These are the two poster children for the MLO fisheye: 56717.2500116 (Orion) and 56717.5972338 (Antares).

Here are the two scenes from Chile, taken a day later: ut030114.0170 (Orion, 56718.051296) and ut030114.650 (Antares, 56718.255312). This is a 15mm fisheye.

This is starting to be OK, actually. On the left is a vector field of the astrometric residuals from generic image 9082, where the arrows have been magnified by a factor of 100. On the right is a comparison of the tphot photometry with Hipparcos. (Hipparcos appears to be a lot better than Yale.)

The RMS scatter of the photometry is approximately 0.1 mag without much effort, and it is cut off at m=6 for a total of 1346 stars.

Photometry Comparison

I see that tphot does not have terrifically looking good residuals when fitting the stars in the MLO Lena pictures. It's really not a problem with fit parameters, unfortunately, it's just that the fisheye lens is a bit out of focus, making mesa-topped PSFs, and tphot's PSF just doesn't match very well. When run on the CP images the tphot residuals look really great.

I asked Chris to fire up sextractor on all the Lena images so that we could do a flux comparison. Obviously comparing the sextractor and tphot fluxes directly will not tell us who's better than whom, so instead I elected to use a little colmerge and awkery to make ".mch" files that replace the tphot instrumental magnitudes with those from sextractor. I can then use the "photo" mongo command to plot photometric zeropoint (comparison between instrumental magnitude with Hipparcos catalog magnitude) as a function of radius.

Note that the scatter of the MLO Lena images is mixed: there is a tight core and a fluffy halo. The reason for this is that MLO had a lot of haze north of Dec +40 that night and most all the scattered points lie there. The core at Dec below +40 has pretty much uniform zeropoints over the sky.

Here are the images from MLO: from left to right tphot and sextractor first epoch, tphot and sextractor second epoch.

Here are the images from CP: from left to right tphot and sextractor first epoch, tphot and sextractor second epoch.

There doesn't seem to be much question that tphot is about x2 better than sextractor, both in cases of difficult PSF and cases of rather Gaussian PSF. Note that the zeropoints for both photometers is very close to the same, so at least they have consistent mean flux estimates.

How about vignetting? I last looked at this 140412, and had
two big problems:

Not clear how to combine everything
There are clouds! Between 9070 and 9079 the instrumental mags changed
by 0.24 mag with small scatter. What the? That's when I think I
rolled the focus.
Anyway, here's the sequence, organized into adjacent pairs as best I
can, matching a vertical to off-vertical:
{{{
IMG_9060 V IMG_9062 S IMG_9070 V IMG_9076 S
IMG_9060 V IMG_9063 S IMG_9079 V IMG_9077 S
IMG_9065 V IMG_9064 S IMG_9079 V IMG_9078 S
IMG_9066 V IMG_9067 W IMG_9083 V IMG_9084 S
IMG_9069 V IMG_9068 W IMG_9083 V IMG_9085 S
IMG_9070 V IMG_9071 N IMG_9092 V IMG_9086 S
IMG_9070 V IMG_9072 N IMG_9092 V IMG_9087 S
IMG_9079 V IMG_9073 N IMG_9083 V IMG_9088 N
IMG_9079 V IMG_9074 N IMG_9083 V IMG_9089 N
IMG_9070 V IMG_9075 S IMG_9092 V IMG_9090 N
IMG_9092 V IMG_9091 N
}}}
Get all the points:
{{
Unknown macro: { exec 10<log.pairs while read LINE<&10 ; do img1=`echo $LINE | awk '
Unknown macro: {print $1}
'` img2=`echo $LINE | awk '
Unknown macro: {print $3}
'` colmerge 1,2 IMG_$img1.mch 1,2 IMG_$img2.mch -tol 0.001 | awk -v i=$img1 -v j=$img2 'NR>1
Unknown macro: {printf "%5.2f %5.2f %7.1f %6.2f %5.2f %7.1f %6.2f %5.2f %s %sn",$11,$16,$17,$18,$19,$38,$39,$40,i,j}
' done > fish.pairs exec 10>&-}}
}

Collect median differences as a function of r1,r2:
{{{
rm fish.diffs fish.errs fish.nums
for ((j=1150; j>=0; j-=50)) ; do
for ((i=0; i<=1050; i+=50)) ; do
d=(`awk -v i=$i -v j=$j '$3>=i && $3<i+50 && $6>=j && $6<j+50

Unknown macro: {print $7-$4}

' fish.pairs | median verb`)
printf " %6.2f" $

Unknown macro: {d[0]}

printf "%4d %4d %6.2f\n" $i $j $

>> fish.diffs
printf "%4d %4d %6.2f\n" $i $j $

Unknown macro: {d[1]}

>> fish.errs
printf "%4d %4d %6d\n" $i $j $

Unknown macro: {d[2]}

>> fish.nums
done
printf "\n"
done

monsta
rpts 1 fish.diffs xpix=(50,0) ypix=(50,0)
rpts 2 fish.errs xpix=(50,0) ypix=(50,0)
rpts 3 fish.nums xpix=(50,0) ypix=(50,0)
cop 4 3
sqrt 4
inverse 4
mi 4 2
}}}
Here's the diff and error images, stretched by -0.5 to +0.5 and 0 to 0.1.
The dark blue is errors of order 0.01 to 0.015.

The result is really remarkably linear in dmag in the r1-r2 direction,
just about 5e-4 mag/pixel. What makes more sense is to look for a
solution that goes as r^2^: here's a plot of dm as a function of r^2^
that has a fitted slope of 0.35 mag/(1000pix)². The scatter is about
0.1 mag.

This fisheye vignetting model is dm = 0.35 mag (r/1000pix)².

Browser not supported