An Alternating Algorithm for Uplink Max-Min SINR in Cell-Free Massive MIMO with Local-MMSE Receiver

K. B. Shashika Manosha; M. Latva-aho; N. Rajatheva; W. A. Chamalee Wickrama Arachchi

REVIEW 4 major objections 5 minor 16 references

An Alternating Algorithm for Uplink Max-Min SINR in Cell-Free Massive MIMO with Local-MMSE Receiver

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read An alternating algorithm that jointly optimizes transmit powers and CPU weighting coefficients raises the minimum user SINR in cell-free massive MIMO with local-MMSE combining, beating the fixed-power weighting baseline.

desk verdict A plausible, cleanly-presented alternating algorithm for max-min SINR in cell-free massive MIMO with L-MMSE, but the power-update step optimizes an unverified approximation and the paper's central empirical claim rests on that approximation. read the letter →

arxiv 1908.03187 v3 pith:4KUKXRHK submitted 2019-08-08 cs.IT eess.SPmath.IT

W. A. Chamalee Wickrama Arachchi , K. B. Shashika Manosha , N. Rajatheva , M. Latva-aho This is my paper · ORCID

classification cs.ITeess.SPmath.IT

keywords cell-freemassiveMIMOmax-minSINRlocal-MMSEcombininggeometricprogramminggeneralizedeigenvalueproblemuplinkpowercontroluserfairnessspectralefficiency

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

In a cell-free massive MIMO uplink, the users with the worst channels set the network's effective rate, so fairness is the performance bottleneck. The paper attacks this by maximizing the minimum SINR across users, optimizing both the transmit power of each UE and the weighting coefficients applied at the central processing unit, under per-UE power limits. Because the combined problem is not jointly convex, the authors split it into two subproblems: weighting coefficients come from a generalized eigenvalue problem, and power allocation is approximated as geometric programming. An alternating algorithm cycles between these steps, and numerical experiments show its min-user uplink spectral efficiency exceeds that of the fixed-power weighting scheme used as a baseline. The stated result is empirical, with convergence demonstrated only through simulations.

What carries the argument

The load-bearing object is the effective SINR expression (11a)-(11c), written as a generalized Rayleigh quotient in the CPU weighting vector $a_k$; this is what turns weighting design into a generalized eigenvalue problem for the matrix pair $(p_k\mathbb{E}\{g_{kk}\}\mathbb{E}\{g_{kk}\}^H, B_k)$. Since the numerator matrix is rank-one, the optimal weight takes the closed form $a_k = B_k^{-1}\mathbb{E}\{g_{kk}\}$. The power-allocation half relies on the approximation (19), which removes the desired-signal term $-p_k\mathbb{E}\{g_{kk}\}\mathbb{E}\{g_{kk}\}^H$ from the denominator; the resulting constraint is a posynomial inequality, so the power subproblem becomes a geometric program. The alternating execution of these two steps, under the L-MMSE local combiners of (8), is the mechanism that produces the reported min-user SE gains.

What would settle it

Run Algorithm 1 on many correlated Rayleigh fading user drops, then evaluate the true SINR (15) at the returned powers and weights: if the worst-user SINR at the output ever falls below the fixed-power baseline of [9] in a tested regime, or if the approximate SINR (19) deviates substantially from the true SINR at those points, the central claim is refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that the nonconvex max-min SINR problem for an uplink cell-free massive MIMO system with local-MMSE combining can be handled by alternating two subproblems, and that the resulting algorithm improves the worst-user experience. With transmit powers held fixed, the effective SINR of each user is a generalized Rayleigh quotient in the CPU weighting vector $a_k$, so the optimal weights are obtained as the generalized eigenvector $a_k = B_k^{-1}\mathbb{E}\{g_{kk}\}$, where $B_k = \sum_i p_i \mathbb{E}\{g_{ki}g_{ki}^H\} - p_k \mathbb{E}\{g_{kk}\}\mathbb{E}\{g_{kk}\}^H + \sigma^2 D_k$. With the weights held fixed, the power subproblem is converted into a geometric program by using the approximate SINR (19), which drops the term $-p_k \mathbb{E}\{g_{kk}\}\mathbb{E}\{g_{kk}\}^H$ from the denominator. Algorithm 1 iterates these two updates, and the numerical results show that the empirical CDF of the min-user uplink spectral efficiency sits to the right of the fixed-power weighting baseline of [9] across pilot reuse factors and for both L-MMSE and MR combining, with convergence observed within a few iterations.

Load-bearing premise

The power-allocation step assumes that the approximated SINR (19), which drops the desired-signal energy term from the denominator, is accurate enough that maximizing it also raises the true minimum SINR (15); the paper provides no bound on the gap between the approximate and true SINR.

Editorial extensions

If this is right

The proposed alternating update gives higher min-user uplink SE than the fixed-power weighting scheme in the simulated cell-free setups with $L=100$, $K=40$, $N=4$ and $L=64$, $K=16$.
The improvement persists across pilot reuse factors $f=1,2,4$ and for both L-MMSE and MR combining, with orthogonal pilots producing the largest min-user SE.
The algorithm converges to a suboptimal solution within a few iterations (empirically after around two iterations in the tested configurations), so the extra computation over the fixed-power scheme is modest.
Because the power subproblem is a GP with $O(K^{7/2})$ per-iteration flops, the added cost of optimizing powers rather than fixing them scales polynomially in the number of users.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A direct test of the power-step approximation would replace (19) with the exact SINR (11c) and rerun Algorithm 1; comparing the two would isolate how much of the gain comes from the approximation rather than from jointly updating powers and weights.
The method's reliance on the approximate SINR suggests its advantage is most likely to hold in interference-limited regimes; in noise-limited or highly correlated channels, the omitted desired-signal term could matter more.
Because the paper demonstrates convergence only numerically, extending the analysis to a monotonicity or fixed-point argument on the true SINR would turn the empirical convergence into a guarantee.
The same generalized-eigenproblem-plus-GP split may transfer to other linear receivers or to downlink fairness problems, where similar quadratic SINR expressions appear.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: this is a conventional but genuinely useful optimization paper. The new thing is that it jointly optimizes transmit powers and CPU weighting coefficients for the uplink of a cell-free massive MIMO system with L-MMSE receivers. That combination is absent from the fixed-power weighting optimization of [9] and from earlier max-min work, so the paper does fill a small, real gap in the literature. The weighting-coefficient subproblem is cleanly derived as a generalized eigenvalue problem, and the alternating structure is natural. The complexity analysis is reasonable. The simulations compare against the fixed-power scheme and show consistent gains across pilot-reuse factors, which is the intended empirical contribution. I believe the paper as written does establish that the proposed algorithm improves min-user SE relative to that one benchmark.

The soft spots are concentrated in the power-allocation step, and they are proportionate to the paper's central claim. Equation (19) approximates the true SINR by deleting a nonnegative term from the denominator, so the GP is maximizing an upper bound on SINR. The paper gives no error bound and no fixed-point analysis connecting the GP solution to the true max-min SINR. The convergence section is also thin: five channel realizations, six iterations, no formal convergence argument. That is enough to suggest a trend, not to prove the method converges. A more serious concern is the Appendix: moving from (22) to (23) defines c_k without the sigma^2 factor that appears in the noise term of (22), so the published GP constraint does not even match the approximate SINR (19). That looks like a typo or an omitted factor, but it matters because a reader cannot reproduce the GP step as written.

I also note the paper does not compare against the closest max-min baseline, [7], which is a missed check: the abstract's claim of "higher min-user uplink SE" is only against fixed-power weighting, not against an existing max-min approach. That is a minor omission, not a fatal one.

On balance, the paper is honest and the core idea is sound enough to warrant engagement. The weighting step is exact and the approximation, while unquantified, is a known and fairly standard move in this subfield. I would send this to referees with a clear request to check the GP derivation and the missing sigma^2 factor, and to add either an error bound or an exact-SINR simulation. The paper is for researchers working on cell-free massive MIMO and practical resource allocation, and they will find it worth reading despite the rough edges. Recommended for peer review, conditional on those fixes.

Referee Report

4 major / 5 minor

Summary. The paper considers uplink max-min SINR optimization in a cell-free massive MIMO system with local MMSE combining. The authors decompose the nonconvex joint problem into a weighting-coefficient subproblem, solved exactly for fixed powers as a generalized eigenvalue problem, and a power-allocation subproblem, approximated as geometric programming using the SINR approximation in Eq. (19). An alternating algorithm (Algorithm 1) is proposed, and simulation results compare the minimum uplink SE against the fixed-power weighting scheme of [9]. The central empirical claim is that the proposed algorithm achieves higher min-user uplink SE than the fixed-power benchmark, and that it converges within a few iterations.

Significance. If the central claim is established, the paper offers a practical, low-complexity fairness mechanism for cell-free massive MIMO with local MMSE combining, going beyond the fixed-power weighting optimization of [9]. The weighting subproblem is cleanly derived and solved, and the overall algorithmic structure is plausible. The paper also includes useful numerical comparisons across pilot reuse factors and receiver types. However, the load-bearing power-update step relies on an unquantified approximation of the SINR, and the published GP constraint contains a missing noise-power factor. Because the main comparison is with a benchmark that is not optimized over powers, the validity of the empirical claim depends on closing this gap; the convergence evidence is also too limited to support a general conclusion. With corrections and additional verification, this would be a useful contribution.

major comments (4)

[Section III-B, Eq. (19)] The power allocation subproblem is solved using the approximate SINR (19), which is an upper bound on the exact SINR (11c): the approximation drops the nonnegative self-interference term p_k a_k^H (E{g_kk g_kk^H} - E{g_kk}E{g_kk}^H) a_k from the denominator. No error bound or fixed-point analysis is provided to show that increasing the approximate SINR increases the exact minimum SINR. As a result, the GP steps may not actually maximize the true worst-user SINR, and the empirical comparison against the fixed-power benchmark could be optimistic. Please quantify the approximation error for the simulation settings, or modify the algorithm to evaluate exact SINR, and report both the approximate and exact min-user SINR in the convergence and SE curves.
[Appendix A, Eq. (23)] The GP constraint as printed does not match even the approximate SINR (19). The noise contribution in (22) is sigma^2 a_k^H D_k a_k, but after rearrangement c_k is defined without the sigma^2 factor, so (23) uses c_k/p_k instead of sigma^2 c_k/p_k. Since sigma^2 is not equal to 1 in the simulations (noise power is -96 dBm while powers are in mW), this is numerically significant and makes the published formulation inconsistent with the stated approximation. Please correct the definition of c_k or the GP constraint.
[Section V-C, Figs. 6 and 7] The convergence claim is supported only by five specific channel realizations and six iterations, with no statistical aggregation or theoretical monotonicity result. This does not establish that the alternating algorithm converges for typical realizations. Please provide convergence results over a larger sample, with averaged curves and variation ranges, or a proof that the sequence of min-user SE values is nondecreasing.
[Proposition 1 and Problem (18)] Proposition 1 shows only that the approximate SINR constraint can be written as a posynomial constraint, not that the GP solution is related to the original max-min problem (15). This distinction should be stated explicitly, and the numerical results should clarify whether the displayed min-user SE is computed from the exact SINR (11c) or from the approximate SINR (19). If the displayed SE uses the approximation, the comparison with fixed-power scheme is not against the true SE.

minor comments (5)

[Section II, Eq. (6)] The noise vector n_l is stated as CN(0, p_i), which appears to be a typo; it should be CN(0, sigma^2 I_N).
[Section II, Eq. (11)] The definition of D_k as a tuple (E{||v_k1||^2}, ..., E{||v_kL||^2}) in C^{L x L} should be a diagonal matrix, since it is used in the form a_k^H D_k a_k.
[Section IV] The statement that the overall complexity is O(K^{7/2}) flops per iteration ignores the weighting subproblem complexity O(KL^3) stated in the same section; the per-iteration complexity should be given as a sum of both terms.
[Algorithm 1] The stopping criterion 'until required accuracy' is vague; please define a precise tolerance or a maximum number of iterations.
[References] Several references, e.g., [6] and [12], have incomplete or inconsistent author names and accents; please standardize them.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the SINR model and fixed-power benchmark are external, the GP step is an explicit approximation rather than a self-reduction, and no fitted parameter is renamed as a prediction.

full rationale

The paper's derivation chain is self-contained and does not reduce to its own inputs. The SINR expression in (11) and the optimal weighting vector in (12) are taken from external references [9] and [12], not from the authors' prior work, and the fixed-power comparison baseline is also [9, Corollary 2]. The power-allocation subproblem is formulated by explicitly approximating the exact SINR in (11c) as (19), dropping the nonnegative term p_k E{g_kk}E{g_kk}^H from the denominator; this is an approximation with an unquantified gap, which is a correctness risk, but it is not circular because the approximate objective is not identical by construction to the quantity being optimized. There is no fitted parameter whose fitted value is later called a prediction: Algorithm 1 is evaluated empirically against an external benchmark. The only self-referential element, the numerical convergence study in Section V-C, illustrates the behavior of the algorithm rather than importing a load-bearing claim from the authors' own prior results. Appendix A contains an apparent algebraic slip when moving from (22) to (23), where c_k omits the sigma^2 factor present in (22), but this is an internal consistency issue, not a reduction of the paper's conclusion to its assumptions. Accordingly, no circular step is exhibited, and the circularity score is 0.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. The optimization variables are transmit powers and CPU weights, which are standard degrees of freedom in the system. The approximation in (19) is a modeling simplification, not a new entity. Simulation constants (noise power -96 dBm, pilot reuse factors, path-loss parameters, angular spread 15 degrees) are taken from [9] and standard 3GPP models, not fitted to make the derivation work.

assumptions (4)

domain assumption Correlated Rayleigh block fading with Gaussian local scattering (Eq. 1) and MMSE channel estimation with pilot contamination (Eqs. 2-5).
Invoked in Section II-A; standard modeling assumption in cell-free massive MIMO taken from prior literature, not derived here.
domain assumption CPU has knowledge of channel statistics (E{g_ki g_ki^H}, D_k) and uses them to compute SINR and optimize weights (Eq. 11).
Used throughout Sections II-B and III; this statistical knowledge is what makes the generalized Rayleigh quotient formulation valid.
ad hoc to paper The SINR approximation in Eq. (19), which deletes the desired-signal term from the denominator, is accurate enough for power optimization.
Introduced in Section III-B to make the power subproblem a GP; no error bound or validation against the exact SINR is provided.
ad hoc to paper The alternating algorithm converges to a useful suboptimal fixed point.
Assumed by Algorithm 1 and supported only by numerical examples in Section V-C (5 channel realizations, 6 iterations); no proof of convergence is given.

how reviews work

0 comments

Cite this review

Pith. "Pith review of An Alternating Algorithm for Uplink Max-Min SINR in Cell-Free Massive MIMO with Local-MMSE Receiver." pith.science (2026). https://pith.science/paper/4KUKXRHK

@misc{pith2026190803187,
  author       = {Pith},
  title        = {Pith review of: An Alternating Algorithm for Uplink Max-Min SINR in Cell-Free Massive MIMO with Local-MMSE Receiver},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4KUKXRHK}},
  note         = {Machine review of arXiv:1908.03187}
}

read the original abstract

The problem of max-min signal-to-interference plus noise ratio (SINR) for uplink transmission of cell-free massive multiple-input multiple-output (MIMO) system is considered. We assume that the system is employed with local minimum mean square error (L-MMSE) combining. The objective is to preserve user fairness by solving max-min SINR optimization problem, by optimizing transmit power of each user equipment (UE) and weighting coefficients at central processing unit (CPU), subject to transmit power constraints of UEs. This problem is not jointly convex. Hence, we decompose original problem into two subproblems, particularly for optimizing power allocation and receiver weighting coefficients. Then, we propose an alternating algorithm to solve these two subproblems. The weighting coefficient subproblem is formulated as a generalized eigenvalue problem while power allocation subproblem is approximated as geometric programming (GP). We empirically show that the proposed algorithm achieves higher min-user uplink spectral efficiency (SE) over existing fixed power scheme which is not optimized with respect to the transmit power. Moreover, the convergence of the proposed algorithm is numerically illustrated.

Figures

Figures reproduced from arXiv: 1908.03187 by the authors.

**Figure 2.** Basic idea of the proposed algorithm. Algorithm 1 1. Initialize p (0) = [p (0) 1 , p (0) 2 , · · · , p (0) K ], i = 0 2. Set i = i + 1 3. Set p (i) = p (i−1) and find the optimal weighting coefficients a (i) = [a (i) 1 , a (i) 2 , · · · , a (i) K ] through solving the generalized eigenvalue Problem (16) 4. Compute p (i) through solving Problem (18) 5. Go back to Step 2 and repeat until required accuracy max pk,ak mi… view at source ↗

**Figure 3.** The cumulative distribution of min-user uplink SE with proposed and fixed power schemes ( [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: The cumulative distribution of min-user uplink SE with different pilot reuse factors ( [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: The comparison of cumulative distributions of min-user uplink SE using L-MMSE and MR, with random pilots for [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: The convergence of the proposed algorithm over different set of channel realizations for [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

**Figure 7.** Figure 7: The convergence of the proposed algorithm over same set of channel realizations for [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

16 extracted references · 14 canonical work pages

[9]

Making cell-free massive MIMO competitive with MMSE processing and centralized implementation,

E. Bj ¨ornson and L. Sanguinetti, “Making cell-free massive MIMO competitive with MMSE processing and centralized implementation,” IEEE Trans. Wireless Commun., 2019, To appear

work page 2019
[7]

Enhanced max-min sinr for uplink cell-free massive mimo systems,

M. Bashar, K. Cumanan, A. G. Burr, M. Debbah, and H. Q. Ngo, “Enhanced max-min sinr for uplink cell-free massive mimo systems,” in 2018 IEEE International Conference on Communications (ICC) , 2018, pp. 1–6

work page 2018
[1]

Noncooperative cellular wireless with unlimited numbers of base station antennas,

T. L. Marzetta, “Noncooperative cellular wireless with unlimited numbers of base station antennas,” IEEE Trans. Wireless Commun. , vol. 9, no. 11, pp. 3590–3600, 2010

2010
[2]

The viability of distributed antennas for massive mimo systems,

K. T. Truong and R. W. Heath, “The viability of distributed antennas for massive mimo systems,” in 2013 Asilomar Conference on Signals, Systems and Computers, 2013, pp. 1318–1323

work page 2013
[3]

Coordinated multipoint: Concepts, performance, and ﬁeld trial results,

R. Irmer, H. Droste, P. Marsch, M. Grieger, G. Fettweis, S. Brueck, H. Mayer, L. Thiele, and V . Jungnickel, “Coordinated multipoint: Concepts, performance, and ﬁeld trial results,” IEEE Communications Magazine , vol. 49, no. 2, pp. 102–111, 2011

work page 2011
[4]

Cooperative distributed antenna systems for mobile communications [coordinated and distributed mimo],

X. You, D. Wang, B. Sheng, X. Gao, X. Zhao, and M. Chen, “Cooperative distributed antenna systems for mobile communications [coordinated and distributed mimo],” IEEE Wireless Communications, vol. 17, no. 3, pp. 35–43, 2010

work page 2010
[5]

Cell-free massive mimo versus small cells,

H. Q. Ngo, A. Ashikhmin, H. Yang, E. G. Larsson, and T. L. Marzetta, “Cell-free massive mimo versus small cells,” IEEE Trans. Wireless Commun. , vol. 16, no. 3, pp. 1834–1850, 2017

work page 2017
[6]

Ubiquitous cell-free massive MIMO communications,

G. Interdonato, E. Bj ¨ornson, H. Q. Ngo, P ˚al K. Frenger, and E. G. Larsson, “Ubiquitous cell-free massive MIMO communications,” EURASIP J. Wirel. Commun. Netw., 2019

work page 2019

Show all 16 references

[8]

A new look at cell-free massive mimo: Making it practical with dynamic cooperation,

E. Bjrnson and L. Sanguinetti, “A new look at cell-free massive mimo: Making it practical with dynamic cooperation,” 2019, [Online]. Available: https://arxiv.org/abs/1906.10853

2019 arXiv
[10]

G. H. Golub and Charles F. V . Loan, Matrix Computations (3rd Ed.) , Johns Hopkins University Press, Baltimore, MD, USA, 1996

1996
[11]

Boyd and L

S. Boyd and L. Vandenberghe, Convex Optimization, Cambridge University Press, Cambridge, UK, 2004

2004
[12]

Massive mimo networks: Spectral, energy, and hardware efﬁciency,

E. Bjrnson, J. Hoydis, and L. Sanguinetti, “Massive mimo networks: Spectral, energy, and hardware efﬁciency,” Foundations and TrendsⓇ in Signal Processing, vol. 11, no. 3-4, pp. 154–655, 2017, [Online]. Available: http://dx.doi.org/10.1561/2000000093

2017 doi
[13]

L. N. Trefethen and D. Bau, Numerical Linear Algebra , SIAM, 1997

1997
[14]

Families of algorithms related to the inversion of a symmetric positive deﬁnite matrix,

P. Bientinesi, B. Gunter, and R. A. van de Geijn, “Families of algorithms related to the inversion of a symmetric positive deﬁnite matrix,” ACM Trans. Math. Softw., vol. 35, no. 1, 2008

2008
[15]

A tutorial on geometric programming,

S. Boyd, Seung-Jean Kim, Lieven Vandenberghe, and A. Hassibi, “A tutorial on geometric programming,” Optimization and Engineering , vol. 8, pp. 67–127, 2007

2007
[16]

Nesterov and A

Y . Nesterov and A. Nemirovskii, Interior-Point Polynomial Algorithms in Convex Programming , lecture notes, 1996, [Online]. Available: https: //www2.isye.gatech.edu/∼nemirovs/Lect IPM.pdf. 11

1996

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [9]

Making cell-free massive MIMO competitive with MMSE processing and centralized implementation,

E. Bj ¨ornson and L. Sanguinetti, “Making cell-free massive MIMO competitive with MMSE processing and centralized implementation,” IEEE Trans. Wireless Commun., 2019, To appear

work page 2019

[2] [7]

Enhanced max-min sinr for uplink cell-free massive mimo systems,

M. Bashar, K. Cumanan, A. G. Burr, M. Debbah, and H. Q. Ngo, “Enhanced max-min sinr for uplink cell-free massive mimo systems,” in 2018 IEEE International Conference on Communications (ICC) , 2018, pp. 1–6

work page 2018

[3] [1]

Noncooperative cellular wireless with unlimited numbers of base station antennas,

T. L. Marzetta, “Noncooperative cellular wireless with unlimited numbers of base station antennas,” IEEE Trans. Wireless Commun. , vol. 9, no. 11, pp. 3590–3600, 2010

2010

[4] [2]

The viability of distributed antennas for massive mimo systems,

K. T. Truong and R. W. Heath, “The viability of distributed antennas for massive mimo systems,” in 2013 Asilomar Conference on Signals, Systems and Computers, 2013, pp. 1318–1323

work page 2013

[5] [3]

Coordinated multipoint: Concepts, performance, and ﬁeld trial results,

R. Irmer, H. Droste, P. Marsch, M. Grieger, G. Fettweis, S. Brueck, H. Mayer, L. Thiele, and V . Jungnickel, “Coordinated multipoint: Concepts, performance, and ﬁeld trial results,” IEEE Communications Magazine , vol. 49, no. 2, pp. 102–111, 2011

work page 2011

[6] [4]

Cooperative distributed antenna systems for mobile communications [coordinated and distributed mimo],

X. You, D. Wang, B. Sheng, X. Gao, X. Zhao, and M. Chen, “Cooperative distributed antenna systems for mobile communications [coordinated and distributed mimo],” IEEE Wireless Communications, vol. 17, no. 3, pp. 35–43, 2010

work page 2010

[7] [5]

Cell-free massive mimo versus small cells,

H. Q. Ngo, A. Ashikhmin, H. Yang, E. G. Larsson, and T. L. Marzetta, “Cell-free massive mimo versus small cells,” IEEE Trans. Wireless Commun. , vol. 16, no. 3, pp. 1834–1850, 2017

work page 2017

[8] [6]

Ubiquitous cell-free massive MIMO communications,

G. Interdonato, E. Bj ¨ornson, H. Q. Ngo, P ˚al K. Frenger, and E. G. Larsson, “Ubiquitous cell-free massive MIMO communications,” EURASIP J. Wirel. Commun. Netw., 2019

work page 2019