Calibrating SC & GL with updated Calib_SC_GL.C
I modified Calib_SC_GL.C to include the alternative calibration fit methods for GridLeak and SpaceCharge. I then ran the macro on files which Alán and Qiye had prepared for Run 12 calibrations. Here I discuss some observations. Please see the descriptions of those alternative fit methods for definitions of the terms I used on this page.
Note on images: clicking on images links to a PDF version (if you want to read the text on the plots better), and the PNG images may actually be larger when viewed on their own (i.e. they are shrunk in the HTML of this page).
Note on fit results: the macro does two independent 3 parameter fits first for fGL of gapf and fSC of sc data respectively before doing a finaly unifying fit of fSCGL on both gapf and sc data simultaneously.
__________________
Stability across use of different calibration parameters:
Here are the results of three groups of three applied parameter sets (i.e. 9 different applied parameter sets total) for UU193, followed by chisq/dof contour plots of the final unfiying fit. The three groups all give results that agree within errors.
SpaceCharge fit results (scaler: zdce+zdcw): Finished outlier removal: 0 (of 93) data points removed (2 iterations) SO : 2521.25 +/- 2525.4 log(SCe) : -13.9974 +/- 0.0716771 log(g5) : 2.43514 +/- 0.127971 RMS_all = 0.0001032 for zdce+zdcw GridLeak fit results (scaler: zdce+zdcw): Finished outlier removal: 0 (of 93) data points removed (2 iterations) g2 : 1.07917 +/- 2.34077 gGL/g2 = SCxGL : 2.92962e-07 +/- 2.42954e-07 GLO : -23668.3 +/- 49653.9 *** FIRST PASS CALIBRATION VALUES: *** SC = (4.736e-08 +/- 2.274e-08) * ((zdce+zdcw) - ( 2521 +/- 2525)) with GL = 6.19 +/- 7.95 USING STDDEV SC = 0.000104 :: GL = 0.015784 SpaceCharge & GridLeak fit results (scaler: zdce+zdcw): g2 : 0.969168 +/- 2.99887 log(g5) : 2.45177 +/- 2.28066 log(SC) : -16.9477 +/- 0.101314 SO : 2490.12 +/- 3584.67 log(GL) : 2.03617 +/- 0.250467 GLO : -18730 +/- 68731.3 *** FINAL CALIBRATION VALUES: *** SC = (4.362e-08 +/- 4.419e-09) * ((zdce+zdcw) - ( 2490 +/- 3585)) with GL = 7.66 +/- 1.92 |
SpaceCharge fit results (scaler: zdce+zdcw): Finished outlier removal: 0 (of 90) data points removed (2 iterations) SO : 2410.76 +/- 2630.65 log(SCe) : -13.9431 +/- 0.0734539 log(g5) : 2.53385 +/- 0.130556 RMS_all = 0.0001033 for zdce+zdcw GridLeak fit results (scaler: zdce+zdcw): Finished outlier removal: 0 (of 90) data points removed (2 iterations) g2 : 1.28046 +/- 4.08483 gGL/g2 = SCxGL : 3.24771e-07 +/- 2.31198e-07 GLO : -14265.5 +/- 36573.5 *** FIRST PASS CALIBRATION VALUES: *** SC = (4.407e-08 +/- 1.99e-08) * ((zdce+zdcw) - ( 2411 +/- 2631)) with GL = 7.37 +/- 8.37 USING STDDEV SC = 0.000104 :: GL = 0.015294 SpaceCharge & GridLeak fit results (scaler: zdce+zdcw): g2 : 1.27919 +/- 5.78729 log(g5) : 2.52205 +/- 1.45551 log(SC) : -16.9371 +/- 0.103622 SO : 2409.05 +/- 3730.94 log(GL) : 1.99852 +/- 0.269841 GLO : -14198.7 +/- 51837.4 *** FINAL CALIBRATION VALUES: *** SC = (4.409e-08 +/- 4.568e-09) * ((zdce+zdcw) - ( 2409 +/- 3731)) with GL = 7.38 +/- 1.99 |
SpaceCharge fit results (scaler: zdce+zdcw): Finished outlier removal: 0 (of 89) data points removed (2 iterations) SO : 2583.93 +/- 2745.24 log(SCe) : -13.9905 +/- 0.0774144 log(g5) : 2.43713 +/- 0.141744 RMS_all = 0.0001095 for zdce+zdcw GridLeak fit results (scaler: zdce+zdcw): Finished outlier removal: 0 (of 89) data points removed (2 iterations) g2 : 1.29112 +/- 3.67433 gGL/g2 = SCxGL : 3.16412e-07 +/- 2.23631e-07 GLO : -16235.1 +/- 37458.4 *** FIRST PASS CALIBRATION VALUES: *** SC = (4.572e-08 +/- 2.136e-08) * ((zdce+zdcw) - ( 2584 +/- 2745)) with GL = 6.92 +/- 7.91 USING STDDEV SC = 0.000110 :: GL = 0.015389 SpaceCharge & GridLeak fit results (scaler: zdce+zdcw): g2 : 1.32024 +/- 5.32895 log(g5) : 2.40959 +/- 1.42456 log(SC) : -16.9045 +/- 0.109115 SO : 2573.48 +/- 3894.28 log(GL) : 1.94915 +/- 0.272678 GLO : -15420 +/- 51390.4 *** FINAL CALIBRATION VALUES: *** SC = (4.555e-08 +/- 4.97e-09) * ((zdce+zdcw) - ( 2573 +/- 3894)) with GL = 7.02 +/- 1.91 |
 |
 |
 |
_______
Validity of forcing SO=GLO:
Using 10 different applied parameter sets, here are fits of 6 parameters vs. 5 parameters (GLO = SO) for UU193, demonstrating that 5 parameters does not work well for some data. I believe this is because the individual fSC and fGL fits give an SO and GLO that differ by too much. The GLO parameter is not well-constrained by the gapf data (large error in the fGL fit), so the single parameter comes to a value which is close to that of SO from the fSC fit. But for this higher value of SO to work in the gapf data, the fSCGL fit must raise the value of GL to make the function go through the data and the intercept (offset), arriving at a very high GL value.
Note: the individual fits (fSC and fGL) are the same regardless of what we do with the parameters of the fSCGL fit, so I've only shown them once in common.
| 6 parameters | 5 parameters (GLO=SO) |
|---|---|
SpaceCharge fit results (scaler: zdce+zdcw): Finished outlier removal: 0 (of 294) data points removed (2 iterations) SO : 2376.2 +/- 7595.99 log(SCe) : -13.9848 +/- 0.212997 log(g5) : 2.45961 +/- 0.382308 RMS_all = 0.0001069 for zdce+zdcw GridLeak fit results (scaler: zdce+zdcw): Finished outlier removal: 0 (of 294) data points removed (2 iterations) g2 : 1.28933 +/- 9.51165 gGL/g2 = SCxGL : 3.28075e-07 +/- 6.53659e-07 GLO : -14069.2 +/- 101562 *** FIRST PASS CALIBRATION VALUES: *** SC = (4.412e-08 +/- 6.035e-08) * ((zdce+zdcw) - ( 2376 +/- 7596)) with GL = 7.44 +/- 24.40 USING STDDEV SC = 0.000305 :: GL = 0.043034 |
|
SpaceCharge & GridLeak fit results (scaler: zdce+zdcw): g2 : 1.24971 +/- 13.0845 log(g5) : 2.46395 +/- 7.79093 log(SC) : -16.9779 +/- 0.299852 SO : 2343.23 +/- 10759.7 log(GL) : 2.10686 +/- 0.709662 GLO : -11673.9 +/- 139816 *** FINAL CALIBRATION VALUES: *** SC = (4.232e-08 +/- 1.269e-08) * ((zdce+zdcw) - ( 2343 +/- 1.076e+04)) with GL = 8.22 +/- 5.84 |
SpaceCharge & GridLeak fit results (scaler: zdce+zdcw): g2 : 0.685682 +/- 8.65123 log(g5) : 2.47992 +/- 1.3724 log(SC) : -17.5258 +/- 0.300813 SO=GLO : 2264.45 +/- 10756.7 log(GL) : 3.12904 +/- 0.456628 *** FINAL CALIBRATION VALUES: *** SC = (2.447e-08 +/- 7.361e-09) * ((zdce+zdcw) - ( 2264 +/- 1.076e+04)) with GL = 22.85 +/- 10.43 |
 |
 |
And a 6 parameter fit vs. 5 for 7 different applied parameter sets of pp500, demonstrating that 5 parameters can work well for some data. Here, I think SO and GLO are close enough to each other in the individual fits that the compromised single offset in the 5 parameter fit is sufficiently close to each of them to not have so big of a side effect. GLO must drop a little bit to approach SO, and so GL must also drop just a little bit to make fSCGL pass through both the gapf data and the intercept (offset), but not much.
| 6 parameters | 5 parameters (GLO=SO) |
|---|---|
SpaceCharge fit results (scaler: zdcx): Finished outlier removal: 4 (of 260) data points removed (3 iterations) SO : 3644.26 +/- 8554.46 log(SCe) : -11.9658 +/- 0.0435477 log(g5) : 2.65234 +/- 0.0719733 RMS_all = 0.001171 for zdcx GridLeak fit results (scaler: zdcx): Finished outlier removal: 0 (of 260) data points removed (2 iterations) g2 : 1.10472 +/- 1.02305 gGL/g2 = SCxGL : 2.67195e-06 +/- 2.07912e-07 GLO : 10089.5 +/- 14934.8 *** FIRST PASS CALIBRATION VALUES: *** SC = (2.598e-07 +/- 3.075e-08) * ((zdcx) - ( 3644 +/- 8554)) with GL = 10.28 +/- 1.57 USING STDDEV SC = 0.002320 :: GL = 0.044169 |
|
SpaceCharge & GridLeak fit results (scaler: zdcx): g2 : 1.10472 +/- 1.44261 log(g5) : 2.65403 +/- 1.91557 log(SC) : -15.1621 +/- 0.0539041 SO : 3768.34 +/- 12150.4 log(GL) : 2.32927 +/- 0.0880414 GLO : 10067.3 +/- 21062.3 *** FINAL CALIBRATION VALUES: *** SC = (2.601e-07 +/- 1.402e-08) * ((zdcx) - ( 3768 +/- 1.215e+04)) with GL = 10.27 +/- 0.90 |
SpaceCharge & GridLeak fit results (scaler: zdcx): g2 : 1.10212 +/- 1.44172 log(g5) : 2.64994 +/- 2.58898 log(SC) : -15.1318 +/- 0.0539347 SO=GLO : 5353.84 +/- 10528.3 log(GL) : 2.27523 +/- 0.0890868 *** FINAL CALIBRATION VALUES: *** SC = (2.681e-07 +/- 1.446e-08) * ((zdcx) - ( 5354 +/- 1.053e+04)) with GL = 9.73 +/- 0.87 |
 |
 |
More to come...
-Gene
- genevb's blog
- Login or register to post comments