Texas Instruments
`RACH Preamble Design
` 8.2
`Discussion
`
`3GPP TSG RAN WG1 Meeting #42bis R1-051058
`San Diego, USA< 10 – 14 October, 2005
`
`Source:
`Title:
`Agenda Item:
`Document for:
`1. Introduction
`The RACH (Random Access Channel) preamble is used in UTRA to allow contention among uplink
`users for usage of the random access channel [1,2]. It is a 4096 chip long code which is modulated by
`one of 16 repeated Hadamard codes of length 16. This contribution investigates whether a similar time
`domain sequence is suitable for E-UTRA or whether a frequency domain RACH preamble is more
`appropriate. A large number of simulations were presented when the RACH preamble was agreed
`upon for UTRA [3]. This contribution does not attempt to duplicate these simulations, but rather
`presents a few simple simulations to make some preliminary comparisons between two techniques.
`Section 2 gives an overview of the two techniques studied, and Section 3 presents simulation results
`comparing the techniques.
`
`2. Two RACH Preamble Structures
`Structure 1 is similar to the current UTRA RACH preamble in that it uses a long code modulated with
`a repeated length 16 Hadamard code. The current UTRA RACH preamble is shown in Figure 1 for
`reference. It has a duration of slightly more than 1 ms so that 15 access slots can be defined within 2
`frames which have a combined duration of 20 ms.
`
`
`
`
`Long scrambling code cL
`
`4096
`
`X
`…
`
`hi
`
`hi
`
`hi
`
`hi
`
`hi
`
`16
`
`256 repeated Hadamard codes
`
`
`
`
`
` Figure 1 – Current UTRA RACH preamble.
`
`Structure 1 that is simulated using the E-UTRA parameters has a duration of 1 ms and has an excess
`bandwidth factor of 0.15. For the 5 MHz bandwidth, Structure 1 is identical to Figure 1, and for the
`1.25 MHz bandwidth the RACH preamble has 1024 samples in order to maintain the 1 ms duration.
`Table 1 gives the simulation assumptions for the RACH preambles simulated in this contribution.
`
`- 1/7 -
`
`SAMSUNG 1009-0001
`
`

`
`Table 1: Simulation Assumptions for RACH preambles for E-UTRA.
`Parameter
`Assumption
`Bandwidth
`1.25 MHz and 5 MHz
`Carrier Frequency
`2 GHz
`Excess Bandwidth Factor
`0.15
`Sampling Rate
`1.024 MHz (BW=1.25MHz) and
`4.096 MHz (BW=5 MHz)
`1 ms = 2 TTI’s = 14 OFDM symbols
`1024 samples (BW=1.25MHz) and
`4096 samples (BW=5 MHz)
`FFT=64, CP=9.14 samples (BW=1.25MHz) and
`FFT=256, CP=36.57 samples (BW=5 MHz)
`TU, with UE speed of 3 kmph
`1 at Transmitter, 2 at Receiver
`Long code
`Long code with repetition for IFDMA
`Time Domain Correlator
`+/- 0.5 OFDM Symbols
`
`RACH Preamble Duration
`RACH Preamble Samples
`
`FFT and CP Sizes
`
`Channel Model
`Antenna Configuration
`RACH Preamble
`Structure 1 (time domain)
`Structure
`Structure 2 (freq domain)
`Receiver Structure
`Search Window Size
`
`
`
`Structure 2 uses IFDMA with a repetition of 4 in order to reduce the bandwidth occupancy of the
`RACH preamble. It is identical to Structure 1 except that for each OFDM symbol the first 1/4 of the
`useful part of the OFDM symbol is repeated 4 times to form the OFDM symbol and then the cyclic
`prefix is inserted. Figure 2 shows the construction of Structure 2. First the long code and repeated
`Hadamard code of Figure 1 are applied. For the 5 MHz bandwidth there are 256 samples in the useful
`part of the OFDM symbol, so the first 1/4 of the OFDM symbol consists of 64 samples which
`corresponds to 4 Hadamard codes of length 16. Thus, while the long code generator produces outputs
`every sample, 64 samples are stored for each OFDM symbol and are used to construct the entire
`OFDM symbol.
`
`
`
`
`For each OFDM symbol duration copy first
`1/4 symbol into CP and other parts of the
`symbol to form IFDMA transmission
`
`
`
`Figure 2: Structure 2 using IFDMA.
`
`
`
`3. Simulation Results
`
`The Node B uses a similar receiver for the RACH preamble as that used for UTRA. A bank of parallel
`correlators is used with half-chip resolution, and the largest correlation output is selected [4]. One
`difference is that two receiver antennas are assumed since this is the baseline assumption for E-UTRA.
`
`- 2/7 -
`
`SAMSUNG 1009-0002
`
`

`
`Figure 3 illustrates the receiver structure. Note that in these simulations the threshold was not
`simulated.
`
`
`Ant 1
`
`Ant 2
`
`Delay=0 x 
`
`Correlator
`
`Delay=1 x 
`
`Correlator
`
`…
`
`…
`
`Delay=N x 
`
`Correlator
`
`Delay=0 x 
`
`Correlator
`
`Delay=1 x 
`
`Correlator
`
`…
`
`…
`
`Delay=N x 
`
`Correlator
`
`1) Combine
`outputs for
`each delay
`noncoherently
`2) Select the
`largest
`correlation
`output
`3) Compare
`the largest
`with a
`threshold*
`
`
`Figure 3: RACH preamble receiver structure. *Threshold was not simulated here.
`
`
`
`Figure 4 shows an example of the correlation output for Structure 1 for the 1.25 MHz bandwidth for
`the AWGN channel with SNR=10 dB. Note that there is a single peak corresponding to the correct
`timing since the long code is effective in suppressing any sidelobes.
`
`
`Figure 4: Example correlation output for Structure 1 (AWGN channel, 1.25 MHz, 10 dB SNR).
`
`
`- 3/7 -
`
`SAMSUNG 1009-0003
`
`

`
`
`Figure 5 shows an example of the correlation output for Structure 2 for the 1.25 MHz bandwidth for
`the AWGN channel with SNR=10 dB. Now there is a peak corresponding to the correct timing but
`also two smaller peaks 1/4 OFDM symbol away. This is caused by the repetition of the same sequence
`4 times within each OFDM symbol. When the correlator aligns with 3 out of the 4 repetitions, there is
`a significant sidelobe. There are also smaller peaks corresponding to an overlap of 2 and 1 of the
`repetitions. These sidelobes decrease the probability of the Node B receiver locking onto the correct
`RACH preamble timing.
`
`
`
`
`Figure 5: Example correlation output for Structure 2 (AWGN channel, 1.25 MHz, 10 dB SNR).
`
`Figure 6 compares the RACH preamble detection performance for 1 RACH preamble for Structures 1
`and 2 for the 1.25 MHz channel. In this simulation the receiver computes the detection metric for all
`16 Hadamard codes, and an error is declared if the wrong Hadamard code has the maximum metric or
`if the timing is off by more than one CP length. Structure 2 (IFDMA) had a loss of between 1 and 2 dB
`because of the timing errors from the sidelobes due to IFDMA.
`
`
`- 4/7 -
`
`SAMSUNG 1009-0004
`
`

`
`
`Figure 6: RACH preamble detection performance for 1.25 MHz channel (TU, 3 kmph).
`
`Figure 7 compares the RACH preamble detection performance for 1 RACH preamble for Structures 1
`and 2 for the 5 MHz channel. The preamble sequence for the 5 MHz channel is 4 times the length of
`the sequence for the 1.25 MHz channel, so there is a reduction of about 6 dB in the required SNR for
`detection.
`
`
`Figure 7: RACH preamble detection performance for 5 MHz channel (TU, 3 kmph).
`
`It may be advantageous for the RACH preamble to occupy only 1.25 MHz of the available bandwidth
`for the 5 MHz channel. This will allow other traffic to be scheduled without interference from the
`RACH. Figure 8 compares the RACH preamble detection performance for 1 RACH preamble for
`LFDMA and IFDMA which occupies 1.25 MHz of the 5 MHz channel. The LFDMA structure shows
`an improvement in the range of about 0.8 to 1.6 dB over the IFDMA structure. While there is a small
`loss in diversity with LFDMA, the IFDMA approach suffers from the multiple sidelobes and timing
`
`- 5/7 -
`
`SAMSUNG 1009-0005
`
`

`
`errors. The current RACH preamble (Structure 1) with a 1.25 MHz bandwidth seems to be a good
`choice for E-UTRA.
`
`
`Figure 8: Comparison of LFDMA (occupies 1.25 MHz out of 5 MHz) and IFDMA (occupies 1.25
`MHz out of 5 MHz using comb) (TU, 3 kmph).
`
`
`
`Figure 9 shows the performance of the RACH preamble detection for the 1.25 MHz channel when the
`effect of timing errors was not considered. This simulation was done to verify that the reason for the
`poorer performance of the IFDMA preamble was due to timing uncertainty because of the time domain
`repetitions. The detection error criterion was modified to declare an error only when the Walsh code
`was misidentified. Note that the two structures perform similarly, so the difference in performance is
`due to timing errors for the IFDMA preamble.
`
`
`Figure 9: Comparison of LFDMA and IFDMA for the 1.25 MHz channel ignoring timing errors (TU, 3
`kmph).
`
`- 6/7 -
`
`SAMSUNG 1009-0006
`
`

`
`4. Conclusion
`
`This contribution presented a comparison of two structures for the RACH preamble: an LFDMA
`structure which is similar to the current UTRA RACH preamble and an IFDMA structure. Since the
`IFDMA structure is formed by making repetitions in the time domain, there is an increased probability
`of incorrect timing recovery with IFDMA. LFDMA does not have the time repetitions and exhibits
`better performance for RACH preamble detection.
`
`
`References
`[1] TS 25.211, “Physical channels and mapping of transport channels onto physical channels (FDD)”
`[2] TS 25.213, “Spreading and modulation (FDD)”
`[3] 3GPP, R1-99893, Motorola and Texas Instruments, “Proposal for RACH preambles”
`[4] Park and Kang, “On the performance of a maximum-likelihood code-acquisition technique for
`preamble search in a CDMA reverse link,” IEEE Transactions on Vehicular Technology, Vol. 42,
`No. 1, pp. 65-74, February 1998.
`
`- 7/7 -
`
`SAMSUNG 1009-0007