English
Language : 

SH7670 Datasheet, PDF (560/1292 Pages) Renesas Technology Corp – 32-Bit RISC Microcomputer SuperHTM RISC engine Family / SH7670 Series
Section 14 DMAC That Works with Encryption/Decryption and Forward Error Correction Core (A-DMAC)
14.3.4 FEC Operation
FEC is an error correction method. This method enables the receiving side to repair the lost packet
without requesting packet retransmission. When repairing the lost packet by using FEC, the
transmitting side uses the original packet group to generate a redundant packet (FEC packet).
When transmitting 100 packets, for example, the transmitting side generates a 10 x 10 packet
matrix, XORs the ten original packets aligned to one row or column, and generates one FEC
packet per row (column). In this example, 20 FEC packets are generated. The transmitting side
transmits the original data packet group and FEC packets to the receiving side. To check whether
the original packets are lost, the receiving side aligns the original packets and FEC packets to the
matrix as in the transmitting side. If a lost original packet is found, the transmitting side can repair
the packet by XORing the other packets in the row and column to which the lost packet belongs
with the FEC packets. Like this, the transmitting side and receiving side need to share the number
of rows and columns of matrix aligned to generate FEC packets before transmitting and receiving
the packets.
The A-DMAC has the XOR calculation function used for FEC operation and supports the
following FEC specifications of RFC2733 and Pro-MPEG:
• XOR calculation of any number of rows (columns)
A variable-length descriptor supports the FEC structure of theoretically infinite length.
• One-dimensional FEC
Not only two-dimensional FEC but also one-dimensional FEC is supported because processing
is performed per row (column).
The CPU must perform the following operations:
• FEC matrix alignment
• Lost packet detection
• Unification of the lengths of packets that constitute a row (column)
(Packets less than the maximum packet length are padded with 0.)
• Repair of a portion of timestamp and payload type from the result obtained from the A-DMAC
Rev. 1.00 Nov. 14, 2007 Page 534 of 1262
REJ09B0437-0100