3GPP TSG RAN WG1 Ad Hoc on LTE
`San Diego, USA, 10 October - 14 October, 2005
`
`R1-051062
`
`Source:
`Title:
`Agenda Item:
`Document for:
`1.
`Introduction
`
`Texas Instruments
`On Uplink Pilot in EUTRA SC-FDMA
`8.2
`Discussion/Decision
`
`1.1
`
`Problem Formulation
`One of the two possible TTI structures for uplink Single Carrier FDMA (SC-FDMA),
`as proposed by drafting group 1, is given in Figure 1 below.
`
`Figure 1: Uplink TTI structure for SC-FDMA.
`
`In Figure 1, LB represents a “Long Block,” which can contain only data symbols, and
`SB represents a “Short Block,” which can contain either pilot or data symbols.
`Therefore, the uplink pilot is always confined inside the SB field. The time duration of
`the SB field is half of the time duration of the LB field. The rest of the numerology for
`the uplink frame structure is given in [1].
`The proposed uplink TTI structure results in the frequency set where the width of
`pilot subcarriers is twice the width of data subcarriers. For example, in the baseline
`case of 5MHz bandwidth, pilot and data subcarriers are as given in Figure 2 below.
`
`Figure 2: Frequency Set for SC-OFDM.
`
`In the case of distributed (IFDMA) uplink transmission, each mobile is
`allocated a set of non-contiguous tones for data subcarriers. In this case, it is unclear as
`to which is the most appropriate allocation of uplink pilot resources. Furthermore, a
`relatively wide-band pilot may have benefits even for localized data transmission,
`because it enables frequency-dependent UE scheduling. The following options should
`be considered.
`
`1.2
`
`Possible Allocations for Orthogonal Uplink Pilot
`a) Time Domain Orthogonality
`Time domain orthogonality is the most obvious alternative for usage of the SB
`field for pilot transmission. However, such a solution may result in a high peak to
`average ratio (PAR) for uplink transmission, which would consequently decrease
`coverage due to the amplifier back-off.
`
`1
`
`APPLE 1034
`
`

`
`b) Frequency Domain Orthogonality
`Frequency domain (IFDMA) orthogonality is another proposed solution [3] for the
`uplink orthogonal pilot, which is a topic of current studies. The main difficulty
`faced by a frequency domain orthogonal pilot is for UE’s near the cell border,
`when the neighbouring cell utilizes the same IFDMA uplink pilot channel. In such
`cases, UEs are likely to have a “dominant interferer” present. Dominant interferer,
`which resides in the neighbouring cell, significantly degrades channel estimation
`performance (for IFDMA pilot), even if his/hers signal is substantially attenuated.
`This claim will be demonstrated below using link – level simulations.
`
`c) Code Domain Orthogonality
`Code domain orthogonality can be achieved with a use of Constant Amplitude Zero
`Autocorrelation (CAZAC) sequences, as we demonstrate in the remainder of this
`document. Furthermore, CAZAC sequences have a flat frequency domain response,
`which makes them attractive for SC – OFDMA systems. Finally, CAZAC pilot is
`very robust to out-of-cell interference, which we show via link – level simulations.
`
`In this contribution, we first demonstrate how to achieve pilot orthogonality in the code
`domain, via cyclic shifts of CAZAC sequences. Second, we demonstrate some of its
`advantages with respect to an alternate proposed solution of IFDMA pilot [3]
`
`1.3
`
`Background on CAZAC Sequences
`An example of CAZAC sequences is given as follows. Let L be any positive integer,
`and let k be any number which is relatively prime with L. Then the n-th entry of the k-
`th Zadoff-Chu CAZAC sequence [2] is given as follows:
`
`
`
`c (n) = exp
`k
`
`c (n) = exp
`k
`
`
`
`n +n
`
`
`
`
`
`
`
`
`
` if L is odd
`
`2
`
`n +
`
` if L is even
`
`n+1
`j2 k
`
`
`
`
`2
`L
`
`
`
`
`j2 k
`
`
`n
`
`
`
`
`L
`2
`
`
`
`
`The set of Zadoff-Chu CAZAC sequences has the following properties:
` Constant magnitude
` Zero circular autocorrelation
` Flat frequency domain response
` Circular cross-correlation between two sequences is low and it has constant
`magnitude, provided that L is a prime number.
`2. Proposal: Allocation of Uplink Pilot Sub-Channels
`In this section we demonstrate how to achieve the uplink orthogonal pilot in the code domain
`with the use of CAZAC sequences. The main idea is to use a single CAZAC sequence per
`sector and exploit the property of zero circular autocorrelation along with the cyclic prefix
`transmission.
`
`2.1
`
`Allocation of Pilot Sub-Channels for a Single Sector
`In order to illustrate how to achieve orthogonality in the code domain, we let the
`CAZAC sequence be “c,” and let its right cyclic shift by Q be specified as SQ(c). Since
`the sequence has zero cyclic autocorrelation, then S0(c), SQ(c), S2Q(c) … SMQ(c) are all
`orthogonal provided that MQ does not exceed the length of the sequence. Furthermore,
`even when S0(c) is cyclically right-shifted by less than Q samples, it remains
`
`2
`
`

`
`orthogonal to the rest of SQ(c), S2Q(c) … SMQ(c). Next, we simply allocate S0(c) to be
`the pilot sequence for UE#0, SQ(c) to be the pilot sequence for UE#1, and proceed
`accordingly until we allocate SMQ(c) to be the pilot sequence for UE#M. Such an
`allocation is illustrated in the following figure.
`
`Figure 3: Proposed Allocation of Uplink Pilot Sequences.
`
`With such an allocation, the arriving multipath signal from each UE will be
`orthogonal, under the assumption that Q is longer than each delay profile. For this
`reason an appropriate choice for Q is the prefix length of the transmission.
`Alternatively, a more conservative allocation would accommodate scenarios where the
`delay profile is longer than the prefix length. In such cases, Q should be longer than
`the transmission prefix.
`
`Allocation of Pilot Sub-Channels in Softer Handover
`For UE’s which are in the Softer Handover, the transmitted signal is received with
`significant power level in two sectors of the Node B. In order to avoid UE self-
`interference, we propose that both serving sectors allocate the same CAZAC sequence,
`with the exact same shift, to UE’s which are shared in the Softer Handover. Hence,
`each sector of a single Node B will utilize the same CAZAC sequence.
`
`Allocation of Pilot Sub-Channels between different Node B’s
`Neighboring Node B’s should utilize different CAZAC sequences for the uplink pilot
`channel in order to achieve interference averaging. For this reason, the most
`appropriate choice for CAZAC sequences are Zadoff-Chu sequences of prime length
`(see Background section above), which have low constant magnitude cyclic cross-
`correlation. Since the number of different Zadoff-Chu sequences is close to the length
`of the sequence itself (hence large), there are no difficulties in constructing the reuse
`pattern for distant Node B’s.
`
`Number of CAZAC sequences
`As stated earlier in the background section, Zadoff – Chu sequences have low constant
`magnitude cross – correlation, provided that their length is a prime number. In this
`section, we present the number of possible sequences, assuming the exact uplink
`numerology from [1], Option2.
`
`2.2
`
`2.3
`
`2.4
`
`3
`
`

`
`Table 1: Number of CAZAC Sequences
`
`1.25MHz
`128
`76
`64
`37
`7
`
`2.5MHz
`256
`151
`128
`73
`15
`
`36
`
`72
`
`5MHz
`512
`301
`256
`151
`31
`
`150
`
`10MHz
`1024
`601
`512
`293
`63
`
`15MHz
`1536
`901
`768
`449
`95
`
`20MHz
`2048
`1201
`1024
`601
`127
`
`292
`
`448
`
`600
`
`288
`
`576
`
`1200
`
`2336
`
`3584
`
`4800
`
`LB Samples
`Used Subcarriers
`in LB
`SB Samples
`Used Subcarriers
`in SB
`CP Samples
`# of distinct
`CAZACs not
`including shifts
`# of distinct
`CAZACs including
`8 shifts
`
`Table 1 is derived as follows. Rows 2 and 4 are from the uplink proposal in [1],
`Option2. Row 3 hasn’t been agreed upon yet (for the uplink), which is why we
`assumed the downlink numerology from [1]. Row 5 is proposed to be the prime
`number which is closest to half of the Row 3. Row 6 is directly from [1]. Row 7 is
`derived based of properties (see background section) of Zadoff – Chu sequences.
`Finally, Row 8 is 8 * Row 7, since the SB (Row 4) accepts 8 distinct circular shifts by
`the cyclic prefix (Row 6).
`
`4
`
`

`
`2.5
`
`Link – Level Simulation Comparison of IFDMA and CAZAC pilot: Dominant
`Interferer Present
`
`Table 2: Simulation Assumptions
`
`Parameter
`Bandwidth
`Channel Model
`Data Channel Turbo Coding
`Data Modulation
`Uplink Numerology
`Dominant out–of–cell interferer
`
`Data Channel
`
`Antenna Configuration
`CAZAC
`
`Pilot
`
`Distributed FDMA
`which occupies each
`6th tone. Number of
`Sub-carriers = 24
`Pilot Modulation
`
`Channel
`Estimation
`
`Time Interpolation
`
`Frequency Interpolation
`Interpolation Method
`
`Assumption
`5 MHz (2.6 GHz)
`TU, Velocities: 3kmh, 120kmh, 360kmh.
`Rate ½
`16QAM
`Option 2 in [1] (Table 9.1.1.2)
`C/I = 10dB, 16dB, 26dB.
`Distr. FDMA which occupies each 6th tone.
`Number of Sub-carriers = 48. Dominant
`interferer uses same channel in neighboring cell
`1 at Transmitter, 2 at Receiver
`Dominant interferer from the neighboring cell
`uses different CAZAC (length = 151)
`Dominant interferer from the neighboring cell
`uses same IFDMA pilot channel, with
`different pilot sequence
`
`QPSK
`Doppler dependent filter coefficients
`MF – Wiener Matched Filter
`ZF – Wiener Zero Forcing Filter
`Least Squares
`Past, Current and Future TTI
`
`Figure 4: BLER with CAZAC Pilot and IFDMA Pilot. UE speed = 3kmh. Dominant interferer present.
`
`5
`
`

`
`Figure 5: BLER with CAZAC Pilot and IFDMA Pilot. UE speed = 120kmh. Dominant interferer present.
`
`Figure6: BLER with CAZAC Pilot and IFDMA Pilot. UE speed = 360kmh. Dominant interferer present.
`
`6
`
`

`
`From the above link – level simulation results (Figure 4, 5, 6) we conclude that the use of
`CAZAC pilot offers much more robust channel estimation, when compared to distributed
`FDMA (IFDMA) pilot. This basically occurs because dominant out-of-cell interferer
`concentrates its power within the same IFDMA pilot channel as the primary in-cell UE,
`thereby creating substantial interference. On the other hand, with CAZAC pilot, out-of-cell
`interferer uses a different long CAZAC sequence, and the interference becomes automatically
`averaged over all UEs. Performance of CAZAC pilot and IFDMA pilot becomes comparable
`only in the case that dominant interferer is severely attenuated (e.g., -26dB with respect to
`main signal).
`3. Conclusion
`
`The set of Zadoff-Chu CAZAC uplink pilot sequences presents an attractive solution for
`the uplink pilot design in LTE. In this document we presented a method for reuse of a
`single CAZAC sequence with cyclic shifts in order to achieve orthogonality in the uplink
`pilot channel. Furthermore, interference management between different cells is fairly
`simple and robust because it reduces to assigning different CAZAC sequences to
`neighbouring cells.
`
`4. References
`[1] TR 25.814 v 0.1.1 “Physical Layer Aspects for Evolved UTRA”
`[2] K. Fazel and S. Keiser, “Multi Carrier and Spread Spectrum Systems,” John Willey and
`Sons, 2003.
`[3] R1-050851, “Uplink Orthogonal Pilot Channel in the Same Node B in Evolved UTRA
`Uplink.”
`
`7