GMINOS for the Mock Data Challenge

In summary the major changes include:

• Change to using RANLUX for the random number generator.
• Connection to Hugh Gallagher's Neugen3 event generator; an updated and revised version of NEUGEN(2).
• New flux method (MFLX 5) that reads and reweights gNuMI column-wise output files directly.
• New module based ('M') planes

RANLUX

Neugen3

gNuMI Flux

To do or check:

• re-re-check event rates. Early test showed right number of events in FarDet for a given # of POTs, but that near det rate seemed ~14% low. Some of this is attributable to falloff of flux as NearDet doesn't see point source (which changes the spectrum over the volume). But test using MC integration could account for only some of this. Perhaps with more statistics.

M-planes: module based planes

Det	Coverage	U plane	V plane
Near	Partial
	Full
Far	Complete

Click on a image to get full scaled version

common features and deficiencies

The database constants appear to have incorrect values for the WLS fiber used in the bypass region -- it seems to be equal to the length of the missing region rather than accounting for the actual path of the swoop around the coil.

At the time of this writing (2003-12-22) the region near the coil for both NearDet full and FarDet complete planes has active strips right up to the coil. This, of course, is not true in reality. Modelling a good approximation of the shape of the bypass inserts is difficult. Having the mass of the strips in place is probably not too incorrect, if only they wouldn't generate hits. This may lend itself to a post-hoc approximation where hits in this region are simply removed (and those straddling the boundary are appropriately truncated). But code for this has not been fully developed, though recent changes to UgliStripHandle should now make the necessary info available via PartialLength().

Detector specific improvements/deficiencies:

NearDet

The beam center location relative to the plane has been reverified ( x0 = 148.84cm, y0=40.9cm in the first steel plane ).

To do or check:

• the exact placement of modules on the plane is still unverified -- it looks approximately correct but isn't necessarily based on reality.
• the vertical placement of the detector relative to the hall floor is correct, but the east-west and north-south placement is still unadjusted.
• The shape of the hall is still a box, with no accounting for the overhang on the east side.
• verify the collar dimensions

FarDet geometry

The files below lack important corrections including changes to the flux (ie. the high-energy tail is too large) and for the overlay file the instantaneous POTs is too high (4.0 vs. 2.4). There are probably other differences as well. Besides, the true MDC set has superior statistics.

Data files using all the above features are now available on /afs/fnal.gov/files/data/minos/:

dir	file	size
d17/detsim03/mock_data	gm_geom_far.preMDC.reroot.root	843MB
d17/detsim03/mock_data	gm_geom_far_nue.preMDC.reroot.root	227MB
d17/detsim03/mock_data	gm_geom_near.preMDC.reroot.root	341MB
d17/detsim03/mock_data	gm_geom_overlay.preMDC.reroot.root	245MB

I don't really know why the near file is larger than the overlay file (which should contain all the near events as well as the rock events. This could be because of an error in the digitization process on the near file. Since all of FLSDigit is superceded with the new PhotonTransport + DetSim code that wouldn't matter. A quick look did show that the records did contain roughly the right number of pileup events. The reduction could also come in part from the elimination of numerous redundant copies (one per record) of the FLSParam table.

The files were generated on minos-rhatcher.fnal.gov which is a 2 CPU 3.2GHz P4 with 1GB memory.

	processed	written	POTs	POTs per written evt	seconds per gen evt	total time
FarDet	36538	36538	3.8e21	1.04001e17	1.2894	13.1hr
FarDet (nue)	9766	9766	1.0e21	1.02396e17	1.3258	3.6hr
NearDet	42016	40713	4.0e16	9.82487e11	1.0910	12.7hr
NearRock	257471	20877	4.0e16	1.91598e12	0.5765	42.2hr
NearOverlay	1000	1000	4.0e16	---	---	~3hr

The FarDet file represents 10yrs of data at our nominal POTs/yr. The Near files represent 1000 spills (8min, 20sec) at the nominal spill intensity; for the NearDet files only those events that generated hits in an active volume (i.e. scintillator strip) were written out.

One should note that the average 21 rock events/spill does not necessarily translate into 21 muons through the detector. Nor will a significant fraction of the interactions with vertices in the detector be reconstructable due to their location in the uninstrumented regions (partial planes and the spectrometer). These serve just to add the correct smearing and confusion to the reconstruction.

Processing the files

These files are much too large to process in a single job both in terms of CPU time and candidate output file size. This can be easily dealt with by modifying the script to take two arguments: a number of records to skip and a number of records to process.

void process_partial(int nskip=0, int nrun=1000000)
{
  gSystem->Load("libDigitization");
  gSystem->Load("libDetSim");
  gSystem->Load("libPhotonTransport");
  gSystem->Load("libDataUtil");
  ...
  JobC jc;
  ... 
  jc.Path.Create("MyPath",
                  "RerootToTruthModule::Get "
                  "ScintHitToDigiPE::Get "
                  "DigiPEtoRawDigitModule::Get "
                  
                  //"OtherModules::Reco "
                  //"Output::Put "
                  );
 
  // Set up the input format as a reroot file.
  jc.Input.Set("Format=reroot");
  ...
  <more job configuration>
  ...
  jc.Input.Next(nskip);          // skip some records
  jc.Path("MyPath").Run(nrun);   // process a limited number more
  jc.Path("MyPath").Report();
 
}

  $ loon -bq 'process_partial.C(10000,100)' file.reroot.root