u with Low-Precision Converters d e . l l e n Christoph - PowerPoint PPT Presentation

Reliable Communication in Massive MIMO u with Low-Precision Converters d e . l l e n Christoph Studer r o c . e c e . p i v

Smartphone traffic evolution needs technology revolution u d e . l l e n r o c . e c e . p i v Source: Ericsson, June 2017

Fifth-generation (5G) may come to rescue u d e . l l e n r o c . e c e . p i v Source: Ericsson, June 2017

5G has a wide range of requirements u d e . l l e n r o c . e c e . p i v

Massive MIMO may provide solutions to all these u d e . l l e n r o c . e c e . p i v

Multiple-input multiple-output (MIMO) principles u d User1 e . BS l l e User2 n r o c ✓ Multipath propagation offers “spatial bandwidth” . e ✓ MIMO with spatial multiplexing improves throughput, coverage, c e and range at no expense in transmit power . p ✓ MIMO technology enjoys widespread use in many standards i v Conventional small-scale point-to-point or multi-user (MU) MIMO systems already reach their limits in terms of system throughput

Massive MIMO*: anticipated solution for 5G u d e Equip the basestation (BS) with hundreds . l or thousands of antennas B l e n Serve tens of users U in the same r o time-frequency resource c . e Large BS antenna array enables high array c gain and fine-grained beamforming e . p i v *Other terms for the same technology: very-large MIMO, full-dimension MIMO, mega MIMO, hyper MIMO, extreme MIMO, large-scale antenna systems, etc.

Promised gains of massive MIMO (in theory) u d ✓ Improved spectral efficiency, coverage, and range e . ➜ 10 × capacity increase over small-scale MIMO l l e ➜ 100 × increased radiated efficiency n r o ✓ Fading can be mitigated substantially → “channel hardening” c . e ✓ Significant cost and energy savings in analog RF circuitry c e . p ✓ Robust to RF and hardware impairments i v ✓ Simple baseband algorithms achieve optimal performance

Short “history” of massive MIMO u 2010: Conceived by Tom Marzetta (Nokia Bell Labs) [1] d 2012: First testbed for 64 × 15 massive MIMO system [2] e . 2013: Samsung achieves > 1 Gb/s with 64 BS antennas [3] l l e 2016: ZTE releases first pre-5G BS with 64 antennas [4] n r 2017: Sprint & Ericsson field tests with 64 antennas [5] o c . Google Scholar search for “Massive MIMO” yields 13,300 results... e c e . [1] T. Marzetta, “Noncooperative cellular wireless with unlimited numbers of base station antennas,” IEEE p T-WCOM, 2010 i [2] C. Shepard, H. Yu, N. Anand, L. E. Li, T. Marzetta, R. Yang, and L. Zhong, “Argos: practical many-antenna v base stations,” ACM MobiCom, 2012 [3] H.Benn, “Vision and key features for 5th generation (5G) cellular,” Samsung R&D Institute UK, 2014 [4] “ZTE Pre5G massive MIMO base station sets record for capacity,” ZTE Press Center, 2016 [5] “Sprint and Ericsson conduct first U.S. field tests for 2.5 GHz massive MIMO,” Sprint Press Release, 2017

u d e . l l e n Practical challenges r o c . e c e . p i v

Practical challenges u d e ✗ The presence of hundreds or thousands of high-quality RF . l chains causes excessive system costs and power consumption l e n r o ✗ High-precision ADCs/DACs cause high amount of raw c baseband data that must be transported and processed . e c e ✗ The large amount of data must be processed at high rates . p and low latency and often within a single computing fabric i v

Power breakdown of a single, high-quality RF chain u d e . l l e n r o c . e c e Analog circuit power of a single RF chain in a picocell BS in Watt [1] . p i v Data converters & amplifiers consume large portion of BS power [1] C. Desset et al., “Flexible Power Modeling of LTE Base Stations,” IEEE WCNC, 2012

u d e . l l e n r We will show that massive MIMO enables reliable o c communication with low-precision data converters . e c e . p i v

Why should we use low-resolution ADCs/DACs at BS? u Lower resolution → lower power consumption d Power of ADCs/DACs scales exponentially with bits e . Massive MIMO requires a large number of ADCs/DACs l l e n Lower resolution → reduced hardware costs r o Remaining RF circuitry ( amplifiers , filters, etc.) needs to c operate at precision "just above" the quantization noise floor* . e Extreme case of 1-bit data converters enables the use of c high-efficiency, low-power, and nonlinear RF circuitry e . p Lower resolution → less data transported from/to BBU i v Example: 128 antenna BS and 10 -bit ADCs/DACs operating at 80 MS/s produces more than 200 Gb/s of raw baseband data *terms and conditions apply

u Uplink: users → basestation d e . l l e n r o c . e c e . p i v

Quantized massive MIMO uplink u ADC map. RF ADC d narrowband data detection e CHEST and channel ADC map. RF . ADC l . . . . . . l e n ADC map. RF ADC r o c . We consider infinite-precision DACs at the user equipments (UEs) e and low-precision ADCs at the basestation (BS) side c e . p ➜ Is reliable communication with low-precision ADCs possible? i v ➜ How many quantization bits are required? ➜ Do we need complicated/complex baseband algorithms?

System model details u d ADC map. RF ADC e (Narrowband) narrowband data detection CHEST and . channel ADC map. RF channel model: l ADC l e . . . . . . y = Q ( Hx + n ) n ADC map. RF r ADC o c . y ∈ Y B receive signal at BS; Y quantization alphabet e c Q ( · ) describes the joint operation of the 2 B ADCs at the BS e . H ∈ C B × U MIMO channel matrix p i x ∈ O U transmitted information symbols (e.g., QPSK) v n ∈ C B noise; i.i.d. circularly symmetric Gaussian, variance N 0

How can we deal with quantization errors? Model 1 u d e . l l e n r o c Assume that input Y is a zero-mean Gaussian random variable . e Simple model: Z = Q ( Y ) = Y + Q c e Quantization error Q is statistically dependent on input Y . p An exact analysis with this approximate model is difficult i v

How can we deal with quantization errors? Model 2 u d e . l l e n r o c Assume that input Y is a zero-mean Gaussian random variable . e Model input-output relation statistically [1] c e Probability distribution p ( Z | Y ) has a known form . p Exact model but a theoretical analysis is difficult i v [1] A. Zymnis, S. Boyd, and E. Candès, “Compressed sensing with quantized measurements,” IEEE SP-L, 2010

How can we deal with quantization errors? Model 3 u d e . l l e n r o c Assume that input Y is a zero-mean Gaussian random variable . e Bussgang’s theorem [2]: Z = Q ( Y ) = gY + E c e Quantization error E is uncorrelated with input Y . p This decomposition is exact → theoretical analysis possible i v [1] A. Zymnis, S. Boyd, and E. Candès, “Compressed sensing with quantized measurements,” IEEE SP-L, 2010 [2] J. J. Bussgang, “Crosscorrelation functions of amplitude-distorted Gaussian signals,” MIT Research Laboratory of Electronics, technical report, 1952

Consider linear channel estimation and detection Bussgang’s theorem linearizes the system model : u d y = Q ( Hx + n ) = GHx + Gn + e e . where G is a diagonal matrix that depends on the ADC and l l e error e is uncorrelated with x n r o Using Bussgang’s theorem, we derive a linear channel estimator: c g � P t =1 y t x H . ˆ t H = e g 2 P · SNR + g 2 + (1 − g 2 )( U · SNR + 1) c e P = number of pilots; g = Bussgang gain that depends on ADC . p i x = (ˆ H ) † y v Zero-forcing (ZF) equalization: ˆ Do such simple receive algorithms work for coarse quantization?

Uncoded BER vs. SNR: ZF with QPSK 10 0 u 1-bit d 2 bit 10 − 1 e bit error rate (BER) 3 bit . ∞ bit l 10 − 2 l e n 10 − 3 r o 10 − 4 c . 10 − 5 e − 20 − 15 − 10 − 5 0 5 10 c SNR [dB] e . p B = 200 antennas, U = 10 users, P = 10 pilots, Rayleigh fading i v Markers correspond to simulation results; solid lines correspond to Bussgang-based approximations [1] S. Jacobsson, G. Durisi, M. Coldrey, U. Gustavsson, and CS, “Throughput analysis of massive MIMO uplink with low-resolution ADCs,” IEEE T-WC, 2017

u d e . l l e n r Are these results still valid for realistic o c wideband massive MIMO-OFDM systems? . e c e . p i v

Full-fledged massive MIMO-OFDM system model [ 1 ] u frequency-selective ADC FEC IDFT P/S RF RF S/P DFT dec. wireless channel ADC d data detection CHEST and e ADC FEC IDFT P/S RF RF S/P DFT dec. ADC . l . . . . . . . . . . . . . . . . . . . . . . . . . . . l e n ADC FEC IDFT P/S RF RF S/P DFT dec. ADC r o c We consider quantized channel estimation and data detection . e We compare two methods: c e Exact model (model quantization statistically) . p Approximate model (treat as unorrelated noise) i v ➜ How many bits are required for reliable uplink transmission? [1] CS and G. Durisi, “Quantized massive MIMO-OFDM uplink,” IEEE T-WCOM, 2016