R1-060867
`
`3GPP TSG RAN WG1 #44-bis
`Athens, Greece, 27 - 31 March, 2006
`Source:
`Texas Instruments
`Title:
`A new preamble shape for the Random Access preamble in E-UTRA
`Agenda Item:
`10.2.3
`Document for:
`Discussion
`1.
`Introduction
`It is assumed that in E-UTRA, a UE uses a contention-based channel, the random access channel, to initiate a (new)
`resource allocation request. A random access attempt contains a preamble which design is driven by the primary purpose of
`allowing Node B to detect the random access attempt with targeted detection and false alarm probabilities, and minimize
`collisions impact on the contention-based channel. Optimization of random access preamble bandwidth/duration with
`respect to that aspect has already been addressed in [1]. In addition, Node B uses the random access preamble to perform
`UL synchronization of an out-of-sync UE. Finally, the random access preamble can be used by the Node B to derive the
`best UL resource allocation for the upcoming random access message or further L2/L3 data. The present contribution
`addresses this issue by proposing an alternate shape for the random access preamble.
`
`2. Random Access usage for scheduling requests
`RAN2 has agreed that the random access usage in E-UTRA is restricted to UE requesting (new) resources on SCH [2], [3].
`A scheduling request may result from different UE situations: initial network access (DETACHED  ACTIVE state
`transition), new data to transmit in ACTIVE state, handover procedure, etc. However, it is assumed that for all these
`different cases, the information required and being conveyed by the scheduling request fits in a small payload referred to as
`random access message. Two options exist for transmitting this message:
`1. The message is sent along with the preamble in the random access burst. In this case, the Node B demodulates/decodes
`the scheduling request and sends back to the UE the associated allocated resource on UL SCH, along with timing
`advance information and C-RNTI [2], if necessary.
`2. The message is sent only after the Node B acknowledged back the random access preamble. In this case, only preamble
`transmission uses a contention-based channel, and the Node B sends back to the UE the associated allocated resource
`on UL SCH for the message, along with timing advance information and C-RNTI [2], if necessary.
`
`In both cases, the Node B must allocate some UL resource to the UE based on known needs. In the following, UEs are
`assumed to be scheduled over localized frequency resource blocks, taking advantage of multi-user diversity. In absence of
`any knowledge of the UL frequency response over the system bandwidth, the Node B can only allocate UL resource blindly
`to the UE based only on the amount of UL resource that is requested and is available.
`It was shown in [1] that a narrow band preamble of 2.5MHz provides optimum detection performance over the SNR range
`required for detection probability and false alarm targets of [0.9 – 0.999] and 10-3 respectively. Such narrow band also
`brings the advantage of leaving room in the spectrum for other preambles or scheduled data (TDM or TDM/FDM option).
`Unfortunately, it allows the Node B to estimate the UL channel quality for that UE within the preamble bandwidth only, so
`the Node B cannot perform UL resource allocation beyond the preamble bandwidth. Optimization of the UL resource
`allocation with respect to the channel frequency response calls for means of estimating the latter beyond the narrow
`preamble bandwidth.
`We propose to address this issue by adding a wideband pilot that can be either attached or embedded to/in the preamble,
`and that the Node B will use for channel estimation upon preamble detection.
`
`3. Random Access preamble with attached wideband pilot (option 1)
`The ‘T’-shape preamble shown in Figure 1 includes a wideband pilot attached at the preamble end. The pilot block should
`be kept short, typically of the same duration as the short block (SB) in the sub-frame format specified in [4].
`Both pilot and preamble use sequences chosen randomly by the UE, but linked so that the Node B can derive the wideband
`pilot code from the preamble signature.
`The preamble uses CAZAC sequences which discrete autocorrelations are zero for all nonzero lags [5]. Let’s have a look at
`the orthogonal sequence space with an example. It was shown in [1] with a simple link budget analysis that a preamble
`duration of 1165 µs is needed to achieve a detection probability of 0.999 with 10-3 false alarm over a 5km cell. With such
`cell size, a maximum round-trip delay of 33.33 µs is expected, thus allowing for 35 cyclic shifted orthogonal versions of the
`sequence. The wideband pilot cannot use any cyclic shift of a given CAZAC sequence, given its duration is the same as the
`maximum round-trip delay. This calls for sequence sets with good auto- and cross-correlation properties such as ZCZ-GCL
`
`1
`
`APPLE 1012
`
`

`

`[6] for the wideband pilot. However, this will reduce the sequence space for the wideband pilot compared to that of the
`preamble, resulting in a higher collision probability, given the preamble can also benefit from randomly choosing a
`frequency chunk in the access slot. On the other hand, the wideband pilot is expected to benefit from the coarse time
`estimation of the preamble. For example, for 2.5 and 1.25 MHz preambles, UEs with the same wideband pilot sequence will
`not collide if they are spaced apart by more than 60 and 120 meters, respectively, in the direction of the Node B. Moreover,
`they will also be orthogonal if they use CAZAC sequences.
`Figure 1 shows examples of T-shape preambles for both the TDM/FDM and TDM-only multiplexing option. Note in the
`former case the wideband pilot is limited to the RACH access slot allocated bandwidth so that the concept best applies to
`the TDM-only option. This option adds a minimum overhead on the RACH burst length, given those already envisioned for
`coverage purpose (1-2ms, [1]), not accounting for the message when attached. Finally, with this option, pilot to other UE’s
`preambles interference (and vice versa) is limited to the overlapping region due to non-aligned UEs (Figure 1).
`
`Figure 1: T-shape RACH preamble in FDM/TDM and TDM multiplexing options
`
`4. Random access preamble with embedded wideband pilot (option 2)
`The ‘+’-shape preamble shown in Figure 2 includes a wideband pilot merged with the preamble. The wideband pilot
`crosses the preamble at a time position depending on the frequency resource block used by the preamble, so as to allow
`time multiplexing pilots of different frequency multiplexed preambles, therefore offering the same collision avoidance
`performance. The pilot block should be kept short, typically of the same duration as the short block (SB) in the sub-frame
`format specified in [4].
`The signature sequence of both the preamble and the pilot are such that they “coincide” in the crossing region: down-
`sampling the wideband pilot to the preamble sampling rate yields the preamble signature in this time interval, resulting in
`the wideband pilot bandwidth being an integer multiple of the preamble bandwidth.
`It is shown in the Appendix how the “crossing region” requirement restricts the sequence space and the bandwidth ratio
`between wideband pilot and preamble when CAZAC sequences are chosen for both the preamble and wideband pilot [5].
`Moreover, at cell edge, a UE may need to send the random access burst at maximum power, i.e. the same power for both
`
`2
`
`

`

`preamble and pilot. As a result, the preamble samples picked from the wideband pilot will experience a power decrease
`compared to the non-pilot preamble samples, thus breaking the “constant amplitude” and therefore autocorrelation property
`of CAZAC sequences. As a result, more study needs to be done to check which of CAZAC sequences or other PN
`sequences are the most appropriate for this scheme.
`Other options could be envisioned for the crossing region such as gating the preamble to let the wideband pilot alone,
`which would bring the benefit of letting preamble and pilot parameters being chosen independently.
`Similar to Section 3, Figure 2 shows examples of preambles/pilots for both the TDM/FDM and TDM-only multiplexing
`option. The same comment applies, i.e. concept best applies to the TDM-only option. Note that this option yields no
`overhead on the random access burst length, as the wideband pilot duration is not added on top of the preamble duration.
`
`Figure 2: Random access preamble with embedded wideband pilot in FDM/TDM and TDM multiplexing options
`
`5. Comparison of attached versus embedded pilot
`Table 1 compares attached and embedded options with respect to different performance criteria.
`
`3
`
`

`

`Table 1: Comparison of attached versus embedded wideband pilot
`Attached
` Embedded
`Good (CAZAC)
`Medium
`
`and
`
`Wideband pilot is:
`Preambles
`orthogonality
`autocorrelation property
`Wideband pilots collision avoidance
`
`Overhead
`Wideband Pilot interference on other
`preambles
`
`Low, not as good as preambles
`collision avoidance
`Low
`Low
`
`As good as preambles collision
`avoidance
`None
`Medium
`
`6. Other benefits of the wideband pilot
`The additional wideband pilot is motivated in Section 3 by the need for providing the Node B with some means for
`estimating the UL channel frequency response over the system bandwidth, thus enabling an appropriate UL resource
`allocation. In this section, we address how the wideband pilot can also improve the detection performance as well as the
`UE’s timing estimation accuracy.
`6.1. Detection performance
`As illustrated in Figure 3, the wideband pilot can be seen as an additional signal that can be further used as a 2nd detection
`stage, to verify the random access reception upon preamble detection. As such it allows for a 2-stage detection process thus
`potentially improving the overall detection performance or reducing the preamble length for the same overall pd and pfa
`performance [7]. Quantitative assessment of the practical benefit if FFS.
`
`Figure 3: Random access preamble with wideband pilot used as a verification stage of the detection
`6.2. Fine timing acquisition
`The time synchronization precision after preamble detection is in the order of the inverse of the preamble bandwidth. As a
`result, exploiting the wideband pilot requires finer time granularity which is achieved by running wideband sample phase
`bins around the initial synchronization estimate on the wideband pilot. This also allows the Node B to send back to the UE
`a finer Timing Adjustment (TA) information.
`
`The time synchronization quantitative improvement over the wideband pilot if FFS.
`
`4
`
`

`

`7. Conclusion
`Given that random access usage in E-UTRA most of the time involves some UL scheduling request by the UE, we
`proposed in this contribution to add to the random access preamble a wideband pilot. This wideband pilot provides, with no
`or little overhead, support to the Node B for UL resource allocation to the UE for either the random access message or
`further L2/L3 data. Two implementation options are proposed: attached or embedded wideband pilot. In addition, this new
`scheme provides more potential benefits such as improved detection performance and finer time acquisition.
`
`References
`[1] R1-060376, “RACH Preamble Design”, Texas Instruments
`[2] 3GPP TR 25.813 V0.6.0, “Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial
`Radio Access Network (E-UTRAN); Radio interface protocol aspects (Release 7)”
`[3] R1-060061, R2-060144, “LTE L1 related questions to RAN1”
`[4] 3GPP, TR-25.814, “Physical layers aspects of evolved UTRA (E-UTRA)”
`[5] D.C. Chu, “Polyphase codes with good periodic correlation properties”, IEEE Transactions on Information Theory,
`July 1972
`[6] R1-060328, “RACH design for E-UTRA”, Huawei
`[7] Stephen S. Rappaport, Donald M. Grieco, “Spread Spectrum Signal Acquisition: Methods and Technology”, June
`1984, Vol 22, N°6, IEEE Communications Magazine
`
`Annex 1 CAZAC sequences for random access preamble with
`embedded wideband pilot
`
`CAZAC Zadoff-Chu sequences of length N whose discrete autocorrelations are zero for all nonzero lags are defined as [5]:
`kMj
`2
`,
`N
`
`kkMj
`N
`
`
`
`exp
`
`exp
`
`
`
` (1)
`
`a
`
`k
`
`a
`
`k
`
`even
`N
`
`,1
`
`
`
`oddN
`
`where M and N are relatively prime.
`
`The signature sequence of both the preamble and the pilot are such that they “coincide” in the crossing region: down-
`sampling the wideband pilot to the preamble sampling rate yields the preamble signature in this time interval. As a result,
`the wideband pilot CAZAC sequence must be chosen so that the sequence resulting from picking 1 element every n is still a
`CAZAC sequence, which restricts the possible pilot sequences to:
`
`nN
`
`;
`
`k
`
`
`
`,2,1
`
`2
`
`)
`
`nk
`Mj
`(
`
`N
`
`
`
`exp
`
`j
`
`Mn
`
`2
`
`k
`2
`
`N
`
`
`
`exp
`
`j
`
`2
`
`kMn
`
`nN
`/
`
`(2)
`
`ak
`
`
`
`exp
`
`with the further restrictions on N, M and n:
`N is even
`
`n is odd
`
` Mn² and N are relatively prime
`
`The condition on n is the most restrictive as it removes some flexibility on the wideband pilot over preamble bandwidth
`ratio.
`
`5
`
`