Lodestar Live - September Update

[Paul81]

Hi All!

It's been a while since I have posted any updates to Lodestar Live, work has been manic recently so I am a little behind on development but have still managed to progress things along so I am happy at the end of the day.

So whats new? Well I have refactored the user interface slightly, and added the camera control, image export and the image processing frame work is in place. I now just need to fill in the blanks as it were. To summarise, right now, its pretty much a glorified capture program, the clever parts come next...

First up some screen shots of the Image Acquisition mode (this is where you capture light frames):

The first screen shot shows a capture (I have put a 12mm TV lens on the camera for the purposes of these shots - it was cloudy last night otherwise I would have some real night images) and the display settings tab. Here you can see a real time histogram of your current live image (which in this case is a single sub, when the stacking stuff is done, it will be a histogram of your current stacked image) and tweak the black and white level to define the display dynamic range. Currently its a linear function, but I'll probably add the option for a log function at some point. You can also play with gamma, brightness and contrast to achieve the image you want. As a note, internally all processing is 16-bit, the image is only scaling to 8-bit when it is displayed. If you maximise, or change the program window size, the image is scaled to match, but is scaled to keep the aspect ratio.

This screenshot is where I have played around a little with the display settings. As a note, the display settings tab is common to all modes.

Switching to the exposure control tab, you can choose your exposure time, start and stop the camera exposing (it automatically stops when you switch modes), and reset the current image stack. I have added a preset exposure times drop down, but you can manually enter a time (in seconds).

The image export tab allows you to set-up well... image export. If selected, the program will save your light and dark frames (using automatic file names) to the directory of your choice. In this way you can stack / process or archive them offline. These are raw FITS exports, with no processing (apart from de-interlacing) applied, and are in native 16-bit format. You can also save a snapshot of the current live view (using the current display settings used to scale to 8-bit PNG) - the intention here is this is the image you would upload to SGL!

On the settings tab you can modify program settings. Currently this is only the display scaling mode. The smooth algorithm is proving pretty good, I had the program running on a HDTV and the image looked very smooth despite it only being 752x580 pixels. This will eventually get populated with settings that control the live stacking.

On to the Dark Frame Acquisition Mode. It is intended you would use this mode before capturing light images - it is here the program builds a master dark to subtract from the lights as you capture them. Its pretty much the same as Image Acquisition mode but for darks... 

And for the final screenshot, I have disconnected the camera. You can still use the program (i.e. play with display settings etc) but it will not acquire any new images until you reconnect it. The little icon in the bottom left corner changes to indicate the current camera state. Along the status bar at the bottom is also a count of the number of dark and light frames you have captured, in addition to a progress bar indicating how long is left on the current exposure. The software is nice and robust - if you keep disconnecting the camera it (shouldn't) crash...

So whats next? Here is my list of things to sort (in the rough order I intend on tackling them) :

1. Dark frame stacking - I needed to get the camera control and image export working to get some test data. I am trialling some rolling median / mean estimators and will select one to use to stack the dark frames into a master dark

2. Dark frame subtraction - pretty much a follow on from #1.

3. Focus, framing and alignment mode - this mode helps you focus, align your mount and frame your target.

4. Live Stacking - have prototyped much of this in matlab, just need to finalise some parts of it and implement it in C++.

5. Zoom / pan etc for the live image

6. Optimisations - Some of the algorithms would benefit from being implemented in SIMD assembler for speed.

I also need to look if there is a way to get the camera to latch the current pixel values - currently I read the odd field first but that takes ~100ms so the even field gets another 100ms of exposure thus you see some banding when the image is de-interlaced.

Thanks for reading, and clear skies to all!

[NickK]

The software is nice and robust - if you keep disconnecting the camera it (shouldn't) crash...

the really annoying one is where someone has a 'delicate' connection - the answer here is not development it is duck tape..

Good to see another capture app just a question - I note that you probably have some interlace issues (alternate rows have different brightness if you look at the shed). Multi-camera support?

[Paul81]

the really annoying one is where someone has a 'delicate' connection - the answer here is not development it is duck tape..

Haha - yeah the simplest solution is always the best!

The issue you spotted in the image is not so much the deinterlacing but is to do with the camera readout. Currently, the camera keeps exposing whilst I read the odd field (which takes ~100ms). When I then read the even field, the even field has had an extra 100ms or so of exposure yielding a slightly brighter image. When this then gets line merged you see the problem as noticed. I am hoping there might be a way of getting the Lodestar to latch the pixel values before I read them to cure the issue, if not I need to get clever on the deinterlacing and perhaps neutrally match levels between odd and even then line merge them. The effect is less noticeable for longer exposures, but is something I am looking to fix.

The lower guts of the code allows multiple cameras, and in fact would allow other SX cameras not just the Lodestar, but I have limited it to a single camera at the upper echelons. Multi-camera is interesting - I was thinking what the effect of using a binoviewer, a mono Lodestar and colour Lodestar to give luminance and colour all at once would be..?! Colour support is coming later - probably next year when I am looking to pick up a filter wheel.

[shirva]

Need to stop messing about and get on wae it..lol...

...multiple camera capture ...big tick.....colour. ..big tick.looking great.......Davy

[nytecam]

Hi Paul - is there a test version for download yet? I'm eager to try it with my Lodestar M and C on my 12" SCT :-)

[Paul81]

Hi nytecam! 

Thank you for the interest! I have not posted anything online as of yet, but I was working on the image stacking the other day and thought some more test data would be a good idea - so far I only have some images I have taken using a 12mm TV lens - some longer FL images would be great. So I think it would be a good idea if I got a beta version online so people can try it out. The program allows you to export dark and light frames to FITS files so any users I would be grateful of receiving some test data to aid in the live image stacking - oh and route out bugs in the current stuff! Lol!

Currently it is a bit of a glorified capture program, but it does dark frame subtract and allow you to apply some image processing live. I think most people will be interested when the live stacking is in place (there are some kinks to iron out with the algorithm first).

What platform are you running? I can build for Windows, Mac and Linux (I have been running Mac). It will probably work with your Lodestar C but the images will not be colour as it is monochrome 16-bit only at the moment. I have plans for colour next year...

Since I started this thread I have got the dark frame stacking and subtraction in place, and I have started work on zoom / pan on the display. Below is an image I took last weekend (before dark frames were in place). Its a single sub, with just some level adjustment. The focus isn't perfect (probably due to no IR block) and my LP is bad, but I was happy with a first light of sorts! :-)

[shirva]

Looking good Paul...what do you mean NEXT year for colour...lol...would be a nice Christmas for us av chaps he he.....think nytecam should be chief tester , his images are amazing and could find no one better for giving your cracking software a good test and review. ...get the thumbs up from nytecam your on to a winner in my humble opinion......Davy

[nytecam]

Hi Paul - still running Win XP on laptop for Lodestar but could migrate to Win7 on the 2nd ---so when you're ready :-)

Last night had a play with my f/3.3 FR running at f/1.8 (severe vignetting solved maybe with perfect flat but other optical aberrations:=/) and f/2.8 ( much better optically) awaits a good flat at dusk tonight. FR for latter shrinks SCT from 3000mm fl to 800mm fl for a wider field. Both tests on the Lodestar-C - getting down to mag 17 and fainter in single 20s exp. Targeted M27, M57 and weird gxy NGC 6745 in Lyra :-)

[Paul81]

Will get something sorted for Windows - I will have to use my wife's Windows 7 computer, but I think it should be OK running on XP. I'll give you a PM when I have a build - probably some point next week.

I may have to consider adding flats into the equation, code-wise I don't think its much different to the handling of the darks. Mag 17 is nice for 20 seconds (colour too). Fast F ratio is the name of the game, my scope is F4.8 so hopefully should serve the purpose. I was looking at the ASA reducer corrector which brings it down to something like F3.5, maybe something for the future.

In terms of colour support, my idea is to get the live stacking working and integrated and any other bugs / gripes sorted and then get colour going. I will need some test data from a colour camera but I can de-bayer it no problem. FYI, my lower level SX code can handle any SX camera, but its currently only allowed to use (the first) Lodestar. Eventually I will open it up to all SX cameras - I need to optimise some of the code first in the image processing chain!

My next concern is 'sneeking' the EQ8 past the wife...

Here are a couple more images from playing last night, I am fairly happy with the work flow and the general use and look and feel. These images are a single 15 sec sub, but dark frame subtracted - using 10 dark frames with my rolling median estimation applied (so a pretty close median based master dark). I did very little display stretching. I collected about 10-15 frames on each target, so I am looking forward to seeing what the outcome of stacking is!

[shirva]

Think it looks a winner. .will the software work with other cameras ie the Samsung scb2000 or the pd1. .Davy

[Paul81]

Think it looks a winner. .will the software work with other cameras ie the Samsung scb2000 or the pd1. .Davy

Thanks for the kind words Davy. At the moment it will only work with SX cameras - the control and capture is integrated so as to make things robust and Mac / Linux friendly. However, it may be possible to open the main system up to other cameras in the future. I like the Lodestar as its just one cable to the laptop (plus SX document their USB protocol on their website which is what sealed it).

I have thought a way forward for other cameras is to take the approach of cutting out camera control and monitor a directory on the filesystem for new images. My issue with this approach is you lose the nice integrated camera control that it currently has. One of my aims with the program is to make it easy and quick to use - I don't want to lose precious clear sky time to resolving computer woes or fiddling with lots of different programs.

[shirva]

Know the feeling, I bought a four channel video grabber and the transfer time from camera to screen was shocking...if I press menu it took a bit for it to display on screen...a nice fast transfer with multiple channel capture would be an asset for widefield capture and close up or luminance channel. ..Davy

[nytecam]

Keep up the good work Paul

Here's my 20s on M57 referred to earlier - I've revised the FR speed down to f/2.8 so still pretty fast with an efl of 850m down from native 3000mm so nice wide fov via the Lodestar cam of 26' x 19'.   Did a decent flat at dusk last night to cope with the f/2.8 vignetting and it seems to work ok.  There is a limit with my severe LP to exp duration before image saturation with faster optics!

 Inset is M57 sharpened [doesn't suit whole image!] and SN2013ev in IC 1296 which is now fading to ~m17.5 in pic so excellent penetration for 20s exp!  The SN was never brighter than m16.

Also below my heavily cropped+enlarge 1st shot with same setup of the "Bird's Head" - a tri-gaxy merger NGC 6745 near Vega.  This is 3x20s manual stack via SX s/ware getting down to m18 via extreme image stretch - good fun

[shirva]

More fantastic images. ..brilliant. ..Davy

[Paul81]

Superb images! It is results like these that made me embark on writing the application! It is amazing what faint fuzzies can be picked out even with short exposures!

I was hoping to a have a test build for Windows but I have hit a minor issue. It turns out the user space access to USB devices on Windows is not so nice - you have to install the user space access driver as the driver for the device - this is a pain and hassle, and will probably cause issues if you already have the official SX driver installed. On Mac and Linux, the user space access is seamless, you simply run my application and bang it works (no need to install any drivers).

No true bother, the SX SDK for Windows will work just fine, I just need to create a wrapper class to interface it to the rest of my code and update the build procedure for Windows. Therefore on Windows my application will use the official SX driver.

I am away for a long weekend this week so probably won't get too much chance to work at it until early next week!

One thing I think I will change at some point is remove the brightness and contrast controls. I have not really found them of any use (in my wide field tests anyway). After reading some articles in AN and online I think some other dynamic range scaling modes would be more useful (currently it is linear between the selected black and white levels).