How to focus manually a lens for which it was not designed.

[xilman]

Originally Posted by fivemack

I'm having considerable trouble getting rid of the background without getting rid of the spiral arms, but attached is the result of a stack in Photoshop of all the M33 images with no attempt at background removal; NGC604 is definitely non-stellar, the faint star at [128,797] next to the brighter binary is recorded by SIMBAD as magnitude 13.8; the blobs at [412,393] [492,437] [587,426] definitely correspond to big nebulae in the AladinLite view (NGC595, NGC592, NGC588).

The star at [215, 732] is I think 2MASS J01343393+3022233 with V=15.1

The SN patrol charts I have date back to 1987 and are designed for visual observers. There are only three comparisons noted. "30" is at 657,496 with V=11.80 B=12.23; "G" is at 235,299 with V=12.66 B=13.17 and "1" is at 630.437 with V=13.36 B=13.82

[jwaltos]

I've been following your discussion also and have found it intriguing. I'm always
up for ingenious solutions to seemingly intractable problems.
I have an Orion XT6 Dobsonian which I am pushing beyond its design specs.

"R" has a nifty GPU facility excellent for stats..I have also used Scilab.. and CelestLab from
CNES may be worth a look if only to get some ideas.

One thing on my `wish` list is to utilize quantum computing tools within an `enhanced`
telescope (see Wikipedia "Action at a distance") not just to view marvelous objects
in the sky but to also test, review and better understand certain theories.

Anyways, I have an up and coming grandchild that has a star named after him. I hope he
will have the curiosity to seek it out since it won't be an easy find but it will be
an educational one.

[jwaltos]

One thing I am cobbling together is to find a means of observing dark matter with simple tools. Combining adaptive optics allowing for observation throughout the EM spectrum with a means of observing gravitational spectra seems to require merging the very, very small with the very, very large.
For any determined amateur there are a lot of great cost effective tools out there but the adage of thinking twice and acting once requires a healthy sprinkling of creativity and resourcefulness.

[xilman]

Originally Posted by fivemack

I'm having considerable trouble getting rid of the background without getting rid of the spiral arms, but attached is the result of a stack in Photoshop of all the M33 images with no attempt at background removal

Here is my latest attempt. 200pixel Gaussian blur to remove the background; two iterations of a 4th-order 13x13 Savitzky-Golay filter to remove noise; conversion to greyscale; brightness normalization and, finally, moderate gamma correction. There are still a few artefacts, notably a swathe of unremoved background in the lower left and a sprinkling of noise all over but it is markedly better than what I've produced so far. Deconvolution with the new PSF might help but that will require more fiddling around than I have time for right now.

The spiral arms look rather pretty IMAO.

[xilman]

Originally Posted by fivemack

I'm having considerable trouble getting rid of the background without getting rid of the spiral arms

Hint: take a look at the blue channel where the Na sky glow is missing.

[xilman]

I know I should investigate this myself but someone might already have done so and can save some repeated effort. Are there any known asteroids in the Capella field of view at the time the image was taken?

IRAF, STSDAS and the GSC 2.3 are wonderful resources. Earlier today I found >900 stars in the Capella image and downloaded the GSC in that region. The former resource has image coordinates and magnitudes in the filter implemented by the green channel with a presently unknown zero offset but which is probably about 8-10 magnitudes fainter than reality. The latter has over 9000 entries within \pm 15 arcmin of Capella and magnitudes in a variety bands. Most stars are much too faint to appear on the image, so I need to learn how to filter properly. For the time being the CSV file will be examined in a spreadsheet.

The first few days of 2017 will be spent fitting image coordinates to RA/Dec and correlating image magnitudes with V magnitudes. When that's in place I'll search for fainter stars (and possibly asteroids, hence my question) and build a PSF model for the image for subsequent image restoration.

[xilman]

Originally Posted by xilman

The first few days of 2017 will be spent fitting image coordinates to RA/Dec and correlating image magnitudes with V magnitudes. When that's in place I'll search for fainter stars (and possibly asteroids, hence my question) and build a PSF model for the image for subsequent image restoration.

Progress. It was a lengthy learning experience but well worth it

The attached capella.png is a 512x512 region of the green channel in Tom's original NEF file. It was centered on Capella and converted to FITS format. After much munging with IRAF tools a RA/Dec coordinate system was found good to around 2.5 arcsec in each axis; the 50-ish brightest stars extracted; photometry performed to (allegedly) 0.1 magnitudes (rising to 0.3 for the very faintest); and a PSF averaged of 17 stars created. The four brightest stars, including Capella, were rejected from the PSF fit because they are effectively saturated.

After that, the image had a background surface removed and then Maximum Entropy reconstruction run for 15 iterations with the noise estimate set to be sigma of the flat image. The resulting image had fitted quite a lot of the background noise (meaning that I'd underestimated it; perhaps I should have used 2 sigma) so I arbitrarily subtracted 1.7 sigma from all pixels. The result is mem15flat.png

If you want to look at these images properly you will doubtless have to download them and fiddle around with contrast stretching because Capella is so bright it hides most everything else. On my monitor, setting gamma=2.5 and then histogram normalization (with the aid of ImageMagick's display utility) give nice results.

Note that there are very pronounced artifacts around Capella and the two next brighter stars. The fourth brightest also shows a few artifacts but nowhere near as noticeable as when the background noise was left in place. I had expected the restoration of Capella to be lousy (which is why the sub-image was chosen, just to see how lousy) but it's a little disappointing that the next two brighter stars didn't restore too well.

The star images are now nicely circular and markedly sharper than in the original. Identifying them and discovering the catalogued magnitudes is a task for another day but I suspect the faintest are around 12th magnitude.

Subsequent tasks include repeating the process on the other two colour channels and attempting to build a plausible RGB image, then doing it all over again with the full 12M pixel RGB image. Once all that is done it will be time to start on the M33 stack, which will doubtless be another lengthy and frustrating experience.

[xilman]

Originally Posted by fivemack

The image scale with this setup is about six arc-seconds per pixel

According to one particular run of the MSCRED package in IRAF

Reference point: 2032.787  1419.384  (pixels  pixels)
X and Y scale: 5.904  5.916  (arcsec/pixel  arcsec/pixel)
X and Y coordinate rotation: 138.529  318.595  (degrees  degrees)

From other runs, I estimate the scale to be 5.91 \pm 0.01 arcsec per pixel.

The rotation is specific to this particular image and is estimate to be 138.3 \pm 0.3 degrees. Note that this corresponds to a 41.7 degrees clockwise rotation from the RA/Dec to the XY axes and a reflection in the Y-axis (i.e. increasing X coordinates correspond to decreasing RA).

[fivemack]

The mount I was using for the smaller lens ( http://www.ioptron.com/product-p/3302b.htm ) has the lens on a ball-head, meaning that it will be at some arbitrary rotation with respect to the sky.

For the larger lens, and the telescope, I use a proper equatorial mount ( http://www.orionoptics.co.uk/MOUNTS/...hereq5syn.html ) on which up-down on the image corresponds to north-south on the sky, and left-right to east-west, though there tend to be a few degrees of deviation because I can't leave the telescope aligned and polar alignment is fiddly.

In particular, the instructions for aligning the mount assume that you have something mounted on it with a reticle eyepiece that you can look through, and I haven't found a good way to get the DSLR to display an indicator for the centre of the frame that I can line up stars against.

This isn't so much a problem with 6-arcsec-per-pixel or 3-arcsec-per-pixel lenses, where I have enough of a field of view that I can see the object despite a few degrees of misalignment, but with the 0.75-arcsec-per-pixel telescope a few degrees of misalignment places the object invisibly outside the FoV.

I think the answer is to align very assiduously using a shorter lens and then switch to the telescope without adjusting the mount at all, but that's not something I can do during a brief break in the clouds. I got not a single usable image from a couple of hours with the telescope on 26 December, and now the moon is up.

[fivemack]

I'm impressed at what maximum entropy reconstruction has done for that image!

Using sky.esa.int and the "Mellinger color" background image (because Capella is not over-exposed enough to blow out a one-degree radius in that one), I can produce the following identifications:

106,53 = Tycho 3358-1284-1 [5 16 27.08 +46 24 57.82 B=7.1 V=6.9]
115,277 = Tycho 3358-2414-1 [5 17 48.73 +46 07 41.09 B=9.2 V=7.4]
141,249 = Tycho 3358-2973-1 [5 17 26.28 +46 08 06.88 B=9.5 V=9.4]
211,182 = Tycho 3358-3148-1 [5 16 31.80 +46 08 27.90 B=8.5 V=8.1]
201,146 = Tycho 3358-1074-1 [5 16 22.10 +46 11 45.19 B=11.3 V=10.3]
203,457 = Tycho 3358-3067-1 [5 18 18.07 +45 48 32.68 B=10.6 V=10.5]

I can't see in the image Tycho 3358-599-1 (5 17 30.64 +45 44 48.80 B=11.3 V=11.1] which would be around 300,432.

[xilman]

Originally Posted by fivemack

I'm impressed at what maximum entropy reconstruction has done for that image!

I can't see in the image Tycho 3358-599-1 (5 17 30.64 +45 44 48.80 B=11.3 V=11.1] which would be around 300,432.

TYC 3358-351-1 is at 05 15 11.314 +45 45 20.15 (473, 259) and is B = 12.46, V =11.51. My rough photometry puts it at 11.96 (\pm 0.14 random, unknown but perhaps \pm 0.2 systematic) in the G channel.

It's only just above the noise but DAOPHOT seems to find it convincingly enough. I've not yet found anything fainter which convinces me but perhaps when the entire 12M pixel image has been processed something will turn up. There are fainter ones in the DAOPHOT output but I've yet to convince myself that they are not just noise.

Added in edit: summing all three channels might give enough S/N improvement to lower the limit by another half magnitude or so, as 2.5* log(sqrt(3))= 0.6.