Transcription of Time--Domain Compression of ComplexDomain Compression …
1 TimeTimeTimeTime----Domain Compression of ComplexDomain Compression of ComplexDomain Compression of ComplexDomain Compression of Complex----BBBB aaaasssseeeebbbbaaaannnndddd LLLLTTTTEEEE SSSS iiiiggggnnnnaaaallllssss ffffoooorrrr CCCC lllloooouuuudddd RRRR aaaaddddiiiioooo AAAA cccccccceeeessssssss NNNN eeeettttwwwwoooorrrrkkkkssssMr. Karl F. Nieman and Prof. Brian L. Evans, Wireless Networking and Communications Group, The University of Texas at AustinOur ApproachLTE-A Signal StructureObjective: Achieve 5xcompression of complex-baseband wireless signals to reduce cost of deployments for LTE-Advanced and Link-Level Simulator w/ CompressionCloud Radio Access Networks (C-RANs)Project supported by HuaweiC-RAN with CPRI Link CompressionProject website: ~bevans/papers/2013/LTEcompression/Virtu al PartitionBS1BS2UE2UE4UE3UE1 CPRI Radio Head (RRH)AFEADCBase Station (BS) ProcessingCPRI link(s).
2 Q strip CP / Q-1 / append CPQ-1 CPRI BS-sideRRH-sideCloud Radio Access Networks (C-RANs) Increased energy-efficiency vs. traditional RANs Static coverage cells are virtualized Processing resources are shared in the cloudCommon public radio interface (CPRI) Fiber or copper backbone of C-RANs Transports complex-baseband wireless samples between basestation (BS) processors and remote radio heads (RRHs) >90% of CPRI line bandwidth is used for complex-baseband sample transport Central Limit Theorem shows OFDMA and SC-FDMA samples converge to complex-Normal distribution: Time-domain samples are not are correlated due to oversamplingLloyd-Max Quantization Minimizes the mean-squared error for a given probability density function Can be designed for polar and rectangular complex-Gaussian distributions For an L-level quantizer, quantization levels can be derived in closed-form:Uplink: SC-FDMAD ownlink.
3 OFDMA DownlinkFigure taken from TSI, lte in a nutshell Reduce in-band quantization noise by 9 dBdue to noise shaping Achieve 3x Compression from noise shaping and Gaussian quantization (see plots on far right) Resample and strip CP in downlink Achieve 5xoverall Compression for downlink and uplinkUplinkCompression in LTE-A MIMO DownlinkRectangular:Polar:Noise Shaping Shape quantization noise to guard bands using recursive error-feedback Shaping filter is a fifth order IIR with flat response over passbandResampling After quantization, samples can be resampled to the Nyquist rate and cyclic prefix (CP) can be strippedFit to Normal distribution vs. FFT/IFFT (layer) mapperMIMO precodingresource element mapperresource element mapperOFDM modulation/add CPmodulation mappermodulation mapperOFDM modulation/add CPRemote Radio Controller(s)Remote Radio Head(s)up-conversion, filteringtime-domain decompresstime-domain (layer) de-mapperMIMO decodingresource element demapresource element demapSC-FDMA de-modulation/remove CPmodulation de-mappermodulation de-mapperSC-FDMA de-modulation/remove CPRemote Radio Controller(s)Remote Radio Head(s)
4 Down-conversion,filteringtime-domain compresstime-domain decompressPed-Achan-nelCompression introduced in link simulator heredeveloped using Technical University of Vienna LTE link-level simulators
