3GPP TSG-RAN WG1 LTE Ad Hoc Meeting
`Helsinki, Finland, 23 – 25 January, 2006
`Source:
`NTT DoCoMo, NEC, Sharp
`Title:
`Orthogonal Pilot Channel Structure in E-UTRA Uplink
`Agenda Item: 5.2.2.2
`Document for: Discussion and Decision
`1.
`Introduction
`
` R1-060046
`
`We proposed an orthogonal pilot channel structure associated with intra-Node B coordination such as
`intra-node B transmission timing control and orthogonal radio resource assignment for single-carrier
`(SC)-FDMA based radio access in the Evolved UTRA uplink [1]. In the contribution, we presented that
`orthogonal channel generation using the frequency domain (FDM) and code domain (CDM) is more
`promising than using the time domain (TDM). This paper proposes the actual orthogonal physical
`channel generation method for SC-FDMA radio access in the E-UTRA uplink. The proposed
`orthogonal channel generation employs a combination of FDM and CDM.
`
`2. Proposed Generation Method for Orthogonal Pilot Channel
`
`In the paper, we focus on the orthogonality among UEs in all cells within the same Node B, i.e., all
`sectors. Here, we assume that the OFDM symbol timings of all cells in the same Node B are
`synchronized perfectly and the uplink transmission frame timing is generated based on the downlink
`received frame timing. Moreover, adaptive transmission timing controls among simultaneously
`accessing UEs are performed over all cells within the same Node B, so that the received timings of all
`UE in the same Node B are aligned within the CP (cyclic prefix) duration.
`
`Figure 1 shows the proposed orthogonal pilot channel generation using the combination of FDM and
`CDM. The proposed method to multiplex orthogonal physical channels is as follows.
`
` Multiplexing UEs with Different Transmission Bandwidths Using Distributed FDMA
`We use distributed FDMA in the frequency domain for multiplexing UEs with different transmission
`bandwidths. This is because orthogonality among UEs with different transmission bandwidths is not
`possible using CDM in general. However, FDMA can achieve orthogonality among UEs with different
`transmission bandwidths as shown in Fig. 1. Employing distributed transmission is necessary to obtain
`channel gain over the assigned transmission bandwidth.
`
` Multiplexing UEs with Identical Transmission Bandwidth Using CDMA
`We use CDMA for multiplexing UEs with identical transmission bandwidths, i.e., sampling rates. In
`DS-CDMA, the code with a good auto-correlation property is necessary such as the CAZAC (Constant
`Amplitude Zero Auto-Correlation) sequence. However, the number of sequences is small when the
`CAZAC sequence is used as the pilot sequence, since the scrambling code cannot be used in
`conjunction with the CAZAC sequence. Thus, [2] proposed employing a cyclic shift based CAZAC
`code generation method, in which, multiple CAZAC sequences are generated by cyclic shift of the
`
`- 1/8 -
`
`APPLE 1032
`
`

`

`original CAZAC sequence. It was reported in [2] that the cyclic shift based CAZAC sequences provide
`a good cross-correlation property to mitigate mutual interference within the duration of the cyclic shift
`value. The cyclic shift value, Δ, is set so that it supports the maximum path delay time. We use the
`cyclic shift based CAZAC sequences from the one original CAZAC sequence with priority as shown in
`Fig. 2. Thus, when all cyclic shift based CAZAC sequences from the one original sequence are used up,
`we use the cyclic shift based CAZAC sequences from the second sequence.
`
`
`
`(Example) There are three pilot channel-bandwidths: 1.25, 5, 10 MHz
`1.25-MHz pilot
`5-MHz pilot
`
`10-MHz pilot
`
`1.25 MHz
`
`5 MHz
`
`Frequency
`10 MHz
`
`Used for other 1.25-MHz pilot
`
`Used for other 5-MHz pilot
`
`Number of multiplexed UEs
`
`CAZAC #2, shift #2
`CAZAC #2, shift #1
`CAZAC #1, shift #N
`Orthogonal
`CAZAC #1, shift #2
`CAZAC #1, shift #1
`
`
`Figure 1 – Proposed orthogonal pilot channel generation using combination of FDM and CDM
`
`
`
`Δ
`
`UE #1
`
`UE #2
`
`UE #3
`
`Pilot (CAZAC #1, shift #1)
`
`Pilot (CAZAC #1, shift #2)
`
`Pilot (CAZAC #1, shift #3)
`
`Cyclic shift
`
`
`
`
`Pilot (CAZAC #1, shift #N)
`
`Figure 2 – Cyclic shift based CAZAC sequence generation from one original CAZAC sequence
`
`Time
`
`
`
`UE #N
`
`
`
`
`
`Thus, by combining FDM and CDM multiplexing, orthogonality among UEs with different and
`identical transmission bandwidths is flexibly achieved. When distributed FDMA transmission is used,
`the sequence length of the pilot channel becomes short in the time domain. Thus, in the proposed
`method, we restrict the application of distributed FDMA only to UEs with different transmission
`bandwidths. Accordingly, we can accommodate a large number of cyclic shift based CAZAC
`
`- 2/8 -
`
`

`

`sequences. Note that when all the available radio resources for orthogonality are used in the respective
`environments, we further allow non-orthogonal channel assignment.
`
`3. Maximum Number of Orthogonal Channels in FDM
`
`In this section, we investigate the maximum number of orthogonal channels, Nmux, in FDM. Assuming
`a TDM-base pilot channel structure, let Δfpilot and Δf be the sub-carrier spacing in one pilot channel
`with FDM and the minimum sub-carrier spacing, i.e., 30 kHz, respectively. Then, Δfpilot becomes Δf x
`Nmux. Thus, the relation such that 1 / Δfpilot > largest delay time must be satisfied. We employ
`Greenstein’s r.m.s. delay spread model in which τrms is represented as T1 x dε x y (μsec) [6]. Then, the
`r.m.s. delay spread value at the cumulative distribution function (CDF) of 90 and 95% are calculated as
`shown in Table 1 for the inter site distance (ISD) of 2.8 and 5.0 km. From the r.m.s. delay spread
`values in Table 1, we calculated the largest delay time as listed in Table 2. Here, we assume an
`exponentially-decayed power delay profile model such as p(t) = 1 / τrms x exp(− t / τrms). We define the
`largest delay time as the delay time of the path, which is 10-dB decayed from the path energy of the
`greatest path, since approximately 90% of the total received power is collected. Therefore, we see
`from Table 2 that the largest delay time at the 90% CDF with σy = 2 dB case is approximately 4.2 and
`5.6 μsec for the ISD of 2.8 and 5.0 km, respectively.
`
`
`
`
`
`CDFCDF
`
`
`90%90%
`
`95%95%
`
`
`ISD = 2.8 kmISD = 2.8 km
`
`σy = 2 dBσy = 2 dB
`
`1.8 µsec1.8 µsec
`
`2.2 µsec2.2 µsec
`
`Table 1 – Delay spread
`
`ISD = 5 kmISD = 5 km
`
`σy = 2 dBσy = 2 dB
`
`2.4 µsec2.4 µsec
`
`2.9 µsec2.9 µsec
`
`
`σy = 5 dBσy = 5 dB
`
`4.3 µsec4.3 µsec
`
`6.5 µsec6.5 µsec
`
`
`σy = 5 dBσy = 5 dB
`
`5.7 µsec5.7 µsec
`
`8.7 µsec8.7 µsec
`
`
`
`
`
`
`
`
`CDFCDF
`
`
`90%90%
`
`95%95%
`
`
`ISD = 2.8 kmISD = 2.8 km
`
`σy = 2 dBσy = 2 dB
`
`4.2 µsec4.2 µsec
`
`5.2 µsec5.2 µsec
`
`Table 2 – Largest delay time
`
`ISD = 5 kmISD = 5 km
`
`σy = 2 dBσy = 2 dB
`
`5.6 µsec5.6 µsec
`
`6.8 µsec6.8 µsec
`
`
`σy = 5 dBσy = 5 dB
`
`10.0 µsec10.0 µsec
`
`15.1 µsec15.1 µsec
`
`
`σy = 5 dBσy = 5 dB
`
`13.2 µsec13.2 µsec
`
`20.1 µsec20.1 µsec
`
`
`
`
`
`
`Meanwhile, when Nmux is 1, 2, 3, 4, 5, 6, 7, and 8, the value of 1 / Δfpilot becomes 33.3, 16.7,
`
`11.1, 8.33, 6.67, 5.55, 4.76, and 4.17 μsec, respectively. Thus, we see for supporting the largest delay
`time for the ISD of 2.8 and 5.0 km, the allowable Nmux value becomes approximately four. Here, when
`we use staggered pilot symbol mapping of the FDM-based orthogonal pilot between two pilot blocks
`within the same sub-frame [1], [3] as shown in Fig. 3, we can increase the allowable Nmux value to
`twice equivalently. Consequently, we see that we can accommodate eight FDM-based orthogonal pilot
`channels in the environment with the ISD of up to approximately 5 km. However, eight orthogonal
`channels with FDM represents almost the worst case. Thus, it should be noted that a larger number of
`physical channels is multiplexed in FDM for UEs in the environments with shorter largest delay times.
`
`- 3/8 -
`
`

`

`Sub-frame
`
`Pilot
`
`Pilot
`
`Frequency
`
`Frequency
`
`Figure 3 – Channel estimation using interpolation of orthogonal pilot channels employing distributed
`FDMA (Different comb-spectra are assigned to two pilot blocks)
`
`
`4. Investigation on Pilot Sequence
`
`We compare the average PER performance using the CAZAC sequence and random sequence in the
`pilot channel. Table 3 lists the simulation parameters, which follow the agreed parameters in [4]. The
`transmission bandwidth is 5 MHz. We use the Zadoff-Chu sequence [5] for the CAZAC sequence with
`the length of 151, along with a random sequence with the length of 150. We employ turbo coding with
`the coding rate of R = 1/2. We assume ideal FFT timing detection and ideal noise power estimation for
`the frequency domain equalizer. The channel gain is estimated in the time domain by taking the
`correlation within the pre-decided time window duration between the received signal and the pilot
`sequence replica. The sample points, which provide greater received power than the pre-determined
`threshold value, are selected within the time window in order to reduce the influence of the noise and
`interference. Thus, the correlations of the selected sample points are converted into a frequency domain
`signal by FFT. We employ a frequency domain equalizer employing the LMMSE algorithm, associated
`with two-branch antenna diversity reception.
`
`
`
`Table 3 – Simulation parameters
`
`Transmission schemeTransmission scheme
`
`Localized FDMALocalized FDMA
`
`5 MHz5 MHz
`
`Transmission bandwidthTransmission bandwidth
`
`7.68 Msps7.68 Msps
`
`Sampling rateSampling rate
`
`0.5 msec (data block x 6, pilot block x 2)0.5 msec (data block x 6, pilot block x 2)
`
`Sub-frame lengthSub-frame length
`
`66.67 µsec / 512 samples66.67 µsec / 512 samples
`
`Data block sizeData block size
`
`33.33 µsec / 256 samples33.33 µsec / 256 samples
`
`Pilot block sizePilot block size
`
`15 kHz (data blocks), 30 kHz (pilot blocks)15 kHz (data blocks), 30 kHz (pilot blocks)
`
`Sub-carrier spacingSub-carrier spacing
`
`CAZAC (Zadoff-Chu) seq., Random seq.CAZAC (Zadoff-Chu) seq., Random seq.
`
`Pilot sequencePilot sequence
`
`4.04 µsec / 31 samples4.04 µsec / 31 samples
`
`Cyclic prefix durationCyclic prefix duration
`
`Raised cosine with 3.13 µsec / 24 samplesRaised cosine with 3.13 µsec / 24 samples
`
`Time windowingTime windowing
`
`QPSK, 16QAMQPSK, 16QAM
`
`Data modulationData modulation
`
`Turbo coding (K = 4, R = 1/2)Turbo coding (K = 4, R = 1/2)
`
`Channel codingChannel coding
`
`1, 4 users1, 4 users
`
`Number of usersNumber of users
`
`2-branch receiver reception,2-branch receiver reception,
`
`LMMSE frequency domain equalizerLMMSE frequency domain equalizer
`
`6-ray Typical Urban, Vehicular A6-ray Typical Urban, Vehicular A
`
`
`
`ReceiverReceiver
`
`
`
`Channel modelChannel model
`
`
`
`
`- 4/8 -
`
`
`
`

`

`Figures 4(a) and 4(b) show a comparison of the average PER performance between the CAZAC
`and random sequences for QPSK and 16QAM modulations, respectively, as a function of the average
`received signal energy per symbol-to-noise power spectrum density ratio (Es/N0) per receiver branch.
`We use a six-ray Typical Urban (TU) channel model with the fading maximum Doppler frequency of
`fD = 55.5 Hz (30 km/h at a 2-GHz carrier frequency). The number of simultaneous accessing UEs is set
`to one and four. The PER performance assuming ideal channel estimation is also given as a reference.
`In a four-user environment, we apply different CAZAC sequences, which are generated by cyclic shift
`of the original CAZAC sequence [2].
`
`Clear improvements are observed in the CAZAC sequence compared to the random sequence.
`First, the results in a one-user environment show that the required average received Es/N0 per receiver
`branch for the average PER of 10-2 using the CAZAC sequence is reduced by approximately 0.5 dB
`compared to that of the random sequence for QPSK modulation. Furthermore, as shown in the Fig. 4(b),
`although the average PER of 10-2 is not achieved with the random sequence, it is significantly
`improved by the CAZAC sequence for 16QAM modulation. This is because the accuracy of the path
`timing detection and channel estimation is remarkably improved due to the good auto-correlation
`property of the CAZAC sequence.
`
`Next, looking at the PER performance in the four-user environment, we find a clear benefit to
`the CAZAC sequence with cyclic shift compared to the random sequence. The average PER
`performance using the CAZAC sequence with cyclic shift in the four-user environment becomes
`almost identical to that in the one-user case. This is caused by the good cross-correlation property of
`the CAZAC sequences through cyclic shift based generation.
`
`
`
`
`
`
`
`
`
`
`
`
`1 user
`Ideal channel est.
`Pilot: random
`Pilot: CAZAC
`4 users
`Pilot: random
`Pilot: CAZAC
`with cyclic shift
`
`100
`
`10-1
`
`10-2
`
`Average PER
`
`QPSK, R = 1/2
`6-ray TU
`10-3
`10
`8
`6
`4
`2
`0
`-2
`Average received Es/N0 per receiver branch (dB)
`
`
`
`- 5/8 -
`
`

`

`(a) QPSK modulation
`
`1 user
`Ideal channel est.
`Pilot: random
`Pilot: CAZAC
`
`100
`
`10-1
`
`10-2
`
`4 users
`Pilot: random
`Pilot: CAZAC
`with cyclic shift
`16QAM, R = 1/2
`6-ray TU
`10-3
`16
`14
`12
`10
`8
`6
`4
`Average received Es/N0 per receiver branch (dB)
`(b) 16QAM modulation
`
`
`
`Average PER
`
`Figure 4 – Average PER performance comparison between CAZAC and random sequence
`(6-ray TU channel model)
`
`
`
`
`
`
`
`
`
`Figures 5 (a) and 5(b) show the average PER performance using the CAZAC and random
`sequences in one and four-user environments for QPSK and 16QAM, respectively, assuming the
`Vehicular-A channel model. Similar to the results in the TU channel model, we see that by applying the
`CAZAC sequences with cyclic shifts, the PER performance in the four-user environment is almost the
`same as that for the one-user case for QPSK modulation and slightly degraded for 16QAM modulation.
`On the other hand, the PER performance employing random sequences is significantly degraded
`compared to that with the CAZAC sequences, particularly in the four-user environment. We conclude
`that the CAZAC sequence is very promising as a pilot channel sequence for the SC-FDMA radio
`access from the viewpoint of the achievable PER performance.
`
`
`
`- 6/8 -
`
`

`

`1 user
`Ideal channel est.
`Pilot: random
`Pilot: CAZAC
`4 users
`Pilot: random
`Pilot: CAZAC
`with cyclic shift
`
`100
`
`10-1
`
`10-2
`
`Average PER
`
`QPSK, R = 1/2
`Vehicular A
`10-3
`10
`8
`6
`4
`2
`0
`-2
`Average received Es/N0 per receiver branch (dB)
`(a) QPSK modulation
`
`
`
`100
`
`10-1
`
`1 user
`Ideal channel est.
`Pilot: random
`Pilot: CAZAC
`
`10-2
`
`4 users
`Pilot: random
`Pilot: CAZAC
`with cyclic shift
`16QAM, R = 1/2
`Vehicular A
`10-3
`16
`14
`12
`10
`8
`6
`4
`Average received Es/N0 per receiver branch (dB)
`(b) 16QAM modulation
`
`
`
`Average PER
`
`
`
`
`
`Figure 5 – Average PER performance comparison between CAZAC and random sequence
`(Vehicular-A channel model)
`
`
`
`5. Conclusion
`
`This paper proposed orthogonal pilot channel for SC-FDMA radio access with multiple transmission
`bandwidth in E-UTRA uplink. The features of proposed orthogonal pilot channel are as follows.
`
`
`- 7/8 -
`
`

`

`- Distributed FDMA is used for multiplexing UEs with different transmission bandwidths.
`- CDMA is used for multiplexing UEs with identical transmission bandwidths, i.e., sampling
`rates.
`In DS-CDMA, cyclic shift based CAZAC sequences from the one original CAZAC sequence is
`used with priority. Thus, when all cyclic shift based CAZAC sequences from the one original
`sequence are used up, we use the cyclic shift based CAZAC sequences from the second
`sequence.
`
`-
`
`Uplink reference-signal structure
`
`
`6. Text Proposal (Section 9.1.1.2 in TR25.814)
`
`--------------------------------- Start of Text Proposal --------------------------------------------------
`
`9.1.1.2.1
`
`Distributed FDMA is used for multiplexing UEs with different transmission bandwidths. CDMA is
`used for multiplexing UEs with identical transmission bandwidths, i.e., sampling rates. In DS-CDMA,
`cyclic shift based CAZAC sequences from the one original CAZAC sequence is used with priority.
`Thus, when all cyclic shift based CAZAC sequences from the one original sequence are used up, we
`use the cyclic shift based CAZAC sequences from the second sequence.
`
`--------------------------------- End of Text Proposal -----------------------------------------------------
`
`
`Reference
`[1] 3GPP, R1-051142, NTT DoCoMo, et al., “Orthogonal Pilot Channel in the Same Node B in Evolved UTRA
`Uplink”
`[2] 3GPP, R1-050822, Texas Instruments, “On Allocation of Uplink Pilot Sub-Channels in EUTRA SC-
`OFDMA”
`[3] 3GPP, R1-051033, Motorola, “Further Topics on Uplink DFT-SOFDM for E-UTRA”
`[4] 3GPP, TR-25.814 (V1.0.1), “Physical Layer Aspects for Evolved UTRA”
`[5] D. C. Chu, “Polyphase codes with good periodic correlation properties,” IEEE Trans. Inform. Theory, vol.
`IT-18, pp. 531 – 532, July 1972.
`[6] L. J. Greenstein et.al. "A New Path-Gain/Delay-Spread propagation model for digital cellular channels,"
`IEEE Trans. Veh. Technol., vol. 46, no2, pp.477-485, May 1997.
`
`- 8/8 -
`
`