Massive MIMO in the UL/DL of Cellular Networks: How Many Antennas Do We Need?
read more
Citations
What Will 5G Be
Massive MIMO for next generation wireless systems
Energy and Spectral Efficiency of Very Large Multiuser MIMO Systems
Energy and Spectral Efficiency of Very Large Multiuser MIMO Systems
An Overview of Massive MIMO: Benefits and Challenges
References
Fundamentals of statistical signal processing: estimation theory
Noncooperative Cellular Wireless with Unlimited Numbers of Base Station Antennas
Scaling Up MIMO: Opportunities and Challenges with Very Large Arrays
Energy and Spectral Efficiency of Very Large Multiuser MIMO Systems
How much training is needed in multiple-antenna wireless links?
Related Papers (5)
Frequently Asked Questions (17)
Q2. What future works have the authors mentioned in the paper "Massive mimo in the ul/dl of cellular networks: how many antennas do we need?" ?
These approximations were shown to be accurate for realistic system dimensions and enable, consequently, future studies of realistic effects, such as antenna correlation, spacing and aperture, without the need for simulations. Simulations for a more realistic system model suggest that MMSE/RZF can achieve the performance of the simple MF/BF schemes with a significantly reduced number of antennas. Since massive MIMO TDD systems are a promising network architecture, it seems necessary to verify the theoretical performance predictions by channel measurements and prototypes.
Q3. How many UTs are needed to achieve a given SNR?
4. Degrees of freedom per UT P/K necessary to achieve ηR∞ versus effective SNR ρN for L = 4 and α = 0.1.we need about P/K = 90 DoF per UT with MF/BF to achieve 90% of the ultimate performance R∞, i.e., 0.9×2.2 ≈ 2 b/s/Hz.
Q4. How do the authors distribute the UTs in the circle?
To allow for reproducibility of their results, the authors distribute K = 10 UTs uniformly on a circle of radius 2/3 around each BS and do not consider shadowing.
Q5. How many antennas can be used to achieve the same performance?
If the number of antennas is doubled, the transmit power can consequently be reduced by a factor two to achieve the same performance.
Q6. What is the UL/DL performance of a linear detector?
Assuming a large system limit, the authors have derived asymptotically tight approximations of achievable UL/DL-rates under a very general channel model which accounts for imperfect channel estimation, pilot contamination, path loss, and terminal-specific antenna correlation.
Q7. What is the BS's response to the training signal?
After correlating the received training signal with the pilot sequence of UT k, the jth BS estimates the channel vector hjjk based on the3 observation ytrjk ∈ C N , given as1ytrjk = hjjk + ∑ l ̸=j hjlk + 1 √ ρtr ntrjk (6)where ntrjk ∼ CN (0, IN ) and ρtr > 0 is the effective training SNR.
Q8. What is the reason to speak about a massive MIMO effect?
Based on their previous observations, it is justified to speak about a massive MIMO effect whenever the SINR γjm (in the UL or DL) is close to γ∞, or in other words, whenever noise, channel estimation errors, and interference are small compared to pilot contamination.
Q9. How many UTs could be served simultaneously?
With MMSE/RZF, only 35 DoF per UT are necessary to achieve the same performance and, consequently, 90/35 ≈ 2.5 times more UTs could be simultaneously served.
Q10. What is the inverse path loss of a UT to a BS?
(2) can represent a physical channel model with a fixed number of dimensions or angular bins P as in [8], by letting R̃jlk = √ ℓjlk [A 0N×N−P ], where A ∈ CN×P , 0N×N−P is the N × (N −P ) zero matrix, and ℓjlk denotes the inverse path loss from UT k in cell l to BS j.
Q11. how many antennas are needed to achieve % of the ultimate performance limit?
the authors have determined how many antennas are needed to achieve η % of the ultimate performance limit with infinitely many antennas and how many more antennas are needed with MF/BF to achieve MMSE/RZF performance.
Q12. What does the effect of adding more antennas on the SNR of UTs be?
In particular, if P saturates for some N , adding more antennas increases the effective SNR but does not reduce the multiuser interference.
Q13. What is the transmit vector for the UTs in cell l?
The transmit vector sl is given assl = √ λl K∑ k=1 wlkx dl lk = √ λlWlx dl l (4)where Wl = [wl1 · · ·wlK ] ∈ CN×K is a precoding matrix and xl = [ xdll1 · · ·xdllK ]T ∈ CK ∼ CN (0, IK) contains the data symbols for the K UTs in cell l.
Q14. what is the ergodic achievable rate of a UT?
Using a standard bound based on the worst-case uncorrelated additive noise [14] yields the ergodic achievable uplink rate Ruljm of UT m in cell j:Ruljm = E [ log2 ( 1 + γuljm )] (13)where the associated signal-to-interference-plus-noise ratio (SINR) γuljm is given by (14) on the top of the next page and where the authors have used E [·|·] to denote the conditional expectation operator.
Q15. What is the UL/DL performance of a simplified channel model?
For a simplified channel model, the authors have observed that the performance depends mainly on the physical DoF perNumber of antennas NFig.
Q16. How many antennas are needed to achieve a more realistic system model?
Simulations for a more realistic system model suggest that MMSE/RZF can achieve the performance of the simple MF/BF schemes with a significantly reduced number of antennas.
Q17. what is the ergodic achievable rate for a channel?
Remark 2.1: Under a block-fading channel model with coherence time T , one could account for the rate loss due to channel training by considering the net ergodic achievable rates κ(1− τ/T )Ruljm and (1− κ)(1− τ/T )Rdljm for a given training length τ ∈ [K,T ] and some κ ∈ [0, 1] which determines the fraction of the remaining time used for uplink transmissions.