In this documentation:
--> Run 20 AuAu 9.2GeV dataset TPC space charge and grid leak calibration.
--> Dataset has 3 periods.
--> First calibration was performed individually for each periods. We found the fit parameters for period 1 that fits the data, however, period 2 and 3 showed odd behavior. 
--> Combining period 1 and 2 dataset, we are able to find best calibration parameters. Period 3 didn't fit well while combining, so this dataset calibration was done separately. 
--> Below, you can find final combined fit parameters for period 1 and 2 (at the beginning) and period 3 calibration progress and the final calibration parameters (at the end).  
--> Since GL is not well constrained, it is fixed to 0.5 for all periods.  
Period1+2 Combined Fit Final result:    

USING STDDEV sc => 0.000288 :: gapf => 0.011614

sc = (8.591e-08 +/- 1.998e-08) * ((bbce) - ( -1395 +/-   4740))

    Fig: Fit from the calibration macro.                              Fig: fit from SC ntuples with fit parameters.                               Fig: sc vs bbce from ntuple

--> I also tried fitting with 2 scalers.  The "PCA iteration 0" is parameters are in good agreement with the fit, however, the PCA iteration 1 has worst agreement. This is what Gene found as well. Below is the PCA iteration 0 result from the  2 scalers fit:

    Fig: 2 Scaler fit with Gene's parameter. 
--> There is not a good agreement. Here we can see two bands at lower ends.  

Remarks: 
--> The 1 scalar result agrees well with the data. With 2 scalar, the agreement improves, but not by much than in 1 scalar. I am inclined to go with the 1 scalar fit result. 

Following up: 
--> Gene wanted to see the correlation with the scalers combination "pow(vpde,0.3):bbce" from the calibration macro. Still, the PCA iteration 1 fit result and plot has huge offset.  Below is the fit result for PCA iteration 0. 

--> With pow(vpdw,0.3):bbce scalers combination: 

log(g5)     : 10.5393 +/- 1.62636

log(SC)     : 0.00156599 +0.164083/-0.193966 or +/- 0.178026

-->Gene suggested to check for different powers in vpdw to find the best correlation parameters. I started with the 0.01 to 0.5. We can see that linearity degrades as the power increases.  
--> This is the best agreement so far for the 2 scalers. 
--> With pow(vpdw,0.01):bbce, vpdw>0 

--> With pow(vpdw,0.1):bbce, with vpdw>0

*** FINAL CALIBRATION VALUES: ***

sc = (1.0 +/- 0.1715)*(-0.0006845*(pow(vpdw,0.1)-( 3.247 +/-  -0.44))+(8.359e-08*(bbce)))
  with GL =  0.50 +/-  0.00

-->With pow(vpdw,0.2):bbce, vpdw>0

Gene noticed that in some runs vpdw=0, for which the SC calibration parameters doesn't work. Those runs needs to be flagged out.

Calibration macro issue: 
It has been noticed that the Calib_SC_GL.C  macro was giving different PCA parameters for iteration 0 and 1. The 0th parameters used to do the job better, however, the iteration 1 parameters we far different. Gene fixed such issue in the calibration macro, which was tested with the Period 3 nutples. Below is the output from the updated calibration macro, where the iteration 0 and 1 are in good agreement and the parameters are consistent with the parameters what we have found before. Though the difference in parameters from iteration 0 and 1 is fixed, we now see that some parameters are at limit. Those needs to be figured out! 

-->Results with updated Calibration macro for pow(vpdw,0.1):bbce, vpdw>0

*** Running PCA iteration 0 ***

 

Running principal components analysis...

 

SpaceCharge fit results {scaler: (-0.000683357*(pow(vpdw,0.1)-3.2566))+(8.34491e-08*(bbce))}:

SO        :           1.00604e-05      +/- 0.000208434

log(SCe)           :           3.72373 +/- 0.117032

log(g5) :           3.708    +/- 1.7771         (AT LIMIT!)

 

GridLeak fit results {scaler: (-0.000683357*(pow(vpdw,0.1)-3.2566))+(8.34491e-08*(bbce))}:

g2                     :           0.100008          +/- 9.10326       (AT LIMIT!)

g1/g2 = SCxGL :           1.21458 +/- 5.04157

GLO                 :           0.0209163         +/- 0.0812235

 

SpaceCharge & GridLeak fit results {scaler: (-0.000683357*(pow(vpdw,0.1)-3.2566))+(8.34491e-08*(bbce))}:

g2                     :           0.0100009         +/- 0.777246     (AT LIMIT!)

log(g5)              :           4.70799 +/- 1.73374       (AT LIMIT!)

log(SC)             :           0.00199137       +/- 0.169789

SO                    :           7.65032e-06      +/- 0.000298696

 

*** FINAL CALIBRATION VALUES: ***

sc = (1.0 +/- 0.1698)*(-0.0006847*(pow(vpdw,0.1)-( 3.245 +/- -0.4371))+(8.362e-08*(bbce)))

  with GL =  0.50 +/-  0.00

 

 

*** Running PCA iteration 1 ***

 

Running principal components analysis...

 

SpaceCharge fit results {scaler: (-0.000694566*(pow(vpdw,0.1)-3.24445))+(8.35898e-08*(bbce))}:

SO        :           2.78488e-06      +/- 0.000208965

log(SCe)           :           3.72119 +/- 0.117028

log(g5) :           3.70798 +/- 1.87534       (AT LIMIT!)

 

GridLeak fit results {scaler: (-0.000694566*(pow(vpdw,0.1)-3.24445))+(8.35898e-08*(bbce))}:

g2                     :           0.1        +/- 6.7895         (AT LIMIT!)

g1/g2 = SCxGL :           1.21426 +/- 5.05104

GLO                 :           0.0209215         +/- 0.0812495

 

SpaceCharge & GridLeak fit results {scaler: (-0.000694566*(pow(vpdw,0.1)-3.24445))+(8.35898e-08*(bbce))}:

g2                     :           0.0100001         +/- 0.777154     (AT LIMIT!)

log(g5)              :           4.70797 +/- 1.73544       (AT LIMIT!)

log(SC)             :           -0.000551406    +/- 0.169772

SO                    :           3.52115e-07      +/- 0.00029943

 

*** FINAL CALIBRATION VALUES: ***

sc = (1.0 +/- 0.1698)*(-0.0006942*(pow(vpdw,0.1)-( 3.244 +/- -0.4311))+(8.354e-08*(bbce)))

  with GL =  0.50 +/-  0.00

 

            Fig: sc (data) vs sc(fit) from the updated calibration macro(Macro always picks the PCA iteration 1 parameters for the fit plot)