Altera SerialLite II Protocol - Manual (Page 37)

Altera Corporation 37
SerialLite II Protocol Reference Manual
SerialLite II Specification
User Packet Data Flow
A ‘cut-through’ data flow must be implemented for data packets.
Meaning that packet data is transmitted as soon as enough data has been
received to fill a column, without waiting for the entire packet to be
delivered to the transmitter. This approach provides the lowest latency.
Priority packets are broken into segments and sent across the link. The
size of the segment is defined to be 2
N columns with the smallest being
eight columns.
A ‘store-and-forward’ data flow must be implemented for priority
packets segments. Meaning that no packet data is transmitted until the
entire segment has been delivered to the transmitter. However, priority
packets less than, or equal to, segment size are buffered before
transmission. This buffering is required to support the retry-on-error
option, which is only allowed for priority packets (see
“Retry-on-Error
(Optional)” on page 2–64 ).
User Packet Encapsulation
The SerialLite II protocol encapsulates data and priority packets by
wrapping start- and end-of-packet sequences around them. To mark a
start of packet, both the user and priority data have a 2-byte sequence
appended to them. The user data is preceded by a start-of-data packet
sequence {SDP} while the priority data uses the start-of-priority packet
sequence {SPP}. The second byte located in the start-of-packet sequences
is for segment identification and channel number as defined in
Tables 2–6
and 2–7 . To mark the packet valid, the end-of-good packet sequence
{EGP} is appended to the tail of the packet, while the end-of-bad packet
sequence marks the end of a corrupted packet. The second byte in the
{EGP}, {EBP} control sequences is used to verify that the segment
identification and channel number was received without corruption as
defined in
Tables 2–8 and 2–9 .
The IDL characters are inserted to maintain the proper location for control
words. Optionally, a CRC can be generated and appended to the sequence
marking the end of the user packet to protect against errors.
Figure 2–16
on page 2–38 shows an example of user data packet encapsulation.