Architecture Review Meeting

Wednesday 17 Jun 2015, 15:00 → 16:15 US/Central

WH7XW (Woodshed) (Fermilab)

Audio connection will be opened with ReadyTalk: Conference code: 886-7778 Phone number: +1 (866) 740-1260 (U.S.A.)

Present: Herbert Greenlee, Jim Kowalkowsky, Gianluca Petrillo, Brian Rebel, Erica Snider
Connected: David Adams

Geometry extension

After some illustration of the current design, the goal of the change was stated as supporting LAr TPC reconstruction from the "virtual" readout space as opposed to the currently supported geometric space.
The latter is represented in terms of wire position and drift time: reconstruction of hits, clusters and 2D vertices in it is currently supported. Similar support is pined for in the virtual space defined by readout channel number and drift time.
Discussion included how to best represent the new concepts in the code and as data, and how the existing data structures should relate to them.

The following proposal was put together:

1. introduce identification classes for the APA concept (group of TPCs sharing readout channels) and the readout plane concept (group of planes sharing readout channels); keep the current channel ID as an identified for the readout channel itself
2. geometry is taxed with the task of providing mapping between the existing geometry elements, represented by geometry identification classes, and the new readout concept, also represented by identification classes
3. the information of this mapping is coded in experiment-specific classes, and not stored in the geometry description of each detector
4. there will be no object representing the characteristics of the new concepts beside the unique identification classes
5. objects affected by the new concepts are by principle all the ones that are currently associated to a geometry element that can be shared: hits (associated with wires), clusters and 2D end points (associated with a wire plane)
6. hits are currently associated also with a channel ID; this is deemed to be sufficient and the class should not undergo any change
7. an algorithm that produces new hits in the geometry space will assign both a complete and valid wire ID and a matching channel ID to each hit; an algorithm that produces new hits in the readout space will assign a valid channel ID and the valid ID of any of the wires connected to that channel [this interpretation is questioned]
8. clusters will have a readout plane identifier added to the existing data, that will then include both a plane and a readout plane identifiers; this solution was preferred to having the same class being able to express either a geometry or a readout ID, depending on the content, on grounds of clarity. 2D points should be treated in the same way as clusters
9. an algorithm that produces new clusters in the geometry space will assign both a complete and valid plane ID and a matching readout plane ID to each cluster; an algorithm that produces new clusters in the readout space will assign a valid readout plane ID and an invalid plane ID

This final proposal did not raise any objection and will be used as reference for the implementation.

Meeting time

The current time was confirmed acceptable by the participants in the room and connected.
No change is planned for the future meeting, coming tentatively next Wednesday.

15:00 → 15:35 Extension of geometry to accommodate shared readout channels

text David Adams's geometry helper (see also the implementation file)

15:35 → 15:45 Establishing a regular time for this meeting