HSUPA Concepts

HSUPA Concepts

Last updated: January 15, 2009

This section is only applicable to the lab application or feature-licensed test application.

HSUPA Transport and Physical Channels

HSUPA (High Speed Uplink Packet Access) is essentially the equivalent of HSDPA for the uplink (see HSDPA Concepts ). It adds an efficient mechanism for transferring bursty packet data traffic over the W-CDMA uplink. HSUPA is officially referred to as E-DCH (Enhanced Dedicated Channel) in the specifications, although in the industry it is typically referred to as HSUPA. The term HSPA (High Speed Packet Access) has become the predominant term used in the industry to refer to both HSDPA and HSUPA as a single entity. Although they are technically independent features, HSUPA is currently always used alongside HSDPA, hence the popularity of the HSPA term.

HSUPA was added as part of Release 6 of the 3GPP specifications. It was designed to be backwards-compatible with existing functionality, so a non-HSUPA UE can still communicate with a HSUPA-capable base station and vice versa. Similar to HSDPA, HSUPA adds a set of new physical channels that provide efficient transfer of packet data traffic over a new transport channel, the E-DCH (Enhanced Dedicated Channel).

In the uplink the following channels are added:

In the downlink the following channels are added:

The E-RGCH and E-HICH share the same code space in the OVSF tree. Orthogonality between the two channels is provided by the use of orthogonal 40 bit signatures that can be transmitted as-is or inverted to signal +1 or -1. Not transmitting a signature in a TTI results in a 0 being assumed by the UE (this is also called DTX in the specifications). The definition of the E-HICH and E-RGCH channels provide the mapping from -1, 0, +1 to the logical values of Up/Down/Hold or ACK/NACK. As there are a total of 40 orthogonal signatures available it is possible to multiplex up to 20 UEs onto a single E-HICH/E-RGCH OVSF code.

UE Categories

3GPP TS 25.306 Table 5.1g defines a set of HSUPA UE categories that restrict the UE's data rate by specifying attributes such as the maximum number of E-DPDCHs, minimum E-DPDCH spread factor, 10 ms vs. 2 ms TTI support, etc.

Fast Scheduling

Although HSUPA is similar in many respects to HSDPA, HSUPA has one very significant difference which the name of the new transport channel, Enhanced Dedicated Channel, hints at. Unlike HSDPA, HSUPA does not utilize a shared channel for data transfer in the uplink. In W-CDMA each UE already uses a unique scrambling code in the uplink so each UE already has a dedicated uplink connection to the network with more than ample code channel space in that connection. This is in contrast to the downlink where the Node B uses a single scrambling code and then assigns different OVSF channelization codes to different UEs. The shared resource in the uplink is actually the interference level at the Node B, which the network manages through the fast closed loop power control algorithm. The fact that the UE has a dedicated connection to the network in the uplink influences the design of HSUPA quite considerably. The goals of HSUPA were to support fast scheduling (which allows the network to rapidly change which UEs are transmitting and at what rate) and to reduce the overall transmission delay. Transmission delay reduction is achieved through fast hybrid ARQ retransmissions, in a manner very similar to HSDPA, and an optional shorter 2ms TTI. As the primary shared resource on the uplink is the total power arriving at the base station, HSUPA scheduling is performed by directly controlling the maximum amount of power that a UE can use to transmit at any given point in time.

The network has two methods for controlling the UE's transmit power on the E-DPDCH; it can either use a non-scheduled grant or a scheduled grant. In the non-scheduled grant the network simply tells the UE the maximum block size that it can transmit on the E-DCH during a TTI. This block size is signaled at call setup and the UE can then transmit a block of that size or less in each TTI until the call ends or the network modifies the non-scheduled grant via an RRC reconfiguration procedure. The block size deterministically maps to a power level, which is also configured by the network during call setup. The non-scheduling grant is most suited for constant-rate delay-sensitive application such as voice-over-IP.

The more interesting method is the scheduled grant. In this case the UE maintains a Serving Grant that it updates based on information received from the network. The Serving Grant directly specifies the maximum power the UE can use on the E-DPDCH in the current TTI. As E-DCH block sizes map deterministically to power levels, the UE can translate its Serving Grant to the maximum E-DCH block size it can use in a TTI (the mapping of power levels is determined by the E-TFCI Reference Power Offsets that are signaled at call setup).

There are two ways the network can control the UE's Serving Grant. The first is through an absolute grant, transmitted on the shared E-AGCH downlink channel, which signals a specific, absolute number for the Serving Grant. The other way is through relative grants, transmitted using the downlink E-RGCH channels, that incrementally adjust a UE's Serving Grant up or down from its current value. At any given point in time the UE will be listening to a single E-AGCH from its serving cell and to one or more E-RGCHs. The E-AGCH is a shared channel so the UE will only update its Serving Grant if it receives a block on the E-AGCH that is destined for it (the E-RNTI identity signaled at call setup is used on the E-AGCH to direct transmissions to particular UEs). The E-RGCH is also shared by multiple UEs but on this channel the UE is listening for a particular orthogonal signature rather than a higher layer identity. If it does not detect its signature in a given TTI it interprets this as a "Hold" command, and thus makes no change to its Serving Grant.

Typically each cell that the UE is in soft handover with will transmit an E-RGCH (although it is possible for the UE to receive only a DPCH from a cell that it is in soft handover with). HSUPA has the concept of a serving Radio Link Set (RLS) and a non-serving Radio Link Set. The group of cells from which the UE can soft combine E-RGCH commands form the serving RLS. The serving RLS by definition includes the serving cell from which the UE is receiving the E-AGCH. Obviously the cells in the serving RLS must all transmit the same E-RGCH command each TTI so typically the serving RLS consists of those cells that are controlled by the same Node B that also controls the serving cell. All other cells that transmit an E-RGCH to the UE form the non-serving RLS.

Cells in the serving RLS can give E-RGCH commands to raise, lower or hold the current Serving Grant. Cells in the non-serving RLS can only give hold or down commands. Thus from the perspective of cells in the non-serving RLS the E-RGCH is primarily an overload control mechanism. If the UE receives a "Down" command from any cell then it must lower its Serving Grant regardless of what the other cells say, if it receives a "Hold" from all cells it will leave its Serving Grant unchanged and if it receives an "Up" command and no "Down" commands (all the cells either transmit "Up" or "Hold") then it will raise its Serving Grant. Up and Down commands are communicated on the E-RGCH transmitting a 40 bit orthogonal signature. Hold is actually communicated by the absence of the expected signature. Although the 3GPP standards define how the network communicates a Serving Grant to a UE, the algorithm by which the network determines which commands should be sent on the E-AGCH/E-RGCH is not defined. This is left for the network infrastructure vendors and operators to design, much the same as was done with the algorithms that trigger handovers.

However, measurement reporting functionality is defined in the standards to allow the UEs to communicate their current status which can then be fed into the algorithm. UE status reporting takes two forms, Scheduling Information transmitted on the E-DCH along with the user data, and a happy bit transmitted on the E-DPCCH L1 physical channel.

The Scheduling Information provides an indication of how much data is waiting to be transmitted in the UE and how much additional network capacity the UE could make use of (if the UE is already transmitting at full power then it would be wasteful to increase its Serving Grant as the UE would be unable to make use of the additional power). This Scheduling Information is typically tacked onto the end of the user data at the end of the MAC-e PDU that is transmitted over the E-DCH. It is sent both opportunistically in bits that would otherwise be spare and periodically as configured by the RRC layer at call setup.

The other status reporting mechanism is the happy bit. This is a single bit that is transmitted on the E-DPCCH physical channel. It can only indicate "happy" or "unhappy" but it is sent every time the UE transmits to the network. A UE considers itself to be unhappy if it is not transmitting at maximum power and it cannot empty its transmit buffer with the current Serving Grant within a certain period of time. The period of time is known as the Happy Bit Delay Condition and is signaled by the RRC layer during call setup. Thus the happy bit is a crude indication of whether the UE could make good use of additional uplink power.

Fast Retransmit

The fast retransmit goal of HSUPA is achieved through a HARQ scheme that runs inside the Node B and works much the same as it does in HSDPA. There are a number of stop-and-wait HARQ processes that operate in parallel (4 for 10ms TTI and 8 for 2ms TTI). Each time the UE transmits, the receiving HARQ process in the network will attempt to decode the data block. If it succeeds, the base station transmits an ACK to the UE over the E-HICH channel and that HARQ process in the UE will move onto the next data block. If the decode fails then the base station transmits a NACK to the UE on the E-HICH. If the maximum number of retransmissions has not been reached then the UE will retransmit the data block again. The UE will either use chase combining (transmission of the exact same bits again) or incremental redundancy (transmission of a different set of bits) depending on how the RRC layer configured the link at call setup. The HARQ process that transmits in a particular frame is determined from the current SFN (this is unlike HSDPA where each HARQ process transmits in a round-robin fashion).

MAC Layer

HSUPA adds a new transport channel, the E-DCH. This layer 1 transport channel connects up to a new MAC sublayer, MAC-e/es. In the UE MAC-e/es are considered one single sublayer, however in the network side MAC-e and MAC-es are considered separate. This is to allow the more real-time critical functionality of MAC-e to be placed into the Node-B and allow MAC-es to run in the RNC.

In the network the MAC-e sublayer contains the HARQ processes, some demultiplexing functionality and the fast scheduling algorithm. MAC-es primarily provides reordering and combining and also disassembly of MAC-es PDUs into individual MAC-d PDUs. HSUPA supports soft handover so it is possible that one or more Node Bs will successfully receive the same MAC-e block from the UE and forward it on to the RNC. MAC-es is where these duplicates are removed. Similarly due to the parallel nature of the HARQ processes it is possible for MAC-e PDUs to arrive out of order so the MAC-es sublayer performs reordering to ensure MAC-d sees PDUs in the order they were originally transmitted (this is important to avoid triggering of spurious retransmissions at the RLC layer).

In the UE, MAC-e and MAC-es are combined into a single MAC-e/es sublayer. This sublayer contains the HARQ processes, performs selection of the uplink data rate based on maintaining the current Serving Grant and provides the status reporting. In any given frame the MAC-e/es layer in the UE will select the block size that it will use for the next transmission on the E-DCH. The choice of block size is determined by the amount of data waiting to be transmitted, the minimum allowed spreading factor (which is signaled at call setup) and the current Serving Grant. What this means is that in a particular TTI the UE has a large range of block sizes that it can pick from. Each block size is transmitted using a specific amount of power which increases as the number of bits in the block goes up. As the network has no way to predict which block size the UE will choose it also has no way to predict the power level at which the UE will transmit. This ultimately means that the base station has to be able to demodulate and make accurate measurements across a wide dynamic range of power levels.

Related Topics

W-CDMA Concepts

HSDPA Concepts