In this memo, we describe a past general trend of the overclock corrected bias, and a sudden change
occurred around the late August of 2005.
To visualize the trend, we first took 5 day moving averages of the data. Prior to this computation, we removed extreme out-layers. The methods of removing them are: (1). dropped 0.5% of the lowest and 0.5% of the highest data points, (2). fitted a straight line using Robust fit method, then took data points only in the 3 sigma area from the line. The moving averages are plotted with a black line in Figure 2. A 4th degree polynomial fit is also computed on the moving averages (not on the original data) so that a smooth fit can be plotted. The polynomial line is plotted with red.
To add an envelope to the data, the maximum and minimum values of the each period (moving 5 days) are taken, and then polynomial lines are fitted on them. These lines are plotted in blue.
Although it is statistically wrong to try to predict future trend from the polynomial fits, it is still useful to know a potential trend in future. We extended the lines linearly (not a polynomial line) for one year (light blue lines). Although we can extend the line futher, it is probably meaningless, and in some cases, it is totally confusing (see CCD4 Node 0).
The standard deviations of the moving averages are also plotted in the second panel.
This plot can show whether scattering of the data is changing or not. The red line
is a polynomial fit, and a light blue line is linearly extended to show a potential
future trend.
To see another CCD/Node, please select from the following table.
| CCD 0 | Node0 | Node1 | Node2 | Node3 | CCD 5 | Node0 | Node1 | Node2 | Node3 | |
|---|---|---|---|---|---|---|---|---|---|---|
| CCD 1 | Node0 | Node1 | Node2 | Node3 | CCD 6 | Node0 | Node1 | Node2 | Node3 | |
| CCD 2 | Node0 | Node1 | Node2 | Node3 | CCD 7 | Node0 | Node1 | Node2 | Node3 | |
| CCD 3 | Node0 | Node1 | Node2 | Node3 | CCD 8 | Node0 | Node1 | Node2 | Node3 | |
| CCD 4 | Node0 | Node1 | Node2 | Node3 | CCD 9 | Node0 | Node1 | Node2 | Node3 |
After an initial drop (DOM < 500), the bias backgrounds of CCDs were quite steady until the late August 2005. At that point, many CCDs/Nodes displayed a noticeable downward shift of the bias values. Especially CCD 0, node 0 and CCD 1, node 1 exhibited a huge jump (the sudden shift happened on Aug 19, 2005; DOM 2220).
Only CCD 5 shows a trend turning upward after the late August. Another Backside CCD: CCD 7, however, does not show this upward trend.
The ACIS gain also changed significantly around the same period (see MTA ACIS Gain Page or a memo by C. Grant). Although the cause could be same for both, the change of the gain is probably not a cause of the shift of the bias background (or vis versa). This is because the significant gain shifts are observed on CCD 0 and CCD 2 but the large bias shifts are observed in CCD 0, node 0 and CCD 1, node 1 . and they are not the same CCDs/nodes.
There are a few other irregular events happened around that time. Focal plane temperature was quite high (see a long term focal plane temperature plot). CTI was slightly dropped (CTI of CCD 0). However, none seems to be a direct cause of the bias shift. At this point, we cannot pin point the cause of these events.