Third Reconstruction Software Workshop, Ecole Polytechnique, Jan 19 and 20, 1996

Recent Changes

• 21 Jan, 1995 -- Added summaries (Bob)

A two-day joint meeting with simulation was held on Jan 19 and 20, 1996, just after the collaboration meeting at Ecole Polytechnique. Torre is maintaining a page on the simulation sessions.

Slides and handouts are being scanned, and will be put on the web as fast as possible.

Reconstruction Goals

• Updates on current activities, changes since December
• Presentation of new and proposed common code
• Decisions on design of material and geometry models for reconstruction

Agenda, Schedule, Summaries

Friday

9:00 - 10:30 Simulation status and recent developments

• BBSIM status discussion focusing on release status and usability
• Reports on work/developments since the Dec workshop (there will be at least a couple)
• BBSIM development
  Fabrizio Bianchi reported here (and on Saturday morning) about efforts to create an SVT digi that will associate the MC and "hardware" information in a useful way in both bbsim and reconstruction objects.

11:00 - 12:30 Reconstruction status and developments I

• Summary of schedule & goals (Bob Jacobsen)
  The first reconstruction software review is expected to be April 11 at SLAC. The overall list (9kB .ps, 2 pages) of topics and example deliverables were presented. We now have to develop our understanding so that we can get these documents written.
• New common code since December; how to use it; future pieces (Bob Jacobsen)
  New code presented (12kB .ps, 4 pages) includes the new CLHEP release (0.15 from the central distribution) which has some nice matrices and linear algebra support. Progress in tracking and the new Micro Framework were presented elsewhere. New features and use hints for the release tools were shown in the How-To session (37kB .ps, 18 pages) Volunteers were sought and found for implementing an Angle class (to encapsulate the various magic of wrapping at pi), to convert ParticleDataTable classes to our use, and to add error matrix support to the tracking classes.
• Histogram/Ntuple classes (Stephane Plaszczynski)
  Stephane presented a C++ ntuple package (Unterface) that is based on our current FORTRAN "unt" package. It provides row-wise ntuples in multiple files, and several people thought it would be easy to extend it to 1d and 2d histograms. It relies on a method for passing strings between FORTRAN and C++ written by Serge Du. The interface of a single ntuple is very similar to the existing HepTuple package, and file access is similar. Unterface's chief advantages are:
  1. It supports multiple HBOOK files at the same time
  2. It can coexist in the same excutable with unt

HepTuple's advantages are:

1. It currently has 1d and 2d histograms
2. It provides a more flexible interface for handling different underlying types of histogramming packages (although with HBOOK the only clear contender, this is of little use).
3. It works on our five UNIX platforms now, has been integrated into reconstruction applications, and is currently in use in several prototype modules. In the discussion several points became clear:
   1. Several people expressed a strong desire for column wide ntuples, either for efficiency reasons or to get access to variable length, hierarchical data. The lack of this in either unterface or HepTuple is viewed as a drawback to either solution. FORTRAN access is available, but a C++ solution (if not OO) would be valuable.
   2. Simone Giani (CERN) pointed out that HBOOK has reached the effective end of its development lifetime, and few or no new features are to be expected. Instead, future activity will be oriented around database methods of storing this same type of data, most probably OO databases. This will eventually provide much more powerful access, but is on the wrong timescale to be of any value to us right now.
4. EMC prototype status (Xiaorong Shi, Steve Playfer)
   EMC Digis are currently made from GHits in a module, which then feeds an initial clustering module. The structure of the software downstream of that is still under discussion; in particular, its not yet clear just how integrated that should be to other types of particle ID.

14:00 - 15:30 Reconstruction status and developments II

• IFR prototype status (Luca Lista)
• SVT/DCH/tracking status (Bob Jacobsen)
  This summary (14kB .ps, 4 pages; diagrams to be scanned) was very similar to one given at the most recent tracking meeting.
• Alignment issues (Dave Brown) David gave a detailed talk from the perspective of aligning a precision tracking detector like the SVT. Key points include:
  • Bootstrap using consistency, and systematics dominate many measurements
  • nDOF very large, not obvious how to constrain things until you study the physical situation. You have to be prepared to include prior measurements as constraints and make arbitrary choices. In particular, the system boundaries are important; systems should be prepared to to provide aligned samples to others system.
  • Constants should be stored as "Nominal + small corrections", with the corrections in the nominal frame. This makes them linear. Once has to think carefully about nesting too many transformations, as these will result in lots of degenerate alignment constants. Single sets of "local" and "global" constants are probably better.
  • You must be prepared to continually� monitor global position and adjust as function of time in quasi-realtime.
  • The actual alignment process requires that it be fast and accurate to rapidly/easily move pieces by small mounts, add/remove hits, check chisq contributions.

16:00 - 17:30 Geant4

• Geant4 talk by Simone Giani and open discussion on Geant4 and BaBar

Saturday

9:00 - 10:30 Fast simulation

• report from the MCFast workshop at LLNL that followed the Dec software meeting: a scenario for collaborative, complementary fast simulation development between BaBar, MCFast and Geant4
• discussion of fast simulation requirements, priorities, development timescale
• organization of work

11:00 - 12:30 Geometry and material models for reconstruction

• Summary of existing code & capabilities (Bob Jacobsen)
  The presentation was primarily a summary of the preceeding talks, plus some items for discussion.
• Discussion
  The discussion was wide ranging, though limited by time. A clear concensus emerged for a "single underlying source", although there were differences of opinion about what that actually meant. There were no objections to moving toward a single software representation of geometrical objects, although it seems clear from examples that some systems require vastly different support from those objects.

14:00 - 15:30 MC Production & Reco Overflow (<1)

• update on production status, plans, interested participants
• T0 - Steve Wagner
  Steve presented the current status of T0 finding using the drift chamber ab initio. He, Steve Schaffner, Anne-Marie Lutz and Maria Grazia Pia have agreed to develop a proposal for distribution and comments.
• H1 Analysis Example - Forest Rouse