Bjoern: FTK Does LVL2 rejects events while FTK is processing? How do you keep synchronization? - the FTK is very fast: with typical processing time of 25microseconds - the LVL2 will wait for FTK tracks How is duplication of data done for the overlap regions? - the dataformatter boards use the ATCA backplane and cables to share data among boards and crates How does this compare with GPUs, FPGAs and CPUs? - GPUs could be fast enough, but the question is how to bring data into GPUs - About high track multiplicity.... Boudoul: CMS L1 - To access this info at LVL1, what LVL1 latency are you assuming? 6.4microseconds - What is driving the z granularity? occupancy for some of the layers we will mix strip&pixels for Z vertexes at LVL1 - Your rejection factor on clusters/stubs implies an output bandwidth 2 orders of magnitued larger than today. it is not a factor 100 higher plus we should tune the TW cut - Do you form the stubs locally? Yes - the stub rate vs Z plot has a strong dependence on the layer? at small radius the spacing among the sensors is small and provides small resolution Artuso: LHCb - the approach sketched by ATLAS would be applicable by LHCb? We may want to into the AM option. - provokative: is there any intelligence in your tracker? the HLT has a lot of intelligence - we had a kick off meeting last week with GPU computing ... Salvati: CMS - slide 15 what is the peak at 25? 25 is the module number (z=0) - how much do you use the pixel Z information in combination with the beam constraint? I.e. how much you gain by going from strips dublet to pixel-strip doublets? We haven't done strips doublets yet. However we want to have pixels for the inner layers. - slide 21: how much was the offset for the plot with the assumption of (0,0) vertex? I don't know. - What you use to define the tracklet algorithms? FPGA? Lookup table? We plan to use lookup tables that should be ok for up to 2mm. - What is the size of the luminous region in the transverse plane? Abbaneo: CMS - would your phi realignment feature be programmable? yes, we think it will be programmable by block not by channel - have you cosidered adapt the pitch so that they are projective? It would be extremely complicated. Plus without programmable realignement there would be constraints on mechanincs. - the module matieral is 2% X0 without services and mechanics. What is the material used for simulation? We use a value larger than 2% in the simulation plus we include in the simulation the structure of the current tracker. Probably we used about 2.5% X0 per layer. Exacly material on slide 9 + fibers + cooling + support from the current tracker Marino: PET - Did you already achieve 100ps timing resolution with LYSO? Yes, this should be achieved on some setup. See ref [4]. Ddi you try ... bromide crystal. - This will be compatible only with FMRI with rapidly changing field? No, know we are only MRI compitible. - What is the 29ps std deviation? 100ps / sqrt(12) + jitter from clock - when do you expect to have a prototype system? as soon as the ASIC will be ready we will do measurements .... I cannot say when we will have a prototype assembled. - what energy resolution do you achieve? 15% FWHM at 511MeV