NETWORK DIAGNOSTICS Testing HSDPA, HSUPA for 3G mobile apps

Transcription

1 NETWORK DIAGNOSTICS Testing HSDPA, HSUPA for 3G mobile apps By Simon Binar Protocol Monitoring Division Tektronix Inc. The market for broadband cellular data services is rapidly evolving. From its deployment phase several years ago, UMTS is slowly becoming a mainstream technology, with mid-priced handsets becoming accessible to a wider share of the end-user market. Thus, operators and manufacturers are looking for ways to deal with an increasing number of potential 3G subscribers and to improve users experience with these broadband services. This translates into improving the capacity of 3G networks and supporting higher peak data rates Figure 1:, which HSDPA introduces, is a shared channel that can be accessed by several users simultaneously. RLC MAC MAC-hs MAC-c/sh over the 384Kbps supported by Rel. 99 UMTS, while using the available 3G spectrum. The Third Generation Partnership Project (3GPP) has already recognized this need and was tasked to produce 3G standards based on W-CDMA. In 2000, a feasibility study was done in line with the 3GPP Rel-4 standardization activities. It aimed to identify possible techniques for improving capacity and achievable peak data rates specifically in the downlink. This resulted in the approval of a range of enhancements as part of 3GPP Rel-5, collectively denoted as High-speed Downlink Packet Access (HSDPA). Later, increasing interactive services in which content is generated rather than used by the end-user demanded improved uplink usage capacity. The 3GPP Rel-6 feature or FDD Enhanced Uplink, more commonly known as High-Speed Uplink Packet Access (HSUPA) is addressing this requirement. Both HSDPA and HSUPA introduce new functions in UTRAN, the 3G radio access network. Note that node Bs and RNCs must be upgraded with appropriate software. Like other technological advancements in cellular networks, both HSDPA and HSUPA must be thoroughly tested both at the functional and performance levels prior to deployment. Enhanced downlink An evolution of W-CDMA, HS- DPA supports peak data rates of 14.4Mbps in one cell. HSDPA introduces a new transport channel, high-speed downlink shared channel () (Figure 1, Table 1), plus two control channels for uplink and downlink. is a shared channel that can be accessed by several users simultaneously. It impacts several protocol layers, the most significant changes in physical and MAC layers. The new channel enables the high-throughput capabilities of HSDPA through the use of: RLC MAC-d UE Uu Node B Lub DRNC Lur SRNC Figure 2: HARQ is a protocol to handle re-transmissions and to guarantee error-free data transmission. Adaptive modulation and coding (AMC) A scheme in which modulation method and coding rate are selected based on information about channel conditions provided by the terminal and node B. In downlink, HSDPA supports 16QAM as a higher-order modulation method for data 2 transmission under good channel conditions, aside from QPSK. Hybrid automatic repeat re- Quest (HARQ) A protocol to handle re-transmissions and to guarantee error-free data transmission. It is a key element of the new MAC entity, MAChs, located both in node B and in user equipment (UE) (Figure 2). Fast packet scheduling algorithm This allocates HS- DSCH resources, such as time slots and codes, to different users, and is implemented as part of node B functionality. Based on this overview, some functions, previously reserved for the RLC protocol layer and to the Serving RNC (SRNC), have been moved down into the MAC protocol layer as well as into node B. The proximity of time-critical functions, like HARQ processing and packet scheduling, to the air-interface is essential, as transmission time interval (TTI) for HSDPA is specified at 2ms, one fifth of the minimum TTI specified for Rel. 99 W-CDMA. Hence, retransmissions as well as changes in modulation method and coding rate may take place every 2ms. This low TTI allows node B to react faster to varying channel conditions, thus allowing HSDPA to provide better performance for high-throughput applications. HSDPA standards go further than and have two additional transport and physi- 1

2 Figure 3: Controlling the flow of MAC-d PDUs in the being sent from RNC to node B is achieved with dedicated messages exchanged between RNC and node B. Figure 4: For an HSDPA test, the protocol tester simulates an RNC, generating appropriate messages towards node B based on test scenarios developed. cal layer channels high-speed shared control channel (HS-SCCH) and high-speed dedicated physical control channel (HS-DPCCH). The former is a downlink channel that provides control information associated with HS-PDSCH. It includes information such as the identity of the mobile terminal for which the next HSDPA subframe is intended, as well as channel code set information and modulation scheme for decoding HS- DSCH subframes. The former is an uplink control channel used to convey channel-quality information (carried by channel-quality indicator) as well as ACK/NACK messages related to HARQ operation in node B. HSDPA also impacts higherlayer protocols, including the MAC layer. / protocol architecture for HSDPA is in Figure 2. Different MAC entities are identified for different classes of transport channels. In 3GPP Rel. 99, dedicated and common transport channels are differentiated, and consequently, the MAC layer has a MAC-d and a MAC-c entity. The introduction of HSDPA defines a new entity, MAC-hs. As opposed to Rel. 99 specifications in which the MAC layer is implemented in the RNC, the MAC-hs is used in node B to take into account the requirement for a high-performance implementation of the standard. The node B MAC-hs handles functions related to, and includes the following: Handling the HARQ protocol, including generation of ACK and NACK messages. Re-ordering out-of-sequence subframes. This is actually a

3 Figure 5: Tektronix K1297-G35 protocol tester shows a visualization of messages. Figure 6: HSUPA introduces five new physical layer channels. function of the RLC protocol. However, this protocol layer is not implemented in node B for. Thus, the MAChs must take over some critical RLC tasks. Subframes may arrive out-of-sequence due to re-transmission activity of the HARQ processes. Multiplexing and demultiplexing of multiple MAC-d flows onto/from one MAC-hs stream. Downlink packet scheduling. Control-plane protocols The introduction of HSDPA also requires additions and modifications to control plane protocols used within the UTRAN. Specifically, these protocols are: Radio resource control (RRC) protocol, which is responsible for a multitude of UTRAN specific functions, including (signaling) radio bearer management. Node B application part (NBAP) protocol, which is implemented at the Iub interface found between node B and RNC. NBAP enables RNC to manage resources in node B. constitutes Downlink Uplink an additional type of node B resource, which also needs to be managed using the NBAP protocol. Radio network subsystem application part (RNSAP), implemented at the I ur interface between two RNCs, is also affected by HSDPA, since HSDPA-related resources in node B are managed by an SRNC that is different from node B s Controlling RNC. User-plane protocols frame protocol (HS- DSCH ) is the relevant user-plane protocol used to convey transport blocks within the UTRAN (Figure 2). The frame protocol is responsible for packaging a set of transport blocks, the basic unit of data passed from the MAC to the physical layer, into a format that can be transmitted over UTRAN s transport network, which is ATMbased for Rel. 99 UMTS. Additional functions supported by the Frame Protocol include Node and transport channel synchronization. An important function is related to controlling the flow of MAC-d PDUs in the being sent from RNC to node B. Since node B buffers are limited, some flow control is necessary. This is achieved with dedicated messages exchanged between RNC and node B (Figure 3). RNC sends the Capacity-Request message to node B to indicate that data is ready for transmission. Depending on the current buffer status within node B, this network element says how many (if any) MAC-d Protocol Data Units (PDUs) the RNC is allowed to Abbreviation Name Function High-speed downlink shared channel Common transport channel for U-plane traffic HS-SCCH HS-DPCCH High-speed shared control channel High-speed dedicated physical control channel Common control channel including information such as UE identity Feedback channel for HARQ ACK/NACK messages as well as for channel quality information Table 1: HSDPA has, HS-SCCH and HS-DPCCH as specific transport and physical channels. 3

4 MAC-d MAC-es/ MAC-e MAC-e transmit for a time period, via the Capacity-Allocation message. HSDPA test There is a need to perform a thorough test of network elements within the UTRAN before releasing them for deployment. Verifying the correct implementation of user plane and control plane procedures is only one of the testing issues to be addressed. Since user data is being exchanged at high throughputs between node B and RNC, there is also a need for performance testing of node Bs and RNCs. Thus, a question that must be answered is whether the capacity allocation algorithms used, particularly in node B, can correctly handle user data arriving at the highest possible throughput, up to 14.4Mbps from the RNC. In functional testing, a typical scenario is shown in Figure 4. The protocol tester simulates an RNC, generating appropriate messages toward node B based on test scenarios developed by the test engineer. All underlying protocol layers not subject to but required for the test case are automatically emulated by the tester. This specific scenario includes the emulation of the core network, such as the SGSN, GGSN and HLR. This way, correct implementation of control-plane procedures across the I ub interface can be verified. In this scenario, node B is the DUT. Other scenarios might require the RNC to be tested, while node B and user equipment E-DCH (UE) would be simulated by the protocol tester. In another test, the goal is to verify node B s internal-memory management and to test whether its implementation can cope with high-throughput user traffic from one or more terminals. In this case, the protocol that needs to be simulated is the (Figure 3). Other protocols needed to set up the required radio bearers, RRC and NBAP, are emulated by the protocol tester. Actual traffic is delivered by an external FTP server simulator, which, for example, could be connected via Ethernet to the protocol tester. In this scenario, the test case can consist of gradually increasing the rate of U-plane traffic up to 14.4Mbps, to determine if buffer overflow occurs at any point in time in node B. Another test case can consist of simulating traffic to multiple UEs to verify node B s MAC-d MAC-es E-DCH UE Uu Node B I ub DRNC I ur SRNC Figure 7: E-DCH introduces new MAC entities for UE, known as MAC-e and MAC-es that are mapped onto network elements. resource-allocation function. Enhanced uplink HSUPA aims to improve capacity and data throughput, and reduce delays in dedicated channels in the uplink. The main enhancement offered by 3GPP specifications is the definition of a new transport channel denoted as enhanced dedicated channel Figure 8: Two separate devices that help visualize how protocol tester works for HSUPA. (E-DCH). The maximum theoretical uplink data rate achievable is 5.6Mbps. With HSDPA, E-DCH relies on improvements implemented both in the and MAC layers. However, one difference is that HSUPA does not introduce a new modulation scheme but relies on QPSK, the existing modulation scheme for W-CDMA. Thus, HSUPA does not implement AMC. At the layer, the E-DCH introduces five new -layer channels. These include enhanced dedicated data channel (E-DPDCH), enhanced dedicated physical control channel, absolute grant channel (E- AGCH), relative grant channel (E- RGCH) and HARQ acknowledgement indicator channel (E-HICH) (Figure 6, Table 3). E-HICH has a similar function to HSDPA s HS- DPCCH, as it provides HARQ feedback information (ACK/NACK). However, it does not have CQI information, since HSUPA does not support AMC. As with HSDPA, node B has an uplink scheduler for HSUPA. However, the goal of this scheduling operation is different from that of HSDPA. HSDPA aims to allocate resources (in time slots and codes) to multiple users. The uplink scheduler aims to allocate only as much capacity (in transmit power) to individual E-DCH users as is necessary to ensure that node B does not have a power-overload. Transmit power of a UE is directly related to the data rate at which it transmits information as a result of the spreading operation inherent with W-CDMA. High-bit-rate transmissions require a low spreading factor to fill the 5MHz bandwidth of W-CDMA, resulting in a higher transmit power than for low-bit-rate applications that require a high spreading factor. Additionally, the more UEs transmitted at once, the more interference they cause for each other. Node B can only tolerate a maximum interference, and beyond that, it may not decode transmissions of individual UEs. 4

5 Feature HSDPA HSUPA Peak data rate 14.4Mbps 5.6Mbps Modulation scheme(s) QPSK, 16QAM QPSK TTI 2ms 2ms (optional)/10ms Transport channel type Shared Dedicated Adaptive modulation and Yes No coding (AMC) HARQ Packet scheduling Soft handover support (U-plane) Since E-DCH is a dedicated channel, it is very likely that multiple UEs will be transmitting at the same time, thus causing interference at node B. Then, node B must regulate the power of individual UEs transmitting in the E- DCH to avoid power ceiling. This transmit-power regulation is thus equivalent to scheduling uplink capacity for each UE transmitting with E-DCH. In other words, the uplink scheduling mechanism is nothing more than a very fast power control mechanism. Two physical scheduling channels, E-RGCH and E-AGCH, tell a UE how to regulate its transmit power level. For E-RGCH, the UE is instructed to either increase or decrease the transmit power level by one step, or to keep the current transmit power level unchanged. For E-AGCH, node B provides an absolute value for the power level of the E-DCH at which the UE must transmit. MAC protocol enhancements Aside from introducing new physical channels, E-DCH also introduces new MAC entities for UE, node B and SRNC. These MAC entities, known as MAC-e and MAC-es (Figure 7) are mapped onto network elements. The MACe is implemented both in the UE and in node B. It mainly handles HARQ retransmissions and scheduling. This is a low-level MAC layer that is very close to the layer. HARQ with incremental redundancy; Feedback in HS-DPCCH Downlink scheduling (for capacity allocation) No (in the downlink) Table 2: HSDPA and HSUPA share a QPSK modulation scheme, a 2ms TTI and HARQ features. Uplink Downlink Meanwhile, the MAC-es entity is implemented in the UE and the SRNC. In the UE, it is partially responsible for multiplexing multiple MAC-d flows onto the same MAC-es stream. In the SRNC, it takes care of in-sequence delivery of MAC-es PDUs, de-multiplexing of the MAC-d flows and distributing these flows into individual queues based on their QoS characteristics. These MAC-d flows may correspond to individual PDP contexts at the I u -PS interface with different QoS profiles, like streaming vs. background. E-DCH, as opposed to HS- DSCH, supports soft handover. This explains why the MAC layer for E-DCH is split between node B and the SRNC. While node B takes care of time-critical functions such as HARQ processing and scheduling, the associated MAC-es entity HARQ with incremental redundancy; Feedback in dedicated physical channel (E-HICH) Uplink scheduling (for power control) Yes in the SRNC takes care of in-sequence delivery of MAC-es frames emanating from different node Bs currently serving the UE. Another difference with is that E-DCH may support both a TTI of 2ms and 10ms as TTI depends on the EU category. mandates a TTI of 2ms. A possible test scenario for HSUPA is highlighted in Figure 8. In this scenario, the DUT is the RNC. The test scenario aims to verify that the MAC-es layer correctly demultiplexes the individual MACd flows from the MAC-es stream and re-orders the individual MACd PDUs into individual re-ordering queues. A further goal of the test is to verify that this re-ordering is implemented according to QoS requirements of individual MAC-d flows. In one test case, only one UE can be simulated to verify Abbreviation Name Function E-DPDCH Enhanced dedicated physical data channel This is the physical channel used by E- DCH for the transmission of user data. E-DPCCH Enhanced dedicated physical Control channel associated with the E- control channel DPDCH providing information to the node B on how to decode the E-DPDCH. E-AGCH Absolute grant channel Provides an absolute power level above the level for the DPDCH (associated with a DCH) that the UE should adopt. E-RGCH Relative grant channel Indicates to the UE whether to increase, decrease or keep unchanged the E-HICH HARQ acknowledgement indicator channel transmit power level of the E-DCH. Used by node B to send HARQ ACK/NACK messages back to the UE. Table 3: The five new layers introduced in HSUPA help it achieve higher capacity and data throughput as well as reduced delays in dedicated channels for the uplink. that the basic MAC-es function is correctly implemented in the RNC. In another test, multiple UEs can be simulated to verify that the MAC-es layer in the RNC can correctly differentiate all UEs from each other. In this test configuration, the protocol tester acts on the one hand as a UE and node B simulator. It simulates U-plane traffic based on the engineer s test-case definition, and emulates all necessary signaling protocols before establishing a MAC-es stream between the UE and the RNC. The traffic generated may correspond to the transmission, for instance, a video file together with an with a large attachment. Note that this all depends on the specific test case. On the other side of the RNC, the protocol tester can implement a core network emulation to ensure that the required PDP contexts are established. Moreover, the U-plane traffic across the Iu-PS interface must be decoded and visualized on the protocol tester. Mapping between MAC-d flows and PDP contexts is possible since the allocation of user traffic on MAC-d flows can also be visualized on the protocol tester. Note that UE/node B Simulator and the CN Emulation/Iu-PS Monitor can be implemented through one single test device.