This note documents the current estimate of the systematic error in the Epoch 3 L7_240 predicted background, derived from data that was not used in the construction of the model. The systematic error is at least 0.18 cts/s/3pcu in the 2-10 keV band for data selected from the front layer. This is approximately three times as large as the statistical error on a given data point.

The background data are accumulated from five different pointing directions (table 1). Data from pointing directions 2-5 were used in the construction of the model, and data from the other pointing direction were used to estimate the systematic error in the predicted background. Light curves were extracted from the data using the standard selection criteria for faint sources (table 2).

Table 1: Pointing Directions

Pointing			Rates vs Bin number	Distribution of Observed-Model rates	Residuals vs Time
1a,b	5.00	-67.00	pt1_plot1.ps	pt1_plot2.ps	pt1_plot3.ps
2a	60.00	+2.00	pt2_plot1.ps	pt2_plot2.ps	pt2_plot3.ps
3a	138.00	+15.00	pt3_plot1.ps	pt3_plot2.ps	pt3_plot3.ps
4a	235.00	+10.00	pt4_plot1.ps	pt4_plot2.ps	pt4_plot3.ps
5a	345.00	-18.00	pt5_plot1.ps	pt5_plot2.ps	pt5_plot3.ps

Table 2: Selection Criteria

Criterion	Description
(BKGD_THETAa < -70.0 or BKGD_THETA > 150.0) or TIME_SINCE_SAAb > 30.0 or TIME_SINCE_SAA < 0.0	Outside the South Atlantic Anomaly
PCUn_ON 0.5	PCUn fully on
ELV > 10.0	Angle between the satellite pointing direction and the Earth's horizon
OFFSET < 0.01	Angle between the satellite pointing direction and the nominal field of view
VpX1L, VpX1R < 6.25	Veto rate between the propane layer and the left/right halves of the Xenon layer
(VpX1L+VpX1R)/Q6c < 0.1	Normalized propane layer veto rate

Light curves are made available showing the observed rate (sky plus internal background), R_i, the model predicted rate, M_i, and the observed-model rate, X_i = R_i - M_i , for each of the five background pointing directions (column 4 in the above table). Time is not shown here as the data are taken from many orbits spread over many years. Instead, the x-axis is a running number of successive orbital bins rather than time. The statistical error for individual points is simply the Poisson error, _i = R_i .

A second set of figures show the distribution of observed-model rate for each background pointing direction (column 5). In the absence of systematic errors, the sigma of the distribution is expected to be the average error.

Finally, plots of the observed-model rate versus time are given for each background pointing direction.

For a particular light curves, the total variance is given by

where the individual data values are weighted by a factor

which includes both systematic and measurement errors. The latter is given by

and the derived systematic error is

Values of the standard deviation, average error, and derived systematic error are given for each of the five background pointing directions (table 3). The large values of systematic error are due, in part, to a small number of outliers, which may (or may not) be representative of those that will also occur in source data. In all but one of these orbital points, the individual 16s count rate rise toward the end of the orbit, and the effect occurs on one particular day (97/01/01), and in all PCU. This suggests that the effect is non-cosmic in origin, and further investigations are under way. Therefore, we also calculate the systematic error excluding points greater than 7 from the mean (i.e. the aforementioned outliers). These values are also given (table 4).

Table 3: Systematic Errors

Pointing	su	< err >	syst
1a	0.335	0.069	0.328
2b	0.160	0.069	0.145
3b	0.280	0.069	0.271
4b	0.211	0.067	0.200
5b	0.212	0.068	0.201

Table 4: Systematic Errors excuding Outliers

Pointing	su	< err >	syst
1a	0.193	0.071	0.179
2b	0.160	0.069	0.145
3b	0.153	0.069	0.137
4b	0.159	0.067	0.144
5b	0.162	0.068	0.147