TLK2226_09 Datasheet, PDF (47/52 Pages) Texas Instruments – 6 PORT GIGABIT ETHERNET TRANSCEIVER

www.ti.com
TLK2226
SLLS689D – JANUARY 2006 – REVISED DECEMBER 2006
APPENDIX B – SGMII 10/100 ODD NIBBLE PACKET/IPG HANDLING
SGMII 10/100 Nibble Processing Assumptions
Case
1
Input (from Mac)
Odd nibble/cycle EXTEND during burst
packet, 3 cycles or more
clk
Output(Rate adaptation)
clk
en
err
Assumptions
Odd nibble CARRIER EXTEND (3 or more)
possible during burst. We will stuff or delete to
octet boundary for PCS encoding. Packets are not
corrupted. ⁽¹⁾
en
or
clk
err
en
2
Odd nibble/cycle regular packet
clk
en
err
Always deleted to Octet
clk
en
err
Packet ends on non octet boundary. We drop the
nibble to form last octet for PCS encoding. (non
valid packet anyway)
err
3
Odd nibble EOP EXTEND
clk
clk
en
err
en
or
err
clk
en
err
4
Odd nibble ERROR during packet
clk
clk
en
err
en
or
err
clk
en
err
5
1 nibble/cycle EXTEND during burst packet clk
clk
en
err
en
or
err
clk
en
err
Odd nibble EOP EXTEND. We do not know if this
is valid thus output to the PCS will be either NO
EXTEND or EXTEND.⁽¹⁾
TX_ERR asserted only during 1 nibble. We do not
know if this is a real error, thus output to PCS
encoding will either have an error indicated on a
byte(2 cycles) or no error.
End of Packet extend only 1 nibble during burst.
Output to PCS will either be a corrupted burst with
nibble extended or joined into 1 big corrupted
packet. We cannot always extend as there is no
CTC on the TX side. Following packets will not be
affected
(1) Note that in the case of an odd EXTEND and if it is stuffed, the PCS TX state machine will embed an |V| (error). The last nibble alone
does not contain enough bits to build a meaningful encoding byte (8’h0F). The packets themselves are not corrupted.
Submit Documentation Feedback
47