PN.S - Tests on Simulated Data

Sophisticated software is needed to exploit the matched images produced by the PNS. We are developing the required procedures with the help of simulated data. For the moment, we have tested the idea with conventional software such as DAOPHOT, which gives a minimal performance estimate.

An IRAF script was used to estimate the completeness of the PNe detection out to various distances in the fixed integration time of 28,000s for a 4m telecope (but now for a bandwidth of 20Å):

D	PN*	SNR	Detection Completeness
Mpc	counts	PN*	PN*   +1.0  +1.5  +2.0  +2.5

25	3390	 31	100%   70%   44%   13%   10%
15	9414	 70	100%   100%  88%   56%   38%
10     21190    122	100%   100%  100%  100%  88%	

PN* are the brightest PNe and completeness was checked to 2.5 magnitudes below PN*.
The background level (B) was taken as 8772 counts, and SNR=S/sqrt(S+B) has been shown for PN*. Here `completeness' was estimated by running numerous trials with randomised noise.

2. Receiver Operating Curves

In the previous table, no account was made of the number of accompanying spurious detections, which is an important factor. Roughly speaking, if each 2k x 2k image contains 100 false detections, then there is a 1/20 chance in finding a corresponding peak in the second image for each false detection (based on the fact that PNe pairs must occur along the same row). Thus, there will be 5 spurious final identifications. The relationship between detection efficiency and false detections been investigated by TSA. In the accompanying figure, for three values of the SNR in the image, a line is traced out from left to right corresponding to the lowering of a detection threshold. As one does this the number of detections increases but so does the number of spurious detections (here, for 2000x2000 image cells). At a SNR of 5, if the threshold is adjusted so that the number of spurious detections is < 1, the detection efficiency is around 85%.

For a fixed value of the final SNR, the second figure shows the performance of a 2 arm (50:50) spectrograph versus a 3 arm spectrograph. The distribution of photons between the three arms is a variable which has been optimised here. There is no improvement.

program listing

N.D. Aug 2000