Study with JLAB LED and Caen VX1290A TDC

CFD with F1TDCLED with 1290 TDC
Good news: Switch to LED from CFD has eliminated multiple timing peaks. We confirmed the presence of these peaks with both CAEN 812 VME CFD and Phillips 715 NIM CFD. Peak positons were strongly correlated with the value of internal CFD delays. Apparently, CFD method of inverting, delaying and subtracting part of the signal does not work quite well with signals of certain magnitude/shape.

Problem: While time differences with LED do not show extra timing peaks, their shape for 2 short trigger paddles are somewhat puzzling, with a dip in the middle rather than at both ends. This results in highly non-gaussing timing distributions. Time-walk corrections improve overall timing resolution from 180ps to 115ps but we hope to improve this number if the shape of distribution is better understood.

Continous TDC mode

In this mode, there is no trigger at all. All hits are recorded. LED output signal width is 40ns, and VX1290A double-hit resolution is 5ns. Long-paddle time difference make sense: it shows decline of PMT efficiency toward both ends (due to 90mV discriminator threshold). Short(trigger) paddle distribution still has mysterious dip in the middle.

Monte Carlo

A simple Monte Carlo somewhat reproduces such shape. It includes light attenuation, time walk, signal smearing and a probability for light reflected from the opposite end of the paddle to produce a signal instead of the direct light. However, it requires 30% probability of reflected signal and a smearing with sigma=1.2ns to reproduce the observed shape. This doesn't sound too realistic.