
The four lines we used are 1.48 keV. 4.51 keV, 5.89 keV and 6.49 keV.
For each line we fit a "Gaussian" and a "Gaussian + constant". There
is no significant difference between the two fits. The mean and the
\sigma of the Gaussian are used for various comparisons.


CCvsTE.xls  : Spreadsheet with all the Cal data analyzed and their basic
              properties. One of the concern was that the FP temperature
	      was changing in the initial CC mode Ext Cal Source observations
	      and that could affect the results.


fig01.ps    : The difference between the PHA values for the four lines
              and those for a reference TE mode observation are plotted
	      as a function of energy.  The horizontal lines near "zero"
	      are TE-TE and the sloped lines are for CC-TE.  The CC mode
	      data shows larger PHA values corresponding to the 4 lines
	      than the TE data.  This can be accounted by a gain change
	      in the CC mode.

tables.ps   : Tables of the Gaussian fits to the four lines separately for
              each node.  Lower half of the table in each page shows the
	      results after removing 96 columns close to readout in each node
	      due to potential (?) problems with BIAS (see the note by Peter
	      Ford on the BIAS analysis of CC vs. TE data)

	      Table 1 : TE vs. TE.  First Column is the line energy. Then for
	      each node, I list the mean and \sigma for the Gaussian fit for
	      the two observations being compared.

	      Tables 2 & 3 : CC vs TE (See table caption for details of the
	      data sets used)

	      Table 4 : Same as Table 3 but using PI instead of PHA. Near the
	      high energy end, the PI values are similar but differ at the low
	      energy end.  The gain in conversion from PHA to PI is tuned at
	      the high energy end. Also note that the PI values for various
	      nodes are very similar, so large part of node-to-node variations
	      have been accounted for in conversion to PI.




CTI Correction for CC mode data:
===============================

In order to study the effect of CTI correction on the CC data, we use the PSU
CTI corrector (Townsley et al.) and use that on the Ext Cal Source Data in
both TE and CC modes.


TE mode data (ObsID 60991):
--------------------------

fig02_60991_S2_CHIPY_vs_PHA_TE.ps   Effect of CTI for FI CHIP S2.

fig02_60991_S3_CHIPY_vs_PHA_TE.ps   Effect of CTI for BI CHIP S3.

Note that for S3 the CTI does not seem to affect much because most of the BI
CTI is serial CTI.




CC mode data (ObsID 60993):
--------------------------

In CC mode the CHIPY values recorded in the event list are not position but
a way to record the event readout times.  Thus, the plot of PHA vs. CHIPY
shows average characteristics of the PHA distribution.

fig03_60993_S2_CHIPY_vs_PHA_CC.ps  Effect of CTI for FI CHIP S2 in CC mode.

fig03_60993_S3_CHIPY_vs_PHA_CC.ps  Effect of CTI for BI CHIP S3 in CC mode.


In case of S2 CC mode observations, we see that the lines are broadened due to
CTI, but for S3 the lines are narrow.



To compare the data in CC vs. TE mode, we plot the spectra of data integrated
over CHIPY.  The CTI corrector uses the default value of CHIPY in the event
list, therefore for TE data, it uses the real value of CHIPY and for CC data
it essentially uses a random number between 1 and 512 for CHIPY as the value
recorded in the CHIPY column has nothing to do with the Y position of the
event in the chip.



fig04_S2_spec.ps

This figure shows the observed spectrum of the Ext Cal Source on S2 in TE mode
(black line : ObsID 60991) and CC mode (green line : ObsID 60993) in the PHA
space.

Notice:

(1.)  This is the spectrum integrated over the whole range of CHIPYs and
that's why it is ugly.

(2.) The CC mode spectrum (green line) shows that it is LESS AFFECTED by CTI.
The CC peak is narrow and less extended towards lower energies.  Which means
that the charge loss in CC mode is less. One reason could be that charge loss
due to short time constant traps would be minimized is CC data where the
charge transfer timescale is 2.85 millisecond vs. about 40 microsecond for the
transfer timescale for the TE  mode data.

(3.)  In some senses, it implies that CC mode is more robust in terms of CTI
degradation that the TE mode.



fig04_S2_spec_CTI_COR.ps

Now we blindly apply Leisa Townsley's  CTI corrector to this data (using the
values of CHIPY in the event list), do the standard filtering etc. and plot
the spectra in PHA space again along with the pre-CTI correction spectra.

In the TE mode data (where we know the CHIPY positions for the events) the CTI
corrector does a fantastic job.  It takes the  raw TE data (black line) and
gives the corrected spectrum (green line) which is what would be expected.

In the CC mode data -- since we do not know the CHIPY and use essentially a
random number -- we start from the raw CC spectrum (blue line) and the CTI
corrector gives he corrected (red line).  The CTI corrector recovers the mean
energy for the lines quite accurately (as compared to the CTI corrected TE
data), but the lines are very broad.



fig05_S3_spec.ps

This figure shows the spectra of TE and CC mode observations of the Ext Cal
Source on S3 integrated over CHIPY.  Notice that the TE data (black line :
ObsID 60991) and CC data (green line : ObsID 60993) show very similar spectra
with the CC data showing a slightly larger PHA value than the TE data at the
high energy end.  (This is what was discussed at the start of this
presentation that the CC data shows larger slope in the PHA vs. energy compared
to TE).

Since for BI chip S3, the dominant CTI is the serial CTI, it is not surprising
that CTI has same effect on both CC and TE data as the serial transfer
timescale is the same for both CC and TE modes.


fig05_S3_spec_CTI_COR.ps

We apply CTI corrector to the S3 data using the values of CHIPY in the event
list.  For TE data, we go from pre-CTICOR (black line) to CTI corrected (green
line).  In TE mode, since we know both CHIPX and CHIPY positions for each
event, the CTI corrected spectrum is what would be expected.

For CC mode data, we go for pre-CTICOR (blue line) to CTI corrected (red
line).  Again an improvement.

Comparing the CTI corrected TE spectrum (green line) and the CC spectrum (red
line), we can conclude that the CTI correction on S3 data in CC mode does a
good job.  I believe this is because most of the CTI on S3 is serial CTI and
the CTI corrector accounts for that as we know the CHIPX positions of all
events in CC mode (serial CTI depends on CHIPX position).  There is some
parallel CTI which CTI corrector does not fully correct as the spectrum was
integrated over the full range of CHIPY.


I believe that for a point source observed on S3 in CC mode (where we can
calculate the point source position along CHIPY from the aspect solution) and
use that CHIPY value for CTI correction, the PSU CTI corrector would do a good
job at mitigating the effects of CTI and I would strongly recommend the
scientist to do so.

Also the CXC calibration group should plan to do some tests to verify or
reject this hypothesis about CTI correction of the CC mode data observed on BI
chips.


Note: Preliminary test on S1 data in CC and TE mode give results similar to
those obtained for S3.

Regarding the CC mode observations on FI chips, I feel that the CTI corrector
might work if we use the expected source position, but I am not sure of that.


Divas Sanwal
