BCP for DCI

Hi Rahul,

The usage of BCP for PDCCH transmit chain processing is shown in section 2.3.2 of the BCP user guide. As shown in Figure 2-3, BCP can be used to perform CRC (single PDCCH case only), Encoding and Rate matching. We don't have a PDCCH example available, but the following may be helpful pointers towards configuring the BCP for PDCCH handling.

1. CRC sub-module: The CRC sub module can perform block concatenation after attaching CRC to each block, but only one METHOD2_ID can be specified which is used to mask all the attached CRCs. The METHOD2_ID of a PDCCH depends on the corresponding RNTI so each PDCCH requests a different METHOD2_ID, therefore the CRC sub module cannot be used to perform CRC attachment for multiple PDCCHs with a single input packet. For this reason, BCP CRC can be used to handle only single PDCCH case and CRC attachment for multiple PDCCHs should be performed by the CPU (as shown in figure 2-3).

2. ENC: ENC is a block based sub module. It supports encoding at most three different blocks in one packet. The PDCCHs transmitted in one TTI may have more than three payload blocks. For example, in the case of an FDD 20MHz cell with 4 antenna ports, as per section 5.3.3.1 of 3GPP TS-36.212 we can figure out that there are at most six types of DCI payload size - 15 (format 1C), 28 (format 0/1A/3/3A), 33 (format 1B/1D), 39 (format 1), 50 (format 2A), 54 (format 2). So the BCP ENC needs 2 input packets to perform encoding for 6 PDCCHs in one TTI.

3. RM: RM is also a block-based sub module. It supports Rate matching of two code blocks in one input packet. RM needs both the payload block size (block size before rate matching, denoted by K) and channel block size (block size after rate matching, denoted by E) as it’s input parameters. There are four PDCCH formats and each format corresponds to one E value. In the above example scenario (FDD 20MHz cell and 4 antenna ports), there are at most 6x4=24 different [K, E] pairs so we need 12 BCP input packets to perform rate matching in one TTI with each packet processing the PDCCHs with two [K, E] pairs.

With the above considerations, to keep simplicity of the framework we recommend using one input packet to process the CRC attachment, convolutional coding and rate matching of a single PDCCH. The BCP input packet will contain the following headers: TM, CRC, ENC and RM, and PKT_TYPE will be Normal packet.

Regards

-Nitin