I have a day to night (sunset) timelapse that I can't seem to process properly with LRTimelapse 4.4. There are these strange jumps in luminosity after processing the key frames.

I attached a screenshot to this post, and the XML files for the keyframes are linked to below:

[Image: https://dl.dropboxusercontent.com/u/1073...6%20PM.png]



https://dl.dropboxusercontent.com/u/1073...3A3906.xmp

https://dl.dropboxusercontent.com/u/1073...3A4080.xmp

https://dl.dropboxusercontent.com/u/1073...3A4254.xmp

https://dl.dropboxusercontent.com/u/1073...3A4428.xmp

First: Please upgrade to 4.5.1. It does not make sense to not work with the latest version.
Second: please upgrade your dng-converter. Then try again. Best do the whole workflow from scratch. Make sure to not delete or add gradients, use only the provided ones. Make sure to use the Sync Script too, see: http://forum.lrtimelapse.com/Thread-how-...ync-script

If the problems persist, please show a screenshot of the regular curves, with "Visual Previews" turned off too.

Updated Adobe DNG and LRTimelapse to their latest versions

The problem still persists, though the jump is slightly less obvious

[Image: https://dl.dropboxusercontent.com/u/1073...3%20AM.png][Image: https://dl.dropboxusercontent.com/u/1073...4%20AM.png]

So, what happened there when shooting where the jump is? Unfortunately you didn't scroll to that position in the table.
But I can see, that you messed up at least twice when adjusting the camera settings. The first time (right after the 2nd blue keyframe) you adjusted more then once subsequently and this happened again in the middle of the sequence. You can see it from the small orange triangles that mark the manual adjustments.
It's advised that you adjust then leave the camera alone for a couple of frames.

Another guessing would be, that you might have skipped some frames at that position, but I cant say, without seeing that position in the table.

But anyway - I'm pretty confident that this is nothing where LRTimelapse is doing anything wrong. LRTimelapse can try to help fixing those errors when shooting - just use the visual deflicker for this. Apply a couple of "refine" steps until the pink curve is smooth.
And next time take more care when shooting.
Another tip: make smaller and more consistent adjustments, 1/3 stops is better then full stops.

The first spike you see is when I accidentally changed the ISO from 100 to 800 when I only meant to change the exposure time from 0.3 sec to 2 sec. Since this is just one frame, I deleted it from the sequence.

The second adjustment you refer to is when I decided 2 secs is still too dark, so I changed it to 2.5 secs. In between the two changes (from 2 to 2.5 sec), there was about 11 frames. Is this enough room between changes? If now, in the future, how many frames should I leave in between changing exposures?

I'd leave at least 3 frames between each adjustment.
And changing ISO from 100 to 800 or Exposure from 0.3 to 2 ist way to much. Use single small steps of 1/3 stops then you will get great results much easier.

This post was last modified: 2016-02-26, 01:56 by xoda.

Ok, the large jumps in exposure may not not ideal, but to minimize the disruptions and the ability to recover shadows, I mostly tried to minimize the number of changes.

Is there anything in post processing I can do now?

I’m not sure if there’s anything I’m doing wrong in terms of workflow and using LRTimelapse. With another time-lapse, there are weird gyrations of luminosity both up and down.

  [Video: https://youtu.be/TzWIj9e0YLo]

Yes, I know this time-lapse’s exposure changes from 1/125 to 1/8 sec. 

But in theory, what’s the difference between a large exposure jump and a smaller one? Why is the video so strange (both ups and downs), and is there anything I can do now to fix this besides visual deflickring? That’s time consuming, especially if one-pass doesn’t work, it takes a lot of time to re-render the visual previews. 



[Image: https://dl.dropboxusercontent.com/u/1073...8%20AM.png]

[Image: https://dl.dropboxusercontent.com/u/1073...4%20AM.png]

[Image: https://dl.dropboxusercontent.com/u/1073...3%20AM.png]

[Image: https://dl.dropboxusercontent.com/u/1073...4%20AM.png]

[Image: https://dl.dropboxusercontent.com/u/1073...2%20AM.png]

I have the impression that the exif data that your camera produces might not be very accurate, since the holy grail wizard still leaves those ups and downs.
Using smaller adjustments will help staying inside the dynamic range of the camera especially if you have scenes like this with very fragile colors where you have to take care not to blow one of the channels. So do 1/3 adjustments next time, this has only advantages. You can automate the ramping with qDslrDashboard to do the ramping automatically without touching the camera: http://lrtimelapse.com/gear/dslrdashboard/

For this sequence, I fear, visual deflicker is your best option. Just do it a couple of passes. It takes some time, but it will deliver the results you expect. Next time make it better when shooting, and you will save time when processing.

Regardless of the camera exif data are accurate in absolute terms,  they are accurate in relative terms. That is, the 1/125 shots are a lot darker than the 1/8 shots, and the 1/8 shots look about the same as each other in terms of brightness.

What I can do to test the exif accuracy of my camera?

This post was last modified: 2016-02-26, 12:16 by Gunther.

How do you know, if they are in relative ways? How do you know if 1/125 in exif is really 1/125 in reality?

Try loading both images in Lightroom (1/8 and 1/125), select both and use "Settings/Match Total Exposure" in the develop module. This will try to bring both image to the same brightness, considering the Exif data. If something with Exif vs. Reality is wrong they won't look the same.