Agilent Technologies J-BERT N4903B, User Manual, Page: 248 / 572

Note that there is a gap where the BER is so close to 10 -12 that we cannot decide.
For example, if we compared 3 × 10
12
bits and got two errors (a measured BER of
0.667 × 10
-12
), we are in the "uncertain" white area on the graph.
In such a case, we need to transmit more bits until the number of bits either reaches
the upper limit (6.296 × 10 12 ), or until we see more errors. If the actual BER is very
close to 10 -12 , however, we are unable to apply a lower or upper limit to the BER,
no matter how many bits we transmit. Whether such a test fails or passes depends
entirely on the application.
Once we are able to decide for a measured point whether its BER is above or below
the BER threshold, we can determine the total jitter at the intersection of the BER
threshold with the bathtub curve.
Since we are unable to find a single point on the slope where the BER is exactly
10
-12
, we search for an interval that brackets the point at which the BER is equal
to 10
-12
. This is illustrated for the left-hand slope in the following figure.
6
Advanced Analysis
248 Agilent J-BERT N4903B High-Performance Serial BERT
The Bracketing Approach