• 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5

Magic Lantern ETTR and LRTimelapse

Offline
#1 yannick_c
Hi,

I open this thread in order to discuss the way we use ETTR with LRTimelapse. I think it could be a good place to share ETTR attemps with the settings used and also discuss if we could have had a better result with differents settings. This could help people to better understand it with exemples using different settings based on different situations and results. We also could find usefull tips.

I'll start with one of my first attemps where I tried to achieve a night to day holy grail timelapse while seeing stars at the begining.

Gear :
Camera : Canon 550D/T2i
Lens : Canon EF-S 10-18 mm f/4,5-5,6 IS STM
Tripod : Manfrotto 055XPROB
Head : Manfrotto 804RC2

Camera and Lens Parameters :
Focal Length : 10mm
Aperture : f/4.5 (I wish I had a f/2.8 lens but that is the wider I could go)
Maximum ISO : 1600 (550D is not good in high ISO, noise starts at 800 and ISO 3200 doesn't help getting more light, it just crop even more the highlights. If I had a 6D I would have put 3200 or 6400)
Image Review : On, 2sec (needed to enable it for ETTR to works)
After starting shooting, I hid the camera display (with the DISP button) to preserve battery (it only shows the previews after each shot). The cameras stops after 7hours of shooting which was enough for me.
I also used the Lens Twist Trick to avoid aperture flicker (don't really necessary I think, cause ETTR will add a lot of flickering and you'll need to unflicker the sequence in LRTimelapse either way)

Parameters in Magic Lantern
Beeps : OFF (wanted to get some sleep without hearing beeps all the time)
Interval : 26 sec (the best compromise I could find between getting the most stars I can at night and not a too horrible effect on the day)
ETTR Parameters :
- Trigger Mode : Always On
- Slowest Shutter : 20 sec (which seems to be in reallity 22.6 seconds. With the ETTR process and everything, I should have put 27sec interval to be more secure)
- Exposure Target : -0.5EV (default value was -1EV, -0.5EV seems to expose a little more to the right, that will let me more time before it starts changing the ISO and shutter speed down and loosing some crucial light info)
- Hightlight Ignore : 0.2% (not sure if it was the best value, it's the more difficult parameter to choose)
- Midtone SNR Limit : 6 EV (default value)
- Shadow SNR Limit : 2 EV (default value)
- Link to DUAL ISO : OFF

Here is the final video after many tweaks to get it usable :

[Video: https://www.youtube.com/watch?v=xf4lc1nIzxk]

My Camera and Lens are not good enough for shooting stars (but it forces me to try to understand how I could use it to maximize the quality and minimize the noise). I've done also a lot of post prod (with Lightroom, LRTimelapse and After Effects) on it to get rid as more as I can of the light pollution and to see the most stars as I can. I cheated a lot on After effect Wink and speed up the video at 2x (but I used all the pictures to add some sort of stacking to increase brightness and decrease noise). That sequence will be the first one of the video I'm working on, and there will be a big title over the sky in the day part so the weird effect on the clouds won't be that annoying.

Here is the difference between the original (untouched pictures out of the camera) and the final result (you can see the great job of the visual deflicker in LRTimelapse) :

[Video: https://www.youtube.com/watch?v=0IIk9IkWsEo]

The original pictures don't seems so overexposed as I think it would be with ETTR (on the day for exemple, they are not that bright), maybe I missed something.

Now let's add more ETTR attemps to the thread and discuss how we can get the best out of it.

Yannick
Offline
#2 Gunther
Could you please post a screenshot of the preview panel (the regular, not visual) so that we can see the blue curve with the out of cam previews?
Have a close look at where ML changed the Camera settings (Exif Values in the table) and compare with the brightness changes. Afaik (we had that topic here a couple of times alread, maybe you'd like to use the forum-search), ML tends to write wrong exif data so using the Holy Grail Wizard might fail or introduce even more flicker. I don't use ML but other users observed this. If this is the case, I'd skip the HG-Wizard and just use the visual deflicker for the sequence.
Subscribe to: LRTimelapse Newsletter, Youtube Channel, Instagram, Facebook.
Offline
#3 c_joerg
Hello Gunther,

do you know, that it is possible, to change Tv and ISO values in 1/96 EV steps and not only in 1/3 EV steps?

In CHDK it works very well, so I would expect that is the same in Magic Latern. Unfortunately you cannot see this in the EXIF data. EXIF Data shows you only in 1/3 EV steps.

So for example you have not only Tv=1/30s and Tv=1/40s, you have also 32 steps between (32 APEX 96 values). So you can change Tv and ISO values very smoothly without any jumps. This is really advantage from CHDK an ML.

Do you think it could also possible with DslrDashboard?


I was also woundering, why I not see steps in the blue curve, when I made time-lapse with CHDK.

Great Joerg
Offline
#4 yannick_c
Here is the luminance curve with the out of camera pictures.

   

And yes, I have wrong EXIF data (it was me in fact with that problem on the other thread, I had to tweak EXIF data with some other value - Makernotes:ExposureTime - with Exiftool to make them right) but I realize later that the visual deflicker will do his great job even with the wrong EXIF. Maybe using that "Makernotes:ExposureTime" which seems to be more precise instead of the EXIF one could help in the Holy Grail workflow on LRTimelapse but not sure this value exists on all cameras.

Here is some exemple where LRTimelapse put the HG keyframes :

   
Offline
#5 Gunther
@Joerg if you have smaller steps then 1/3 then use the visual deflicker. Cameras are normally not able to do smaller adjustments and thus won't write those to the exif data. LRTimelapse expects 1/3, 2/3 or 1 stop steps for holy grail. Everything else: use visual deflicker, skip Holy Grail Wizard in LRT.
Subscribe to: LRTimelapse Newsletter, Youtube Channel, Instagram, Facebook.
Offline
#6 Craigular.B
@yannick_c and @gwegner, for Magic Lantern-created files, where are you seeing the incorrect EXIF data written? I just checked a couple of my files and it seemed like the Shutter Speed field was correct.

I had no issues with Magic Lantern ETTR created files and using the HG workflow.
Offline
#7 yannick_c
That problem doesn't seem to happen to everyone, it must be due to camera model, or maybe some other hidden setting. In fact, Canon DSLR camera can take either 1/3 or 1/2 stop interval pictures but should use the value set in Camera (in the custom settings, 1/3 for me, default value and the one expected by LRTimelapse). In my case, ETTR uses the 1/3 and 1/2 interval but writes only 1/3 format interval in the EXIF data even the picture what shot with a 1/2 interval.

If you look on the HG keyframes in my second screenshot (more visible on the right part), there are changes not repercuted in keyframes (theses are the 1/2 interval)

If you don't have lumincance bumps while the same shutter speed value is written, your EXIF data should be right.

In my case, I used pyExifToolGUI to look at the values where I had these bumps. In these cases (for the two pictures before and after the bump), the ExposureTime in EXIF was the same but the ExposureTime in the Makernotes (the good one) was different .

But I think, it's not that a big deal. LRTimelapse does the job well even with wrong data after deflickering.
Offline
#8 yannick_c
Here is another test with ETTR while shooting with the sun in front of the camera behind big clouds. Could have done it without ETTR but didn't know at the start how long I will be shooting. Also, with ETTR, I was sure my highlight won't be blown out (wich in final gives me a lot of underexposed shadows and darks)

[Video: https://www.youtube.com/watch?v=XK-ujA_2LzY]

Gear :
Camera : Canon 550D/T2i
Lens : Canon EF-S 10-18 mm f/4,5-5,6 IS STM
Tripod : Manfrotto 055XPROB
Head : Manfrotto 804RC2

Camera and Lens Parameters :
Focal Length : 10mm
Aperture : f/4.5 (f/8 or f/11 would have been better for image quality and sharpness, but I chose the wide open aperture in case I stayed longer with the luminace going down and I should need that wide aperture)
Maximum ISO : 1600 (the ISO stayed at 100)
Image Review : On, 2sec

I didn't used the Lens Twist Trick to test if LRTimelapse will get rid of the aperture flicker easily (to be honest, I forgot to do it lol)

Parameters in Magic Lantern
Interval : 7sec

ETTR Parameters :
- Trigger Mode : Always On
- Slowest Shutter : 2 sec
- Exposure Target : -0.5EV
- Hightlight Ignore : 0.2% (you can look at differents values tested later in the post)
- Midtone SNR Limit : 6 EV (default value)
- Shadow SNR Limit : 2 EV (default value)
- Link to DUAL ISO : OFF

Here is the luminance curve in LRTimelapse :

   

Here are the different tests with the Highlights Ignore Value (you can see the blow out highliths in red)

0.2%
   

1.5%
   

3%
   

6%
   

As a conclusion to the post, I'll say that you have to make a compromise with the highlights you want to preserve and the noise you will add on the postprocess part (the most part will come from getting the exposure right in the shadows and darks). I think that all of this just apply to that kind of shot with sun in front of the camera and behind clouds. But we'll see that with some other tests. The "Shadow SNR Limit" may also have an impact on all of this but don't really know how to use it yet.

Feel free to share what you think about it.
Offline
#9 Gunther
Pretty good result! But in the end I think for that sequence, you could have achieved the same result by just setting the camera to M and shooting with constant settings taking care not to over expose the highlights.
Subscribe to: LRTimelapse Newsletter, Youtube Channel, Instagram, Facebook.
Offline
#10 yannick_c
You're completly right Gunther. Could have done it in M mode. ETTR was just for test purpose for a sunset (or maybe holy grail) sequence I'll be shooting on that spot. I didn't even know that I would go there 30 minutes before and for how long I would be shooting, but the rain was coming, and the sunset was too far in time to stay longer.

...also check out: