11.4.9 Downlink channelization codes

Time multiplexing of data and control channels discussed in the previous subsection means that only one dedicated physical channel (DPCH) is necessary to transmit both these information streams. When a single such channel meets the requirement for the data rate the BS forms it by means of a user-specific channelization code sequence, which, of course, is unique over the cell (or sector) and cannot be utilized to contact any other MS. The scenario of multicode transmission arises when the only physical channel is incapable of transmitting data at the demanded rate. Then the BS involves several parallel physical channels to contact the same MS. These channels always operate with the same spreading factor, but with different channelization codes, which cannot be reused by the same BS in connections with other MSs. There is no need to repeat command messages in all multicode channels, so control information is sent to the MS over only one of them.

As such, the downlink and uplink channelization codes are of the same type. The Walsh functions family or OVSF tree of Figure 11.8 is used with spreading factor in the range from 4 to 512. Some channelization sequences are not available for DPCH, being allocated to common channels, like CPICH. For example, under the minimal spreading factor only three code sequences may serve DPCH, providing maximal gross data rate 2.88 Mbps. Assuming channel code rate 1/2 and unavoidable overhead (control commands etc.), this gross rate does not comply with 3G demands on the ‘pure’ data rate (up to 2 Mbps), which is one of the reasons why the specification stipulates high-speed uncoded transmission, too.

When many users are operating at different rates their channelization codes should preserve orthogonality despite the difference in values of the spreading factors, or, which is the same, in code lengths. As one may conclude on inspecting Figure 11.8, two Walsh functions of different lengths are orthogonal over the minimal length interval if and only if neither of them is a descendant of the other. It follows, then, that—unlike the uplink, where every MS, being isolated from the others by its unique scrambling law, has the whole set of channelization codes at its exclusive disposal—the management of the downlink channelization is much more complicated. Indeed, different MSs should be assigned different subsets of Walsh sequences, containing no descendants of sequences currently serving other MSs. This problem of dynamic resource coordination is being solved at the upper layers of the network protocol stack.

11.4.10 Downlink scrambling codes

As was mentioned earlier, downlink scrambling codes secure separation of signals of different BSs. Each scrambling sequence has in its basis a Gold sequence of length L ¼ 218 1 ¼ 262 143. Two LFSR with the feedback polynomials f1(x) ¼ x18 þ x7 þ 1 and f2(x) ¼ x18 þ x10 þ x7 þ x5 þ 1 generate initial m-sequences related to each other as required to produce a Gold ensemble (see Section 7.5.2). Although there are 218 þ 1 ¼ 262 145 Gold sequences of the length above, the specification limits the number of those utilized to 213 ¼ 8192. Two segments of length 38 400 are cut from every allowed sequence: the initial one and the one offset by 217 ¼ 131 072 chips, converted then to the f 1g alphabet in the usual fashion (6.15). The resulting binary sequences are, respectively, the real and imaginary parts of a downlink QPSK scrambling code.