The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC'07)

Uplink Control Channel Design for 3GPP LTE

Amitava Ghosh, Rapeepat Ratasuk, Weimin Xiao
Motorola Networks 1501 West Shure Drive, Arlington Heights, IL 60004, USA

Brian Classon, Vijay Nangia

Motorola Labs 1301 E. Algonquin Rd, Schaumburg, IL 60196, USA

Robert Love, Dale Schwent, David Wilson

Motorola Mobile Devices 600 North US Highway 45, Libertyville, IL 60048, USA

Abstract Long term evolution (LTE) of the UMTS Terrestrial Radio Access and Radio Access Network is aimed for commercial deployment in 2010. Goals for the evolved system include support for improved system capacity and coverage, high peak data rates, low latency, reduced operating costs, multi-antenna support, flexible bandwidth operations and seamless integration with existing systems. To reach these goals, a new design for the air interface is currently being specified in the 3GPP standards body. The Uplink (UL) for LTE is based on Single Carrier Frequency Division Multiple Access. The UL control channel carries non-data associated control signaling like CQI, ACK/NACK, Scheduling request etc. To maintain the low PA power de-rating, the single carrier property of the UL has to be maintained. As such, special consideration should be given to the UL control channel design. This paper discusses in detail the LTE UL control channel design and its performance. I. Introduction Evolved UMTS Terrestrial Radio Access (E-UTRA) is aimed at commercial deployment around 2010 timeframe. Long term goals for the system include support for high peak data rates (100 Mbps downlink and 50 Mbps uplink), low latency (10ms round-trip delay), improved system capacity and coverage, reduced operating costs, multi-antenna support, efficient support for packet data transmission, flexible bandwidth operations (up to 20 MHz) and seamless integration with existing systems. To reach these goals, a new design for the air interface is envisioned. Single-Carrier Frequency Division Multiple Access (SC-FDMA) is selected to efficiently meet E-UTRA performance requirements for Uplink (UL). SCFDMA has many similarities to OFDM, chief among them is that frequency domain orthogonality is maintained among intra-cell users. SC-FDMA also has a low power amplifier derating (Cubic Metric / PAPR) requirement, thereby conserving battery life or extending range. The UL control channel is designed to carry non-data associated control signaling like CQI, ACK/NACK, Scheduling Request (SR) etc. To maintain the single carrier property of the UL special consideration should be given to UL control channel design. This paper provides a detailed view on LTE UL control channel design and its performance. The paper is organized as follows. In Section II, an overview of E-UTRA uplink structure and frame format is provided. Section III, provides a summary of the uplink control signaling. In Section IV, multiplexing details of uplink control

signaling in the presence of data is discussed. conclusions are drawn in Section V. II. Overview of SC-FDMA UL

Finally,

In the uplink, Single-Carrier Frequency Division Multiple Access (SC-FDMA) is selected to efficiently meet E-UTRA performance requirements. SC-FDMA has many similarities to OFDM, chief among them for the uplink is that frequency domain orthogonality is maintained among intra-cell users to manage the amount of interference seen at the base station. SCFDMA also has a low power amplifier de-rating (Cubic Metric / PAPR) requirement, thereby conserving battery life or extending range. The baseline SC-FDMA signal is DFT-Spread OFDM (DFT-SOFDM) [2] as shown in Figure 1. The only difference from OFDM is the addition of the M-point FFT (DFT) in the figure which spreads M symbols onto the M subcarriers selected by the symbol to subcarrier mapping. The selected subcarriers must also be either adjacent or evenly spaced to maintain the low PA power de-rating. The signal is considered single carrier as the first M-point FFT and the larger N-point IFFT cancel each other resulting in a single carrier signal in the time domain. The receiver can use simple frequency domain equalization.

Figure 1 - Block diagram for DFT-SOFDM. An advantage for DFT-SOFDM as a SC-FDMA technique is that the numerology can match the OFDM downlink, with excellent spectral occupancy due to the IFFT providing pulse shaping of the signal. Table 1 provides the uplink subframe numerology for different spectrum allocations. The OFDM numerology provides for an additional vacant DC subcarrier to simplify some receiver architectures; a vacant subcarrier cannot be used with DFT-SOFDM without affecting the low PA de-rating property of DFT-SOFDM. In this case, normal cyclic prefix is applied as shown in the table. An alternate frame structure with extended cyclic prefix length of 16.67 s is available to handle large cells.

1-4244-1144-0/07/$25.00 2007 IEEE.

The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC'07)

Table 1. Parameters for UL transmission scheme.

Spectrum Allocation (MHz) 20 15 10 5 3 1.4 SC-FDMA (s/#of occupied subcarriers /samples) 66.67/1200/2048 66.67/900/1536 66.67/600/1024 66.67/300/512 66.67/144/256 66.67/72/128 (4.69 s) 12, (5.21 s) 2 CP duration (s)

III. UL Control Signaling In principle, uplink control signaling can be divided into two categories: data-associated and data non-associated control signaling. Data-associated control signaling is always transmitted with and used in the processing of data packet. Examples of this control signaling include transport format, new data indicator, and MIMO parameters. In LTE it was agreed that this type of control signaling is not necessary. Control signaling not associated with data is transmitted independently of uplink data packet. Examples of this control signaling include ACK/NACK, CQI, and MIMO codeword feedback. When users have simultaneous uplink data and control transmission, control signaling is multiplexed with data prior to the DFT to preserve the single-carrier property in uplink transmission. In the absence of uplink data transmission, this control signaling is transmitted in a reserved frequency region on the band edge as shown in Figure 3. Note that additional control regions may be defined as needed.

The physical uplink shared channel is defined by one subframe and the parameters NTx and k0, used in the generation of the SC-FDMA signal. The variables NTx and k0, determining the transmission bandwidth and the frequency hopping pattern, respectively, are under control of the uplink scheduler and may vary on a per-sub-frame basis. The number of SCFDMA symbols in a slot depends on the cyclic prefix length configured by higher layers. The uplink slot format (a subframe consists of two slots) with normal cyclic prefix (CP) is shown in Figure 2 with seven SC-FDMA symbols. For frames with extended cyclic prefix, only six SC-FDMA symbols are present. The uplink supports QPSK, 16-QAM and 64-QAM modulation.
Tcp Td LB Data LB RS 0.5 ms

Figure 3. Control regions for uplink. Allocation of control channels with their small occupied bandwidth to carrier band edge resource blocks reduces out of carrier band emissions cause by data resource allocations on inner band resource blocks and maximizes the frequency diversity benefit for frequency diverse control channel allocations while preserving the single carrier property of the uplink waveform. This FDM allocation of control resources to outer carrier band edge allows an increase in the maximum power level as shown in Figure 4 as well as maximizes the assignable uplink data rate since inserting control regions with consecutive subcarriers in the central portion of a carrier band requires that the time+frequency resources on either side of the control region to be assigned to different UEs.
28.0

Figure 2. Uplink slot format. Two types of reference signals (RS) are supported on the uplink - (a) demodulation reference signal, associated with transmission of uplink data and/or control signaling and (b) sounding reference signal, not associated with uplink data transmission used mainly for channel quality determination if channel dependent scheduling is used. Orthogonality of reference signals is obtained via frequency domain multiplexing onto distinct set of sub-carriers. The RS sequence length is equal to the number of sub-carriers in the resource blocks. The RS sequence is generated either by truncation or cyclic extension of ZC (Zadoff-Chu) sequences depending on the allocation size. It was observed that for a given size, neither truncation nor cyclic extension was the best. Many options exist for selecting either truncation or cyclic extension RS construction method, including: 1. Choose the method that for a given resource block (RB) size minimizes the amount of truncation or cyclic extension, 2. Choose the method that for a given RB size maximizes the number of sequences with Cubic Metric <= the target data modulation (e.g., QPSK).

Max Power Level (dBm)

QPSK - 5MHz, Band Edge RBs for Data 27.0 26.0 25.0 24.0 23.0 22.0 0 2 4 6 8 10 12 14 16QAM - 5MHz, Band Edge RBs for Data QPSK - 5MHz, Band Edge RBs for Ctl 16QAM - 5MHz, Band Edge RBs for Ctl
Max Power Practical Limitation due to EVM and other considerations

#RBs of size 25 subcarriers

Figure 4. Increase in maximum power level if control is mapped to band edge.

The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC'07)

Table 2 provides the required quality targets for uplink control signaling. Table 2. Uplink control signaling target quality. Event ACK miss detection DTX to ACK error NACK to ACK error CQI block error rate A. Channel Quality Information The CQI structure is shown in Figure 5. The transmission spans the entire 1ms sub-frame and up to six users may be multiplexed within the sub-frame via different cyclic shifts of a Constant Amplitude Zero Auto-Correlation (CAZAC) sequence, e.g. Zadoff-Chu sequence. Data is modulated on top of the CAZAC sequence using QPSK modulation. Target quality (1e-2) (1e-2) (1e-4) FFS (1e-2 1e-1)

In Figure 5, only one reference signal per slot is shown. In general, the number of reference signals required depends on the feasibility of using an advanced receiver to perform joint channel estimation and decoding. This in turn depends on the number of CQI bits to be supported. Two types of receivers are possible Type 1: Channel estimation is first done based on the reference signals, and then CQI decoding is performed based on these channel estimates. Type 2: Channel estimation and decoding is done jointly using all possible CQI codewords. While this receiver is more complicated than Type 1 receiver, complexity is manageable for the CQI codeword length being considered (10 bits or less). Performance comparison between the two receiver types is shown in Figure 4 with Type 2 outperforming Type 1 receiver by approximately 2-3 dB. This is because, for this receiver, channel estimation is aided by CQI codeword detection.
10
0

10-bit CQI, TU, QPSK, Non-Ideal Chan Est 3 km/h 120 km/h 350 km/h

10

-1

BLER

10

-2

Receiver Type 2 (24,10), 1 RS

Receiver Type 1 (20,10), 2 RS

Figure 5. CQI channel structure. The number of CQI bits may vary between 5-10 bits depending on whether wideband or narrowband CQI reports are transmitted. However, larger CQI reports may be transmitted using multiple subframes. In addition, repetition may be used to ensure reliable reception from cell edge users. An example of CQI performance is shown in Figure 6 for various coding schemes.
10
0

10

-3

-15

-10 -5 SNR (dB) per antenna

Figure 7. CQI performance with advanced receiver. B. ACK/NACK Figure 8 illustrates the ACK/NACK channel structure. Note that in this case only acknowledgment is present (no CQI or data). To provide the maximum number of multiplexed users, both frequency domain and time domain code multiplexing are used. In the frequency domain, different cyclic shifts of a CAZAC sequence are used to differentiate users. For instance, with sequence length of 12 corresponding to one resource block, 6 available cyclic shifts are possible. In the time domain, block spreading is used to further multiplex additional users. For instance, within each cyclic shift of a Zadoff-Chu sequence, reference signals are multiplexed using DFT code of length 3 while the acknowledgments are multiplexed using Walsh-Hadamard code of length 4. As a result, acknowledgments from 18 different users may be multiplexed within one resource block. The ACK/NACK is then modulated onto the frequency and time-spread sequence. Both 1-bit and 2-bit acknowledgements are supported using BPSK and QPSK modulations.

CQI (5-bit, 10-bit), TU, QPSK, Receiver Type 2, Non-Ideal Chan Est

10

-1

BLER

10

-2

10

-3

5-bit CQI, (32,10) Reed-Muller 5-bit CQI, Convolutional 5-bit CQI, (24,12) Golay 10-bit CQI, (32,10) Reed-Muller 10-bit CQI, Convolutional 10-bit CQI, (24,12) Golay -20 -15 -10 SNR per antenna (dB) -5 0

10

-4 0017

Figure 6. CQI performance with various coding schemes.

The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC'07)
10
0

5-bit CQI, TU (3 km/h), Receiver Type 2, Non-Ideal Chan Est CQI BLER - (20,5) Code 1-bit ACK/NACK SER

10

-1

Error Rate

10

-2

Figure 8. ACK/NACK structure - users are multiplexed using different cyclic shifts and time-domain spreading. Figure 9 shows performance of 1-bit acknowledgments from 18 multiplexed users. Although not shown here, for 2-bit acknowledgments the performance is approximately 3dB worse.
10
0

10

-3

10

-4

-12

-10

-8

-6 -4 SNR (dB) per antenna

-2

ACK/NACK Performance - 18 users, GSM-TU (3 km/h) 3 km/h 350 km/h

Figure 10. Performance of 5-bit CQI and 1-bit ACK/NACK (BPSK) at TU 3 km/h. IV. Multiplexing of Control and Data To preserve the single-carrier property of uplink transmission, L1/L2 control signaling must be multiplexed with data prior to the DFT when both data and control are to be transmitted in the same TTI. This may be performed as shown in Figure 11 where uplink data is uniformly punctured to provide room for control signaling. Naturally, in case of turbo coding, puncturing is only performed on the parity bits. Since the Node B has prior knowledge of uplink control signaling transmission, it can easily de-multiplex control and data packets. In addition, a power boosting factor may be applied when data is punctured to ensure similar data packet performance to when control is absent. This is especially important in the case of re-transmission since the data MCS cannot be changed due to synchronous H-ARQ operation in the uplink. This appropriate power boosting factor (in the order of 0.5-1.5dB) can be calculated based on the coding rate reduction resulting from puncturing. With appropriate power adjustment there should be little effect on the H-ARQ performance at the receiver. Of course, power boosting is not possible when the UE is power-limited (e.g. at the cell edge).

10

-1

BER

10

-2

10

-3

10

-4

-20

-18

-16

-14 SNR (dB)

-12

-10

-8

Figure 9. Performance of 5-bit CQI and 1-bit acknowledgments (BPSK) at TU 3 km/h. A. CQI + ACK/NACK When CQI and ACK/NACK are to be transmitted simultaneously, they are coded separately and multiplexed in a TDM fashion. This allows greater control of CQI and ACK/NACK error requirements, and the ability to multiplex ACK/NACK into CQI reports that are transmitted in multiple sub-frames (either for large CQI report or through the use of repetition) once CQI transmission has started. Figure 10 illustrates the performance of 5-bit CQI + ACK/NACK under TU 3 km/h channel with realistic channel estimation. In this case, one SC-FDMA symbol per slot was used for ACK/NACK. As an alternative, scheduling restriction may be used to ensure that CQI and ACK/NACK will not be transmitted in the same sub-frame. However, this may place unnecessary and complicated constraint on the scheduler. Alternately, only ACK/NACK can be transmitted (CQI is not transmitted in the sub-frame). This may result in some scheduling and resource allocation efficiency loss as some CQI reports will be lost.

Data

Puncturing / Insertion
N TX symbols

DFT

Sub-carrier Mapping

IFFT

CP Insertion

Control

Gain Factor

Size-NTX

Size-N FFT

Figure 11. Multiplexing of control signaling with data. Figure 12 illustrates typical performance degradation due to turbo-code puncturing for both QPSK and 16-QAM. From the figure, it is seen that the performance loss depends on the initial coding rate. However, it may be observed that in general the amount of resources required to accommodate control information is small and less than 1dB degradation can be expected. As a result, appropriate power boosting should be comfortably accommodated unless the UE is already in a power-limited situation (e.g. cell edge transmission).

The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC'07)
4 3.5 3 Puncturing Penalty (dB) 2.5 2 1.5
ACK

QPSK 16-QAM

variable size which must be taken care of by the rate-matching algorithm.

CQI

Coding

Repetition

MUX

Modulation

Coding

Repetition

1 0.5 0 -0.5

Figure 14. Mapping to multiple codewords. Since control is multiplexed with data prior to the DFT, appropriate modulation and coding selection for control is required for reliable reception. As a result, the amount of coded data to be punctured is variable based upon the MCS selected for control. In this case, rate matching may be done in one step. With one-step rate matching, the number of bits punctured for control is factored in when computing the effective coding rate.
VI. Conclusions

0.35

0.4

0.45

0.5

0.55 0.6 Code Rate

0.65

0.7

0.75

0.8

Figure 12. Performance loss due to puncturing (turbo code). Since both control and data must be transmitted with the same power, reliable reception of control information can be achieved through appropriate selection of modulation and coding. Since these control fields are generally small, codeword mapping is use to provide additional protection. Subsequent to codeword mapping, repetition (if necessary) and modulation selection are performed according to information about the channel. Obviously, this selection can be tied to the MCS of the data block to aid in the decoding. In addition, it should also depend on the uplink data transmission method (LFDMA or L-FDMA with hopping). This is because these two localized transmission methods have different target error rates for the same selected MCS. As a result, control power requirement relative to the two transmission methods is different. Two possible codeword mappings for the control signaling are as follows (a) Single codeword: In this case, all control fields are mapped into a single codeword (i.e. jointly coded) as shown in Figure 13. If all fields are not present, dummy input values are inserted which are then ignored at the Node B. Alternatively, the UE may use the available fields to transmit some additional information based on an agreed upon methodology (e.g. UE that does not support MIMO may transmit wideband CQI in the MIMO field). This results in codeword of the same length which may simplify the multiplexing and de-multiplexing process. However, with this approach it may be difficult to satisfy performance requirements of different control fields. Also, overhead is higher.
(CQI, ACK, MIMO, ...)

This paper provided an overview of the UL control channel design for 3GPP LTE. REFERENCES [1] 3GPP TR 25.913, Requirements for Evolved UTRA (EUTRA) and Evolved UTRAN (E-UTRAN), v.7.3.0, March 2006. [2] 3GPP TR 25.814, Physical Layer Aspects for Evolved UTRA, v.2.0.0, June 2006.
[3] R1-070777, E-UTRA Multiplexing of UL Control Signaling with Data, Motorola, RAN1#48, St. Louis, USA, Feb 2007.

[4] R1-070394, Multiplexing of L1/L2 control signals between UEs in the absence of data, Nokia, RAN1#47bis, Sorrento, Italy, Jan. 2007. [5] R1-070782, Multiplexing of UL L1/L2 control signals in the
absence of data, Motorola, RAN1#48, St. Louis, USA, Feb 2007.

[6] R1-070162, EUTRA UL L1/L2 Control Channel Mapping, Motorola, RAN1#47bis, Sorrento, Italy, Jan. 2007.
[7] R1-070778, CQI Feedback Overhead with CDM Uplink Control Channel Region, Motorola, RAN1#48, St. Louis, USA, Feb 2007.

Coding

Repetition

Modulation

Figure 13. Mapping to one codeword.

(b) Multiple codewords: In this case, each control field is individually mapped to a codeword with its own repetition factor as shown in Figure 14. This allows individual adjustments of transmission energy using different coding and repetition so that performance of each control field can be controlled. However, this results in a control portion of

[8] R1-070275, Ack/Nack transmission without reference signal overheadin E-UTRA UL, TI, RAN1#47bis, Sorrento, Italy, Jan. 2007. [9] R1-070472, Uplink control Signaling Summary of email discussions, Ericsson, RAN1#47bis, Sorrento, Italy, Jan. 2007.
Note 3GPP documents may be downloaded from ftp://ftp.3gpp.org