Exposure adjustment in auto ramping

Anything else that doesn't fit in the other more specific Forums
Post Reply
Posts: 22
Joined: August 27th, 2020, 8:20 am

October 25th, 2020, 2:06 am

Hi there,

If I want the exposure to be 1 stop lower during the entire timelapse sequence with auto ramping. How can I achieve that?

Can I just set the initial exposure 1 stop lower? Will unleased consider the initial settings and make all the following shots 1 stop lower?

What is the Auto Ramping Adjust percentage exactly mean? like -96%, is that around 1 stop lower? If that's the case, we cannot adjust to -2 stops?

Posts: 487
Joined: October 9th, 2018, 4:17 pm

October 26th, 2020, 5:57 pm

Autoramp adjustment is a little tricky, especially with the LRT Algorithms.

Your first question is easy to answer:
for the LRT Algorithms: Yes - just set the exposure for the first image as you want it, and this will be the reference for the entire sequence
for the EV Smoothing Algorithm: use either the cameras EV compensation, or the Autoramp Adjustment value to underexpose the image. The Algorithm will always try to adjust settings so that the the EV-meter is at ±0.0 So if the Exposure compensation of the camera is set, while Mode is M, you'll see that the value of the EV meter changes.
If you set an adjustment value, the Autoramp algorithm will try to keep the EV-meter centered around the Adjustment value. ie if you set it to -1.3, all images will be underexposed (per the camera's ev-meter) by 1.3 stops.

With the EV Smoothing Algorithm, this means you can adjust during the Autoramp fairly intuitively by changing the adjustment value.
With the LRT Algorithms, it's not so straight forward. The Algorithm is based around a luminance value, calculated from the histogram. A reference luminance is calculated on the first image, and whenever a threshold of reference+10% is exceeded, we adjust the camera settings by 1/3rd of a stop. Unfortunately, the luminance value has no units, and is not at all relatable to EV values. In other words, 10% adjustment of the reference value can be equivalent to 1/10th of a stop in one scene, or 2 stops in another scene. We can't know, and it will almost certainly change over the course of an Autoramp. This means that the Algorithm only works properly in one direction, because in many scenes, after making an adjustment of -1/3rd of a stop, the calculated luminance jumps from lets say reference+11% (ie just over the 10% threshold) to ref-30%. ie if we let it adjust in both directions, it would simply flipflop and alternatingly adjust by +1/3 then -1/3 stops. This in turn has a negative effect on the adjustment method: the adjustment changes the reference value by + or -10%, but because of the way the algorithm works, it does not have an immediate effect on the next image in the one direction! In the one direction it will have result in "next change will be sooner" and in the other direction "next change will be later than it would have been". Because the adjustment steps are the same as the threshold, "sooner" is likely to mean "after the next image" but it's impossible to say when "later" will be. It's a bit complicated and means it's not all that usable in the one direction.
Just for clarification: on a sunset, it's fairly quickly visible if you do a positive adjustment (eg +10%) resulting in a brighter image, but negative adjustments just mean that the next settings change will take place much later (when the scene has become darker, and therefore also resulting in a darker image). The opposite is true for sunrises.

We tried a lot of ideas of how to decouple making adjustments from the rest of the algorithm, but because of the nature of the luminance value not being in relation to any EV value steps, we only ended up making things worse, so we left it at the intention of the original author of the Algorithm.

I hope this helps, even if it's a little complicated.

PS. We are working on new Algorithms which we hope to release in the future.
Founder & CEO of Foolography, Hardware & Firmware developer.
Posts: 22
Joined: August 27th, 2020, 8:20 am

October 28th, 2020, 12:01 am

I appreciated your detailed explanation! A lot of information to digest.

First It's glad to know the initial image exposure settings will affect the entire sequence for the LRT algorisms. On the other hand, the initial image EV settings wont affect the entire sequence for the EV24h algorism.

For now, I guess I better not touch the percentage adjustment for any of those LRT algorisms, as I'm not confident what the actual adjusted stops it will yield to.

I will stick with below two options when I need an auto ramp adjustment:

1) For LRT algorisms, I will set an initial desired EV with 1 stop lower in camera before I start the auto ramping.

2) For EV 24h algorism, I can make some adjustment by the actual stops (or 1/3, 1/2 stops). In my case, its -1. This is very useful when shooting the gold color cast on top of a maintain during sunrise/sunset, as my photos always got over exposed due to the bright sky in this situation.

By the way, there's no exposure compensation in M mode. But I got your point, that the EV metering will always yield to ±0.0 no matter what the initial exposure settings are when using the EV24h algorism, and this is totally different than those LRT algorisms.

Thank you!
Post Reply