Experimental & modelling on risetime/jitter/mask
We spent a little time in the lab to see if the methodology of the standard
was working in practice.
Jitter bathtub measurements could be fitted to the equations in the draft
standard.
Sensitivity penalty vs. risetime fitted the model reasonably. The
significance of DCD was shown too.
We tried adding sinusoidal jitter and fitting to an enhanced link model with
a DJ (inc. SJ) jitter penalty term. The sensitivity penalty vs. DJ was
quite a curved function: low DJ had little effect, high DJ had a large
effect. We are seeing on the order of 1.5-2 dB for 0.3-0.35 UI.
We varied both risetime and DJ (SJ). Measured penalties of different eyes
which just passed the eye mask were around 2 dB. There was quite a lot of
experimental scatter, but the penalties seemed 0.5-1 dB bigger than the
approx. 1.5 dB ISI penalty in the current link model - not surprising, as
that has no allowance for jitter.
This work was done as much as possible with standard test equipment, filters
etc. to establish general principles, and should be seen as very
preliminary. The sensitivity measurements were taken "in the middle of the
eye". Measurements deliberately exercising the ~0.2 UI TP4 eye opening
would be interesting.
I am comfortable with the trends these results are showing. We could not
have hoped to have all that jitter without penalty, and the penalties seem
bounded and manageable.
Piers