Resolving M3 to the core with lucky imaging

A place to share images that you have taken with SharpCap.
Forum rules
Please upload large images to photo sharing sites (flickr, etc) rather than trying to upload them as forum attachments.

Please share the equipment used and if possible camera settings to help others.
timh
Posts: 218
Joined: Mon Aug 26, 2019 5:50 pm

Resolving M3 to the core with lucky imaging

Post by timh »

The large image is a composition of M3 comprising.

Three separate integrations of 52x40s, _90x20s and 180x10s_exposures at gain 124 (near unity) combined into a single 64 bit composition in PixInsight, processed and then stretched using high dynamic range B spline transformation to flatten down the brightness of the centre intensity so that stars can be resolved across the whole cluster.

PDS200 SW Newtonian, Baader flattener (F = 5.0) on a CEM70 Ioptron mount with PDS2 guiding (f = 160 mm guide scope and ASI120MM camera)
ASI 294MC PRO camera (4.63 uM) pixels at -10C

Bortle 6, Moon 2 days after full

M3 has a particularly compact core with stars reportedly packed as close as 1 per square arc sec and thus impossible to resolve under normal seeing. The image scale was 0.95 arcsec/ pixel and the average FWHM of the 10s frames was about 2.6 arcsec. The small picture top right shows a low stretch image of the core of the main image resolved to about 2.6 arc sec.

I was interested to see if it was possible to go further and resolve more stars using lucky imaging?

Therefore, with the same set up as above but on a later date (no moon) I used Sharpcap (64 bit beta =4.0.7728.0) and a mono camera with smaller pixels - ASI 294 MM PRO with 2.315 uM pixels - to capture a 12 min 640 x 320 pixel 16 bit SER file of 5570 x 130ms exposures at gain 570 (max). These were quality sorted in PIPP and then the best 57 (1%) stacked in Autostakkert to produce a TIFF file that was stretched in PI.

The resultant image - image scale ~ 0.47 arc sec/ pixel and therefore sampled at a level that could - in principle -show resolution down to about 1.2-5 arc sec - is shown at the bottom right. While the image is not very deep (only equivalent to 6.5s total exposure time) it does indeed show improved resolution versus the 2.6 arcsec image obtained conventionally and uncovers some stars that could not otherwise be detected.

The 130 ms time was a bit too long for lucky imaging - this was a compromise to get bright enough frames even at maximum gain. The camera sensitivity, time scale and brightness of the object were probably all at the edge of where deepsky lucky imaging just about becomes possible. Doing this in a better quality way will probably take a bigger scale telescope --both longer focus and reduced F number- as well as stacking the best frames properly and carrying out cosmetic correction or using a bad pixel map. But while the deepsky applications are limited it is encouraging to see the method finally work at all for something other than the moon or a planet.
Attachments
M3 luckyimaging.JPG
M3 luckyimaging.JPG (128.74 KiB) Viewed 471 times
User avatar
oopfan
Posts: 1002
Joined: Sat Jul 08, 2017 2:37 pm
Location: New York
Contact:

Re: Resolving M3 to the core with lucky imaging

Post by oopfan »

Hi Tim,

With regards to the "lucky" stack, perhaps there is something else at work here. In all of my AP shots, I am familiar with faint stars appearing smaller in diameter than bright, unsaturated stars. Could it be that your sub-one-second exposure is making otherwise bright stars appear fainter and therefore smaller in diameter?

EDIT: May I suggest another experiment? Use just one camera, the one with the smaller pixel scale. Choose an exposure for the lucky stack, say 50ms. Next, choose a gain where the peak pixel value of a star (of your choice) in the core is 1000 ADU. Capture several hundred frames. Next, choose a long exposure, say 20s. Choose a gain where the peak pixel value of that same star is 1000 ADU. Capture some number of frames which makes the integration time of the two stacks equivalent. Calibrate, register, and stack each of the two sets of exposures. Perform a side-by-side comparison.

Brian
User avatar
oopfan
Posts: 1002
Joined: Sat Jul 08, 2017 2:37 pm
Location: New York
Contact:

Re: Resolving M3 to the core with lucky imaging

Post by oopfan »

Tim,

Here is a close double star in Hercules useful for experimentation. Both stars are nearly equal in magnitude and color (7th and G2V).

Brian

Star data courtesy of Sky & Telescope
Star map courtesy of Computer Assisted Astronomy (C2A)
Orbit courtesy of Richard Dibon-Smith at http://www.dibonsmith.com
Attachments
Double Star ADS 11483 (Her).jpg
Double Star ADS 11483 (Her).jpg (110.33 KiB) Viewed 431 times
Last edited by oopfan on Mon May 03, 2021 11:28 pm, edited 1 time in total.
timh
Posts: 218
Joined: Mon Aug 26, 2019 5:50 pm

Re: Resolving M3 to the core with lucky imaging

Post by timh »

Hi Brian,

Thanks once again for the interest and the highly pertinent questions and thoughts. I did carry out a number of control experiments - not exactly as you described - but that convinced me that the lucky imaging had genuinely worked more or less as expected. See what you think...

You are right. I was worried at first that the improved resolution might be more apparent than real and simply due to oversaturation and/or star blooming in the process of stretching the linear image. So a mini series of experiments - seemed a bit too arcane to originally include but now you have provided the excuse :-).

First just to note that the two ASI 294 cameras are essentially exactly the same (same chip I think) except that the mono version has no Bayer matrix in front - a slightly higher QE therefore - and also allows you to unlock the standard 4.63 uM pixels into a 4 X 2.315 uM pixel mode. In the standard mode the ADC is 14 bit but only 12 bit in the 2.135 uM pixel mode. So it seemed valid to work controls across the two cameras since the differences between them are slight and predictable but it could have been done more rigorously.

All the experiments were carried out on the same night (30/04 to 01/05) switching over from first using the ASI 294 MC colour camera and then to the ASI 294MM mono camera. With either camera fitted focus seemed equivalent on the night - estimated FWHM 2.5-2.8 arc sec.

The image scale for captures with the MONO camera in 2.315 uM mode was 0.475 arc sec/ pixel. For the colour camera it was 1.05 arc sec/ pixel (not exactly double the mono because the colour camera had a 0.9X reducer/ flattener afixed).

Expt 1. (PIPP quality selected and AUTOstkrt stacked) 200 x 50ms (10s total) exposures at gain 570 using the ASI 294MC sized 240 x 320 pixels)

Expt 2. (PIPP quality selected and AUTOstkrt stacked) 100 x 100ms (10s total) exposures at gain 570 using the ASI 294MC sized 240 x 320 pixels)

Expt.3 (PIPP quality selected and AUTOstkrt stacked) 57 x 130ms (7.4s total) exposures at gain 570 using the ASI 294MM sized 480 x 640 pixels)

Expt.4 (PIPP quality selected and AUTOstkrt stacked) 570 x 130ms (74s total) exposures at gain 570 using the ASI 294MM sized 480 x 640 pixels)


Running the PI statistics analysis on the raw linear stacks indicated that none of the pixels were fully saturated in experiments 1) and 2) -- i.e the ADU maximum of 16384 was not exceeded - and similarly none of the pixel scores in 3) and 4) exceeded 4096.

Probably the most compelling evidence though is simply to blow the pictures up and count the number of pixels separating the peaks that correspond to stars? How many pixels separate the closest clearly separate stars?

In expt 1) 50ms was clearly too short an exposure time and there is lots of background noise. Nevertheless it was pretty clear in both 1) and 2) which were both sampled at a resolution of 1.05 arc sec/pixel that the resolution is somewhere around 3 pixels which accords with the FWHM estimate and expectation based on the sampling rate needing to be about 2.5 to 3 ? So the resolution of experiments 1) and 2) in 'lucky imaging' appeared to be about the same as that in single 10s frames - and in stack of 10s frames that comprised the full image of M3.

Similar analysis of experiment 3) and 4) does indicate - I think anyway 3-4 pixels - and thus an improved resolution probably somewhere between 1.5 and 2 arcsec - which again is about in line with expectation based on a resolution of 0.475 arcsec/pixel and sampling of about 3.

Of course you are right that a good way to test it all out is to take double stars -at known separations - tighter than the seeing - and confirm that you can indeed separate them this way. Anyway --all good fun doing this- tight double stars another potential application alongside globular cores.

TimH
Attachments
EXPT4.JPG
EXPT4.JPG (64.1 KiB) Viewed 423 times
EXPT3.JPG
EXPT3.JPG (82.31 KiB) Viewed 423 times
EXPT2.JPG
EXPT2.JPG (54.88 KiB) Viewed 423 times
EXPT1.JPG
EXPT1.JPG (56.04 KiB) Viewed 423 times
User avatar
oopfan
Posts: 1002
Joined: Sat Jul 08, 2017 2:37 pm
Location: New York
Contact:

Re: Resolving M3 to the core with lucky imaging

Post by oopfan »

Tim,

Sorry, but my brain broke when you switched cameras from color to mono. In my opinion, experiments 3 and 4 using the MM show the best star separation. In the MM image, I see two clearly separated stars, but the MC image shows an elongated blob. Isn't the MC image supposed to be "lucky"?

EDIT: I do understand that the resolution of a color cam is not as good as mono due to the Bayer matrix. And I understand what you are doing by re-sampling, but I become concerned when asked to count the number of pixels between the ends of an elongated blob. I feel like I need more proof that the stars are cleanly separated.

Brian
timh
Posts: 218
Joined: Mon Aug 26, 2019 5:50 pm

Re: Resolving M3 to the core with lucky imaging

Post by timh »

Hi Brian,

Thanks once again for the challenge (that forces me to think more deeply). Yes images 3 and 4 show the best resolution.

Now I may have got something wrong in my logic - and am very happy to be put right ?

I considered the colour camera experiments to be the controls because although the 'lucky' technique used was just the same as for the mono camera (i,e stacking lots of short exposures) the colour camera experiments were inherently limited to no better than 2.5 to 3 arc sec resolution because the image scale at 1.05 arc sec per pixel was too high. i.e. Assuming that an image needs to be sampled at > 2.5 then these experiments should result in a best resolution of 2.5 - 3 arc sec. They appeared to do this and so I considered had therefore worked as predicted. This resolution also happens to be about the same as the resolution that I get normally from stacking 10s frames. So 'lucky' imaging at this resolution scale didn't stand to gain me anything ?

The real experiment was 3) and 4). In this case the image scale was small enough that sampling at > 2.5 X corresponded to less than 1.5 arcsec - and so in this case lucky imaging could provide an opportunity for visible improvement over the 2.5-3 arc sec that I normally see

I think that in reality the improvement was not as good as it might have been because of limitations in the precision of alignment and due to the fact that short high gain exposures tend to be noisy. But it is a step in the right direction and 3) and 4) are better resolved than anything I have achieved before. It would be better probably to align in PI or DSS rather than autostackkert.

PS

You are also right to state that the pixel gap assesment method is a bit dubious and especially in fuzzy images. I lined up 153 of the 130ms frames using star align in PixInsight which did seem to do a better job and present this blow up (attached) which may make counting easier? It is still subjective of course because some of the (putative) stars are relatively faint --- my judgement would be that probably there are at least some star pairings here that are partly resolved (i.e by at least at a half height maximum) that are certainly within 5, most probably within 4 and possibly in a couple of cases within 3 pixels of eachother -- which would correspond with resolution in the range 1.4 - 2.4 arc sec.

Overall though I can't see any other reasonable explanation for improved resolution when doing nothing other than changing the imaging scale down from 1.05 to 0.475 arc sec / pixel other than the lucky imaging must be working? i.e. there is surely no way that my seeing is good enough for it otherwise to have made a difference?


TimH
Attachments
M3Core.JPG
M3Core.JPG (48.55 KiB) Viewed 410 times
Last edited by timh on Mon May 03, 2021 7:28 pm, edited 1 time in total.
User avatar
oopfan
Posts: 1002
Joined: Sat Jul 08, 2017 2:37 pm
Location: New York
Contact:

Re: Resolving M3 to the core with lucky imaging

Post by oopfan »

To everyone joining the conversation,

Here is a graphic that I borrowed from Wikipedia to basically show what Tim's technique is hoping to accomplish. Imagine that the top and bottom image are the same double star separated by "x" arc-seconds. The top image shows how the two stars appear when the camera's exposure is set to a typical value used for astrophotography. They appear merged due to the turbulence of the atmosphere. Next, consider the bottom image. It is the same two stars but now using a short exposure commonly used for lunar and planetary photography, also known as "lucky imaging".

Brian
Attachments
Lucking Imaging for stars.jpg
Lucking Imaging for stars.jpg (10.5 KiB) Viewed 410 times
User avatar
oopfan
Posts: 1002
Joined: Sat Jul 08, 2017 2:37 pm
Location: New York
Contact:

Re: Resolving M3 to the core with lucky imaging

Post by oopfan »

Hi Tim,

I grabbed your photo and annotated it here:
M3Core-two-stars-resolved.jpg
M3Core-two-stars-resolved.jpg (36.3 KiB) Viewed 409 times
Perhaps I do not properly understand atmospheric dynamics, but I would hazard to guess that what we call "seeing" is an average of two extremes. On the one end, we enjoy very brief moments where the atmosphere is rock steady, and on the other it is torn apart by turbulence. My belief is that "seeing" is the average of those two extremes. With lucky imaging we are striving to identify and stack only those moments where the atmosphere is steady. If that is possible, then the only limiting factor is your aperture.

EDIT: For example, my 71mm aperture can almost resolve two stars separated by 1.5 arc-seconds. However, my session suffered from poor seeing of 4 to 5 arc-seconds. It seems entirely possible to me that there will be at least a few frames out of a thousand where I am aperture-limited. If I can identify and stack only those few frames, then that is my goal.

Brian
timh
Posts: 218
Joined: Mon Aug 26, 2019 5:50 pm

Re: Resolving M3 to the core with lucky imaging

Post by timh »

Hi Brian

Yes that question of just how to select the 'good-seeing' images from the stack is a really critical point and one that I am not at all clear about. ..or indeed what percentage of goodies to expect at different exposure times

At the moment I am taking about 5000 or so exposures and then entirely relying on PIPP to work some magic to assess and order them in terms of 'quality'. Whatever that program is doing to make the call is key and I really don't understand that well enough. The algorithm might not be right at all for stars as opposed to surface details and that is something to look at. Visually though the first frame does look better resolved than further down the stack..so it is doing something right. But it might be better to make that quality call some other way? - maybe PIs subframe slector?

I take the 'top' quality 200 or so and align them versus the first (nominally highest quality) frame - some fail to align so that is another point of selection. Finally integrate the stack .

But it is hard to know how many frames to go for and where the quality cut off should be? I assumed for example that when I was taking 130 ms frames - which is rather long- that there would be relatively fewer 'good' frames in that stack than when the exposure was 100ms. But I am just guessing whether that means take 2%, 5 % or 10% for example. I need to look at the frames more closely .

For the 200 mm Newtonian the theoretical Dawes limit is 0.57 arc sec I believe. But to realise that resolution I would also need to have set up at an imaging scale of ~ 0.2-0.25 arcsec/ pixel as well as to have selected only those images taken under near perfect seeing.

No doubt lot to think about when you get into this more deeply :-)

P.S. A link on what the PIPP quality selection algorithms do. https://astronomy.stackexchange.com/que ... tary-image

Tim
Last edited by timh on Mon May 03, 2021 10:23 pm, edited 1 time in total.
User avatar
oopfan
Posts: 1002
Joined: Sat Jul 08, 2017 2:37 pm
Location: New York
Contact:

Re: Resolving M3 to the core with lucky imaging

Post by oopfan »

Hi Tim,

You said:
For the 200 mm Newtonian the theoretical Dawes limit is 0.57 arc sec I believe. But to realise that resolution I would also need to have set up at an imaging scale of ~ 0.2-0.25 arcsec/ pixel as well as to have selected only those images taken under near perfect seeing.
Hmm, I doubt that your cam's pixel scale is 0.25 arcsec at prime focus, so it seems unreasonable to expect to achieve aperture-limited performance with a normal set up. But I am certain that you could resolve that double star in Hercules that is separated by 1.7 arcsec. That would be a worthy objective: to clearly separate it on a night with 3-4 arcsec seeing.

EDIT: I just ran your scope and cam through the CCD Suitability Calculator: Resolution 0.96 arcsec/pixel (using 1000mm FL and 4.64 micron pixels). So, you should be able to resolve two stars separated by 1.7 arcsec. That is, of course, if you can tame the atmosphere. (I wish I could lend you my 290M with 2.9 micron pixels! That would give you a pixel scale of 0.6 arcsec/pixel.) Maybe I'll look to see if I can find a double star separated by 3 arcsec.

Brian
Post Reply