Starlight Live V3.0 Now Available

[Paul81]

Starlight Live V3.0 is now available to download via the SX website:

OSX: Starlight Live for Mac OSX
Win32: Starlight Live for Windows

From the change log:

- Exposure Tab now stores separate exposure properties for focus and dark / image acquisition modes. User selections are saved upon program exit and restored upon next load.
- Added camera binning. When used with colour cameras, this will automatically switch internal processing to be monochrome. Live stacking can combine multiple exposures of different (binned) resolutions, but will always re-scale newer exposures to the resolution of the first exposure that was stacked (after reset).
- Added support for cameras with Thermo-electric cooling. The cooler set point temperature is set to the specified value upon start of image acquisition. The current sensor temperature is read and the user interface updated every 3 seconds after camera connection.
- Added the ability to modify the stacking mode during a sequence without discarding image data. The stacked image data is transferred to the new stacking mode as the first exposure. Data in the old mode is discarded.
- Added 'sum and mean stack' mode whereby N number of frames are first stacked using the sum operator into a 'super sub-exposure' and 'super sub-exposures' are mean stacked together. In this mode, the display is only updated every N exposures, but with the benefits of both sum and mean stacking.
- Added defective pixel removal during live stacking process. This can be enabled via the 'Defective Pixel Removal' checkbox on the Stacking tab.
- Added support for SX filter wheel. Filters can be selected on the Exposure tab.
- Modified CYGM to RGB colour decoding algorithm to yield better colour reproduction.
- Modified main dialog tab bar to keep tabs top justified.
- Modified star detection algorithm to increase the ability to detect more stars in each image for stacking.
- Fixed bug: Keyboard controls not working for display processing tab (user needs to first click on the tab controls to give them focus, then keys are available).
- Fixed bug: First time install stacking settings are sub-optimal (stacking was disabled by default and max pixel displacement was set to 0). This has now been corrected such that by default stacking is enabled and max displacement is 16.
- Fixed bug: Issue in core function which effects scaling of values in Display Processing tab sliders.

Known Bugs / Defects:

- The defective pixel removal algorithm is work in progress. In some situations it can yield ‘green’ stars and image artefacts around bright edges in deep sky objects. This is being investigated and corrected for a future release.

 

The next major capability update will be the live recording of sessions to a video file suitable for sharing on the web (V4). Any bug fixes or minor updates available beforehand will be in V3.X releases.

Enjoy!

Paul

 

[seo]

Thanks a lot for sharing Paul. I'll test it asap.

[Ain Soph Aur]

Hi Paul. I already have it downloaded and am impatiently awaiting sunset. Thankfully I have a clear forecast here tonight. plus a hour or two of moonless skies.

There isn't much I can test now, but the filter wheel controller is handling my SX USB wheel perfectly. As I am ordering a SX-825 on the 1'st, the TEC control will come in handy very soon.

I had forgotten about the upcoming changes to the CYGM to RGB conversion. If I understand correctly, this should improve the color reproduction of my X2 Colour?

Hopefully some folks across the pond can share some initial observations soon!

Thanks Paul!

Edited March 26, 2016 by Ain Soph Aur
typo

[MrOD]

You Rock! Thanks

PS. Sum&Mean is awesome.

Edited March 26, 2016 by MrOD

[aparker]

Very excited.  Clear skies here tonight so I'll be out there giving it a test drive.  Already bench-tested the filterwheel control.  Really like the fact you can name the filter positions - no more needing to remember that 2=H-alpha, 3=OIII, etc...  I'm working with the AT72 tonight, so probably won't have reason to make use of the binning, but Tuesday night (assuming the forecast holds) I'll be back to the C8 and after the Owl Nebula, so 2x2 binning will be perfect.

Thanks as always for this super piece of software.

One question re: binning.  I noticed that there are 1x1, 2x2, and 4x4 options.  Is there some aspect of chip architecture or SX firmware that makes only powers of 2 possible, or could 3x3 be implemented?

 

[bwana]

Does SLL support any ASCOM devices, i.e.: an ATIK filterwheel with an ASCOM driver?

The reason I ask is because I'm seriously looking at acquiring an Ultrastar CCD and already have a full suite of RGB/Narrowband filters in a 9 position ATIK filterwheel.  And for that matter an ATIK 428EX CCD camera, also with an ASCOM driver.

I really do like the features in SLL; a major decision point in going with a Starlight Express camera.

[aparker]

I am still unclear as to how the sum-and-then-mean feature is going to do something that mean stacking does not.  As long as pixel values are stored as sufficiently precise floating point decimals, then scaling a plain mean stack appropriately should give the same result.

[RobertI]

Thanks for the new release Paul, a fantastic list of new features, I'm looking forward to seeing what people can do with the new features, particularly the improved colour of the X2C and binning with the Ultrastar.

[aparker]

5 hours ago, aparker said:

Very excited.  Clear skies here tonight so I'll be out there giving it a test drive.  Already bench-tested the filterwheel control.  Really like the fact you can name the filter positions - no more needing to remember that 2=H-alpha, 3=OIII, etc...  I'm working with the AT72 tonight, so probably won't have reason to make use of the binning, but Tuesday night (assuming the forecast holds) I'll be back to the C8 and after the Owl Nebula, so 2x2 binning will be perfect.

Thanks as always for this super piece of software.

One question re: binning.  I noticed that there are 1x1, 2x2, and 4x4 options.  Is there some aspect of chip architecture or SX firmware that makes only powers of 2 possible, or could 3x3 be implemented?

 

So I only had about an hour before the combination of high clouds and a rising moon put paid to my evening, but I was very happy with SLL v3.0 for that hour.  The stacking seems qualitatively better (faster).  It's really nice that focus/frame saves its own exposure time.  And the filterwheel control worked perfectly.  I had hoped to capture some longer H-alpha subs of M42, but even the shift since last Saturday had it down in the trees before I could get much done.  Orion may have to wait until next winter now.  I was able to shift and get some H-alpha on the Rosette before the clouds got ridiculous...

This was 10x 30sec, 6nm H-alpha

From the background fuzziness you can see I was already shooting through cloud at this point.  Oh well.  The software worked great.  Hopeful that Tuesday night works out as a good time to try out binning.

[Bill S]

Thanks for the great Easter present.

Bill S

[Hibou]

Good work Paul. I did a brief test of 5s dark frames on a Lodestar-X2 and can clearly see that isolated hot pixels are almost completely removed with "Defective Pixel Removal". The "smooth" high quality display was better for seeing that; eventually you might include an auto display range toggle as in Nebulosity. BTW a saved FITS dark frame without pixel removal was essentially identical to that obtained with Nebulosity.

I wonder why you only allow bad pixel filtering with stacking, and why only stacked frames are exported as FITS ? Non-stacked frames should just be a special case, and might also be filtered and exported as FITS. Binning also works well, but as aparker wrote, there is no obvious reason why 3x3 binning shouldn't be possible (as in Nebulosity). Cooling control is also an important addition, though I myself don't have a cooled SX camera. 

Thanks for all your hard work. Alan.

[Astrojedi]

Thanks Paul for another great release.

Here is a quick capture from last night. Transparency was poor but seeing was not bad. Note the color in M82. Required no tweaks for color balance or saturation.

 

 

 

And another with slightly higher color saturation.

 

Edited March 27, 2016 by Astrojedi

[Paul81]

Thank you all for the kind words, glad the app is working well out in the wild! Thank you too for the work by the beta testing in helping getting the release ready for the wider audience!

 

To answer a few questions that have popped up on this thread:

 

On 25 March 2016 at 15:43, Ain Soph Aur said:

I had forgotten about the upcoming changes to the CYGM to RGB conversion. If I understand correctly, this should improve the color reproduction of my X2 Colour?

Hopefully should do - testing with some daylight images showed significant improvement! 

22 hours ago, aparker said:

Is there some aspect of chip architecture or SX firmware that makes only powers of 2 possible, or could 3x3 be implemented?

Nope, the chip should support it. There should be no real issue with supporting 3x3. Certain parts of the processing in SL use optimised assembler for speed which requires certain data alignment / size constraints, so I need to check these are not violated (or if so make provisions) for 3x3. I don't think this will be a big issue so will look to get a minor V3.1 update with 3x3 binning in it soon. I'll perhaps wait a week or two incase any bugs surface.

22 hours ago, aparker said:

 

I am still unclear as to how the sum-and-then-mean feature is going to do something that mean stacking does not.  As long as pixel values are stored as sufficiently precise floating point decimals, then scaling a plain mean stack appropriately should give the same result.

 

I am curious too! I haven't gone through any formal mathematical proof to look at the pro / cons, so will just have to see what happens in the wild :-). It kinda feels like it will be useful for keeping exposures short (say <15s) but taking some advantage of lesser noise of mean. I think time and experimentation will tell.

7 hours ago, Hibou said:

I wonder why you only allow bad pixel filtering with stacking, and why only stacked frames are exported as FITS ?

The defective pixel filtering is performed as part of the stacking process. You can export non stacked frames too - 'Export Raw Images' on the Image Export tab. These are raw FITS files with no processing applied (apart from de-interlacing for Lodestar), the idea being is I can use them for debugging and development (its basically the raw data from the camera) and also you guys can use them if you want to post process your raw exposures in more AP oriented software (neb, pixinsight etc). I deliberately apply no processing before the data is saved to FITS. You can also save the stacked 16-bit data to FITS before it is processed for display.

 

Looking forward to seeing more images and observing reports as they come in - the images posted on this thread are looking good :-)

 

Paul

 

[Astrith]

Many thanks Paul....Just keeps on getting better !! Alan.

[MrOD]

I've Managed to get some test shots also with my 200mm lens.
IC405 Ha


M97 OIII


 

[MrOD]

And my main test field lately - NGC2247
Ha Mean

Ha Median

Ha Sum+Mean

Edited March 28, 2016 by MrOD

[MrOD]

Is it just me or is there something wrong with my stacks? Brighter starts seems to get some kind of compression artifacts after stacking.
I have also tried to capture it in multi-spectrum SII Ha OIII. I was doing that with sum+mean. I have started with Ha 4*15 for all channels, than switched to SII for Red (bin2 4*10s), than for Ha for Green (bin2 4*10s) and than for OIII for blue (planned to do also bin2 4*10s) and here wen't something wrong - it looked like all data from R and G went into B channel, nothing left in R and G.

[Paul81]

I have been investigating the occurrence of black halos around bright stars after stacking (what I think you are referring to as compression artefacts). My primary school of thought is that this is undershoot in the lanczos interpolation I use when new exposures are transformed to the reference frame (i.e. rotated and translated). Ringing artefacts due to undershoot is a know issue with lanczos. I have been investigating clamping to help with this. Its also an issue with bicubic interpolation too. If I use bilinear interpolation, although the quality is not as good, the artefact is not present.

With regards to the multi-spectrum sum+mean - do you have the raw FITS from that sequence? If so that can help debugging of any issues. Its entirely possible there is a bug somewhere in the code for the new mode.

[MrOD]

Yes I do have fits files.
Archive below contains all data and two stacked fits. I remember that i have made one OIII/Blue with blank filter (this one i have substracted with undo last function) but from what i have checked there are few more so Red#40 Red#41 Green#60 Blue#80 seems to be wrong.
On the total stack for main+channels (NGC2237_100x10s_sum-mean_2016.3.26_21.03.49) You can see the halos You talk about.
Data [27Mb]: https://onedrive.live.com/redir?resid=6EF43B6B1304390B!31657&authkey=!AJDbGhBote6-d-E&ithint=file%2czip

I've had also one weird behavior but i am quite sure it was described way earlier here on SGL just need to find a while to see what You guys have figured out about it. Anyway 10th frame caused this (even if i made undo and than new frame for stack it looked the same):

Edited March 28, 2016 by MrOD

[Paul81]

Perfect - downloaded the data so when time allows will debug what is going on with the sum+mean in multispectral. I believe it works fine in 'ALL' mode - so mono or OSC sum+mean should be OK (it was tested in mono). 

What value do you use for max displacement? Also 'Frames to sum before mean'? I will need those for debugging so I can replicate your session.

Also, if the max displacement value is too high you can get the stacking issue observed in picture in your last post. I tend to use 16-32.

Paul

[MrOD]

I've had:
- Max Pixel Displacement = 2
- Frames to sum before mean = 4 (I am not sure if I changed that for anything else for individual channels)
Is there a possibility to stack fits files in SLL? There is no direct option but maybe like loading fit file there are some magical command line parameters?

[Paul81]

Excellent, thank you!

I'd increase your max pixel displacement, otherwise you could be creating a lot of key frames which will increase noise. The mean displacement in the stacking stats should hopefully be a guide - set it to 2-4x the average value you get from your equipment (might be small given the short focal length so 2 could be fine in this case - will let you know based on your data).

 

I have a test mode and test harness I use for debugging and development which allows me to use the FITS files as a simulated camera.

[lukebl]

Thank you so much for this software, which I just downloaded!

This is a quick capture of M108 with just 30s (no darks or flats) stacked exposures with my ancient SXVF-H9 and 250mm f/4.8 Newt.

It's great to be be able to quickly produce 'live video' images with a relatively old cam, captured briefly between cloud gaps, without faffing around with elaborate processing or long exposures. Plenty of flaws in the image, but there are distant 18th mag. galaxies in the original image.

 

 

Edited March 28, 2016 by lukebl

[roelb]

M101 - Pinwheel Galaxy NGC 5457

Evo 8 @ 0.5 x - Lodestar X2 Mono - Suburban area with veal clouds - Without LP filter - No dark flats
StarLight Live V.3: 31 x 15 s stack sum 2-mean-linear

Because of the veal clouds, I let ran the stacking up to 31.

I was surprised that SLL was able to produce an acceptable image through the veal clouds!

(PS: false indication M102 on image)

 

[Martin Meredith]

I had a chance to try out some of the new features of v3 last night too with my new travel setup, mainly looking at the many open clusters in Puppis. Here's a quick shot of M47, showing better colours than previously (although I had the saturation far too high as I wanted to distinguish the blue-white and orangey stars). This was done with sum-plus-mean stacking (5s subs, 3 of each summed, then mean stacked). I didn't take darks and instead relied on defective pixel removal, which worked well. This was taken at about f6.6 using my older (less sensitive) colour Lodestar which has lots of hot pixels. Stacking was very fast and accurate. On one occasion I also got the psychedelic vortex-like image that MrOD showed an example of. It reminded me of saddle points in maths classes long ago... 

Martin