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

Re: 64/66 Frame Synch Performance

Dear Bob,

> For the plots of frame sync performance in the 64b/66b PCS presentation at
> Ottawa, is the BER being used the error rate for frame synchronization or
> the link BER?  At least, it seems to me that the two are different.  Is
> that the case?  Thanks.

The probability of a sync error (defined as a BER event converting a
valid "01" or "10" to either "11" or "00" is dominated by single error
events and would be the probability of one error in either position +
the probability of an error in both positions:

    Psync_error = 2*(1-pe)*(pe) + (pe)^2

For low ber, this is dominated by the first term, which is very close to

Actually I'm not sure if we used 2*ber for the graphs, or just ber. 
When you are playing with an equation that has a slope of 32x on a log
scale, it doesn't make much of a difference. 

For the out of sync condition, we assume the effective sync BER is 0.5.

This is because the scrambler makes all the states "00", "01", "10" and "11"
equally likely.  2 of the 4 are considered valid sync and the others are

So the sync algorithm boils down to going out of sync reliably and quickly
on 0.5 BER, but staying in sync very tenaciously on a 10e-5 ber or lower.