Freescale, the Freescale logo, AltiVec, C-5, CodeTEST, CodeWarrior, ColdFire, ColdFire+, C-Ware, the Energy Efficient Solutions logo, Kinetis, mobilegt, PowerQUICC, Processor Expert, QorIQ, Qorivva, StarCore, Symphony and VortiQa are trademarks of Freescale Semiconductor, Inc., Reg. U.S. Pat. & Tm. Off. Airfast, BeeKit, BeeStack, CoreNet, Flexis, Layerscape, MagniV, MXC, Platform in a Package, QorIQ Qonverge, QUICC Engine, Ready Play, SafeAssure, the SafeAssure logo, SMARTMOS, TurboLink, Vybrid and Xtrinsic are trademarks of Freescale Semiconductor, Inc. All other product or service names are the property of their respective owners. 2012 Freescale Semiconductor, Inc. TM
Course web site: Contact e-mail addresses: Presenter: andrei.enescu@freescale.com HR: roxana.dragomir@freescale.com Third lecture: Quiz with two prizes and other give-aways for participants Kwikstik K-40 http://www.freescale.com/webapp/sps/site/prod_summary.jsp?cod e=kwikstik-k40&fsrch=1&sr=1 Attending certificate for each student taking the quiz and attending at least 2 lectures TM 2
LTE is a good guy Superior technologies Great capacity: 3Gbps/1.5Gbps promised in LTE-A Responsibility lies in L1 Flat network Smaller latency Smaller costs Different deployment scenarios Macro/Metro (BIG) Pico/Femto (SMALL) TM 3
What is Layer 1 (L1)? Block diagram (Tx + Rx) Technologies OFDMA SC-FDMA LTE radio interface TDD/FDD Throughput calculation TM 4
Transforms packets of bits into analog signal which is transmitted onto antenna and viceversa Performs processing in order to deal with the radio channel What do we do inside? Transmission Packet bits get some redundancy for protection over the channel Bits are mapped onto symbols Symbols are transposed into samples (a digital signal) Digital signal is converted into analog Reception Analog signal is downconverted and digitized Samples are transposed into symbols Symbols are decoded into bits Bits are decoded into raw bits Signal repairments TM 5
Bit-rate Symbol-rate Signal processing -Channel coding (Conv, Turbo) -Interleaving -HARQ -Digital modulation (QPSK, 16QAM, 64QAM etc.) -Apply high-level modulation (OFDM, SC-FDMA etc.) -MIMO -Digital signal processing (interpolation, scaling, advanced algorithms) -D/A conversion -Filtering -Upconversion -Amplification TM 6
Signal processing Symbol-rate Bit-rate -Amplification -Downconversion -Filtering -A/D conversion -Digital signal processing (downsampling, scaling, advanced algorithms) -Equalization, synchronization -MIMO decoding -Demodulation (OFDM, SC-FDMA) -Digital demodulation (QPSK, 16QAM, 64QAM etc.) -Channel decoding (Conv, Turbo) -De-Interleaving -HARQ TM 7
Typical single-carrier case Assume BPSK (1 bit/symbol) T b 1, 0, 1, 0 1 Multipath -1 Time B=1/T b Frequency Received Signal = sum of delayed and weighted replicas of transmitted signal Channel acts just like a filter In-band distortion Intersymbol Interference Could be resolved via deconvolution with an inverse filter Or TM 8
Let s say that we divide input stream into 4 separate substreams, over periods of 4T b. B = 4 x 1/(4T b ) = 1/T b 1, 0, 1, 0 1 0 1 0 1-1 1-1 1-1 1 Frequency -1 Time 4T b Each substream modulates one of 4 adjacent carriers Each substream will have 1/4 th of the original bandwidth If subbands are adjacent, then the total bandwidth remains the same Need to make sure they don t interfere one with another (e.g. they are orthogonal) It can be proven that orthogonality is achieved if subcarriers are spaced by 1/(4T b ) = Size of a sub-band TM 9
Receiver decodes substreams independently If channel varies slowly over one sub-band, it can be considered virtually constant for each substream Assumption is based on: Number of sub-bands Bandwidth Channel frequency variation Each received substream = Transmitted substream x Weight Receiver is trivial No deconvolution Simple division Frequency OFDM: Generalize this example for any number (N) of subbands (Orthogonal Frequency Division Multiplexing) B=1/T b TM 10
Frequency Frequency T sym = 4T b Time DF B=4DF OFDM T sym = T b Time B SC TM 11
Frequency Each aggregated OFDM symbol has a cyclic prefix in front of it used to avoid interference between symbols and subcarriers Cyclic prefix length is given by the multipath delay spread in the channel T sym = 4T b Time DF B=4DF T OFDM = T sym + T g TM 12 T g T sym
Subcarriers It s the same principle, applied for multiuser In OFDMA, the grid is partitioned among different users Users get to have different slices of the grid OFDMA = Orthogonal Frequency Division Multiple Access User 1 Symbols User 4 User 2 User 3 RE = Resource Element TM 13
Subcarrier spacing is always 15kHz Depending on the number of subcarriers, we get the total bandwidth Bandwidth flexibility Used subcarriers Guard subcarriers Example B = 10MHz N = 1024 sc 212 sc Fs = 1024 * 15kHz = 15.36MHz Bandwidth TM 14 600 sc + DC 211 sc 9MHz ½*Sampling freq
Channel bandwidth is the occupied bandwidth Plus some guard bands Subbands have sidelobes that may produce interference Bandwidth [MHz] Occupied bandwidth [MHz] Sampling frequency [MHz] Number of subcarriers Number of data subcarriers 1.4 1.08 1.92 128 72 3 2.7 3.84 256 180 5 4.5 7.68 512 300 10 9 15.36 1024 600 20 18 30.72 2048 1200 TM 15
Not all of the resource elements actually carry information Some of them are predefined symbols used to carry: Control information Describe L1 parameters of the data channels Generated and reported back to L2 Reference signals Used for mitigating any impairment occurred on the channel They are generated and interpreted at L1 level Synchronization preambles L1 Throughput is affected by the transmission of these signals TM 16
Independent RSs are sent for each transmit antenna 1 Tx Overhead = 4.76% 2 Tx Overhead = 9.52% 4 Tx Overhead = 14.29% 8 Tx Overhead = 30% (roughly) TM 17
High PAPR (Peak to Average Power Ratio) Typical of 10-12dB (depending on number of subcarriers) PAPR [db] Peak power level in dbm Link performance depends on average power The higher, the better Average power level in dbm Amplifiers should be chosen according to peak power Higher => more expensive, larger size or subject to potential distortions It becomes even more problematic for User Equipment TM 18
Over uplink, a slightly different approach is used A typical OFDM is employed, but with a DFT precoding DFT codes make sure power is spread across the whole bandwidth It s less likely that subcarriers are summed co-phased and produce peaks This works especially if one single user does not fill the whole bandwidth Users are separated in frequency SC-FDMA and OFDMA roughly achieve the same capacity TM 19
T frame = 10ms SF0 SF1 SF2 SF3 SF4 SF5 SF6 SF7 SF8 SF9 T subframe = 1ms Slot 0 Slot 1 T slot = 500ms T CP = 5.21ms T CP = 4.69ms Tdata = 1/15kHz = 66.67ms T CP = 16.67ms Tdata = 1/15kHz = 66.67ms OR Normal cyclic prefix Extended cyclic prefix 7 symbols per slot 6 symbols per slot TM 20
TDD = Time Domain Duplexing enodeb and UEs use the same band for transmission DL and UL must be separated in time FDD = Frequency Domain Duplexing enodeb and UEs use separate bands for transmission DL and UL are active 100% in time TM 21
There are 7 possible frame formats Each frame format imposes a fixed scope for each of the 10 subframes in a frame DL subframe UL subframe Switching subframe Between a DL subframe and an UL subframe Which configuration would you choose for a video surveillance app? TM 22
Frame format #1 SF 0 SF 1 SF 2 SF 3 SF 4 SF 5 SF 6 SF 7 SF 8 SF 9 DwPTS (Downlink Pilot Time Slot) GP (Guard period) UpPTS (Uplink Time Slot) DL UL No Tx DL gets 2 subframes (20% x capacity) UL gets 6 subframes (60% x capacity) TM 23
We have two types of slots Slot with normal cyclic prefix 7 symbols of data Slot with extended cyclic prefix 6 symbols of data Bandwidth [MHz] Number of data subcarriers Number of REs per slot Ext CP Normal CP 1.4 72 864 1008 3 180 2160 2520 5 300 3600 4200 10 600 7200 8400 20 1200 14400 16800 TM 24
Each RE carries a symbol from a digital constellation: QPSK: 2 bits/symbol 16-QAM: 4 bits/symbol 64-QAM: 6 bits/symbol Max throughput is achieved for 64-QAM Bandwidth [MHz] Number of data subcarriers Number of kbits per slot (64QAM) Ext CP Normal CP 1.4 72 2.592 3.024 3 180 6.480 7.560 5 300 10.800 12600 10 600 21.600 25.200 20 1200 43.200 50.400 14% overhead for ext CP TM 25
Count out RSs, consider entire frame and only normal CP MIMO 4x4 brings 4 times capacity increase At the expense of ~15% pilot overhead MIMO 8x8 brings 8 times capacity increase At the expense of ~30% pilot overhead Bandwidth [MHz] Number of kbits per slot (64QAM) Total capacity [Mbps] MIMO 4x4 capacity [Mbps] MIMO 8x8 capacity [Mbps] 1.4 3.024 5.184 3 7.560 12.96 5 12600 21.6 10 25.200 43.2 20.5632 33.8688 51.408 84.672 85.68 141.12 171.36 282.24 20 50.400 86.4 TM 26 UL 342.72 564.48 DL
Throughputs DL: 565Mbps / UL: 342Mbps We are still far away from the envisaged peak rate of 3/1.5Gbps (DL/UL) LTE-A systems will operate in 100MHz bands Provides x5 capacity increase DL: 2.82Gbps / UL: 1.7Gbps *This is only valid for FDD For TDD, UL and DL share the same capacity *There are some other physical channels that limit the L1 capacity (sync signals: PSS, SSS, PRACH) TM 27
Not all UEs in a cell benefit from the max peak rate In order to enable it, you must: Be close to enodeb Use 64QAM Use MIMO 8x8 / 4x4 Experience a good channel May use normal CP Be alone in a cell All this capacity is shared among all UEs in the cell TM 28
Platforms B4860 B4420 Implementation Kernel Specs Design Implementation Unit testing What is QA? Profiling Integration Framework Profiling Integration testing System Integration testing Quiz Survey TM 29
TM