In my opinion, a strange 2-bump shape of Delta(t)=t1-t2 timing distribution in the short trigger counters is currently the main reason we cannot improve timing resolution beyound 120ps.
We moved the paddle into the black box and did a number of tests on it. TDC was configured to run in free mode (without any triggers) accepting all hits (apart from 40ns dead-time of JLAB LED).
First, we used Cesium source located at 3 positions (both ends and the middle) for exactly 1 minute at each position. The observed spectrum makes sense: the highest efficiency is in the middle, and Delta(t) converted into geometrical position along the paddle also makes sense
The 4 plots below show variations in t1-t2 distribution shapes for different LED tresholds.
It seems, there is also a small change in the shape of the bumps depending on terminating or not terminating signal cables at PMTs as shown on the next 2 plots. I'm still not sure what to conclude out of this.