1

 Adaptive modulation is a transmission scheme in digital communications where the

transmitter adapts its transmission mode in accordance with the channel. Depending on
the condition of the channel, the transmitter could be adapting one or more of the
following: constellation size, code rate, and power.

 Optimization criterion:

 Maximize throughput

 Minimize average power

 Minimize average BER

2
 Adaptive modulation systems invariably require some channel information at the transmitter.

 This could be acquired in time division duplex systems by assuming the channel from the

transmitter to the receiver is approximately the same as the channel from the receiver to the
transmitter.

 Alternatively, the channel knowledge can also be directly measured at the receiver, and fed back to

the transmitter.

 Adaptive modulation systems improve rate of transmission, and/or bit error rates, by exploiting

the channel information that is present at the transmitter.

 Especially over fading channels which model wireless propagation environments, adaptive

modulation systems exhibit great performance enhancements compared to systems that do not
exploit channel knowledge at the transmitter.

3
 In adaptive transmission we estimate the power gain or received SNR at time i and

adapt the modulation and coding parameters accordingly.

 The most common parameters to adapt are the data rate R[i], transmit power S[i], and

coding parameters C[i].

4
 An estimate ˆg[i] of the channel power gain g[i] at time i is available to the receiver after

an estimation time delay of ie and that this same estimate is available to the transmitter
after a combined estimation and feedback path delay of id = ie + if .

 The availability of this channel information at the transmitter allows it to adapt its

transmission scheme relative to the channel variation.

 We assume that the feedback path does not introduce any errors, which is a reasonable

assumption if strong error correction and detection codes are used on the feedback
path.

 The rate of channel variation will dictate how often the transmitter must adapt its

transmission parameters, and will also impact the estimation error of g[i].

5
 There are many parameters that can be varied at the transmitter relative to the channel

gain .

 This adaptive techniques associated with variation of the most common parameters:

 data rate, power, coding, error probability, and combinations of these adaptive

techniques.

6
 In variable-rate modulation the data rate R[γ] is varied relative to the channel SNR

γ.

 This can be done by fixing the symbol rate Rs = 1/Ts of the modulation and using

multiple modulation schemes or constellation sizes, or by fixing the modulation

(e.g. BPSK) and changing the symbol rate.

 Symbol rate variation is difficult to implement in practice since a varying signal

bandwidth is impractical and complicates bandwidth sharing.

 In contrast, changing the constellation size or modulation type with a fixed symbol

rate is fairly easy, and these techniques are used in current systems.
7
 If a discrete set of modulation types or constellation sizes are used, each value of γ is

mapped to one of the possible modulation schemes. This is done to maintain the BER below a
given value.

8
 Adapting the transmit power alone is generally used to compensate for SNR variation due

to fading.

 The goal is to maintain a fixed bit error probability or, equivalently, a constant

received SNR.

 The power adaptation thus inverts the channel fading so that the channel appears as an

AWGN channel to the modulator and demodulator.

S(γ) σ
 The power adaptation for channel inversion is given by = ,
S γ


σ equals the constant received SNR.

 The average power constraint S implies that,

9
 Solving above equation for σ yields that σ = 1/E[1/γ].
 σ is determined by p(γ) which in turn depends on the average transmit power S’
through γ.

10
 It is possible adapt the instantaneous BER subject to an average BER constraint P b’.

 In fading channels the instantaneous error probability varies as the received SNR γ varies,

resulting in an average BER of Pb’ = Pb(γ)p(γ)dγ.

 This is not considered an adaptive technique since the transmitter does not adapt to γ.

 Thus, in adaptive modulation error probability is typically adapted along with some other

form of adaptation such as constellation size or modulation type.

11
 In adaptive coding different channel codes are used to provide different amounts of
coding gain to the transmitted bits.

 For example. a stronger error correction code may be used when γ is small, with a
weaker code or no coding used when γ is large.

 Adaptive coding can be implemented by multiplexing together codes with different

error correction capabilities.

 An alternative technique to code multiplexing is rate-compatible punctured

convolutional (RCPC) codes.

 RCPC codes consist of a family of convolutional codes at different code rates Rc = k/n.

 The basic premise of RCPC codes is to have a single encoder and decoder whose error
correction capability can be modified without transmitting certain coded bits
12
(puncturing the code).
 RCPC codes have a rate-compatibility constraint so that the coded bits associated with a high-

rate (weaker) code are also used by all lower-rate (stronger) codes.

 Thus, to increase the error correction capability of the code, the coded bits of the weakest code

are transmitted along with additional coded bits to achieve the desired level of error
correction.

 The rate compatibility makes it very easy to adapt the error protection of the code, since the

same encoder and decoder are used for all codes in the RCPC family, with puncturing at the
transmitter to achieve the desired error correction.

 Decoding is performed by a Viterbi algorithm operating on the trellis associated with the

lowest rate code, with the puncturing incorporated into the branch metrics.

 Puncturing is a very effective and powerful adaptive coding technique, and forms the basis of

adaptive coding in GSM’s EDGE protocol for data transmission .

13
 Hybrid techniques can adapt multiple parameters of the transmission scheme,

including rate, power, coding, and instantaneous error probability.

 In this case joint optimization of the different techniques is used to meet a given

performance requirement.

 Rate adaptation is often combined with power adaptation to maximize spectral

efficiency.

 We can apply this joint optimization to different modulations in subsequent

sections.

14
Adaptive MQAM varies both the rate and the transmit power
to maximize the throughput under the constraint of fixed
average transmit power and the constraint of fixed
instantaneous BER

Seminar: 1. Adaptive MQAM (Variable rate -

variable power MQAM)

2. Impact of finite constellations

15
 Quadrature amplitude modulation (QAM) is a popular scheme for high-rate, high bandwidth

efficiency systems.

 QAM is a combination of both amplitude and phase modulation.

 Mathematically, M-ary QAM is described by

𝑠𝑚𝑛 (𝑡) = 𝐴𝑚 cos 2𝜋𝑓𝑐 𝑡 + 𝜃𝑛 , 𝑚 = 1,2, … , 𝑀1

𝑛 = 1,2, … , 𝑀2

 The combined amplitude and phase modulation results in the simultaneous transmission of

log2 M1 M2 bits/symbol

16
17
 Adaptive modulation where the rate and power of MQAM is varied to maximize spectral

efficiency while meeting a given instantaneous Pb target.

 Basic idea: adapt at transmitter relative to channel fade level (borrows from capacity ideas).

 Parameters to adapt (degrees of freedom) include constellation size, transmit power,

instantaneous BER, symbol time, coding rate/scheme, and combinations.

 Optimization criterion for adaptation is typically maximizing average rate, minimizing

average power, or minimizing average BER.

18
 Adaptive modulation where the rate and power of MQAM is varied to maximize spectral efficiency
while meeting a given instantaneous Pb target.

 Basic idea: adapt at transmitter relative to channel fade level (borrows from capacity ideas).

 Parameters to adapt (degrees of freedom) include constellation size, transmit power, instantaneous
BER, symbol time, coding rate/scheme, and combinations.

 Optimization criterion for adaptation is typically maximizing average rate, minimizing average
power, or minimizing average BER.

 Constellation size and power adapted to maximize average throughput given an instantaneous BER
constraint.

 BER bound BER(γ) = .2 exp[−1.5γP(γ)/((M − 1)S)] inverted to get adaptive constellation size M[γ]
below with K = −1.5/ln(5 · BER) that meets the BER constraint for any adaptive power policy P[γ]:

19
 Optimal power adaptation P(γ) found by maximizing average throughput E[log2 (M[γ])] =
E[log2 (1 + KγP(γ)/P) relative to P(γ).

 Optimal power adaptation is the same waterfilling as the capacity-achieving strategy with an
effective power loss K.

 Optimal rate adapation found by substituting optimal power adaptation into M(γ), yielding
R(γ) = log2 (γ/γK), γ > γK, where γK is cutoff value for the water-filling power policy.

 Same optimal power and rate adaptation as the capacity-acheiving strategies with an
effective power reduction K = −1.5/ ln(5 · BER).

 Throughput is within 5-6 dB of channel capacity. • Different modulations and BER bounds
result in different adaptive policies.
20
 Constellation restricted to finite set {M0 = 0, M1, . . ., MN−1}

 Divide the fading range of γ into N discrete fading regions Rj.

 Within each region “conservatively” assign constellation Mj : Mj ≤ M(γ) ≤ Mj+1, where

M(γ) = γ/γ∗ K for some optimized γ ∗ K.

 Power control based on channel inversion; maintains constant BER within region Rj.

 Using large enough constellation set results in near-optimal performance.

 Additional power penalty of 1.5-2 dB if each constellation restricted to a single transmit

power.

 Restricting the size of the constellation set in adaptive modulation leads to negligible
performance loss.

21
 Additional coding gain can be achieved with adaptive modulation by superimposing trellis

codes or more general coset codes on top of the adaptive modulation.

 Specifically, by using the subset partitioning inherent to coded modulation, trellis or lattice

codes designed for AWGN channels can be superimposed directly onto the adaptive
modulation with the same approximate coding gain.

 The basic idea of adaptive coded modulation is to exploit the separability of code and

constellation design inherent to coset codes.

 Coded modulation is a natural coding scheme to use with variable-rate variable-power

MQAM, since the channel coding gain is essentially independent of the modulation.

 Therefore adjust the power and rate (number of levels or signal points) in the transmit

constellation relative to the instantaneous SNR without affecting the channel coding gain
22
23
 Let Gc denote the coding gain of the coset code.

 The source coding (modulation) works as follows. The signal constellation is a square lattice

with an adjustable number of constellation points M. The size of the MQAM signal
constellation from which the signal point is selected is determined by the transmit power,
which is adjusted relative to the instantaneous SNR and the desired BER.

 Optimal power adaptation policy for this modulation scheme:

where γ0 is the cutoff fade depth, and Kc = KGc

 The resulting spectral efficiency is, where γKc = γ0/Kc.

24
25
Narrowband MIMO Model
 A narrowband point-to-point communication system of Mt transmit and Mr receiver

antennas is shown in figure .

 This system can be represented by the following discrete time model:

26
 Simply as y = Hx + n. Here x represents the Mt-dimensional transmitted symbol, n is the Mr-

dimensional noise vector, and H is the Mr × Mt matrix of channel gains hij representing the
gain from transmit antenna j to receive antenna i.

 We assume a channel bandwidth of B and complex Gaussian noise with zero mean and

covariance matrix σ²nIMr , where typically σ²n = N0B.

 For simplicity, given a transmit power constraint P we will assume an equivalent model

with a noise power of unity and transmit power P/σ²n = ρ, where ρ can be interpreted as
the average SNR per receive antenna under unity channel gain.

 This power constraint implies that the input symbols satisfy

 or, equivalently, that Tr(Rx) = ρ, where Tr(Rx) is the trace of the input covariance matrix

Rx = E[xxT].

27
 Different assumptions can be made about the knowledge of the channel gain matrix H at the

transmitter and receiver, referred to as channel side information at the transmitter (CSIT) and
channel side information at the receiver (CSIR), respectively.

 For a static channel CSIR is typically assumed, since the channel gains can be obtained fairly

easily by sending a pilot sequence for channel estimation.

 If a feedback path is available then CSIR from the receiver can be sent back to the transmitter to

provide CSIT.

 When the channel is not known at either the transmitter or receiver then some distribution on

the channel gain matrix must be assumed.

 The most common model for this distribution is a zero-mean spatially white (ZMSW) model,

where the entries of H are assumed to be i.i.d. zero mean, unit variance, complex circularly
symmetric Gaussian random variables.
28
 When both the transmitter and receiver have multiple antennas, there is another mechanism for
performance gain called multiplexing gain.

 The multiplexing gain of a MIMO system results from the fact that a MIMO channel can be
decomposed into a number R of parallel independent channels. By multiplexing independent data
onto these independent channels, we get an R-fold increase in data rate in comparison to a system
with just one antenna at the transmitter and receiver. This increased data rate is called the
multiplexing gain.

 In this section we describe how to obtain independent channels from a MIMO system.

 Consider a MIMO channel with Mr × Mt channel gain matrix H known to both the transmitter and
the receiver.

 Let RH denote the rank of H. From matrix theory, for any matrix H we can obtain its singular value
decomposition (SVD) as H = UΣV, where U and V are unitary matrices
29
 The parallel decomposition of the channel is obtained by defining a transformation on the

channel input and output x and y through transmit precoding and receiver shaping.

 In transmit precoding the input to the antennas x is generated through a linear transformation

on input vector 𝒙
෥ as x = 𝑽𝑯 𝒙
෥.

 Receiver shaping performs a similar operation at the receiver by multiplying the channel

output y with 𝑼𝑯 , as shown in figure

 The transmit precoding and receiver shaping transform the MIMO channel into RH

parallel single-input single-output (SISO) channels with input 𝒙

෥ and output 𝒚
෥
30
 where., -

31
 Channel capacity depends on what is known about the channel gain matrix or its distribution at

the transmitter and/or receiver.

 Throughout this section it is assumed that the receiver has knowledge of the channel matrix H.

 2 types of channels: Static and Fading channels

1. STATIC CANNEL CAPACITY

 The capacity is given in terms of the mutual information between the channel input vector x

and output vector y as

 H(Y) and H(Y|X) the entropy in y and y|x,

 The definition of entropy yields that H(Y|X) = H(N), the entropy in the noise.

 Since this noise n has fixed entropy independent of the channel input, maximizing mutual
32
information is equivalent to maximizing the entropy in y.
 The mutual information of y depends on its covariance matrix, which for the narrowband

MIMO model is given by,

 The mutual information

 Entropy of y is maximized when Ry is maximized which in turn means that Rx has to be

maximized

 The MIMO capacity is achieved by maximizing the mutual information over all input
covariance matrices Rx

 We now consider this maximizing under different assumptions about transmitter CSI.

33
a) Channel Known at Transmitter: Waterfilling

 The MIMO channel capacity for a fixed channel matrix H known at the transmitter and receiver.

 The capacity equals the sum of capacities on each of the independent parallel channels with the

transmit power optimally allocated between these channels. This optimization of transmit power
across the independent channels results from optimizing the input covariance matrix to maximize
the capacity formula.

 The MIMO capacity with CSIT and CSIR as

 Since ρ = P/σ²n, expressing capacity in terms of power Pi to i-th parallel channel

 The capacity can also be expressed in terms of the power allocation Pi to the ith parallel channel as

where ρi = Pi/σ²n and γi = σ²i P/σ²n is the SNR associated with the ith channel at full power.
34
 The water-filling power allocation is given by ,

 For some cutoff value γ0. The resulting capacity is then,

 Capacity under perfect CSIT and CSIR can also be defined on channels where there is a
single antenna at the transmitter and multiple receive antennas (single-input multiple-
output or SIMO) or multiple transmit antennas and a single receive antenna (multiple-input
single-output or MISO). These channels can only obtain diversity gain from the multiple
antennas.

b) Channel Unknown at Transmitter: Uniform Power Allocation

 The receiver knows the channel but the transmitter does not. Without channel information,
the transmitter cannot optimize its power allocation or input covariance structure across
antennas.

 The best strategy should be to allocate equal power to each transmit antenna, resulting in an
ρ
input covariance matrix equal to the scaled identity matrix: Rx = IMt .
Mt 35
 Mutual information given by,

 Using the SVD of H, we can express this as,

 where and RH is the number of nonzero singular values of H.

 Capacity,

 Defining M = min(Mt, Mr), this implies that as M grows large, the MIMO channel capacity in the

absence of CSIT approaches and hence grows linearly in M .

 In a static channel if the average mutual information is not known then the rate of transmission is

unknown.

 In such case capacity is given by Outage capacity ( probability that transmitted data will not be

received correctly)

36
 As SNR grows large, capacity also grows linearly with M = min(Mt,Mr) for any Mt and Mr.

 These results are the main reason for the widespread appeal of MIMO techniques: even if the
channel realization is not known at the transmitter, the capacity of MIMO channels still grows linearly
with the minimum number of transmit and receiver antennas, as long as the channel can be
accurately estimated at the receiver.

 Thus, MIMO channels can provide very high data rates without requiring increased signal power or
bandwidth.

 At very low SNRs transmit antennas are not beneficial: capacity only scales with the number of
receive antennas independent of the number of transmit antennas.

 Lack of CSIT does not affect the growth rate of capacity relative to M, at least for a large number of
antennas, it does complicate demodulation.

 Specifically, without CSIT the transmission scheme cannot convert the MIMO channel into non-
interfering SISO channels. 37
2. FADING CHANNELS

 Now that the channel gain matrix experiences flat-fading (H varies with time), so the gains hij

vary with time.

 As in the case of the static channel, the capacity depends on what is known about the channel

matrix at the transmitter and receiver.

a) Channel Known at Transmitter: Water-Filling

 Capacity is given in terms of ergodic capacity of the channel.

 There are two possibilities for allocating power under ergodic capacity.

 Short term power constraint (Power associated with each channel realization must equal

average power constraint P )

38
  A less restrictive constraint is a long-term power constraint, where we can use different

powers for different channel realizations

 Short-term power constraint give rise to water-filling in space

 Long-term power constraint allows for two-dimensional water-filling across both space and

time.

b) Channel Unknown at Transmitter: Ergodic Capacity and Capacity with Outage

 Consider now a time-varying channel with random matrix H known at the receiver but not the

transmitter.

 Ergodic Capacity: Optimization problem at Tx (Finding optimum input covariance matrix)

 For scalar channels, 39

 Channel Unknown at Tx: Capacity with Outage

 Outage capacity can be improved by not allotting power to one or more Tx antennas,

especially when outage probability is high.

b) No CSI at the Tx or Rx

 Linear growth in capacity as a function of number of antennas disappear

 Capacity depends on the underlying channel model

 Capacity depends on the structure of the fading process

 For general fading process, no multiplexing gain associated with multiple antennas when

there is no Tx or Rx CSI
40
 In the case of flat fading channel ergodic capacity or the outage capacity is used

 Frequency selective fading capacity (more degrees of freedom)

 Capacity Calculations are complex

41
 MIMO Systems can provide two types of gain

Spatial Multiplexing Gain Diversity Gain

• Maximize transmission rate • Minimize Pe (conservative

(optimistic approach) approach)
• Use rich scattering/fading to • Go for Reliability / QoS etc
advantage
• Counter fading

 As expected, there is a tradeoff

 System designs are carried out to achieve a little bit of both.

42
 The multiple antennas at the transmitter and receiver can be used to obtain diversity gain

instead of capacity gain.

 In this setting, the same symbol, weighted by a complex scale factor, is sent over each transmit

antenna, so that the input covariance matrix has unit rank. This scheme is also referred to as
MIMO beamforming*.

 As indicated in the figure, the transmit symbol x is sent over the ith antenna with weight vi.

 On the receive side, the signal received on the ith antenna is weighted by ui.

 Both transmit and receive weight vectors are normalized so that ||u|| = ||v|| = 1. The resulting

received signal is given by

y = u∗Hvx + u∗n,

*Unfortunately, beamforming is also used in the smart antenna to describe adjustment of the 43
transmit or receive antenna directivity in a given direction.
 Beamforming provides diversity gain by coherent combining of the multiple signal paths.

y=u*Hvx + u*n

 where if n = (n1, . . . , nMr ) has i.i.d. elements, the statistics of u∗n are the same as the

statistics for each of these elements

 If H is known, the received SNR is optimized by choosing u and v as the principal left and

right singular vectors of the channel matrix H.

 The corresponding received SNR can be shown to equal γ = λmaxρ, where λmax is the

largest eigenvalue of the Wishart matrix, W= H𝑯𝑯 . 44

 The resulting capacity is corresponding to the capacity of a
SISO channel with channel power gain λmax.

 When the channel is not known at the transmitter, the transmit antenna weights are all equal.

45
 The previous sections suggest two mechanisms for utilizing multiple antennas to improve

wireless system performance.

 One option is to obtain capacity gain by decomposing the MIMO channel into parallel channels

and multiplexing different data streams onto these channels. This capacity gain is also referred
to as a multiplexing gain/Spatial multiplexing.

 However, the SNR associated with each of these channels depends on the singular values of the

channel matrix.

 Alternatively, beamforming can be used, where the channel gains are coherently combined to

obtain a very robust channel with high diversity gain.

 It is not necessary to use the antennas purely for multiplexing or diversity.

46
 The diversity/multiplexing tradeoff or, more generally, the tradeoff between data rate,

probability of error, and complexity for MIMO systems.

 It has primarily focused on block fading channels with receiver CSI only .

 When both transmitter and receiver know the channel the tradeoff is relatively
straightforward: grouped for diversity gain and then the multiplexing gain corresponds to the
new channel with reduced dimension due to the grouping.

 For the block fading model with receiver CSI only, as the blocklength grows asymptotically

large, full diversity gain and full multiplexing gain can be obtained simultaneously.

 For finite blocklengths it is not possible to achieve full diversity and full multiplexing gain

simultaneously, in which case there is a tradeoff between these gains.

 A simple characterization of this tradeoff is given for block fading channels with blocklength

T ≥ Mt + Mr − 1 in the limit of asymptotically high SNR. 47

 In this analysis a transmission scheme is said to achieve multiplexing gain r and diversity gain

d if the data rate (bps) per unit Hertz R(SNR) and probability of error Pe(SNR) as functions of
SNR satisfy

 where the log in second equation can be in any base.

 For each r the optimal diversity gain dopt(r) is the maximum the diversity gain that can be
achieved by any scheme.
 If the fading blocklength exceeds the total number of antennas at the transmitter and receiver,
then
dopt(r) = (Mt − r)(Mr − r), 0 ≤ r ≤ min(Mt,Mr).

48
 The above function is plotted in the below given figure.

49
 Since a MIMO channel has input-output relationship y = Hx + n, the symbol

transmitted over the channel each symbol time is a vector rather than a
scalar, as in traditional modulation for the SISO channel. Moreover, when
the signal design extends over both space and time, it is typically
referred to as a space-time code.

 Our major concentration will be on Space time block code

(STBC),space time trellis coding(STTC), spatial multiplexing and
Blast architectures.

50
 MIMO systems makes use of space-time block codes to allow the transmission
of many number of copies of an information stream over a number of
antennas and to utilize the multiple received data units to enhance the
capability of data transfer.

 Space time coding helps to combat major problems like fading and thermal
noise because of the use of number of copies of data .

 In the space time block coding scheme, the stream of data is encoded in the
form of blocks before the transmission takes place.

 These blocks of data are segregated over multiple antennas and spacing in time
domain is also done.

 The general representation method used for representing space time block
51
code is by a matrix as shown below:
 Time slot is represented by each row and one antenna’s transmission in time domain is

represented by each column.

52
 Space-time trellis codes makes use of a scheme in which the
transmission of redundant copies of data in done in both time and
space by making use of multiple numbers of antennas.

 These copies of the redundant data sets are utilized by the receiver

to recover the original transmitted data.

 Space-time trellis codes are able to give coding and diversity gain along

with improved bit error rate performance.

 Complexity of encoding and decoding increases as it makes the use of

trellis codes. 53
 Spatial multiplexing is a technique widely used in MIMO communications

to send independent data streams which are separately encoded by

using the multiple transmit antennas.

 The multiplexing of space dimensions is done or reutilization is done

multiple times hence it is called as spatial multiplexing.

 If the transmitter consists of Mt transmit antennas and Mr receive

antennas then the maximum spatial multiplexing is given by:

Ms =min (Mt ,Mr) (in case of a linear receiver)

54
 Bell Labs Layered Space Time (BLAST) architectures are a family of
architectures which came into view for systems which employs multiple antenna
arrays at both the transmitter and receiver end.

55
 The overall spectral efficiency can be thus increased as the multiple numbers of streams

are transmitted in parallel approach.

56
 In this article, we only consider the overview of the various BLAST architectures. The

BLAST architectures are categorized into three types:

1) D-BLAST architecture

2) V-BLAST architecture

3) Turbo BLAST architecture

57
 Bell Labs Layered Space Time (BLAST) architectures are a family of architectures which
came into view for systems which employs multiple antenna arrays at both the
transmitter and receiver end.

 It is the original scheme of BLAST architecture family with diagonally layered

architecture i.e. it uses diagonally layered coding sequence.

 The layered design of D-BLAST architecture has number of layers equal to the number
of antennas in which each layer is coded independently.

 The major advantage of the D-BLAST architecture is that it fully utilizes the spatial and
time diversity.

 Due to full utilization of spatial and time diversity it requires multiple numbers of
decoders which introduces the drawback of complexity.
58
V-BLAST Architecture

 The Vertical BLAST architecture is a much simpler architecture as compared

to the D-BLAST architecture with reduced computational complexity which in

turn lowers its transmit diversity.

59
Turbo BLAST Architecture

 The Turbo BLAST architecture uses random layered space-time coding

scheme (RLST) before transmission at the encoder. The encoder & decoder
of turbo BLAST architecture is shown in the below mentioned figures
respectively

60
61