Known limitations of the simulation¶
Listed in order they occur in the simulation chain:
Severity, in brackets after the entry:
1 We can probably get away with this
2 We should probably fix or understand this
3 We absolutely have to fix this
After listing your issue below, please also fill-in a new issue with the keyword "Tracker: NOvA Simulation Problem" by going to https://cdcvs.fnal.gov/redmine/projects/novaart/issues/new and choose the "NOvA Simulation Component" affected by your problem.
- Currently unexplained ~20% flux loss (2)
- No simulation of varying intensities (2)
- Status of SKZP? (2)
- Energy loss of ND cosmic muons in rock is only mean energy loss, and doesn't allow for scattering (1)
- Status of rock overlays? (2)
- Weights assigned to the GSimpleNtpFlux input files work on a relative basis but screw up the p.o.t. accounting. See docdb-7839. Not an issue unless we want this functionality. (1)
- EventGenerator sometimes tags nue/electron scattering as NC (see docdb-8023). Which causes the wrong thing when we oscillate it, because numu lack a W-exchange diagram, so oscillations should have an effect. Tagging them as CC is always safe. This might be a noticeable effect on the background to the nue analysis. (1.5)
- Beam spill window is the same for all detectors in MC, but not in reality. Also the beam spill window moved around a little during early running. All of these effects are <1 microsecond or so. (1)
- Vacuum instead of air between cells (1.5)
- Perfect alignment (1)
- None? Well, the problems with air...
- Various approximations in the transport (1)
- Doesn't simulate rolloffs observed in data (2)
- All channels give same light level (1.5)
- Long attenuation length is set to ~8.5m (based on a benchtop test at Caltech?). Measurements at the module factory seem to favour ~7.5m (eg docdb 8094) (1.5)
- Photo-electrons don't necessarily correspond to reality. (1.5)
- Current data-based nature of empty cell noise histogram and inconsistency with the noise added to traces in hit cells (2.5)
- Use of this same histogram in ND and FD simulation (3)
- All channels have same PEMaxADC. Also no variation in time (1.5)
- Readout window is a hair shy of (500+50) microseconds since we don't simulate clockticks outside of that window and thus can't make fully formed RawDigits right up to the edges of the window. (1)
- DCS noise distribution generated is gaussian, doesn't include non-gaussian tails observed in data (2)
- MC bad channels are all taken from a single date (2)
- Birks suppression is now handled by summing the visible energy at each step calculated by G4EmSaturation into a new data member of FLSHit.
- Number of neutrino interactions in GENIE spills isn't Poisson distributed - Fixed by M. Baird on 09/10/12 (see docDB # 7915)
- Tilt and stagger in FD planes - New geometries available - produced by X. Tian (see DocDB-8033) and validated by E. Niner
- GENIE does not decay tau and charm particles by default. Need to fix this in simulation to allow tau file production for analysis - Fixed by R. Hatcher (see DocDB-8142). A more complete solution using Pythia and Tauola will be implemented over the next few months.