3GPP TSG RAN WG1#48 R1-070734
`Saint Louis, USA, 12 – 16 February, 2007
`Source:
` Texas Instruments
`Title:
`ACK/NAK Channel Transmission in E-UTRA Downlink
`Agenda Item:
` 6.9.2
`Document for:
`Discussion and Decision
`1. Introduction
`ACK/NAK transmission in E-UTRA downlink (DL) has been previously considered in [1-3]. Several
`principles are generally accepted.
`
`The first principle is that the UE ID should not accompany the ACK/NAK transmission as the
`corresponding overhead (e.g. 16-bit UE ID) cannot be possibly accepted. Instead, implicit mapping
`should be used to derive the UE ID from the ACK/NAK signaling.
`
`The second principle is that ACK/NAK transmission should exploit the frequency diversity the channel
`provides and should provide interference randomization/averaging. Two basic schemes have been
`proposed; FDM [1] and CDM [2, 3] with Walsh-Hadamard (WH) orthogonal spreading. Each of the
`proposed schemes has certain trade-offs.
`
`Individual FDM for the ACK/NAK transmission [1] has the following attributes:
`a) it allows for frequency diversity (with ACK/NAK repetition)
`b) it allows for individual transmission power control (TPC)
`c) it does not allow for interference randomization/averaging as each ACK/NAK is located on few
`sub-carriers making it susceptible to interference in general and especially with use of TPC
`d) it does not allow for joint power balancing among several ACK/NAK signals which can lead to
`significant power variations per ACK/NAK sub-carrier and ineffective TPC for cell edge UEs.
`
`
`CDM for the ACK/NAK transmission [2, 3] has the following attributes:
`a) it allows for frequency diversity
`b) it allows for both individual TPC and joint power balancing of multiple ACK/NAKs
`c) it allows for interference randomization/averaging (spreading gain)
`d) it relies on WH orthogonality to achieve previous properties. However, this orthogonality is
`destroyed in frequency selective channels assuming transmission substantially over the entire
`bandwidth for frequency diversity. Then, the ACK/NAK signals interfere (transmission is no
`longer orthogonal) and, with the application of TPC, this leads to “near-far” effects which
`severely limit the ACK/NAK detection performance (as demonstrated in this contribution).
`
`
`To exploit the advantages of FDM and CDM while avoiding their corresponding shortcomings, hybrid
`CDM/FDM is considered for the ACK/NAK transmission for the same reasons it has been adopted for
`the ACK/NAK transmission in the UL.
`
`2. Hybrid CDM/FDM
`Figure 1 shows the hybrid CDM/FDM transmission for DL ACK/NAK. Its basic principles are:
`a) Localized WH spreading over a narrow enough BW for the channel to have no frequency
`selectivity, thereby preserving CDM orthogonality.
`
`- 1/9 -
`
`BlackBerry Exhibit 1005, pg. 1
`
`

`
`b) Localized WH spreading is repeated in frequency several times to capture the frequency
`diversity of the channel. The frequency separation between two repetitions should be large
`enough for the frequency response of the channel to substantially change and small enough to
`capture all or most such changes.
`
`POWER
`
`
`
`Walsh 12
`
`CDM ACK/NAK
`
`...
`
`Other L1/L2 Control
`
`Walsh 4
`
`Walsh 3
`
`Walsh 2
`
`Walsh 1
`
`CQI – based
`-
`TPC
`
`FREQUENC
`
`
`
`Walsh 12
`
`...
`
`Walsh 4
`
`Walsh 3
`
`Walsh 2
`
`Walsh 1
`
`Other L1/L2 Control
`
`Walsh 12
`
`...
`
`Walsh 4
`
`Walsh 3
`
`Walsh 2
`
`Walsh 1
`
`Walsh 12
`
`...
`
`CDM ACK/NAK
`
`Other L1/L2 Control
`
`Walsh 4
`
`Walsh 3
`
`Walsh 2
`
`Walsh 1
`
`Figure 1: Downlink ACK/NAK Transmission.
`
`Figure 1 considers WH sequences with length 12 which may be appropriate for 5 MHz operating BW
`having a maximum of 6 simultaneously scheduled UEs and a maximum of 6 VoIP groups. If VoIP does
`not support HARQ, a smaller WH sequence length (e.g. 8) can be used instead.
`
`The WH sequence length, and therefore the number of occupied sub-carriers, can be static, semi-static,
`or dynamic. A constant (static) WH sequence length is selected to always accommodate the maximum
`possible number of ACK/NAKs (Figure 1). However, the larger the WH sequence, the larger the
`number of required sub-carriers to exploit the frequency diversity of the channel. This increase in
`bandwidth is offset by a corresponding reduction in power. Nevertheless, as bandwidth is more
`valuable than power (relative to capacity), the WH sequence length should only be as large as required
`to convey a certain number of ACK/NAK signals. Figure 2 shows a possible semi-static or dynamic
`WH sequence length adaptation between two sub-frames.
`
`The WH sequence length adaptation can be based on Cat0 of the L1/L2 control channel. For example,
`not considering VoIP groups for clarity (assuming VoIP employs HARQ), if Cat0 specifies that the
`maximum number of UL grants during a sub-frame is 7, a WH sequence of length 8 can be used for the
`ACK/NAK transmission. Similarly, if Cat0 specifies a maximum number of 4 UL grants during a sub-
`frame, the corresponding WH sequence length at a given subsequent sub-frame does not need to be
`larger than 4. The WH sequence length can vary between Cat0 transmission periods. Figure 2
`illustrates an exemplary application of this concept.
`
`
`- 2/9 -
`
`BlackBerry Exhibit 1005, pg. 2
`
`

`
`Other L1/L2 Control
`
`CDM ACK/NAK
`
`W44
`
`W43
`
`W42
`
`W41
`
`FREQUENC
`
`CQI – based
`-
`TPC
`
`W88
`
`...
`
`CDM ACK/NAK
`
`Other L1/L2 Control
`
`W84
`
`W83
`
`W82
`
`W81
`
`CQI – based
`-
`based
`TPC
`
`
`
`W44
`
`W43
`
`W42
`
`W41
`
`W88
`
`...
`
`W84
`
`W83
`
`W82
`
`W81
`
`Other L1/L2 Control
`
`TTI N
`
`Other L1/L2 Control
`
`W44
`W43
`
`W42
`
`W41
`
`W88
`
`...
`
`W84
`W83
`
`W82
`
`W81
`
`Other L1/L2 Control
`
`POWER
`
`CDM ACK/NAK
`W44
`
`W43
`
`W42
`
`W41
`
`POWER
`
`W88
`
`...
`
`CDM ACK/NAK
`
`Other L1/L2 Control
`
`W84
`W83
`W82
`
`W81
`
`TTI N+M
`
`FREQUENCY
`
`Figure 2: Adaptive Selection of the WH Sequence Size for ACK/NAK transmission.
`
` A
`
` suggestion for implicit mapping of the ACK/NAK to the respective UE is to have the number of the
`WH sequence used to convey the ACK/NAK be the same as the number of the first RB where the UE
`was scheduled [2]. This approach has the following two shortcomings:
`a) it cannot be used with MU-MIMO where a conflict exists
`b) it assumes a fixed, maximum, WH length regardless of the number of scheduled UEs.
`
`
`To improve the flexibility of the implicit ACK/NAK mapping, Cat0 can again be used and the mapping
`between the WH sequence conveying the ACK/NAK signal and the intended UE can be based on the
`code-word number where the UE finds its UL grant. For example, if Cat0 specifies that the number of
`UL grants during a sub-frame has a maximum value of P, a number of P code-words conveying UL
`grants is expected by each UE. If a UE finds its UL grant in code-word Q (with Q being smaller than or
`equal to P), it can then expect the WH sequence number Q to be used to convey the associated
`ACK/NAK at a specified subsequent sub-frame. Clearly, the WH sequence used should have length
`larger than or equal to P. Typically, if Cat0 is transmitted in every sub-frame to dimension that
`remaining L1/L2 control channel, all P L1/L2 control channel code-words have an UL grant. If Cat0 is
`transmitted less often than every sub-frame, one or more of these P L1/L2 code-words may not have an
`UL grant (no signal is transmitted if the number of UL grants is less than P (“empty” sub-carriers) – the
`UL grants between two consecutive Cat0 transmissions cannot exceed the maximum value specified by
`the most recent Cat0 transmission).
`
`An alternative to using larger WH sequences to convey a larger number of ACK/NAK signals is to use
`multiple shorter WH sequences. This is applicable for the larger operating bandwidths (e.g. 5 MHz and
`above) where the maximum number of ACK/NAK signals would necessitate longer maximum WH
`
`- 3/9 -
`
`BlackBerry Exhibit 1005, pg. 3
`
`

`
`sequences thereby creating orthogonality issues and “near-far” effects in frequency selective channels
`even with localized spreading. Then, the multiplexing in Figure 1 can be equivalently accomplished
`with the multiplexing in Figure 3. The implicit mapping rule and occupied sub-carriers do not change
`and the only tradeoff is the somewhat reduced interference suppression capability which is offset by
`the absence of any orthogonality or “near-far” effects issues. The simulation results in the next section
`suggest that a WH sequence length of 4 or 8 and localized spreading completely avoids “near-far”
`effects in TU channels while some issues begin to appear for length 12.
`
`POWER
`
`CDM ACK/NAK
`W4 W4 W4
`W3 W3 W3
`W2 W2 W2
`
`Other L1/L2 Control
`
`W1 W1 W1
`
`W4 W4 W4
`W3 W3 W3
`W2 W2 W2
`
`W1 W1 W1
`
`Other L1/L2 Control
`
`W4 W4 W4
`W3 W3 W3
`W2 W2 W2
`
`W1 W1 W1
`
`Other L1/L2 Control
`
`W4 W4 W4
`W3 W3 W3
`W2 W2 W2
`
`W1 W1 W1
`
`CQI – based
`-
`TPC
`
`FREQUENC
`Figure 3: Using Multiple Localized Short WH Sequences for ACK/NAK transmission.
`
`
`
`
`
`- 4/9 -
`
`BlackBerry Exhibit 1005, pg. 4
`
`

`
`3. Performance Results
`The simulation assumptions are provided in the Appendix. Antenna diversity (2 Tx) and actual channel
`estimation are applied. The operating bandwidth is 5 MHz and the TU6 channel is assumed (it should
`be noted that no performance difference was observed for flat channels, such as the PA, as expected).
`The ACK/NAK transmission is assumed to occur at the beginning of the sub-frame (as for L1/L2
`control) to minimize latency. Several cases are considered including:
`a) Multiplexing length 4 and length 12 WH sequences (4 and 12 ACK/NAK signals)
`b) TPC application with various power levels for the ACK/NAK signals
`c) Distributed and localized with various repetition factors CDM
`
`
`Figures 4-6 provide the ACK/NAK BER when 4 and 12 ACK/NAK signals are transmitted having
`various power levels. The total power per sub-carrier is kept constant and the SNR is normalized for
`the WH sequence length (and hence normalized by the number of sub-carriers used for ACK/NAK
`transmission or equivalently the number of ACK/NAK signals). TPC is applied as:
`a) 4 ACK/NAK case: No power variation for 2 ACK/NAK, 3 dB boost for 1 ACK/NAK (cell edge
`UE with low G) and 3 dB reduction for 1 ACK/NAK (e.g. cell interior UE with high G).
`b) 12 ACK/NAK case: No power variation for 4 ACK/NAK, 3 dB power boost for 4 ACK/NAK,
`and 3 dB power reduction for 4 ACK/NAK.
`
`
`The following can be observed:
`a) The orthogonality loss for distributed CDM transmission in frequency selective channels leads
`to error floors which are particularly severe for signals transmitted with reduced power (“near-
`far” effect) and the larger the number of distributed sub-carriers, the more severe the impact.
`b) Frequency diversity is highly beneficial to performance. Most gains are achieved with 4
`repetitions of localized CDM transmission, which for 5 MHz transmission bandwidth implies
`repetition every 1.25 MHz (for the TU channel).
`c) CDM over 4 sub-carriers (60 KHz) is somewhat preferable to CDM over 12 sub-carriers (180
`KHz), especially for the ACK/NAK with reduced transmit power, as some channel selectivity
`does exist for the larger bandwidth and some manifestation of “near-far” effects is observed.
`
`
`
`
`Figure 4: BER for ACK/NAK without power variation.
`
`
`
`- 5/9 -
`
`BlackBerry Exhibit 1005, pg. 5
`
`

`
`
`
`
`Figure 5: BER for ACK/NAK transmitted with 3 dB reduced power.
`
`
`
`
`
`
`
`Figure 6: BER for ACK/NAK transmitted with 3 dB boosted power.
`
`
`Figures 7-9 present the ACK/NAK BER for the same setup as before but with a larger power boosting
`and reduction of 6 dB in order to examine rather extreme situations for “near-far” effects (stronger
`signals have 12 dB more power than weaker ones). Two additional observations apply for this setup:
`a) For the weaker signals, even a small loss in CDM orthogonality due to some channel selectivity
`results in noticeable losses (BER for ACK/NAK with 6 dB reduced power with WH sequence
`length of 12 versus a WH sequence length of 4). It should be noted that a more complex
`LMMSE receiver could substantially suppress the BER floor but would still not eliminate the
`performance loss particularly for UEs with high speed or low SINR.
`b) For the stronger signals having limited impact of “near far” effects and interference due to
`orthogonality destruction (for 4 ACK/NAK signals), frequency diversity with distributed CDM
`outweighs the interference avoidance due to orthogonality with localized CDM and only two
`repetitions (limited frequency diversity).
`c) BER less than 1% can be achieved for the 5% geometry CDF (< -5 dB).
`
`- 6/9 -
`
`BlackBerry Exhibit 1005, pg. 6
`
`

`
`
`Figure 7: BER for ACK/NAK without power variation.
`
`
`
`Figure 8: BER for ACK/NAK transmitted with 6 dB reduced power.
`
`
`Figure 9: BER for ACK/NAK transmitted with 6 dB boosted power.
`
`- 7/9 -
`
`
`
`
`
`
`
`BlackBerry Exhibit 1005, pg. 7
`
`

`
`Finally, for simulation control, the BER is provided in Figure 10 when all ACK/NAK signals have the
`same transmission power. For localized CDM, only a minor fractional dB difference exists between
`transmission over 12 sub-carriers (180 KHz) and 4 sub-carriers (60 KHz) implying that the less perfect
`CDM orthogonality over 180 KHz is not large enough for the interference effects to be significant.
`
`
`
`
`Figure 10: BER when all ACK/NAK signals have the same power.
`Additional simulation results (not included for brevity) showed that the ACK/NAK BER for
`transmission with a WH sequence of length 8 exhibits characteristics between those observed for WH
`sequence lengths of 4 and 12, as expected, but being closer to the ones for length 12 (sensitivity to
`“near-far” effects). Also, splitting the CDM distributed transmission over smaller bandwidths (e.g. two
`transmissions, each over one of the two 2.5 MHz bands, instead of one over 5 MHz) did not generally
`provide any noticeable benefits as CDM orthogonality is still practically destroyed.
`
`4. Discussion
`This contribution considered aspects for DL ACK/NAK signaling. Pure FDM is not recommended as
`without significant overhead and a large number of sub-carriers per ACK/NAK, the transmission is
`susceptible to interference. Also, TPC is possible only per individual ACK/NAK which results to
`interference spikes/valleys per sub-carrier which is detrimental particularly for cell edge UEs.
`Distributed CDM achieves interference randomization and averaging through the spreading gain but
`suffers from orthogonality destruction in frequency selective channels. To maintain CDM orthogonality,
`a hybrid CDM/FDM approach was considered with localized CDM repeated over several equally
`spaced frequency bands. This is practically the concept as the one adopted in the UL for this purpose.
`The hybrid CDM/FDM ACK/NAK transmission was shown to have robust performance under a
`variety of conditions, including rather extreme power differences in the transmitted signals, and
`provides all the desirable properties for the ACK/NAK transmission for any number of ACK/NAK
`signals.
`Frequency diversity is highly beneficial and the larger the number of repetitions for the localized CDM
`ACK/NAK transmissions, the better the performance. The tradeoff is that a larger number of repetitions
`require larger BW occupancy, although the power of transmissions can be correspondingly decreased.
`Nevertheless, as bandwidth is more valuable than power in terms of capacity, it is preferable that
`repetitions are over 1.25 MHz as nearly all frequency diversity gains (for the TU channel) are obtained
`for this repetition granularity. Also, as in typical operating environments a mixture of channels will be
`experienced with many having smaller frequency selectivity than the TU channel, a smaller granularity
`
`- 8/9 -
`
`BlackBerry Exhibit 1005, pg. 8
`
`

`
`in repetitions will provide even smaller gains than indicated by the simulation results. Moreover,
`repetition over 1.25 MHz provides immediate scalability for the considered operating bandwidths.
`To completely avoid any impact of the channel selectivity on the CDM orthogonality, CDM should be
`over 4 sub-carriers (superposition of 4 ACK/NAK signals). A larger number of ACK/NAK signals can
`be supported with additional CDM over 4 sub-carriers (as in Figure 3). The tradeoff is the somewhat
`reduced interference randomization/averaging due to the smaller spreading gains as the transmission
`power is distributed over a small number of sub-carriers. Alternatively, if extreme differences (e.g.
`above 10 dB) in the transmission power of ACK/NAK signals are avoided, CDM over 8 sub-carriers or
`even 12 sub-carriers (for smaller power differences) can be used.
`
`5. Conclusions
`Hybrid CDM/FDM is recommended for DL ACK/NAK signaling as it provides all desirable properties
`(frequency diversity, individual power adaptation for coverage and enhanced reliability for NAK,
`interference averaging and randomization, power balancing per sub-carrier among multiple ACK/NAK
`signals, applicability to all E-UTRA BWs and for any number of ACK/NAK signals) while avoiding
`the shortcomings of pure CDM or FDM. The following attributes are suggested for hybrid CDM/FDM
` Localized spreading over 4 or 8 sub-carriers; more than 4 or 8 ACK/NAK signals, respectively,
`can be accommodated with additional WH sequences (of length 4 or 8, respectively).
` Repetitions of localized CDM every 1.25 MHz. ACK/NAK overhead is less than 1% of the sub-
`frame (e.g. for 5 MHz and 8 ACK/NAK signals, the overhead is about 0.75% of the TTI).
`
`Implicit ACK/NAK mapping using the respective UL grant codeword number indicated by Cat0.
`Note that in the UL, hybrid CDM/FDM ACK/NAK transmission was chosen over the pure FDM one
`for the same reasons.
`
`References
`[1] R1-063207, “Downlink ACK/NACK Mapping for E-UTRA”, NEC
`[2] R1-063074, “Downlink Acknowledgment and Group Transmit Indicator Channels”, Motorola
`[3] R1-063326, “ACK/NACK Signal Structure in E-UTRA Downlink”, NTT Docomo, et. al.
`
`Appendix
`The simulation assumptions are provided in Table A1.
`
`
` Table A1: Link Level Simulation Assumptions.
`Parameter
`Assumption
`Bandwidth
`5 MHz (2.0 GHz)
`Channel Model
`TU6
`UE Speed
`3 Kmph
`Tx Antenna Configuration
`2 (SFBC)
`Rx Antenna Configuration
`2
`Time Interpolation/Averaging
`Linear – Doppler dependent coefficients
`Channel
`Estimation
`Frequency Interpolation
`Least Squares
`RS used for CE
`Previous Sub-frame, first Slot of Current Sub-frame
`
`
`
`- 9/9 -
`
`BlackBerry Exhibit 1005, pg. 9