Contributions and Proposals of UPC Department 1 NEWCOM

Contributions and Proposals of UPC - Department 1 NEWCOM Antonio Pascual. Iserte Universitat Politècnica de Catalunya Barcelona, 9 March 2005 A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM

Outline 4 Design of MIMO systems: a practicaloverview: • MIMO robust designs • Design of limited feedback 4 Some proposalsfor future work: • Other robust designs • Impact of MIMO channels on system level aspects A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 2

Outline 4 Design of MIMO systems: a practicaloverview: • MIMO robust designs • Design of limited feedback 4 Some proposalsfor future work: • Other robust designs • Impact of MIMO channels on system level aspects A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 3

Robustness 4 Robust. Designs: • They take into account the errors in the CSI • Sources of errors in the CSI for MIMO channels: o Estimation noise o Channel variability o Quantization of the channel estimate (feedback) 4 Robustness. Strategies: • Bayesian approach: statistical solution • Maximin approach: best worst-case A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 4

Robustness. Strategies 4 f: cost function to minimize, B: TX, A: RX 4 Bayesian. Design • Best statistical mean value 4 Maximin. Design • Optimization of the worst-case A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 5

Maximin. Robust. Transmitter- MIMO 4 OSTBC + power allocation + beamforming: 4 Receiver: linear operations A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 6

Notation 4 SNR: 4 Notation: • Error in the channel estimate: o Convex uncertainty region: • Beamformers: Beamformers estimated eigenmodes of • Power distribution: distribution • Cost function to maximize: A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 7

Problem. Formulationand Solution 4 Maximin robust design: 4 The problem can be solved using standard software packages for convex optimization problems: A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 8

Uncertainty. Regions(I) • Several sources of errors can be modeled (a) Gaussian estimation errors (TDD) (b) Errors from scalar quantization (FDD) (c) Combination of both (FDD) Quadratic convex problems… A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 9

Uncertainty. Regions(II) • Also vector quantization… • In general… any error that can be modeled by a convex uncertainty region A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 10

Robust. Adaptive. Modulation A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 11

Outline 4 Design of MIMO systems: a practicaloverview: • MIMO robust designs • Design of limited feedback 4 Some proposalsfor future work: • Other robust designs • Impact of MIMO channels on system level aspects A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 12

Limited. Feedback 4 Introduction : • Design of the feedback strategy: Receiver Transmitter channel estimate ? ? ? Feedback Channel Quantization B bits of feedback A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 13

Limited. Feedback - MIMOChannels(I) 4 Scalar Quantization (SQ): • Direct quantization of the channel: o Non-iterative: uniform vs. non-uniform o Iterative quantization: delta modulation • Quantization of the strongest eigenvectors: o Direct quantization o Quantization in a parameterized space (e. g. the set of orthonormal vectors) A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 14

Limited. Feedback - MIMOChannels(II) 4 Vector Quantization (VQ): • Distortion measure: o Suboptimum: Suboptimum distortion = error in the channel estimate o Optimum: Optimum distortion = system performance • Transmitter architecture: STC F (space-time code) (linear precoder) B bits of feedback N = 2 B: F 1, … , FN A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 15

Limited. Feedback - MIMOChannels(III) • Precoder codebook: {F 1, …, FN} • Precoder selection function F=f (H): performance metric (SER, MSE, SNR, mutual information, etc…) 4 Example of design of the precoder codebook: • Use a set of linear precoders F 1, …, FN spanning maximally spaced subspaces Grassmannian packaging • Random precoder: easier, asymptotically equivalent to Grassmannian A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 16

Illustrative. Results A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 17

Outline 4 Design of MIMO systems: a practicaloverview: • MIMO robust designs • Design of limited feedback 4 Some proposalsfor future work: • Other robust designs • Impact of MIMO channels on system level aspects A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 18

Other Robust. Designs 4 Extensionsof the robust designs: • Other transmitter architectures • Multiuser scenarios • Limited channel knowledge: o Only gain o Only phase • Different degrees of knowledge at the transmitter and the receiver • Robustness against implementation errors A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 19

Impactof MIMO on System Level 4 Analysisof the benefitsprovidedby MIMOat the system level: • Identification of simulation scenarios • Identification of appropriate figures of merit • Identification of a channel quality metric for interface between the physical/link and higher layers o MIMO channel capacity o maximum eigenvalue of the MIMO channel matrix o others… ? ? ? A. Pascual Contributions and Proposals of UPC to Department 1 - NEWCOM 20