Barrel HCal Meeting
(compiled by Anjaly)
GENERAL
• Eric mentioned the possibility to accommodate Elke in the meeting.
• Stefan suggested sending a poll to decide on a new date and time.
PROTOTYPE STATUS AND PLANS
• Norbert showed the plot from the phase scan comparing 50 cm and
500 cm long cables (signal from preamp-shaper).
• Eric suggested performing a SPICE simulation to estimate impedance
matching, signal reflection, ringing etc. For this, the input
characteristics of the HGCROC need to be known. He noted that
free Linux versions of SPICE are available.
• Martin questioned whether we still need to worry about signal
reflection once waveform analysis can be done to quantify
degradation.
• Norbert added that in cases of high multiplicity, long signals might
interfere with the following signals, which should be checked in
the simulation. He mentioned he would ask his engineers to do a
quick calculation on this.
• Anjaly reported technical issues from last week with the setup.
specifically, ASIC2 was not recording data. Power cycling the
board (turning it off and on) seemed to resolve the issue. The
board is now successfully taking data.
• Martin suggested building wide scintillators so that cosmic rays can
hit anywhere across the surface. This would help collect more
statistics. He offered to put together a few slides outlining this
idea.
SIMULATION
• Derek gave a detailed update on major software priorities, focusing
on completing TDR analyses and preparing a Pre-TDR update.
• David updated on his dropout study in neural networks: (Dropout is
the percentage of neurons that are randomly ‘dropped’
(deactivated) in each epoch during training. This prevents the
model from overly relying on the same paths and adds
non-linearity, though it makes training harder. More details
given during a previous update on the multi-layered perceptron
network)
His conclusion is that 50% dropout is too high, while
using 10% or less might be appropriate. Further testing is needed
to determine whether or not to use this and what value to be
used.