching between DL & UL. In the centre
of the special subframes a guard period is provided which allows UL
timing to be advanced.
Signal Structure Physical layer translate data into reliable
signal for transmission between eNodeB and UE. Each block of data
is first protected against transmission errors, first with a Cyclic
Redundancy Check (CRC), and then with coding; The initial
scrambling stage is applied to all DL channels and helps
interference rejection. Scrambling sequence uses order- 31 Gold
code, which are not cyclic shifts of each other. Scrambling
sequence generator is re-initialized every subframe (except PBCH),
based on cell-id, subframe number (within a radio frame), UE
identity and codeword id. Scrambling sequence generator is similar
to pseudo-random sequence used for Reference Signals, only
difference is the method of initialization. A fast-forward of 1600
places is applied at initialization to ensure low cross-correlation
between sequences used in adjacent cells. Following scrambling,
data bits from each channel are mapped to modulation symbols
depending on modulation scheme, then mapped to layers, precoded,
mapped to RE, and finally translated into a complex-valued OFDM
signal by IFFT. To communicate with eNodeB cells, UE must first
identify the DL from one of these cells and synchronize with it.
This is achieved by means of special synchronization signals
embedded into the OFDM structure by cell search and
synchronization. Then UE estimates DL radio channel to perform
demodulation of received DL signal, based on pilot signals
(reference signals) inserted into DL signal. The channel designs
are explained next.
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
14
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
15
1.2 Synchronization and Cell Search Cell Search executes
synchronization for time and frequency parameters, necessary to
demodulate DL and to transmit UL with correct timing and acquires
some critical system parameters. Three major synchronization
requirements:
1. symbol timing determines correct symbol start position and
sets the FFT window position; 2. carrier frequency synchronization
to reduce frequency errors by oscillator & Doppler shift; 3.
sampling clock synchronization.
1.2.1. Synchronization Sequences and Cell Search in LTE Two
relevant cell search procedures exist in LTE:
1. Initial synchronization,
UE detects a cell and decodes all information required to
register. This is required, for example, when UE is switched on, or
it has lost connection to the serving cell.
2. New cell identification,
When UE is already connected and is detecting a new neighbour
cell. UE reports new cell measurements to Serving cell for
handover. Procedure is repeated periodically until either Scell
quality becomes satisfactory again, or UE moves to another
cell.
Fig 2.2.1.1 FDD and TDD Synchronization Signalling
Synchronization procedure detects specially designed Primary
Synchronization Signal (PSS) and Secondary Synchronization Signal
(SSS). This enables time and frequency synchronization, provides
the UE with physical cell identity (PCI) and CP, and informs UE
whether cell uses FDD or TDD. Here is figure explaining relative
location of PSS and SSS in frame structure of FDD and TDD
respectively.
Fig 2.2.1.2 Cell Synchronization Process
In initial synchronization, UE proceeds to decode PBCH for
critical system information (SI). For new cell identification, UE
does not need to decode PBCH; it makes quality-level measurements
(RSRP/RSRQ) and reports to the serving cell.
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
16
The sync signals are transmitted periodically, twice per 10 ms
radio frame.
In FDD cell, PSS is always located in the last symbol of the 0th
and 10
th slots of each frame,
thus enabling UE to acquire the slot boundary timing
independently of the CP. SSS is located in the symbol immediately
preceding PSS, for coherent detection of SSS relative to PSS.
In TDD cell, PSS is located in 2nd
symbol of the 2nd
and 12th slots, while the SSS is located 3
symbols earlier and falls in previous slot. Position of SSS
changes depending on CP length of the cell. At this stage, CP
length is unknown and SSS is blindly detected by checking for SSS
at expected positions.
PSS in a given cell is same in every subframe, SSS may change
& thus UE knows the position of the 10 ms radio frame boundary.
PSS and SSS are transmitted in the central six Resource Blocks
(RBs), irrespective of the system BW (6 to 110 RBs), without
knowing BW. The PSS and SSS are each comprised of a sequence of
length 62 symbols, mapped to the central 62 subcarriers around d.c.
subcarrier which is left unused. Five REs at each extremity of each
sync sequence are not used. Thus a UE can detect the PSS and SSS
with size-64 FFT and a lower sampling rate if all 72 subcarriers
were used. In case of MIMO at eNodeB, PSS and SSS are always
transmitted from same antenna port in a subframe, while between
different subframes they may be transmitted from different antenna
ports for diversity. PSS and SSS sequence indicate one of 504
unique PCI, grouped into 168 groups of three identities. The three
identities in a group are assigned to cells under same eNodeB.
Three PSS sequences are used to indicate the cell identity within
the group, and 168 SSS sequences are used to indicate the identity
of the group. PSS uses ZadoffChu sequences
1.2.2. ZadoffChu Sequences ZadoffChu (ZC) sequences (Generalized
Chirp-Like (GCL) sequences) are non-binary unit-amplitude
sequences, which satisfy a Constant Amplitude Zero Autocorrelation
(CAZAC) property. The ZC sequence of odd-length NZC is given by
aq(n) = exp [j2q (n(n + 1)/2 + ln)/ NZC ]
where q {1, . . . , NZC 1} is the ZC sequence root index, n = 0,
1, . . . , NZC 1, l N is any integer (In LTE l = 0). ZC sequences
have the following important properties. Property 1. A ZC sequence
has constant amplitude which limits PAPR and generates bounded and
time-flat interference to other users, and its NZC-point DFT also
has constant amplitude. Property 2. ZC sequences of any length have
ideal cyclic autocorrelation (correlation with circularly shifted
version of itself is a delta function). ZC periodic autocorrelation
is exactly zero for 0 and it is non-zero for = 0, whereas PN
periodic autocorrelation shows significant peaks, some above 0.1,
at non-zero lags. CAZAC sequence allows multiple orthogonal
sequences to be generated from the same ZC sequence. Indeed, if the
periodic autocorrelation of a ZC sequence provides a single peak at
the zero lag, the periodic correlation of the same sequence against
its cyclic shifted replica provides a peak at lag NCS, where NCS is
the number of samples of the cyclic shift. This creates a
Zero-Correlation Zone (ZCZ) between the two sequences. As a result,
as long as the ZCZ is dimensioned to cope with the largest possible
expected time misalignment between them, the two sequences are
orthogonal for all transmissions within this time misalignment.
Property 3. The absolute value of the cyclic cross-correlation
function between any two ZC sequences is constant and equal to
1/NZC, if |q1 q2| (where q1 and q2 are the sequence indices) is
relatively prime with respect to NZC . Selecting NZC as a prime
number results in NZC 1 Zaddoff-Chu sequences which have the
optimal cyclic cross-correlation between any pair. Cyclic extension
or truncation preserves both the constant amplitude property and
the zero cyclic autocorrelation property for different cyclic
shifts. The DFT of a ZC sequence xu(n) is a weighted
cyclicly-shifted ZC sequence Xw(k) such that w = 1/u mod NZC. This
means that a ZC sequence can be generated directly in the frequency
domain without the need for a DFT operation.
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
17
1.2.3. Primary Synchronization Signal (PSS) Sequences There are
0 to 503 (total 504) PCI are available to be assigned. Every cell
will have a PCI where
PCI=3* NID,1 + NID,2.
The NID,1 is the PCI group ID - , NID,1 can have values 0 to 167
and
NID,2 is the PCI local (may be sector) ID - NID,2 can have
values 0 to 2.
How does the PCI optimization affect my Network? Well, this is
the main parameter, by which the
PSS, SSS and reference signals will be generated. Even your
scrambling code with which every DL
and UL signal will be scrambled, will depend on this. So, every
generated signals uniqueness
depends on this parameter. Lets understand how some of the
signals are generated based on PCI.
If PSS, SSS, RS and other generated signals are not unique, then
my every operation will be affected
and it may reflect as latency in Synchronization detection,
Interference and lower SINR values for
signals, which will end up in low CQI.
1.2.4. PSS Generation The PSS represented by d(u,n) is generated
from a frequency-domain Zadoff-Chu sequence
according to
61,...,32,31
30,...,1,0)(
63
)2)(1(
63
)1(
ne
nend
nnuj
nunj
u
where the Zadoff-Chu root sequence index u is given by following
table.
(2)IDN
Root index u
0 25
1 29
2 34
The mapping of PSS to resource elements depends on the frame
structure, FDD or TDD. The
sequence d(u,n) is mapped to the resource elements according
to
231
61,...,0 ,
RBsc
DLRB
,
NNnk
nnda lk
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
18
For FDD, the PSS is mapped to the last OFDM symbol in slots 0
and 10. For TDD, the PSS is apped
to the third OFDM symbol in subframes 1 and 6. Resource elements
(k,l) in the OFDM symbols used
for transmission of the primary synchronization signal where
66,...63,62,1,...,4,5
231
RBsc
DLRB
n
NNnk
are reserved and not used for transmission of the primary
synchronization signal. N=0,1,61 are used with above formulae.
PSS is constructed from a freq-domain ZC sequence of length 63,
with middle element punctured to avoid transmitting on d.c.
subcarrier. This set of roots for ZC sequences was chosen for its
good periodic autocorrelation and cross-correlation properties.
These sequences have a low-frequency offset sensitivity (maximum
undesired autocorrelation peak /desired correlation peak) at a
certain frequency offset, giving best robustness. Also the ZC
sequences are robust against frequency drifts. Thus, PSS can be
easily detected during the initial synchronization with a frequency
offset up to 7.5 kHz. The selected root combination satisfies
time-domain root-symmetry, sequences 29 and 34 are complex
conjugates of each other and can be detected with a single
correlator. UE must detect PSS without any prior knowledge of the
channel, so noncoherent correlation is required for PSS timing
detection.
1.2.5. Secondary Synchronization Signal (SSS) Sequences SSS
maximum length M-sequences, can be created by cycling through every
possible state of a shift register of length n, resulting in
M-sequence of length 2n 1. Two length-31 BPSK secondary sync codes
(SSC1(even (di)) and SSC2(odd d(i)) are interleaved to construct
SSS sequence in frequency-domain. Two codes are two different
cyclic shifts of a single length-31 M-sequence. Cyclic shift
indices are derived from a function of PCI group. Two codes are
alternated between the first and second SSS in each radio
frame.
5 subframein )(
0 subframein )()12(
5 subframein )(
0 subframein )()2(
)(11
)(0
)(11
)(1
0)(
1
0)(
0
10
01
1
0
nzncns
nzncnsnd
ncns
ncnsnd
mm
mm
m
m
with
30,30
2)1(,2)1(
31mod131
31mod
(1)ID
(1)ID(1)
ID
01
0
NqqqN
qqqNm
mmm
mm
Thus UE determines the 10 ms radio frame timing from a single
observation of a SSS. SSC2 is scrambled by a sequence that depends
on the index of SSC1. Sequence is then scrambled by a code that
depends on the PSS. Scrambling code is mapped to the PCI within the
group corresponding to the target eNodeB. The resource mapping is
done as per the following:
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
19
2 typestructure framefor 11 and 1 slotsin 1
1 typestructure framefor 10 and 0 slotsin 2
231
61,...,0 ,
DLsymb
DLsymb
RBsc
DLRB
,
N
Nl
NNnk
nnda lk
SSS sequences are spectrally flat. PSS, the SSS can be detected
with a frequency offset up to 7.5 kHz. Channel is known based on
the PSS sequence first and then SSS detection is done. However, in
the case of synchronized neighbouring eNodeBs, coherent detector
performance can be degraded. If an interfering eNodeB employs the
same PSS, phase difference between them can have adverse impact on
estimation of the channel coefficients. If BW of the channel is
less than the six RB for SSS, impact may be bad, hence minimum 6 RB
legth is chisen. M-sequence and WalshHadamard matrices are similar
and index remapping is done. This reduces complexity of SSS
detector, as complexity = N log2 N with N=32, complexity= 32
log2 32 = 160.
1.2.6. Cell Search Performance A new cell detection delay for UE
and report it to S-eNB should be less than acceptable threshold.
eNodeB uses the reports to prepare intra- or inter-frequency
handover. A multicell environment with three cells with different
transmitted powers with synchronized and unsynchronized eNodeBs
should be analysed. For the propagation channel, various multipath
fading, at least two receive antenna, UE speeds, (5 km/h, 300 km/h)
should be considered. Cell search performance is measured as
90-percentile(maximum time required to detect a target cell 90%of
the time) identification delay. After detection of PSS-SSS, RSRP is
measured. For initial synchronization case the time taken to decode
PBCH is adapted, and not just of reporting of measurements on
RS.For inter-frequency handover, performance can be derived from
the intra-frequency performance timing.
Coherent Versus Non-Coherent Detection A coherent detector uses
knowledge of the channel, while a non-coherent detector uses an
optimization metric of average channel statistics. In PSS,
non-coherent (No channel estimation available) detection is used,
while for SSS, coherent (channel estimation) or non-coherent
techniques can be used.
1.2.7. Reference Signals and Channel Estimation In any
communication system signal x transmitted by A passes through a
radio channel H (exhibit multipath fading, causing ISI) and suffers
additive noise before being received by B. To remove ISI,
equalization, detection algorithms and knowledge of Channel Impulse
Response (CIR) is used. OFDMA is quite robust against ISI by CP
which allows very good equalization at receiver. Coherent detection
uses amplitude and phase information exchanged between eNodeBs and
UEs. This comes at a price of overhead of channel estimation by
exploiting known signals which do not carry any data, sacrificing
spectral efficiency. Known reference signals are inserted into the
transmitted signal structure. Reference signals(known) are
multiplexed with data symbols (unknown at receiver) in either
frequency, time or code domains. Time multiplexing, known
preamble-based training transmission also is another technique.
Orthogonal RS multiplexing is the most common technique. OFDM
transmission is a two-dimensional lattice in time and frequency,
which helps multiplexing of RSs mapped to specific REs according to
a specific pattern. Since RS are sent only on particular OFDM REs
(particular symbols and subcarriers), channel estimates for non-RS
REs have to be computed via interpolation.
1.2.8. Design of Reference Signals in LTE In DL, three different
types of RS are provided:
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
20
1. Cell-specific RSs ( common RSs). 2. UE-specific RSs, in the
data for specific UEs. 3. MBSFN-specific RSs, for MBSFN.
1.2.9. Cell-Specific Reference Signals (CRS) In an OFDM-based
system, an equidistant arrangement of RS in the lattice structure
diamond shape achieves the Minimum Mean Squared Error (MMSE)
estimate. In time domain, RS spacing in is governed by maximum
Doppler spread (highest speed-540km/h(150m/s)) to be supported.
Doppler shift is fd = (fc v/c) where fc is the carrier frequency, v
is UE speed in m/s, and c= 3 * 10
8 m/s. Considering fc = 2 GHz (2*10
9Hz) and v = 500 km/h, fd
=(2*109*150/3*10
8) =1000 Hz. According to Nyquists sampling theorem, minimum
sampling
frequency to reconstruct the channel is Tc = 1/(2fd)= 0.5 ms.
This implies that two RS/slot are needed to estimate channel
correctly.
Fig 2.2.8.1 Cell RS for 1, 2 and 4 antenna
In frequency domain, there is one RS every six subcarriers on
each symbols including RS symbol, but staggered so that within each
RB there is one RS every 3 subcarriers. This spacing is goverened
by expected coherence BW of channel, governed by channel delay
spread. The 90% and 50% coherence BW are given respectively by
Bc,90% = 1/50d=20kHz and Bc,50% = 1/5d=200kHz where d is the r.m.s
delay spread=1000ns. In LTE spacing between two RS in frequency, is
45 kHz (3 symbols), enough to resolve expected frequency domain
variations of the channel. RS patterns are designed to work with
MIMO antennas defined for multiple antenna ports at eNodeB. An
antenna port may be either a single physical antenna, or a
combination of multiple physical antenna elements. The transmitted
RS in a given antenna port defines the antenna port from the point
of view of the UE, and enables UE to derive channel estimate for
that antenna port.
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
21
Fig 2.2.8.2 Antenna ports and Physical antennas
Up to eight cell-specific antenna ports may be used by eNodeB,
requiring UE to derive up to eight separate channel estimates. For
each antenna port, a different RS pattern is designed, to minimize
intra-cell interference between multiple transmit antenna ports. Rp
is used for RS Tx on antenna port p. Also, when a RE is used for RS
on one antenna port, corresponding RE on other antenna ports is set
to zero to limit interference. Mark that, density of RS for third
and fourth antenna ports is half of the first two, to reduce
overhead. In cells with a high prevalence of high-speed users, use
of four antenna ports is unlikely, RSs with lower density can
provide sufficient channel estimation accuracy.
Fig 2.2.8.3 Antenna port example of port 0 and port 5
All the RSs (cell-specific, UE-specific or MBSFN specific) are
QPSK modulated to ensure low PAPR. The signal can be written as
r(l,ns,m) = 1/2[1-2c(2m)] + j1/2[1-2c(2m+1)] where m is RS index,
ns = slot number and l =symbol number within slot, c(i) is
length-31 Gold sequence, with different initialization values
depending on type of RSs. RS sequence carries unambiguously one of
the 504 different cell identities, Ncell ID. For the cell-specific
RSs, a cell-specific frequency shift (Ncell ID mod 6) is also
applied. This shift avoids collisions between common RS from up to
six adjacent cells. Transmission power of RS is boosted, up to max
6 dB relative to surrounding data symbols, designed to improve
channel estimation. If adjacent cells also transmit high-power RS
on same REs, interference will prevent the gain.
1.2.10. UE-Specific Reference Signals(URS) UE-specific RS may be
used in addition to CRSs, embedded only in a specific UEs scheduled
RBs, using a distinct antenna port. UE is expected to use them to
derive the channel estimate for demodulating data in PDSCH RBs.
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
22
Fig 2.2.9 Port 5, UE specific Reference Signals
A typical usage of URSs is beam-forming of data transmissions to
specific UEs. Rather than using physical antennas of other CRS
antenna ports, eNodeB may use a correlated array of antenna
elements to generate a narrow beam in the direction of a particular
UE. Beam will experience different channel response requiring URSs
to demodulate the beamformed data coherently. The URS structure is
chosen not to collide with CRSs, and hence URSs does not affect
CRSs. URSs have a similar pattern as CRSs allowing re-use similar
channel estimation algorithms. Density is half of CRS, minimizing
the overhead.
1.2.11. RS-Aided Channel Modelling and Estimation Channel
estimation problem is related to physical propagation, number of
transmit and receive antennas, BW, frequency, cell configuration
and relative speed.
1. frequencies and BW determine the scattering. 2. Cell
deployment governs multipath, delay spread and spatial correlation.
3. Relative speed sets time-variations.
Propagation conditions characterize the channel in three
dimension (frequency, time and spatial) domains. Each MIMO
multipath channel component can experience different scattering
conditions across the three domains. LTE specifications do not
mandate any specific channel estimation technique, and there is
therefore complete freedom in implementation provided that the
performance requirements are met and the complexity is
affordable.
1.2.12. Frequency Domain Channel Estimation The natural approach
to estimate the whole CTF is to interpolate its estimate between
the reference symbol positions. As a second straightforward
approach, the CTF estimate over all subcarriers can be obtained by
IFFT interpolation. More elaborate linear estimators derived from
both deterministic and statistical viewpoints are proposed -Least
Squares (LS), Regularized LS, Minimum Mean-Squared Error (MMSE) and
Mismatched MMSE. It is seen that IFFT and linear interpolation
methods yield lowest performance. The regularized LS and the
mismatched MMSE perform exactly equally. Optimal MMSE estimator
outperforms any other estimator. MMSE-based channel estimation
suffers the least band-edge degradation, while all the other
methods presented are highly adversely affected.
1.2.13. Time-Domain Channel Estimation Time-Domain (TD) channel
estimation requires sufficient memory for buffering soft values of
data over several symbols while the channel estimation is carried
out. However, correlation between consecutive symbols decreases as
UE speed increases. TD correlation is inversely proportional to the
UE speed sets a limit on the possibilities for TD filtering in
high-mobility conditions.
Finite and Infinite Length MMSE-(TD-MMSE) The statistical TD
filter which is optimal in terms of Mean Squared Error (MSE) can be
approximated in the form of a finite impulse response filter. It
can be observed that, unlike Frequency-Domain (FD) MMSE filtering,
the size of the matrix to be inverted for a finite-length TD-MMSE
estimator is independent of the channel length L but dependent on
the chosen FIR order M. Similarly to the FD counterpart, the
TD-MMSE estimator requires knowledge of the PDP, the UE speed and
the noise variance.
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
23
Normalized Least-Mean-Square(NLMS) An adaptive estimation
approach can be considered which does not require knowledge of
second-order statistics of both channel and noise. A feasible
solution is the Normalized Least-Mean-Square (NLMS) estimator. It
can be observed that the TD-NLMS estimator requires much lower
complexity compared to TD-MMSE as no matrix inversion is required,
as well as not requiring any a priori statistical knowledge. Other
adaptative approaches could also be considered such as Recursive
Least Squares (RLS) and Kalman-based filtering.
1.2.14. Spatial Domain Channel Estimation(SD-MMSE) LTE UE is
designed for MIMO. Consequently, whenever the channel is correlated
in the spatial domain, the correlation can be exploited to provide
a further means for enhancing the channel estimate. If it is
desired to exploit spatial correlation, a natural approach is again
offered by Spatial Domain (SD) MMSE filtering.
-
LTE Physical Layer- 3PCA-L1 Certification email: @yahoo.com Page
24