List of Errata to

Massive MIMO Networks: Spectral, Energy, and Hardware Efficiency

Updated: February 26, 2020.

This documents lists typos detected in the published manuscript of:

Emil Björnson, Jakob Hoydis and Luca Sanguinetti (2017), “Massive MIMO Networks: Spectral, Energy,
and Hardware Efficiency”, Foundations and Trends R in Signal Processing: Vol. 11: No. 3-4, pp 154-655.
http://dx.doi.org/10.1561/2000000093

This document will be updated if further typos are discovered and the latest version is available at:
https://massivemimobook.com

All the typos listed in this document have been corrected in the authors’ version of the manuscript, which
you can download as a free PDF from https://massivemimobook.com.

If you discover a potential typo that is not listed in the latest version of this document, we appreciate if
you report it to us. You can send an email to emil.bjornson@liu.se

List of Errata
1. In the paragraph before Eq. (1.23), “horizontal ULA with antenna spacing dH ” should be “horizontal
ULA with antenna spacing dH ∈ (0, 0.5]”.

2. In Section 1.3.5, the statement “the same as that of sending M additional UL pilot signals” is only true
if M ≥ K. To make it more accurate, the statement should instead be “the same as that of sending
max(M, K) additional UL pilot signals” and the following sentence should be added to the footnote on
the same page: “More precisely, with the multiplexing gain min(M, K) of SDMA, we need max(M, K)
symbol transmissions to feed back the M K channel coefficients.”
Consequently, the average pilot overhead of the FDD protocol is M +K+max(M,K)
2 and not M + K 2
(which is only correct for M ≥ K). This error is found at several places in this section. Moreover,
Figure 1.22 is shown as

50
Number of UEs (K)

40
FDD
30
TDD
20

10
Shaded area: M> 4K
0
0 20 40 60 80 100
Number of antennas (M)

but should be
 50

Number of UEs (K)

40

30
TDD
20
FDD
10
Shaded area: M> 4K
0
0 20 40 60 80 100
Number of antennas (M)

2

2


tr (Rjjk ) tr (Rjjk )
3. In Eq. (2.17), the final expression j 2
(Mj βlk )
should be j 2
(Mj βjk )
.

4. Figure 2.6 is shown as

10

Reference: Uncorrelated fading

0
Normalized eigenvalue [dB]

−10

−20

−30

Laplace
−40 Uniform
Gaussian
−50
0 20 40 60 80 100
Eigenvalue number in decreasing order

but should be

10

Reference: Uncorrelated fading

0
Normalized eigenvalue [dB]

Laplace
−10 Uniform
Gaussian

−20

−30

−40

−50
0 20 40 60 80 100
Eigenvalue number in decreasing order
 This error was a consequence of insufficient accuracy in the computation of the covariance matrices.
The related sentence
“In fact, a uniform angular distribution makes 66% of the eigenvalues 50 dB smaller than in the reference
case, while this happens for around 40% of the eigenvalues with Gaussian and Laplace distributions.”
should read as
“In fact, a uniform angular distribution makes 68% of the eigenvalues 30 dB smaller than in the reference
case, while this happens for 40% of the eigenvalues with Gaussian distribution and 19% with Laplace
distribution.”

5. In the first paragraph of Section 3.3.3, hjli ∼ NC (0M , Rjli ) should be hjli ∼ NC (0Mj , Rjli ).

6. In Eq. (3.36), IM should be IMj . This typo also appears on the row right above Eq. (3.37).

 −1  −1
7. In Eq. (3.38), 1
pli Ψjli should be 1
pli τp Ψjli .

8. In the paragraph after Theorem 4.6, the statement “reduce the number of samples τd used for DL
data” should be “reduce the number of samples τu used for UL data”.

2 2
9. In Eq. (4.29), σUL should be σDL .

10. In Section 5.3, the statement “The efficiency of a PA is defined as the ratio of input power to output
power” should be “The efficiency of a PA is defined as the ratio of output power to input power”.

11. In Section 6.1.2, after Eq. (6.7), the statement “LTE only requires EVM ≤ 0.0175” should be “LTE
only requires EVM ≤ 0.175”

12. The subsection title “7.4.1 Physical Array Size and Antenna Spacing” should be “7.4.1 Preliminaries
on Physical Array Size”.

13. Figure 7.26 is shown as

0.25
Cell radius: 350 m
Cell radius: 100 m
0.2
Variance in (2.17)

0.15

0.1

0.05

0
1 2 4 8 16 32 64 128 256
Number of subarrays

but should be
 0.5
Cell radius: 350 m
Cell radius: 100 m
0.4

Variance in (2.17) 0.3

0.2

0.1

0
1 2 4 8 16 32 64 128 256
Number of subarrays

14. Figure 7.27 is shown as

1
S =1
S =16
0.8 S =256

0.6
CDF

0.4

0.2

0
−120 −100 −80 −60 −40 −20
Average channel gain [dB]

but should be

1
S =1
S =16
0.8 S =256

0.6
CDF

0.4

0.2

0
−160 −140 −120 −100 −80 −60
Average channel gain [dB]
 In the paragraph that describes this figure, the sentence “However, going from S = 1 to S = 16
improves the median of β by around 9 dB; increasing this number to S = 256 adds another 5 dB.”
should read as: “However, going from S = 1 to S = 16 improves the median of β by around 12.5 dB;
increasing this number to S = 256 adds another 9.5 dB.”
15. The proof of Lemma B.14 only holds when B is a Hermitian matrix. A correct proof that applies to
any deterministic B is:
1
“Note that a = A 2 w for w = [w1 . . . wN ]T ∼ NC (0N , IN ), thus we can write
1 1
E{|aH Ba|2 } = E{|wH (AH ) 2 BA 2 w|2 } = E{|wH Cw|2 } (B.23)
1 1
where we defined C = (AH ) 2 BA 2 . Let cn1 ,n2 denote the element in C on row n1 and column n2 .
Using this notation, we can expand (B.23) as
N
X N
X N
X N
X
2
H
E{|w Cw| } = E{wn? 1 cn1 ,n2 wn2 wm1 c?m1 ,m2 wm
?
2
}
n1 =1 n2 =1 m1 =1 m2 =1

N N X
N
(a) X X
= E{|wn |4 }|cn,n |2 + E{|wn |2 }E{|wm |2 }cn,n c?m,m
n=1 n=1 m=1
m6=n

N
X N
X
+ E{|wn1 |2 }E{|wn2 |2 }|cn1 ,n2 |2
n1 =1 n2 =1
n2 6=n1

N N X
N N N
(b) X X X X
= 2|cn,n |2 + cn,n c?m,m + |cn1 ,n2 |2
n=1 n=1 m=1 n1 =1 n2 =1
m6=n n2 6=n1

N
X N
X N
X N
X
= cn,n c?m,m + |cn1 ,n2 |2
n=1 m=1 n1 =1 n2 =1
(c)
= |tr(C)|2 + tr(CCH ) (B.24)

where (a) utilizes that circular symmetry implies that E{wn? 1 wn2 wm1 wm
?
2
} is only non-zero when the
terms with conjugates have matching indices to the terms without conjugates. The first expression is
given by n1 = n2 = m1 = m2 , the second term is given by n1 = n2 and m1 = m2 with n1 6= m1 , and
the third term is given by n1 = m1 and n2 = m2 with n1 6= n2 . In (b), we utilize that E{|wn |2 } = 1 and
E{|wn |4 } = 2. In (c), we write the sums of elements in C using the trace. The resulting expression is
equivalent to (??), which is shown by replacing C with A and B and utilizing the fact that tr(C1 C2 ) =
tr(C2 C1 ) for any matrices C1 , C2 such that C1 and CT2 have the same dimensions.”
16. The second paragraph of Definition B.8 should read as follows:
“If the random variable x, with support in X and PDF f (x), is given and the conditional PDF is
f (y|x), then the conditional differential entropy is
Z Z


H(y|x) = − log2 f (y|x) f (y|x)f (x)dxdy.” (B.57)
Y X

17. In the paragraph after (C.63), A = τp Ψjjk should be A = τp (Ψjjk )−1 .