Test Run to Compare Canon 14-bit Raw Vs 8-bit .Mov

[Xiga]

Well seeing as it was cloudy all weekend here in N.Ireland, i didn't get a chance to test out my new proposed method for capturing Jupiter, namely, to try and use Magic Lantern to capture 14-bit Raw video clips instead of using the camera's heavily compressed 8-bit .mov files. 

So instead i decided to play about with the Magic Lantern (ML) menus so i know exactly what to use when the time comes. As it turns out, these settings are, 640x480 60fps Movie Crop Mode in the Canon menus, and then in the ML menus to decrease the capture settings to 640x426 @ 48fps. The reason for this is because the SD card slot of the 60D is hobbled to write only 20MB/s, and only with these reduced settings i can get about 2 mins 15 secs (give or take) until the camera starts skipping frames, but this should hopefully be enough for my needs. I am a little concerned though at having only 426 of vertical pixels to play with, but with 480 i can keep Jupiter in frame pretty well so i hope this won't be an issue. 

So i wanted to test 2 things:

1) how to get the raw video into Autostakkert2! without losing any data

2) all other things being equal, how much better does 14-bit look compared to 8-bit on the 60D

So i started by taking a 10 sec 14-bit video (640x426 @48fps) which gave me a 238 MB .RAW file. I then took an 8 sec 8-bit video which gave me a 21 MB .MOV file. 

The next problem was how to convert the .RAW file into a format i could work with. For this part i had to visit the ML forums and find a program that would convert the raw file into multiple .DNG files. As it turns out, the 60D is not particularly well supported for ML so i had a bit of bother finding a program that would produce .DNG files without covering them in hot pixels. But eventually i found one called raw2dng_escho.exe that did the job perfectly. Depending on your particular model, you will need to search around the ML forums for a similar program, but i suspect it will be easier than it was for me! So, once extracted, i had 522 .DNG files.

The next stage was to convert them into .TIFF files. I first tried PIPP, but it gave me files tinted in red. Then i tried Photoshop and it worked perfectly. Just go File->Scripts->Image processor, selext the input & output directories, and save as .TIFF, and let PS work through all the files. (i did a test run here based on a full 2 min 15 sec clip, which was 6,600 frames, and it took my powerful computer with an SSD 80 mins to convert them all). 

It was at this point that i had a Eureka moment. I always thought you had to use an .avi or .ser file for AutoStakkert, but as it turn out, if you have a folder full of .TIFF files, then you can simply drag and drop the folder onto AS and it will use them just fine! So i went ahead and processed the 522 .TIFF files and saved the stacked image. I stacked 100% of them btw, seeing as they were all identical!

I then converted the 21 MB .MOV file into an .AVI using VirtualDub and loaded it into AS, which gave me 505 frames, which i stacked pretty much exactly the same. 

So, how do the 2 compare?

8-bit

14-bit

The image sizes are slightly different, but even with the 14-bit image being slightly bigger it still shows up a heck of a lot more detail! 

Now i know it's a pretty crude experiment, but with there being no seeing to contend with, and a similar number of frames at play, i'm hoping this shows up the maximum expected improvement one can expect in moving from 8-bit to 14-bit imaging. 

Not owning a dedicated planetary camera (and a computer by my scope is a no-no for now) I for one am excited by this and cannot wait to try it out on the real thing! 

Interested to hear your thoughts on this guys. Am i missing something obvious here and just being thick?! lol

Clear Skies!!!

[Xiga]

Woops, just realised I forgot to mention how these images were actually taken!  (it's late here). I simply downloaded a picture of Jupiter from Google and made it my desktop background. I then shot it from a couple of metres away in the dark. I guess I was slightly closer for the raw image. 

[sharkmelley]

That's a very interesting test.  I have done similar tests myself on planetary cameras, by displaying an image of Jupiter on a laptop screen for a target!

There is a clear difference between the two results but I don't understand why one gives a bigger Jupiter than the other.  Was something moved between the 2 runs?  If so is there any possibility that focusing was not so spot on for the 8bit run?

Having said that, the difference you are seeing (between compressed and uncompressed runs) is more or less what I would expect to see.  Under perfect seeing conditions your results should be applicable since Jupiter would appear stable and perfectly sharp. The only question is whether any difference would be noticeable in conditions limited by the seeing.  Unfortunately, that is a very difficult question to answer.

Mark

[Xiga]

The reason for one being slightly bigger than the other must be down to me being a bit closer to the screen in one of the shots. Unfortunately i wasn't being very scientific at this point, i simply wheeled my chair across the room and snapped away while holding the camera, so as well as distance there is also the possibility that focus was slightly different as well. My bad!

I will see if i can do another test run later tonight with the camera in a fixed position and with identical focus, for a definitive result, but i suspect the results will be similar.  

[neil phillips]

Woops, just realised I forgot to mention how these images were actually taken!  (it's late here). I simply downloaded a picture of Jupiter from Google and made it my desktop background. I then shot it from a couple of metres away in the dark. I guess I was slightly closer for the raw image. 

Woops I got all excited for a sec there, thought our techniques were rubbish for a second. Glad to know that might not be the case. I was just thinking look how fine the detail in the GRS is.

Glad you remembered to mention it. other wise I would have been searching for cannons

[Xiga]

Woops I got all excited for a sec there, thought our techniques were rubbish for a second. Glad to know that might not be the case. I was just thinking look how fine the detail in the GRS is.

Glad you remembered to mention it. other wise I would have been searching for cannons

Lol, it dawned on me not long after i posted that i forgot to mention that they are not real images, and that someone might get the wrong idea!

No, unfortunately this is the best i have managed so far:

http://stargazerslounge.com/topic/235469-how-well-can-you-process-my-jupiter-tif/

And for what it's worth, here is the link to the imagei grabbed from Google:

http://solarsystem.nasa.gov/images/Jupiter.jpg

[Xiga]

Okay so i took 2 more comparison shots last night, from a little further back from the screen this time:

8-bit

14-bit

All settings were as per the original post.

This time however, the camera was in a fixed position and never moved between shots. Likewise, focus never shifted either, and this seems tobe backed up by the fact that both image sizes now look identical.

Additioal detail is still obvious so all things considered i think it's fair to say there is a significant improvement to be had here for any Canon DSLR users out there willing to delve into Magic Lantern.

[sharkmelley]

A worthwhile improvement.  Thanks for sharing this interesting experiment.  I really must install Magic Lantern and give it a whirl

Mark