Author Topic: Exposure Confusion.  (Read 937 times)

lonetraceur

  • New to the forum
  • *
  • Posts: 10
Exposure Confusion.
« on: March 05, 2017, 07:06:06 PM »
Hi all,

I wondered if someone could explain a couple things in regards to exposure readings with Magic Lantern...

My current setup:

- 5D Mark III.
- Magic Lantern Nightly.2017Jan27.5D3113
- Video mode.
- Metering mode: Eval.
- ISO: 160, Shutter: 1/48 (180deg), Expo. Override : On, ExpSim: Movie.
- FPS Override: 23.976, Shutter fine-tuning: +0.84 ms.
- Canon Expo. simulation: On.
- Camera on tripod, unmoving.

So, the thing that is confusing me is that when I half-press the shutter button or press the 'info' button to have the canon live-view display it's metering ELI (Exposure Level Indicator), the indicator is more than -3 Stops on the scale while in video mode.

However, if I go into photo mode and look at the Live-view with the ELI it shows the indicator at +1 Stop.

Also, in the photo mode, if I look through the viewfinder at its ELI, that shows -2/3 stop.

So, I have three different readings for the exact same shot. I want to get the right exposure for RAW video, but I'm not sure what to trust. I also know I can switch on the histogram and waveform monitor, but I want to be sure that I haven't switched something on/off that's affecting these readings...

Anyone have any advice or want to verbally slap me and tell me where I'm messing up?


Teamsleepkid

  • Member
  • ***
  • Posts: 176
Re: Exposure Confusion.
« Reply #1 on: March 06, 2017, 01:00:21 AM »
I use magic lantern exposure aids like expose to the right and histogram.
EOS M

a1ex

  • Administrator
  • Hero Member
  • *****
  • Posts: 10342
  • 5D Mark Free
Re: Exposure Confusion.
« Reply #2 on: March 06, 2017, 01:26:10 AM »
Expo. Override : On

I don't expect this setting to play nice with Canon's auto exposure algorithms in LiveView. Looks like it does affect their reading (but not ML's raw histogram).

With this setting disabled, and a Canon lens mounted, the indicator is consistent in movie mode vs photo mode LiveView. Metering outside LiveView cannot (and should not) be compared with metering in LiveView, as it uses different hardware and algorithms.

With a manual lens mounted, Canon's exposure indicator and exposure simulation in LiveView unfortunately cannot be trusted in M photo mode, on many models, including 5D3 (in particular, but not only, after shooting with a narrow aperture on a Canon lens). Expo override attempts to fix the latter, but since I don't use Canon's exposure indicator (here's why), I didn't even think about testing it.

To check whether your model is affected by this bug:
- use plain Canon firmware, without ML (just to exclude any possibility of ML interfering)
- use LiveView with a Canon lens at f/22 in M mode (no need to take a picture)
- switch to a manual lens without chip (or just remove the lens)
- take a picture - it will be much brighter than your LiveView image.

lonetraceur

  • New to the forum
  • *
  • Posts: 10
Re: Exposure Confusion.
« Reply #3 on: March 07, 2017, 09:44:12 AM »
Thank you for the great reply a1ex.

That makes a lot more sense now. I do indeed use a Samyang 24mm F1.4 manual lens and I can see why the camera would 'remember' the last Canon lens I've put on it and it's relative aperture. I hadn't even thought of that. I will try the test you've suggested too.

I had considered that perhaps since the movie mode was using a 16:9 frame recording size, rather than the photo's full frame, the 'Eval' metering mode for movies was not using the top and bottom parts of the frame, therefore, using less of the picture to evaluate the metering level. This would be particularly different if, for instance, the picture had lighting at the top of bottom of the frame.

However, on reflection, I surmised that the metering would be done on the raw input of the sensor, before any cropping for video. Is this right?

As for the remembered aperture, is there a way to find out what that value is?

I think the best course of action is, as you've suggested, is to ignore the ELI altogether and use the histogram and ETTR. Since it just can't be trusted.