Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Update to questions on SLP



All,

I have spent some additional time thinking about synchronization of SLP over
four lanes and the probablility of a misdetected idle on one of four lanes.
The assumption is that the T-flag (12 bytes) is split equally amongst the 4
lanes (3 bytes each) and that detection is done on all 24 bits of each lane.
I have randomly selected one of the proposed control characters (which is
used shouldn't affect these calculations). There is no assumption about any
additional errors due to bit flips.

The following 8 lines indicated the positions where the idle stream might be
located within a data frame on any one of the four lanes:

0) XXXXXXXX 101100011011000110110001 ........
1) XXXXXXX1 01100011011000110110001X ........
2) XXXXXX10 1100011011000110110001XX ........
3) XXXXX101 100011011000110110001XXX ........
4) XXXX1011 00011011000110110001XXXX ........
5) XXX10110 0011011000110110001XXXXX ........
6) XX101100 011011000110110001XXXXXX ........
7) X1011000 11011000110110001XXXXXXX ........

The probability that three byte idle pattern will exist, perfectly aligned
from a byte perspective (line 0), is 
1/2^24. 

The probability for line 1 is: 1/2^23 (for the center set) times 1/2 (the
probablility that the previous byte ends with a "1") = 1/2^24.

Etc.

The probability of the three byte idle pattern existing within the data on
one lane is therefore: 
8 * 1/2^24 = 1/2^21.

Assuming that a similar case can be made for an idle pattern with an EOP in
it, the probability of the three byte idle pattern existing within the data
on one of the for lanes is 
1/2^21 * 4 = 1/2^19 = 1/524288.

At 10Gb/s, this would be a idle pattern detected within the data patten
(where it causes grief) on one of the four lanes, on an average, 
1e9/524288 = 1907 times per second.

So, in light of this, how will the byte, and lane synchronization reliably
take place and how long will it take?

jonathan

Jonathan Thatcher, 
Chair, IEEE 802.3ae (10 Gigabit Ethernet)
Director of Engineering, World Wide Packets
PO BOX 141719, Suite B; 12720 E. Nora, Spokane, WA 99214
509-242-9000 X228; Fax 509-242-9001; jonathan@worldwidepackets.com

application/ms-tnef