1. On the Multiple Access Schemes for IEEE 802.16m: Comparison of SC-FDMA and OFDMA Document Number: C802.16m-08/045 Date Submitted: Jan 16, 2008 Source: Yang-Seok ChoiIntel corpE-mail:yang-seok.choi@intel.comyang-seok.choi@intel.com RongzhenYangIntel corpE-mail:rongzhen.yang@intel.comrongzhen.yang@intel.com JiachengWangIntel corpE-mail:jiacheng.wang@intel.comjiacheng.wang@intel.com Tom HarelIntel corpE-mail:tom.harel@intel.comtom.harel@intel.com YuvalLomnitzIntel corpE-mail:yuval.lomnitz@intel.comyuval.lomnitz@intel.com HujunYinIntel corpE-mail:hujun.yin@intel.comhujun.yin@intel.com Venue: TGm Call for contribution on SDD, Levi, Finland Base Contribution: C80216m-08/045 Purpose: For discussion of comparison between OFDMA and SC-FDMA, and approval of OFDMA system by IEEE 802.16 Working Group Notice: This document does not represent the agreed views of the IEEE 802.16 Working Group or any of its subgroups. It represents only the views of the participants listed in the Source(s) field above. It is offered as a basis for discussion. It is not binding on the contributor(s), who reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an IEEE Standards publication; to copyright in the IEEEs name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEEs sole discretion to permit others to reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this contribution may be made public by IEEE 802.16. Patent Policy: The contributor is familiar with the IEEE-SA Patent Policy and Procedures: and.http://standards.ieee.org/guides/bylaws/sect6-7.html#6http://standards.ieee.org/guides/opman/sect6.html#6.3 Further information is located at and.http://standards.ieee.org/board/pat/pat-material.htmlhttp://standards.ieee.org/board/pat

2. 2 SC-FDMA structure and Link level comparison

3. 3 SC-FDMA TX Structure Spreading by DFT –Signal at each subcarrier is a linear combination of all M symbols –~2 dB gain in PAPR Low PAPR Low PAPR High PAPR Signal at each subcarrier is a linear combination of all M symbols Duality

4. 4 OFDMA in SISO Received signal after FFT The channel matrix H is orthogonal even in frequency selective channel –No inter-carrier interference –No ISI due to CP One-tap linear equalizer is sufficient Channel Matrix :

5. 5 SC-FDMA in SISO Received signal after FFT The channel matrix is NOT orthogonal in frequency selective channel –Inter-subcarrier interference due to the spreading matrix –No ISI due to CP Spreading Matrix : Channel Matrix :

6. 6 SC-FDMA in SISO (contd) MMSE Equalizer One tap equalizer followed by De-spreading Equalizer output :

7. 7 SC-FDMA in SISO (contd) Post-MMSE SINR OFDMA: SC-FDMA : From above

8. 8 SC-FDMA in SISO (contd) Note is a harmonic mean of Thus, –where the equality holds if and only if is constant regardless of l (i.e. flat fading) – is constant irrespective of m Steeper PER curve, Diversity gain –As delay spread increases, PER curve moves to right Longer delay spread at Cell edge –As M increases, becomes smaller in frequency selective channel –Note As delay spread and/or M increase, the loss of SC-FDMA in link level will be more evident

9. 9 Link-Level Simulation Results In frequency-selective fading, the loss due to loss of orthogonality is noticeable –Delay spread=CP, rms delay spread=CP/4, exponential decaying delay profile 3 dB loss SC-FDMA OFDMA

10. 10 MIMO Use of maximum likelihood detector (MLD) receiver –In large eigen-value spread channel, MMSE does not provide MIMO gain –fast MLD for 2x2 and 4x4 MIMO available –Virtual MIMO, MU-MIMO in UL Significant Gain Other fast MLD

11. 11 KxK MIMO (contd) OFDMA : block diagonal matrix No inter-carrier interference KxK MLD per subcarrier SC-FDMA: Not a block diagonal matrix Inter-carrier interference Cant apply per-subcarrier MLD Need KMxKM MLD (not feasible) K x K matrix

12. 12 Large PAPR in Frequency Domain After spreading : –x can be modeled as Gaussian random variable –means high PAPR in frequency domain SC-FDMA has larger PAPR in frequency domain –Out-of-band emission Though Average ICI power and OOBE are the same as in OFDMA, Larger fluctuation of instantaneous OOB Emission causes worse interference to adjacent carrier –In-band fluctuation Larger ICI power variance in time-varying channel

13. 13 ICI ICI component at k-th subcarrier : ICI power : 4th order moment (assuming flat channel) Variance of ICI power OFDMASC- FDMA QPSK12 16QAM1.322 64QAM1.3812

14. 14 ICI (contd) Normalized variance of ICI power SC-FDMA exhibits higher fluctuation of ICI power

15. 15 TX power improvement of SC-FDMA

16. 16 Simulation assumptions QPSK modulation WiMAX frequency assignment (BW=10 MHz, Nfft=1024, Nused=841, SamplingFactor=28/25) PA model: Rapp-3, saturation power 31dBm Spectral mask FCC BRS (absolute) and ETSI Mobile (relative) WiMAX UL permutation: Distributed (WiMAX-I PUSC), 3 subchannels Localized (WiMAX-I AMC) SC-FDMA modes Distributed diversity mode Localized (adopted in LTE) –TX power shown is the maximum TX power that can be attained with the above PA parameters while obeying FCC masks

17. 17 Transmit power and consumed power Higher TX power with same 1dB compression point, implies higher power consumption Thus, even if with same PA settings, a certain TX power improvement is shown, it is not always feasible due to power consumption A possible fair normalization is to keep the consumed power constant (by changing Vcc of the PA), and measure the improvement in TX power for the same consumed power –With Class-AB amplifier the consumed power is approximately proportional to –Therefore if results show N-dB improvement, there is also N/2-dB penalty in consumed power –In order to normalize to the same consumed power while keeping a constant backoff, the improvement is halved (i.e. the improvement will be N/2-dB)

18. 18 TX power improvement PUSC (distributed OFDMA) AMC (localized OFDMA) Distributed SC-FDMA Localized SC-FDMA 23.3 dBm 25.2 dBm 25.7 dBm 26.1 dBm

19. 19 TX power improvement (contd.) No difference in maximum TX power if resource is allocated at band center Localized SC-FDMA and OFDMA Centered in band 30.53 dBm 30.57 dBm Subcarrier MappingGain of SC-FDMA over OFDMA under same PA size Gain of SC-FDMA over OFDMA under same power consumption and backoff Distributed2.4 dB1.2 dB Localized @Band edge0.9 dB0.45 dB Localized @Band center0.04 dB0.02 dB

20. 20 Modeling PAPR/CM not accurate metric –Indirect method –Need to consider OOBE, EVM requirement, power consumption, and Multipath Effect together Rather, –Pass to RF filter (Tx Mask) –PA –Check OOBE and EVM requirement –Adjust Tx power –Channel –Check PER/Coverage

21. 21 Block Diagram of Joint Simulation Joint simulation (PA model+link level) : automatically include the non-linear distortion

22. 22 Path loss Received signal power 1dB compression point backoff Tx antenna gain Rx antenna gain Path loss With 90% availability of shadow fading ( 8dB standard deviation) Path loss : Urban Macro

23. 23 Simulation conditions 1dB compression: 31dBm (assume the same PA size) Power amplifier backoff: depend on DFT size, subcarrier location, etc. Tx antenna gain: 0dBi Rx antenna gain (include cable loss): 15dBi Carrier frequency: 2.5 GHz System bandwidth: 10MHz Noise figure: 5dB FFT size: 1024 ½ CTC QPSK Channel Model: Urban Macro-cell in 16m EVM document 1x1 SISO/2x2 MIMO (SM, vertical coding) Antenna spacing: Tx = 0.5 lambda, Rx = 4.0 lambda Packet length = 120 bytes SC-FDMA, localized, 32/64/128 DFT OFDMA, localized, 32/64/128 used sub-carriers Band edge and center Rapp power amplifier model, p=2.0 8 times over-sampling 193-order low-pass filter, cut-off frequency = 0.9 Equalization: MMSE (MLD for 2x2 MIMO, OFDMA)

24. 24 OFDMA vs. SC-FDMA: 1x1 SISO, band edge Equalizer loss in SC-FDMA is more dominant than the effect of larger EVM noise and backoff in OFDMA SC-FDMA has higher power consumption

25. 25 OFDMA vs. SC-FDMA: PER CDF @ SNR = 5dB

26. 26 OFDMA vs. SC-FDMA: 2x2 MIMO, band edge MMSE : with larger M, SC-FDMA is better –In correlated MIMO channel the Interstream interference effect is more dominant than the equalizer loss in SC-FDMA –EVM noise effect is more dominant than noise (note operating SNR is higher at MMSE) –The crossing point moves to higher SNR MLD : the gain of MLD is noticeable In cell edge, STBC will be chosen highly likely. In this case, OFDMA will be better than SC-FDMA as in SISO case

27. 27 OFDMA vs. SC-FDMA: 1x1 SISO, band center

28. 28 OFDMA vs. SC-FDMA: 2x2 MIMO, band center

29. 29 Duality OFDMASC-FDMA PAPRHigh in Time Low in Frequency Low in Time High in Frequency SpreadingData spread in Time Data localized in Frequency Data spread in Frequency Data localized in Time

30. 30 Duality (contd) - PAPR PAPR in TD –High in OFDMA Smaller Tx power : Due to higher PAPR, more back-off needed. However, by scheduling the resource at the center of band, no difference compared with SC-FDMA is observed Higher EVM noise : Due to higher PAPR, more non-linear distortion observed. However, in cell edge the thermal noise is dominant. In addition, the operating SNR of OFDMA is lower due to no equalizer loss and advanced receiver such as MLD. Thus, the impact of EVM noise is negligible. –Low in SC-FDMA Larger Tx power PAPR in FD –Low in OFDMA Smaller fluctuation of OOBE Smaller fluctuation of ICI power –High in SC-FDMA Higher instantaneous OOBE –Larger variance of out-of-band power : More ACI to neighboring systems Higher instantaneous ICI power in time varying channel –Larger variance of ICI power

31. 31 Duality (contd) - Spreading Spreading in TD –Data spread in TD in OFDMA More robust to impulse noise and nonlinear distortion –Data localized in TD in SC-FDMA More susceptible to impulse noise and nonlinear distortion Spreading in FD –Data localized in FD in OFDMA Less frequency diversity –Data spread in FD in SC-FDMA More frequency diversity –Steeper PER curve Equalizer loss –PER curve moves to right

32. 32 Pilot design OFDMA –Two dimensional pilot allocation : Time and Frequency –More flexible and potentially lower pilot overhead SCFDMA –Only Time domain : pilot subcarriers@ dedicated symbol Does not allow mix of data and pilot subcarrier Cant optimize the pilot design

33. 33 TDD Duplex Scheme OFDMA can be straightforwardly used to exploit the TDD reciprocal DL/UL channel properties By applying the DL common pilots and UL dedicated pilots or sounding symbol Enable many channel-aware transmission techniques and allow the implementation based enhancement Such as beam-formed MIMO, SDMA SC-FDMA makes it difficult to explore the TDD application/advantage if not possible

34. 34 Conclusions and Remarks SC-FDMA exhibits 0 to 1.2 dB gain in max TX power owing to smaller PAPR –However, by proper scheduling the resource, OFDMA shows no degradation –The typical Tone Reservation/clipping algorithms can achieve similar PAPR of SC-FDMA SC-FDMA exhibits equalizer loss in frequency selective channel –Especially when the delay spread is large and/or the number of subcarriers is large Note that in cell edge the delay spread is larger No future proof for Higher order MIMO in SC-FDMA –Practical MLD for MIMO is not feasible even in 2x2 MIMO –Higher order MIMO is not possible –Practical MLD receiver in OFDMA significantly outperforms SC-FDMA receiver –The limitation of SC-FDMA to evolve for future UL MIMO Capability is clear In SC-FDMA asymmetric resource allocation in UL/DL –OFDMA can be used to exploit the TDD reciprocal DL/UL channel properties SC-FDMA should be ruled out for 16m Multiple access discussion –Adopt OFDMA system in both UL and DL