Xilman

11 hours ago, neil phillips said:

Interesting recording these Satellites. More like professional astronomy work

You make an important point there which deserves deeper consideration. I feel an essay coming on. A later post perhaps.

I find it remarkable that amateurs can now do work which was exclusive to professionals only 20 years ago. Remember that these satellites were not discovered until mid-2000 and then they were found in images taken on the  3.6m CFHT telescope which is located at a superb site on Mauna Kea.

If it were not for the fact that the satellite orbits are very well characterized from 20 years of observation using the Cassini orbiter and ground based professional astrometry, my observations would be useful to refine their orbits. As it is, all I can confirm is that the JPL-provided positions are good to within a second of arc these days.

You could image these objects with your kit, were it not for the fact that Saturn is so far south in the sky right now. You could give it a try in another 4 years or so. The satellites of Uranus and Neptune with a comparable brightness (Sycorax and Nereid at least) should already be possible for you, as are dozens of asteroids and minor planets in the outer solar system.

32 minutes ago, Paul M said:

I was thinking that it is a bit fiddly but just now I've been playing with a feature in ASTAP, designed for stacking on a minor planet/comet. Select "Ephemeris Alignment" and it will stack on the ephemeris position of the selected object. That object need not be visible in individual subs or even the stack. It does have to be an object in its minor planet and comet catalogues which are editable. So I suspect it will be possible to get ASTAP to accept more esoteric objects so long as the MPC has them on file. 

I know, lazy! 

I just used it on a recent run of subs of the fast moving comet C/2020 R4 Atlas. It worked quite well but I screwed the final image by inadvertently including some outliers that were outside the field of the main tranche of subs. 

Anyway, thanks for giving me some ideas!

 

De nada.

I've never heard of ASTAP before. Is it a Windoze-only program? If so, does it work under WINE?

My procedure sounds fiddly but it actually takes very little typing (most of which is cut and paste) and not much more brainpower after the first couple of runs.

Here is Siarnaq, also known as Saturn XXIX. Most of the technical details are as for Albiorix above. It was imaged in 43 1-minute subs at 22:00 UT on 2020-09-13, which is a couple of hours earlier than Albiorix.

This was a tricky one because at the time of observation the V=20.9 magnitude satellite was only a few seconds of arc away from a star 2.5 magnitudes brighter --- a factor of ten in intensity.

The images are (left) stacked on the stars, showing a trailed Siarnaq, and stacked on the predicted motion of the satellite (right), showing noticeably trailed stars and a round satellite, though at poor contrast because of light bleeding in from the star. The quality is not as good as I would hope because there are clear guiding errors on the stars. Two remedies spring to mind. The first is to try again later this year and the other is to deconvolve the image using a bright unsaturated star to estimate the point spread function.

2 hours ago, Deadlake said:

They're cheaper if you go to Japan:

https://www.kyoei-osaka.jp/SHOP/nikon-nav-125hw.html

Still twice what I paid for the entire telescope, and that is without shipping and import duty.

Think I will pass on that one.

1 minute ago, Littleguy80 said:

I’m a little crazy in this respect. I use a Vixen HR 3.4mm in my manual dob (350x mag and 0.12 degree TFOV). You have to work for it but I’m quite well practiced now. I was able to see Neptune’s moon Triton with this combo. I only had orthos for high power eyepieces when I first got my dob so I had no choice but to get used to that. I guess a lot depends on the quality of the mount you’re using too. 

You're a better man than I, Gunga Din. Definitely old-age and incompetence, especially the latter...

1 minute ago, Stu said:

What eyepieces are you using to get to the x200? One benefit of the 100 degree afov eyepieces is significantly more time in between nudges compared with, say a Plossl or 68 degree.

Alternatively, do as I do and get an EQ platform. I use up to x360 with my tall and spindly 8” f8 quite easily with the target remaining centred for long periods.

I don't, as I said. I restrict myself to about x100.

I was commenting on figures presented earlier in the thread where figure > x250 were bandied around, as were sub-5mm focal lengths.

That said, back in the late 80's when I had a 18" Dob, a 10mm EP would give me over x200. Wasn't very useful so I stuck to (IIRC, it's been a long time) a 20mm, a 30mm and 2x Barlow for the odd occasions I thought high power would be useful. Only have a Skywatcher 250 now so the issue isn't as important.

Good point about an EQ platform. I'd like to find one which doesn't cost more than the scope. For me DIY means damage-it-yourself so making one isn't an option.

I don't know about you guys but I find magnifications > x200 or so essentially unusable with a Dob. The speed at which objects traverse the field of view (not to mention having to focus during that short interval) and the difficulty finding them again afterwards makes calculations based on exit pupil rather academic.

Perhaps it's just old age and incompetence but I restrict myself to x150 or lower.

Only if you have a decent drive should you start worrying about diffraction-limited magnification or excessively small exit pupils. Again, IMO.

17 hours ago, Xilman said:

I need to measure the image to get a good estimate for the SNR but I doubt that Albiorix reaches much more than 5 sigma and may be as low as 3 or so.  Whatever it is, the individual subs will have a sigma sqrt(62)  (approximately eight) times smaller and so well under unity. I will try to make those measurements tomorrow and report them here.

It was remarkably difficult to measure the SNR with the tools available to me. My favourite photometry program, Russ Laher's APT, failed completely. Astrometrica took a bit of persuasion and gave values between 2.7 and 3.4 depending on precise positioning of the aperture and its radius, so I conclude that the SNR sigma is about 3 --- as in the original guess.  That for each sub would be around 0.4.

The MPCReport.txt from Astrometrica contains the line

99999         C2021 05 04.00000 19 52 26.10 -21 42 19.0          22.43G      J22

where the position is 19:52:26.10, -21:42:19.0 and the magnitude is 22.43 in Gaia's g-band. The MPC ephemeris has 19:52:26.3, -21:42:18 and V=21.5. Agreement is satisfactory in my opinion. With such poor statistics and measurements in two very different band passes, one should expect positions and magnitudes to differ somewhat.

Rule of thumb is that sigma=3 is sufficient for identification of a known object, where additional information is available, but 5 is needed for discovery. These correspond to a 0.3% and 3ppm chance respectively that it is a false positive assuming the errors follow a normal distribution.

1 hour ago, Paul M said:

That's brilliant work. One thing that I'm trying to work out is how you stacked on the mean motion of the object? No reference object in any frame!

How did you achieve that?

Despite your claim, there were hundreds of reference objects in every frame --- the stars!

Stage 1: plate-solve the subs; this puts a WCS (world co-ordinate system) in the header of each. I use a local installation of astronomy.net for this.

Stage 2: look up the satellite's ephemeris and, in particular, the sky motions in RA & Dec of the satellite. https://minorplanetcenter.net/iau/NatSats/NaturalSatellites.html provides this service.

Stage 3: use a home written Perl script which tweaks the WCS in each sub's header to subtract off the sky motion which has taken place since the first sub was taken.

Stage 4: co-add all the tweaked subs using SWarp. That utility uses the WCS in each sub to align them.

Stage 5: use the ds9 FITS image viewer to stretch contrast, zoom, smooth, etc, until the target is visible at the location given in the ephemeris.

Easy, but fiddly, when you know how!

32 minutes ago, NigeB said:

That's a really nice analysis @Xilman! Would be interested in knowing what level of detection is represented in this image (i.e. sigma), and what levels in the individual subs are. Just curious. Regardless, really nice work, you've inspired me to try something similar. Did you do this from your UK or La Palma site?

Thank you for your kind words. The images were taken from the La Palma site because I have only a 250mm Dobsonian here in the UK. It's not suited for imaging and certainly not for objects that faint.

I need to measure the image to get a good estimate for the SNR but I doubt that Albiorix reaches much more than 5 sigma and may be as low as 3 or so.  Whatever it is, the individual subs will have a sigma sqrt(62)  (approximately eight) times smaller and so well under unity. I will try to make those measurements tomorrow and report them here.

Please try for this kind of imaging. Hardly anyone does it and although the orbits of Saturnian satellites are well known, that is not true of other solar system bodies of similar brightness and you can provide data of real value to refining their orbits. If it wasn't for the declination of the planets these days you would have no real problem picking up sub-21 magnitude satellites with your kit in Rutland. It's a pity our summer nights are so short and bright and the planets are so low in our sub-arctic skies.

Although this image is not of a planet, it is of a planetary satellite and I can't find anywhere more appropriate to post it.

Last summer I imaged (the locations of) some satellites of Jupiter and Saturn but have only just got around to processing them. One target was Albiorix, aka Saturn XXVI. Discovered in 2000, it is only 30km across, roughly half the diameter of the M25 motorway. At the time of observation it was magnitude V=21.5.

62 1-minute subs taken with an unfiltered SX814 on a 0.4m Dilworth were stacked on the mean motion of the satellite and the result compared with the MPC ephemeris and the DSS2 images. The stars are trailed; the faintest one nearby is catalogued at g=20.68 in Gaia EDR3. Its trailed image is marked with the asterisk and red arrow. Despite the low signal to noise, Albiorix shows up untrailed in precisely the correct location; there are no stars of comparable brightness at that position in DSS2 and there were no asteroids thereabouts at that time according to the MPC, so I'm reasonably confident of the identification.

4 hours ago, Froeng said:

...another one of the "simple astrophotographs". Not really a "pretty picture", but the proof that even supposedly faint objects can be quite rewarding with little effort! Unguided C8 at f/10 on Vixen SP-DX with 20x1min exposures using an old Olympus E-PL5

I wasn't sure this was going to work - but I did manage to capture the jet coming out of the black hole...!

Next time, I will try maybe 10sec subs and see if this would produce more detail.

Best Regards

Frank

Looks pretty to me.

Better than anything I have managed.

On 26/03/2021 at 16:15, Macavity said:

Aside from not too difficult (DOT-like!) 3C273, the "Hubble Deep Field".
- Or rather the "faintest" and (in this case) not too attractive an image? 🥳
Most of effort was in finding (multi-step EEVA star hopping) the thing...

[removed word] attractiveness!

If you've got it, flaunt it.

On 26/04/2021 at 07:09, Ouroboros said:

I have been thinking I ought to update my now rather old versions of KStars for some time. I am slightly worried about disrupting a working system.  How does one go about it? Does it update EKOS and INDI at he same time?  

If it aint broke, don't fix it.

That is what I have been told. I tend to follow the advice: "if it aint broke, fix it until it is".

On 03/04/2021 at 18:45, Macavity said:

Worth a go at KBO's maybe? Pluto is only a *couple* of magnitudes
fainter than e.g. Triton. But probably not too "well placed" just now... 🤔

A possibility may be (472651) 2015 DB216 which is a Uranus co-orbital centaur. It is presently in Virgo at mag 20.2.  If you fancy giving it a try, the position at 2021-04-06T00:00:00Z is 12:20:05 -02:06:00 and moving 8.47 as/hr in PA 280.5.

Not easy, but who wants an easy life anyway? By contrast Umbriel (UII) is mag 15.2 and shows up very clearly on your images.

Somebody please go for it. I can't because my observatory is inaccessible.

8 minutes ago, vlaiv said:

Any details on capture and processing for those lovely images?

Obviously I can only comment on the Caliban image, not the others.

Telescope: 0.4m (16" for those attached to imperial) Dilworth relay.

Camera: unfiltered Starlight Xpress Trius-PRO SX814

Good to see some interest so soon after posting!

Can't offer anything other than planetary satellites yet and won't be able to until I return to La Palma later this year.

Here is an image of Caliban, aka Uranus-XVI. Apologies for the low quality but it was magnitude 22.2 at the time.

I also have Nereid (N-II) and Sycorax (U-XVII), as well Triton if anyone is interested. Images of the big four of Uranus have yet to be processed. Not yet managed to get Miranda.

Anyone else here interested in imaging objects in the outer solar system?

It's not obviously "planetary" but I can't find anywhere more appropriate to post. Anything less than a parsec away is not deep sky in my opinion.

I'm thinking of TNOs, centaurs, Plutinos, KBOs, SDOs and many other members of the TLA zoo, as well as satellites of Uranus and Neptune.

It seems to me that this is an area ripe for exploration because hardly any amateur images out there. There be dragons!

Not always easy and often a bloody difficult challenge. Who wants an easy life anyway, and who is scared of a challenge?

On 28/03/2021 at 00:07, JRWASTRO said:

Not to worry, as we say.

I only wish to point out that the filtering algorithm should calculation of the polynomial coefficients at each step. In the example I used if you use a polynomial of order 11 and padded the data file with 5 points on the front and back one will need to compute 2000 polynomial sets.

About the DFT  - it is evaluated using the Fast Fourier Transform - but that is another issue!!

The FT of a signal is complex - it has a magnitude and a phase. For real signals the FT will exhibit Hermitian Symmetry (a very important property). When you multiply the X(f) by its complex conjugate X*(f) you will get a real value representing a power spectrum etc. etc. etc.

FYI I can take two signals x1(t)  => X1(f) and x2(t) => X2(f) and obtain : S12(f) = X1(f) x X2*(f) which is the cross power spectral density and is a complex function which tells us about the correlation between the two functions - especially the phase of the PSD.

Jeremy.

 

 

For your first point, the beauty of the S-G algorithm is that the polynomial sets are computed iteratively and represented as a convolution kernel. For those readers who have not yet learned the theory behind it, I urge you to do so. It really is a beautiful algorithm and the original paper is deservedly one of the most quoted in the scientific literature.

I know about how to compute Fourier transforms, their properties, and how to use them for various convolution, correlation, spectral analysis and so on. I kept my presentation simple and to the subject of smoothing of real valued time series data. Anyone who wishes to learn more can easily find vast amounts of information on the web, ranging from hand-waving to hard-core mathematics and everything in between, including software implementations.

In your particular case, wherein lies the significant noise? x1(t) or x2(t) or both?  Have you characterized the noise (e.g, pink, white or blue)? Have you tried smoothing one or both time series? Sorry I can't help much but that is because I have never needed to address the problem of interest. Presumably you have done a literature search.

8 hours ago, JRWASTRO said:

Very good.

Please tell me how many polynomial coefficient sets do I need to use here:

My noisy data set has c. 22000 points  of a complex waveform. The data is, of necessity, complex it is at 0Hz.  LPF here produces a BPF.

Jeremy.

I don't know is the simple answer. The reason is that I have only ever applied S-G filtering to purely real data. Complex data is outside my experience. Is its Fourier transform pure real?

The original question was for a real time series and my answer was given in that context.

However, a different and interesting question which I may investigate further.

7 minutes ago, Astro Waves said:

Side topic, I did post this is imaging discussion but its also very appropriate here seen as how no one answered it yet. Do the imaging settings need to be the same when imaging on multiple nights? I've managed to get two nights on the Whale galaxy but with different exposure times. They are also slightly framed different ...

Nope, none of exposure times, exposure dates and field centres need be the same.  To see the latter to be true, consider the number of mosaics that are out there: overlapping images which have been stitched together. All you need is co-adding software which can correctly position the subs. Personally I plate-solve first and then use SWarp to stack the subs. SWarp reads the co-ordinate system from the individually solved subs.

SWarp also handles the multiple exposure time issues by weighting each sub by its exposure before co-adding.

Do you get the idea that I really like SWarp?

12 hours ago, JRWASTRO said:

Greetings Steve,

Implement the SG filter and presenting the results will need some software like MATLAB (very expensive) or OCTAVE (free open source software).

Create a list of your raw data (CSV) that can be read by OCTAVE and it will be straight forward to implement a SG filter and or a Moving Average (MA) filter.

Octave is a good program but I dispute the necessary part. Anyone who followed my advice and started investigations with the Wikipedia article would have seen that coefficients up to degree 25 were published as long ago as 1965.

A link in the wikipedia leads to https://sites.google.com/site/chandraacads/resources/sg-filter/db

where you will find tables of coefficients for 1 through 4 dimensional data and for a wide range of polynomial degrees and window sizes.

Once you have the coefficients it is relatively easy to implement the filter in the programming language of your choice, Excel-compatible spreadsheets if you wish though, to be honest in high dimensions it will run rather slowly unless you use FFT convolution.

That said, if the tabulated coefficients are not suitable for your needs, you will need to compute them. Free software exists to do this for you. Another link from the Wikipedia page leads to C source https://zenodo.org/record/1288901

Have fun!

1 hour ago, vlaiv said:

Too much work as you can get the same result in easier way.

By doing that you are effectively interpolating between original data points. There is no more data to be gained - these intermediate stacking steps will have high correlation with other samples - so you might as well just interpolate existing 10 points with some interpolation algorithm to get nice curve.

check out this for example (hit example on the page and play around with data)>

https://tools.timodenk.com/cubic-spline-interpolation

Also look into Savitzky-Golay filtering. That method uses a least-squares polynomial fit to runs of data points. It works very well in practice in reducing noise without greatly broadening genuine features.

A simple variation can be used to find maxima and minima in the data, perhaps from an eclipsing binary.

Descriptions and code for S-G filtering are all over the net. Finding them is left as an exercise. Hint: Wikipedia is a good starting point.

6 minutes ago, AndrewRrrrrr said:

we use MJD at work - the integer part of the MJD is a much smaller number than the JD and the "day" starts at midnight. The JD "day" starts at noon. 

The JD is probably more useful to an astronomer because the integer number does not roll-over at midnight, in the middle of a viewing session.

So it depends what you're doing really. Maybe something to do with being at sea in the old days - the only time you could determine exactly was midday when the sun is at its highest point in the sky. 

Just a few guesses really!!!!

In these days when 64-bit cpus are widespread the extra size of the full JD (of which there are many variants BTW) is not such a big deal as when 32-bit floating point was commonplace. There is a big difference between 6-7 digits precision and 15-16 bit precision!

Your "more useful" comment belies your location. 😉  Someone in Oz or NZ may politely disagree.

44 minutes ago, robin_astro said:

 Ideally it is better to analyse the subs separately, dealing with any outliers at this stage eg with a sigma clip (or if present in all of them with a bad pixel map)

Seconded!