《Reference signal generation and selection for E-UTRA uplink》由会员分享，可在线阅读，更多相关《Reference signal generation and selection for E-UTRA uplink（4页珍藏版）》请在金锄头文库上搜索。
1、-1/4-3GPP TSG RAN WG1 Meeting #49 R1-072083Kobe, Japan, May 7-11, 2007Source: PanasonicTitle: Reference signal generation and selection for E-UTRA uplinkAgenda Item: 7.11.2Document for: Discussion and Decision1. IntroductionIn Sorrento meeting, the followings were agreed Uplink reference signal sequ
2、ence length is equal to the number of sub-carriers in the resource blocks, Either truncation or cyclic extension of ZC sequences is used, depending on the RB allocation size. This paper discusses the reference signal generation method of truncation or cyclic extension per RB in terms of the number o
3、f available ZC sequences, Cubic Metric, cross correlation between different sequence indexes and cross correlation between different cyclic shifts.2. Discussion2.1. Truncation or Cyclic extension2.1.1 Number of available ZC sequences and Cubic Metric5 raised an issue for uplink reference signal that
4、 power-limited UEs suffered from high CM signals. Reserving ZC sequences with low CM for the power limited UEs was also proposed to avoid the power de-rating depending on CM of DM RS. However, the reservation and allocation of ZC sequences taking into account CM would be complex.In order to avoid th
5、is issue, elimination of ZC sequences with CM larger than QPSK can be applied as a simple solution. In this case, we have to take into account the number of available ZC sequences with CM lower than QPSK. According to 3 , Table 1 shows the number of available ZC sequences whose CM is lower than that
6、 of QPSK in case of truncation and cyclic extension.Table 1 Number of ZC sequences with CM lower than QPSK.Truncation Cyclic extension# of RBs # of ZC sequences# of sequence with CM lower than QPSK# of ZC sequences# of sequence with CM lower than QPSK1 12 6 10 62 28 8 22 123 36 18 30 244 52 24 46 24
7、5 60 32 58 326 72 38 70 38For 3 RBs, the number of ZC sequences with CM lower than QPSK is 18 and 24 in case of truncation and cyclic extension, respectively. These numbers would relax cell planning sufficiently. Therefore, elimination of ZC sequence with lager CM can be applied for 3 and larger RBs
8、, irrespective of generation method.Meanwhile, 1RB and 2RBs case give only smaller number of ZC sequences for both truncation and cyclic extension. If limiting the usable ZC sequence by CM, cell planning or hopping pattern planning would be more complex.We consider the two alternative sequence gener
9、ation methods for 1 or 2RBs as follows;Alt. 1: Choosing either truncation or cyclic extensionAlt. 2: Computer calculated CAZAC is applied with low CM and low cross-correlation 4 -2/4-Alternative 1If we have to select either of truncation or cyclic extension scheme, cyclic extension is preferable for
10、 1RB with taking into account the maximum CM properties of all available sequences to mitigate the power de-rating depending on CM of DM RS, as shown in Figure 11 .Raw CubicMetr dBZ sequnc idxTruncatio(N=3) Cylexs1234567891020456 CM ofQPSKFigure 1 Raw Cubic Metric of reference signal of truncation a
11、nd cyclic extension (1RB).Alternative 2CAZAC sequences generated by computer calculation such as 4 is attractive approach for 1 and/or 2 RBs case to increase the number of available reference signal with low CM and low cross-correlation. Therefore, the computer generated CAZAC sequence is applied as
12、 alternative solution.Our preference is Alternative 2 because both cell planning and power de-rating issues can be solved. But exact sequence set and performance comparison has not been shown yet, therefore, careful consideration is necessary before final agreement of the computer generated CAZAC se
13、quence. Moreover, computer generated CAZAC requires memory to keep sequence set. Therefore, computer generated CAZAC should be applied only for 1 or 2 RBs.2.1.2 Cross-correlation between cyclic shift sequencesIn this section, we compare cross-correlation properties between cyclic shift sequences of
14、truncation and cyclic extension. Interference signal is added only long blocks of reference signal part to evaluate the BLER performance degradation due to interference between different cyclic shift sequences, that is, long blocks for data transmission do not have any interference. Figure 2(a) and
15、Figure 2(b) shows BLER performance with QPSK R=1/2 and 16QAM R=1/2, respectively when 5 other different cyclic shift sequences are multiplexed as interference. From the results, there is no difference between truncation and cyclic extension in term of cross-correlation between cyclic shift sequences
16、.Table 2 Simulation condition.Parameter ValueRS generation method Cyclic-extension, TruncationBandwidth 1RB(180kHz), 3RB(540kHz), 6RB(1.08MHz)Number of multiplexed UEs 6 (only RS is interfered)Number of available cyclic-shift 6Cyclic shift separation method Simple rectangular mask is used after IDFT in time domainSIR of Power spectrum density 0dB (only RS part)Channel model Typical Urban 6-path (mobility: 30km/h)Modulation and coding scheme QPSK R=1/2, 16QAM R=1/2HARQ No
