Systematic error of the tower gain determination with MIPs
see also earlier analyses

INPUT:
- E-EMC tower ADC (pedestal subtracted, the same vale for all 4 days)
- L3 tracks
- L3 vertex
- trigger ID

METHOD:
- select prim tracks (L3 tracks passing vertex, delta Z< 4 cm)
- require prim track to pass through front and back face of a tower
- data from towers at fixed eta were added together to increase statistics
- Landau shape was fitted for ADC=[8,60]

CUTS :
"A" ==> all hits
"P" ==> pT >0.5 GeV, retains 50% of towers
"S" ==> delta eta <0.02 and delta phi <1.5 deg ( at depth of SMD), retains 50% of towers
"L" ==> dE/dX=[0.5,2.0] (exist only for long tracks), retains 13% of towers but all at eta bin 11 and 12

TRIGGER ID selection:
"Mid" ==> 1000=MinB + 1101=BHT1 +1102=BHT2
"Fpd" ==> 1003=FPD
"Endcap" ==> 1151=EHT1 + 1152=EHT2

Run selection:
- large files from day 120,122,123,124, total few M events
- trigger mix: most ppTrans-1 + some minB
- total 113,000 towers pointed by a track

Results:
- fig 1 & 2 show variables to which cuts were applied.
- fig 3 illustrates all towers are seen in the data
- fig 4 shows example of ADC spectra for various triggers with fitted Landau
- table 1 contains mean value of Landau with its error
- note, the error of the MIP peak remains unchanged after cuts (throwing away half of the data), meaning the cuts help

Conclusions:

  • Position of the MIP peak depends strongly (up to 2-3 ADC counts) on cuts and trigger IDs.
  • It also depend (just follows) on the left limit of the fitting range
  • the Endcap HT triggers seem to predict on the high side
  • the Fpd trigger is on the low end
  • Increase of statistics does not help, since fits are already good.
  • The MIP position is in reasonable agreement with the M-C results ( using sampling fraction of 0.05 , posted earlier) 
    eta bin=5 MIP ADC=13.0
eta bin=6 MIP ADC=14.2
eta bin=7 MIP ADC=15.5
eta bin=8 MIP ADC=16.7
eta bin=9 MIP ADC=18.0
eta bin=10 MIP ADC=19.5
eta bin=11 MIP ADC=21.0
eta bin=12 MIP ADC=22.5
    However, the eta dependence for the data is too flat (could be a fit artifact).

  • Decisions:
    - declare the expected MIP calibration is no better then ~20%
    - apply moderate cut: pT>0.4 GeV/c and eta/phi to reduce data by 50%
    - fit Landau to each tower. If fit fails due to a low statistics take average ADC=[5-60] and convert it according to the formula
       "peak" = 1.03 * "mean" - 7.98
    (see this plot)
    - load gains to DB on the day 120 (May 1). After this date HV was not changed.

    Piotr & Jan, updated June 19,2003.

    Distribution of variables before any cut has been applied.
    Fig 1. Y/cm vs. X/cm for accepted tracks at the depth of SMD.
    Towers at eta bin>=4 without ADC>4 : 5TC12, 5TD10, 5TD04, 6TA12, 6TB04, 6TC09, 7TB09, 7TD08, 8TD08, 8TD04, 8TE12 are most likely to be dead


    Fig 2. Top-left : delta phi/deg. Top-Right : delta eta. Bottom-left : dE/dX (a.u.) vs. transverse momentum (GeV/c). Bottom-right : transverse momentum (GeV/c). Lines mark various cuts. ( ps )

    Fig 3. Quality of the track matching for individual towers. X-axis enumerates 240 towers, new eta bin starts every 20 [X=20*(etaBin-1)+(sub-'A'+5*(sec-5))]. Top: Y=delta phi/deg. Middle: Y= delta eta. Bottom: number of entries per tower with ADC>4.

    Fig 4. Example of summed over phi ADC spectra with cut=A . ADC spectra for all towers at eta bin=5 are plotted at the top left panel. Black, green, and red histo correspond to triggers selection: Fpd, Mid, and Endcap. Landau shape was fitted for ADC=[8,60]. The top-right panel shows eta bins 6, and so on.

    PostScript for : cut=A , cut=S , cut=P , cut=L , cut=S and P
    log scale : cut=A , cut=S , cut=P , cut=L , cut=S and P

    Table 1: MIP peak position from Landau fit 
    cut=A (all data)

eta   trg=Fpd      trg=Mid        trig=Endcap
bin   mean  err    mean  err      mean  err
 5    13.1  0.4     13.9  0.4     15.8  0.6  
 6    14.0  0.3     14.0  0.4     16.1  0.6  
 7    14.6  0.3     15.0  0.5     16.7  0.5  
 8    15.3  0.4     15.4  0.5     16.9  0.7  
 9    15.5  0.4     15.8  0.5     16.2  0.9  
10    14.7  0.4     14.8  0.5     16.5  0.8  
11    15.8  0.3     15.3  0.4     16.9  0.5  
12    16.8  0.4     17.1  0.6     17.6  0.8  

cut=S  (del eta & phi ) 
 5    13.8  0.5     15.3  0.4     16.4  0.6  
 6    15.0  0.3     14.3  0.5     16.6  0.6  
 7    14.7  0.5     15.4  0.6     17.3  0.6  
 8    15.2  0.5     15.3  0.5     18.2  0.7  
 9    15.9  0.4     15.9  0.6     17.3  0.8  
10    15.3  0.5     14.9  0.6     16.5  0.9  
11    15.9  0.4     15.5  0.5     16.6  0.6  
12    16.3  0.6     17.5  0.6     16.5  1.2  


cut=P (pT>0.5)
 5    13.3  1.0     13.6  1.2     16.4  0.7  
 6    15.1  0.5     15.4  0.5     17.5  0.9  
 7    15.8  0.6     16.9  0.9     17.7  0.8  
 8    16.9  0.6     17.1  0.8     18.2  1.3  
 9    17.2  0.6     17.6  0.8     18.0  1.3  
10    16.9  0.5     16.0  0.9     18.6  0.8  
11    17.2  0.5     17.1  0.6     18.1  0.8  
12    18.9  0.6     18.7  0.8     19.8  1.0  
 
cut=L (dE/dX=[0.5,2.])
10    12.0  6.7     18.6  1.1     16.3  0.9  
11    15.8  0.5     15.4  0.7     17.3  0.5  
12    17.6  0.6     17.9  0.6     17.9  0.8  

cut=S and P ( del eta & phi & pT)
 5    14.3  1.3     17.1  1.3     16.9  0.9  
 6    16.3  0.6     16.9  0.7     17.9  0.9  
 7    16.9  0.8     17.4  0.9     19.7  1.2  
 8    17.7  0.9     17.5  0.9     21.2  1.4  
 9    18.3  0.6     17.3  1.1     20.7  1.1  
10    17.6  0.7     15.6  1.3     19.2  1.1  
11    17.5  0.5     17.7  0.7     18.6  1.0  
12    19.1  0.7     20.2  0.8     18.3  2.1