Video: Dean Schamberger, Serban Protopopescu, Ursula Bassler
This was the initial organizational meeting for the Task Force. Some of the members are still on vacation. The meeting was somewhat free form to air some of the issues.
Marek Z. asked whether the Task Force should consider jet shapes and that this could use the help of the QCD group, and would make an excellent thesis topic.
Serban P. has been looking at single electron MC from Z to ee and sees no difference in resolution for thresholds at 1.5 or 2.5, implying that only noise is being added. Volker B. mentioned that seems to contradict with some of the studies shown in the EM-id meeting last week by Alexis Cothenet and Marie-Claude Cousinou and wondered about the fake rate. Serban sees 3 cells in the mean difference go to about 11 cells when the threshold is lowered. The Delta_pt/pt does not change. Non-linearity effects have been included.
There was general discussion about how the noise is put into the MC and whether this is modeled correctly. Silke D. described what she believed Vishnu Z. is doing, which is to use a flat model in phi of the noise with values in eta and layer coming from data (see her slides). A gaussian with a standard deviation is used to smear the signal but it seems that Vishnu does this twice in an effort to emulate the baseline subtraction, which according to Dean S. is NOT what we want. The effect of this would be to widen the noise by sqrt(2). This term is for the "pure" electronics noise and does not include the pile-up effects from luminosity and multiple or preceeding interactions.
It was generally agreed that pile-up is the most sensible place for these corrections to live and Laurent Duflot will be asked for help in ensuring this is done correctly. It was also agreed that we want to emulate the offline suppression and use a real pedestal file for each channel to get the rms smearing of the pedestals for electronics noise emulation.
A question arose as to whether Robert Zitoun's non-linearity corrections are applied correctly to negative cells, which people thought it most likely was, but needs to be checked.
There was a long discussion of what MC samples would be needed for studies and this was summarized in a message after the meeting. All MC should be plate level with the same version of code. D0gstar files would be generated first to save time until we are comfortable that the d0sim and d0reco steps are correct.