Digital Signal Processing Reference
In-Depth Information
the input for the Rake receiver, which has been implemented on Montium tile 2. Gener-
ated scrambling codes can be mapped efficiently on an FPGA. The three blocks, Montium
tile 1, Montium tile 2, and FPGA, are controlled by a general-purpose processor.
15.3.1.2.1 Block versus Streaming Communication
Two different mechanisms to exchange data with a Montium can be distinguished:
block mode and streaming mode. Some applications require all the input data to be
stored in local memories before execution can be started. This operation mode is called
block mode. The most important feature of this communication mode is that, during
the data transfers, the Montium is halted to make sure the execution is not started until
it is sure that the data are valid. In streaming mode, the Montium processes data while
simultaneously performing data transfers.
The timing properties of the UMTS communication system concern both the data
processing and the control part of the receiver. It seems natural that data processing is
performed according to the streaming communication principle, while control-oriented
functions are partly done according to the block communication. The implemented
WCDMA receiver has been realized according to the streaming communication prin-
ciple because for block communication too many resources (memory for storing the
scrambling code) are required during the baseband processing. According to
Table 15.3
the scrambling code sequence consists of 38,400 samples. Hence, if the WCDMA
receiver operates in block mode, 38,400 samples have to be stored in the local memory
of the Montium. Even when data would be processed on a slot basis instead of a frame
basis, still 2,560 data samples would have to be stored in local memory. Therefore, we
may conclude that the block communication principle in the WCDMA receiver is not
efficient, since blocks are too large.
15.3.1.2.2 Communication Requirements
The implemented WCDMA receiver thus operates according to the streaming commu-
nication principle. The implemented receiver can process four individual paths of the
received signal (see also
Figure 15.6
). Consequently, the receiver requires four complex-
number data streams for the four implemented fingers. All implemented fingers require
the same scrambling code. The scrambling code can be generated using, e.g., an FPGA
tile. The implemented receiver takes the complex-number scrambling code stream as
an input. Before de-spreading starts, the appropriate spreading code is stored in the
local memory of the Montium. The spreading code is stored in local memory because
the code has a maximum length of 512 samples. Thus, a relatively small amount of data
has to be stored in contrast to the scrambling code. Furthermore, the spreading code is
assigned to a particular user in the UMTS communication system, and therefore, the
spreading code will not change frequently. After de-spreading and before de-mapping,
the received symbols of the individual signal paths are combined. During this combin-
ing phase, each symbol is scaled according to a complex-number coefficient before sum-
mation. These complex-number coefficients are provided by the channel estimator. The
receiver outputs a bit stream with the received data. The characteristics of all the input
and output streams are given in
Table 15.4
.
Search WWH ::
Custom Search