0% found this document useful (0 votes)
26 views

Lec 9

The document summarizes Lecture 9 on the diversity-multiplexing tradeoff (DMT) for theoretical wireless communications. It discusses the formulation of DMT and analyzes the optimal tradeoff for scalar, parallel, MISO, 2x2 MIMO, and general MIMO Rayleigh channels. Key information presented includes the definitions of diversity gain and multiplexing gain, and how DMT characterizes the relationship between them.

Uploaded by

khodang
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
26 views

Lec 9

The document summarizes Lecture 9 on the diversity-multiplexing tradeoff (DMT) for theoretical wireless communications. It discusses the formulation of DMT and analyzes the optimal tradeoff for scalar, parallel, MISO, 2x2 MIMO, and general MIMO Rayleigh channels. Key information presented includes the definitions of diversity gain and multiplexing gain, and how DMT characterizes the relationship between them.

Uploaded by

khodang
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 19

Lecture 9

DMT

Ragnar Thobaben
ISE/EECS/KTH

Formulation
Lecture 9: Diversity-Multiplexing Tradeoff
Theoretical Foundations of Wireless Communications1
Scalar Rayleigh
Channel

Parallel Rayleigh
Channel

MISO Rayleigh
Channel Ragnar Thobaben
2×2 MIMO Rayleigh ISE/EECS/KTH
Channel

General MIMO i.i.d.


Rayleigh Channel

Friday, May 25, 2018


09:00-11:30, ”Kansliet”

1
Textbook: D. Tse and P. Viswanath, Fundamentals of Wireless Communication
1 / 19
Overview
Lecture 7+8: MIMO Architectures
Lecture 9
• Transmitter architectures: V-BLAST and D-BLAST.
DMT • Receiver architectures
Ragnar Thobaben
ISE/EECS/KTH • Linear decorrelator
• Linear MMSE
Formulation • Successive cancellation (decorrelator and MMSE)
Scalar Rayleigh
Channel
• Capacity and outage probability
Parallel Rayleigh
Channel Lecture 9: Diversity-Multiplexing Tradeoff (Ch. 9.1)
MISO Rayleigh
Channel 1 Formulation
2×2 MIMO Rayleigh 2 Scalar Rayleigh Channel
Channel
PAM and QAM
General MIMO i.i.d.
Rayleigh Channel Optimal Tradeoff
3 Parallel Rayleigh Channel
4 MISO Rayleigh Channel
5 2×2 MIMO Rayleigh Channel
6 General MIMO i.i.d. Rayleigh Channel
Optimal Tradeoff
Geometric Interpretation

2 / 19
Formulation: Diversity-Multiplexing Tradeoff (DMT)

Lecture 9 Diversity gain d ∗


DMT

Ragnar Thobaben • Important performance measure for slow fading channels.


ISE/EECS/KTH
(Fixed rate R; highest achievable rate depends on channel
Formulation realization and is a RV; outage probability.)
Scalar Rayleigh
Channel
• Example: i.i.d. MIMO, d ∗ = nt · nr .

Parallel Rayleigh • Outage probability: pout (R) ∼ 1/SNRd
Channel

MISO Rayleigh
Channel

2×2 MIMO Rayleigh Multiplexing gain r


Channel
• Fast fading MIMO channels: multiplexing capability.
General MIMO i.i.d.
Rayleigh Channel (Coding over many channel realizations; average capacity.)
• Example: i.i.d. MIMO, C = nmin log SNR + K , with
nmin = min(nr , nt ), i.e., r = nmin .

3 / 19
Formulation: Diversity-Multiplexing Tradeoff (DMT)
• To maximize the diversity gain, the rate R has to be fixed.
Lecture 9 • If we want to communicate at a rate R = r log SNR (i.e., a fraction
DMT
of the fast fading capacity), which diversity gain can we get?
Ragnar Thobaben
ISE/EECS/KTH

Formulation Diversity-Multiplexing Tradeoff (DMT)


Scalar Rayleigh
Channel A diversity gain d ∗ (r ) is achieved at multiplexing gain r if
Parallel Rayleigh
Channel R = r log SNR
MISO Rayleigh
Channel and ∗
2×2 MIMO Rayleigh
Channel
pout ≈ SNR−d (r )
,
General MIMO i.i.d. or more precisely
Rayleigh Channel

log pout (r log SNR)


lim = −d ∗ (r ).
SNR→∞ log SNR
The curve d ∗ (r ) is the diversity-multiplexing tradeoff of the slow fading
channel.

DMT for (space-time) coding schemes: replace pout by pe (and d ∗ by d).


4 / 19
Scalar Rayleigh Channel
– PAM and QAM
• Scalar channel model: y [m] = hx[m] + w [m], with
Lecture 9 • Additive white Gaussian noise, i.i.d., w ∼ CN (0, 1);
DMT • Fading coefficient h ∼ CN (0, 1);
Ragnar Thobaben • Power constraint SNR.
ISE/EECS/KTH
• PAM error probability
Formulation
s !
Scalar Rayleigh
Channel 1 2
Dmin 1 22R
PAM and QAM pe ≈ 1− 2
≈ 2
≈ ,
Optimal Tradeoff
2 4 + Dmin Dmin SNR
Parallel Rayleigh
Channel for high SNR, where
MISO Rayleigh
√ √
• We assumed that the constellation ranges from − SNR to + SNR;
Channel √
• The minimum distance is given by Dmin ≈ SNR/2R .
2×2 MIMO Rayleigh
Channel
• With R = r log SNR, we get
General MIMO i.i.d.
Rayleigh Channel
1
pe ≈ .
SNR1−2r
• Diversity-multiplexing tradeoff for PAM:

dpam (r ) = 1 − 2r , r ∈ [0, 1/2].

5 / 19
Scalar Rayleigh Channel
– PAM and QAM
• QAM: 2R/2 constellation points per real/imaginary dimension.
Lecture 9
DMT • Minimum distance and QAM error probability (high SNR)
Ragnar Thobaben √
ISE/EECS/KTH
SNR 2R
Dmin ≈ and pe ≈ .
Formulation 2R/2 SNR
Scalar Rayleigh
Channel
• Diversity-multiplexing tradeoff for QAM:
PAM and QAM
Optimal Tradeoff dqam (r ) = 1 − r , r ∈ [0, 1].
Parallel Rayleigh
Channel

MISO Rayleigh
Channel

2×2 MIMO Rayleigh


Channel

General MIMO i.i.d.


Rayleigh Channel

6 / 19
Scalar Rayleigh Channel
– PAM and QAM

Lecture 9 Analysis, Case 1: dmax := d(0)


DMT
• Classical diversity gain for a scheme with fixed rate.
Ragnar Thobaben
ISE/EECS/KTH
• Describes how fast the error probability pe can be decreased with
Formulation SNR for a fixed rate R = const.
Scalar Rayleigh
Channel
• Example: Increasing the SNR by 6 dB for fixed rate.
PAM and QAM
Optimal Tradeoff
Parallel Rayleigh
Channel

MISO Rayleigh
Channel

2×2 MIMO Rayleigh


Channel

General MIMO i.i.d.


Rayleigh Channel

→ pe decreases due to increasing Dmin with increasing SNR.

7 / 19
Scalar Rayleigh Channel
– PAM and QAM

Analysis, Case 2: rmax = arg d(r ) = 0


Lecture 9
DMT • Number of degrees of freedom.
Ragnar Thobaben
ISE/EECS/KTH • Describes how fast the rate R can be increased with SNR for a fixed
Formulation
error probability pe = const.
Scalar Rayleigh • Example: Increasing the SNR by 6 dB.
Channel
PAM and QAM
Optimal Tradeoff
Parallel Rayleigh
Channel

MISO Rayleigh
Channel

2×2 MIMO Rayleigh


Channel

General MIMO i.i.d.


Rayleigh Channel

→ For pe = const (i.e., Dmin = const), the gain in SNR can be


utilized for increasing the number of bits.

8 / 19
Scalar Rayleigh Channel
– Optimal Tradeoff

• Outage probability and high-SNR approximation,


Lecture 9 n o
DMT

Ragnar Thobaben
pout = Pr log(1 + |h|2 SNR) < r log SNR
ISE/EECS/KTH
SNRr − 1
 
Formulation = Pr |h|2 <
SNR
Scalar Rayleigh
Channel 1
≈ .
PAM and QAM
Optimal Tradeoff
SNR1−r
Parallel Rayleigh
Channel
(Last step: for Rayleigh fading and small , Pr{|h|2 < } ≈ .)
MISO Rayleigh → Uncoded QAM is DMT optimal.
Channel

2×2 MIMO Rayleigh


Channel
Comment
General MIMO i.i.d. • Assume pe = a · pout , with a constant a, with 0 ≤ log(a)  log SNR
Rayleigh Channel for high SNR. Then
 
log pe log a log pout log pout
lim = lim + = lim .
SNR→∞ log SNR SNR→∞ log SNR log SNR SNR→∞ log SNR

• That a scheme is DMT optimal does not mean that it achieves the
outage probability. But it has the same exponential decay.

9 / 19
Parallel Rayleigh Channel

Lecture 9
• Channel Model: L parallel channels, yl [m] = hl xl [m] + wl [m], with
DMT • Additive noise wl ∼ CN (0, 1);
Ragnar Thobaben • Transmit power constraint SNR per sub-channel.
ISE/EECS/KTH
→ Maximum diversity gain of d ∗ (0) = L.
Formulation
• Assume target rate R = r log SNR bit/s/Hz per sub-channel.
Scalar Rayleigh
Channel • Outage probability,
Parallel Rayleigh ( L )
Channel X 2
MISO Rayleigh pout = Pr log(1 + |hl | SNR) < Lr log SNR .
Channel
l=1
2×2 MIMO Rayleigh
Channel
• Outage occurs if each of the sub-channels is in outage, accordingly
General MIMO i.i.d.
Rayleigh Channel  n oL 1
pout ≈ Pr log(1 + |hl |2 SNR) < r log SNR ≈ .
SNRL(1−r )
• Optimal DMT,

d ∗ (r ) = L(1 − r ), r ∈ [0, 1].

10 / 19
Parallel Rayleigh Channel
• Repetition coding: transmit identical QAM symbols over the
Lecture 9 parallel channels:
DMT

Ragnar Thobaben
ISE/EECS/KTH
drep (r ) = L(1 − Lr ), with r ∈ [0, 1/L].

Formulation • Repetition coding achieves full diversity d ∗ (0) = L.


Scalar Rayleigh
Channel
• Rate is reduced by a factor 1/L.
Parallel Rayleigh • Comparison
Channel

MISO Rayleigh
Channel

2×2 MIMO Rayleigh


Channel

General MIMO i.i.d.


Rayleigh Channel

11 / 19
MISO Rayleigh Channel
• Channel Model: nt transmit antennas,
Lecture 9
DMT y [m] = h∗ x[m] + w [m],
Ragnar Thobaben
ISE/EECS/KTH with
• Additive noise w ∼ CN (0, 1);
Formulation • Overall transmit power constraint SNR.
Scalar Rayleigh → Maximum diversity gain of d ∗ (0) = nt .
Channel
• Assume target rate R = r log SNR bit/s/Hz.
Parallel Rayleigh
Channel • Outage probability,
MISO Rayleigh    
2 SNR
Channel

2×2 MIMO Rayleigh pout = Pr log 1 + khk < r log SNR .


Channel nt
General MIMO i.i.d.
Rayleigh Channel • khk2 is a χ2 random variable with 2nt degrees of freedom, and we
have Pr{khk2 < } ≈ nt such that
. 1
pout = nt (1−r )
.
SNR
• Optimal DMT

d ∗ (r ) = nt (1 − r ), r ∈ [0, 1].
12 / 19
MISO Rayleigh Channel
• Alamouti converts a MISO channel into a scalar channel with the
Lecture 9 same outage behavior.
DMT

Ragnar Thobaben
→ Alamouti plus QAM is DMT optimal.
ISE/EECS/KTH
• Repetition coding: transmitting the same symbol over the two
Formulation antennas, one at a time; DMT for nt = 2,
Scalar Rayleigh
Channel drep (r ) = nt (1 − 2r ).
Parallel Rayleigh
Channel

MISO Rayleigh
Channel

2×2 MIMO Rayleigh


Channel

General MIMO i.i.d.


Rayleigh Channel

13 / 19
2×2 MIMO Rayleigh Channel
Comparison of four schemes (uncoded QAM)
• Repetition coding, Alamouti, and V-BLAST with nulling
Lecture 9
DMT
(decorrelator).
• The schemes convert the channel into scalar channels.
Ragnar Thobaben
ISE/EECS/KTH → DMT can be obtained as in the scalar case.
Formulation • V-BLAST with maximum likelihood (ML) decoding
Scalar Rayleigh • Pairwise error probability (PEP) between two codewords
Channel
(averaged over the channel realizations and with average transmit
Parallel Rayleigh energy normalize to one)
Channel
16
MISO Rayleigh Pr{xA → xB } ≤ (see as well (3.92)).
Channel
SNR2 kxA − xB k4
2×2 MIMO Rayleigh
Channel • Worst-case PEP is of the order (XA and XB differ only in one
General MIMO i.i.d.
dimension by the minimum distance of the QAM constellation)
Rayleigh Channel
16 · 2R
= 16 · SNR−(2−r ) .
SNR2
√ √
(with DAB = SNRkxA − xB k, Dmin ≈ SNR/2R/4 , R = r log SNR)

→ DMT: d(r ) = 2 − r .

• Optimal DMT: piecewise linear joining the points (0, 4), (1, 1), and
(2, 0).
14 / 19
2×2 MIMO Rayleigh Channel
Comparison of the four schemes
Lecture 9
DMT

Ragnar Thobaben
ISE/EECS/KTH

Formulation

Scalar Rayleigh
Channel

Parallel Rayleigh
Channel

MISO Rayleigh
Channel

2×2 MIMO Rayleigh


Channel

General MIMO i.i.d.


Rayleigh Channel

15 / 19
General MIMO i.i.d. Rayleigh Channel
– Optimal Tradeoff

Lecture 9
DMT • Channel model: (nt × nr ) MIMO channel with i.i.d. Rayleigh fading.
Ragnar Thobaben
ISE/EECS/KTH • Outage probability (for R = r log SNR),

Formulation
mimo
pout (r log SNR) = min Pr{log det(Int + HKx H∗ ) < r log SNR}.
Kx :Tr[Kx ]≤SNR
Scalar Rayleigh
Channel → Transmit strategy Kx depends on the SNR.
Parallel Rayleigh
Channel • Assumption: Kx = SNR/nt Int .
MISO Rayleigh (Suboptimal but leads to the same decay rate as the optimal Kx ).
Channel
   
2×2 MIMO Rayleigh SNR
Channel
iid
pout (r log SNR) = Pr log det Int + HH∗ < r log SNR .
General MIMO i.i.d.
nt
Rayleigh Channel
Optimal Tradeoff • DMT for the MIMO channel with i.i.d. Rayleigh fading can be
Geometric
Interpretation shown to be a piecewise linear curve joining the points

(k, (nt − k)(nr − k)), for k = 0, . . . , nmin .

16 / 19
General MIMO i.i.d. Rayleigh Channel
– Optimal Tradeoff

Lecture 9
DMT

Ragnar Thobaben
ISE/EECS/KTH

Formulation

Scalar Rayleigh
Channel

Parallel Rayleigh
Channel

MISO Rayleigh
Channel

2×2 MIMO Rayleigh


Channel

General MIMO i.i.d. • Left: optimal DMT for the MIMO channel with i.i.d. Rayleigh
Rayleigh Channel
Optimal Tradeoff
fading.
Geometric
Interpretation • Right: adding one transmit and receive antenna moves the entire
DMT curve by 1 (and not only the extreme point rmax ).
• Optimal DMT can be achieved by space-time codes with length
l = nt + nr − 1.

17 / 19
General MIMO i.i.d. Rayleigh Channel
– Geometric Interpretation
• Outage probability
   
SNR
Lecture 9
DMT
pout (r log SNR) = Pr log det Int + HH∗ < r log SNR
nt
Ragnar Thobaben (n )
ISE/EECS/KTH min  
X SNR 2
= Pr log 1 + λi < r log SNR .
Formulation
i=1
nt
Scalar Rayleigh
Channel
with
• Random singular values λi of the matrix H;
Parallel Rayleigh
Channel
• nmin eigenmodes with effective SNR, SNRλ2i /nt .

MISO Rayleigh • Modes are ...


Channel
• active if SNRλ2i /nt is in the order of SNR;
2×2 MIMO Rayleigh
Channel
• inactive if SNRλ2i /nt is in the order of 1 or smaller;
General MIMO i.i.d.
Rayleigh Channel • Case 1, r → 0: Outage happens if all modes are inacative (i.e.,
Optimal Tradeoff SNRλ2i /nt ≤ 1); happens if H is close to the zero matrix.
Geometric
Interpretation

→ Outage probability is in the order 1/SNRnr nt . 18 / 19


General MIMO i.i.d. Rayleigh Channel
– Geometric Interpretation
Case 2, r is a positive integer.
• For high SNR, it can be shown that a typical outage event is
Lecture 9
DMT characterized by r modes being fully effective and the remaining
Ragnar Thobaben modes being fully ineffective.
ISE/EECS/KTH
• Outage happens if H is close to the “space” Vr of all rank-r
Formulation matrices.
Scalar Rayleigh
Channel

Parallel Rayleigh
Channel

MISO Rayleigh
Channel

2×2 MIMO Rayleigh


Channel • A rank-r matrix H is described by rnt + (nr − r )r parameters:
General MIMO i.i.d. • rnt parameters to specify nr linearly independent row vectors.
Rayleigh Channel • (nr − r )r parameters to specify the remaining row vectors as linear
Optimal Tradeoff
Geometric
combinations.
Interpretation • Vr is rnt + (nr − r )r dimensional, and the “orthogonal” space is
nt nr − (rnt + (nr − r )r ) = (nt − r )(nr − r )
dimensional, which is precisely the SNR exponent of the outage
probability.
• The outage probability is proportional to the probability that
(nt − r )(nr − r ) parameters collapse (i.e., |hi,j |2 < 1/SNR).
19 / 19

You might also like