Day 2, session 2, topic 2, Present: Bruce B Tom J Jim A. Marta Dave S. (came later) Raw notes from the session. Tom entered quite a lot of information directly into the document. Starting point. LArSoft provide the ability to measure energy of neutrino interactions. How do you incorporate uncertainties and other error conditions into the calculations? * noisy and dead wires add to the errors. Should every algorithm produce uncertainties? (all should have associated errors) * vertex algorithms? yes * track fitting? yes Needs to be a way to calibrating hits. Where should wire signal processing be done? * filter noise * prepare wire signals for hit finding Icarus does something a bit different than the fitting in larsoft (fitting an analytic function. What are hits in a single train? Need to separate nearby charge depositions into district hits. Do you need to be able to generate data product where * signal processing is applied * calibrations are applied What about disambiguity? What about hits that are on the edge of the event time window? What about modifying data in the event? e.g. refining and finding ambiguities? What about multiple views? Is there a natural width to the hit? okay to have services apply energy corrections. uboone - there are shortcuts because of processing times (FFT on 9600 wires is too long - ROI is used instead) Only ROI object are stored after deconvolution. DAQ might be producing ROIs. Helps if zero surrpression DUNE is a bit different, but still uses the ROI (copied it) What about ROIs on neighboring wires. DAQ is worried about this because of caching of hits. flexibility is key: will have many … difference between MC hits and recob::hits What is the MC hit? Validation is usually done on single particle samples that do not have ambiguity. Want to have single particle calculations with multiple-particle samples. Maybe a MC-truth-cluster is appropriate here. Clusters: 2d reco only Want to run the tracking algorithm over data product with MC truth information applied, along with the real stuff. Wes - (validation questions) systematic uncertainties in the: diffusion, energy deposition. Should be be able to generate parallel collections with things applied and not applied with regards to MC information? Apply different energy corrections to different product in the event. Is labeling adequate for this? How are settings managed? requirement? method for tacking provenance set during simulation. are service setting saved now? what about easy of following associations and better support for associations -> navigating from product to product more easier (with better labeling). They want this. What about services for delivery of multiple associated files on input (groups of files delivered)? Not require the carrying forward of products so one file is present. Back to tracking… Polar coordinates are not well-defined in larsoft. Source code must be looked at to find out what they mean. What about the momentum of the track? Only filled in when it is a Kalman fit track. Does momentum depend on the particle ID? KE does. Should the track contain momentum and KE at all? This sounds messy - momentum needs to be filled out later when confident about particle ID, but original track is copied to produce the track with good numbers because they are not know when the track is born. The track needs to provide all the information necessary to do energy calculation. Need to reconstruct the energy of the track. Calorimetry not present yet. Want the hierarchy of particles represented and also to be able to traverse this hierarchy. PFParticles in particular here. Yikes - wants to be able to update this structure - primary and secondary particles. Used Geant4 particle tree as an example. PFParticles - is this unique to Pandora? What about hits that are not assigned to any particle? Is it straightforward to find them? This is not available by default. Should this be required. Need to identify the “Fuzz” to get the calorimetry correct. This is the hits. These are hits not associated with particles. Modules other than Pandora are tagging things like this now. Vertex need location and also the uncertainties. It should also have the clusters associated with it. Need to quickly location the end of a track or shower what it is associated with - there might be a gap here. Should it be the end of the whole track / shower? * Need to be able to measure the charge between the end of the track and the vertex.