On Mon, 11 Dec 2006, Malcolm J. Currie wrote:
> This was the effective command line.
>
> makeflat in="*pre(~2000,~2000)" out=firsflat method=br keepin minpix=1 \
> title='Output from MAKEFLAT' useset=f logto='Both' usevar genvar=f \
> logfile=CCDPACK.LOG clean gamma=3 iter=3 boxsize="[15,15]"
>
>
> Input Parameters: Mean Value %<Contribution>
> ...raw/010806-1113-pn0-pre(1999:3998,2001:4000) 111.17 15.11
> ...raw/011006-1160-pn0-pre(1999:3998,2001:4000) 109.99 48.51
> ...raw/011206-1265-pn0-pre(1999:3998,2001:4000) 159.05 36.27
> ...raw/020806-1117-pn0-pre(1999:3998,2001:4000) 172.71 18.58
> ...raw/020806-1118-pn0-pre(1999:3998,2001:4000) 106.75 45.71
> ...raw/021206-1270-pn0-pre(1999:3998,2001:4000) 123.52 22.61
> ...raw/031206-1275-pn0-pre(1999:3998,2001:4000) 110.87 19.19
> ...raw/031206-1277-pn0-pre(1999:3998,2001:4000) 75.87 20.42
> ...raw/041206-1279-pn0-pre(1999:3998,2001:4000) 120.68 52.36
> .../raw/080306-331-pn0-pre(1999:3998,2001:4000) 113.99 66.66
> ...raw/100706-1092-pn0-pre(1999:3998,2001:4000) 183.50 63.36
> ...raw/181106-1229-pn0-pre(1999:3998,2001:4000) 114.58 37.71
> ...raw/191106-1225-pn0-pre(1999:3998,2001:4000) 291.30 11.17
> ...raw/200806-1131-pn0-pre(1999:3998,2001:4000) 133.62 18.29
> ...raw/200806-1133-pn0-pre(1999:3998,2001:4000) 102.83 34.79
> ...raw/201106-1233-pn0-pre(1999:3998,2001:4000) 203.69 38.71
> ...raw/231106-1241-pn0-pre(1999:3998,2001:4000) 211.89 39.45
> .../raw/260306-481-pn0-pre(1999:3998,2001:4000) 106.44 17.52
> ...raw/281106-1252-pn0-pre(1999:3998,2001:4000) 153.19 17.97
> ...raw/291106-1259-pn0-pre(1999:3998,2001:4000) 233.46 8.80
>
> Data combination method : BROADENED
> Minimum number of contributing pixels : 1
>
> Output Parameters:
> Output NDF: /data/mjc/topsat/raw/firstflat
> Output NDF data type : _REAL
>
>> stats firstflat
>
> Pixel statistics for the NDF structure /data/mjc/topsat/raw/firstflat
>
> Title : Output from MAKEFLAT
> NDF array analysed : DATA
>
> Pixel sum : 3.89519e+06
> Pixel mean : 0.973799
> Standard deviation : 0.0393955
> Minimum pixel value : 0.824687
> At pixel : (2814, 2482)
> Co-ordinate : (2813.5, 2481.5)
> Maximum pixel value : 1.17688
> At pixel : (2711, 3285)
> Co-ordinate : (2710.5, 3284.5)
> Total number of pixels : 4000000
> Number of pixels used : 4000000 (100.0%)
>
>
> The surprising thing is the mean being markedly different from 1.0. I
> tried using the mean. The percentage contributions are far more even in
> the 90+ percent region, yet the mean of the flat was 0.970. Looking at
> the image the left-hand side has a mean below 0.96 with pixel-to-pixel
> fluctuations of about 0.02 and the right half has a mean about 0.99,
> 0.045 noise (as seen from GAIA->X/Y Mean of the whole array).
>
> Between 2 to 5% of 4 million pixels are rejected in the clipping stage
> reducing the sigma by typically 15%.
>
> Does anyone have any ideas why the mean flatfield value isn't very close
> to 1.0?
Hi Malcolm,
must be down to significant shape variations in the data, or non-symmetric
outliers which have not been cleaned, which amounts to the same thing. If
the shapes are quite different you can imagine cases where having data
normalised to one, and then combined, (which is what MAKEFLAT does), give
a mean other than one. The things you're combining for a flat field are
supposed to be more-or-less the same thing (/shape).
Cheers,
Peter.
|