New issue 673: Bulb Ramping
https://bitbucket.org/hudson/magic-lantern/issue/673/bulb-ramping
frame_michael on Wed, 14 Sep 2011 10:27:43 +0200:
Description:
Hi ML Team,
first off its absolutely great what you guys are doing -
it at least tripples the price my T2I is worth.
I do have a suggestion for a further enhancement.
You do have a built in intervalometer which works great.
But if you are doing timelapse from day to night the problem is
as you know - that the aperture is changing dramatically.
So it would be great to have something like the bulp ramper which changes specific values over a period of time.
See:
http://www.thewhippersnapper.com/LittleBramper/Site/Home.html
which is a great but i know you guys could do an even better version inside the firmware!
thanks - Michael
--
This is an issue notification from bitbucket.org. You are receiving
this either because you are the owner of the issue, or you are
following the issue.
- I've started the camera in bulb mode
- I've enabled bulb ramping (default setting)
- I've set bulb timer to 1 second
- I've set aperture to f22 to get correct exposure at 1 second (this
step is important!)
- I've set intervalometer on, 10s, wait
- I've closed ML menu and it worked from the first try.
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> ml-devel+u...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
- I've started the camera in bulb mode
- I've enabled bulb ramping (default setting)
- I've set bulb timer to 1 second
- I've set aperture to f22 to get correct exposure at 30 second (this
step is important!)
- I've set intervalometer on, 10s, wait
- I've closed ML menu and it takes 3 pic, the first and the last are 30s the second pic is 20s.
after that i get a calibration error. The pic are exposed correctly
Why 30 seconds?
My first 3 exposures are (as shown in exif) 2s, 0.5s, 1s. If I set the
bulb timer to 2, I get 4s, 1s, 2s. If the middle image is properly
exposed, calibration is OK.
Can you catch a screenshot of the fine print? I know it has a redraw
bug and disappears quickly. When I've tested the pre-bulbramping
version, I've set the review time at 8 seconds to be able to read it.
There's a bug which triggers ERR70 in LiveView with bulb mode on 550D
only (as far as I could tell). I'm not sure what causes it, but the
shutter already has a different sound, so it may be related (my
shutter count is over 100.000).
Can you tell me if you get any ERR70 in LiveView, photo mode?
Don't set ISO auto; it will result in all shots having the same
exposure, so it won't be able to calculate. What ML does here is to
see the relationship between 1 EV and the brightness level in
percentage. This varies according to picture style, your preference
for exposure... etc. Set it on fully manual, so that image is exposed
well at 1 second, and let it run.
You can do the calibration at some aperture value and then change it
while the intervalometer is running.
1) Expose your image properly for your selected bulb timer value (1s,
2s etc). This is VERY important. Don't underexpose and don't
overexpose. Use the histograms, the zebras... etc (you have them in
Play mode too). Don't use DOF preview, you won't be able to do any
preview for 1-second exposures or longer.
I assume everyone knows how to adjust ISO, shutter and aperture to get
a correct exposure.
2) Make sure ML can measure properly the brightness. Enable image
review and make sure the image in playback mode is full-screen.
3) When the calibration error appears, there is also some fine print.
Read it and make sure brightness level is not 0 or 100. Also make sure
bracketing works properly (e.g. 1s, 0.5s, 2s). Exif will only display
1s for shorter exposures, but you can listen to the shutter or check
the overall brightness.
If I don't know what results you are getting, I can't help you, sorry.
Post some images or videos showing what you are trying to do and what
happens instead.
Next, the percentile is what to meter for (highlights, shadows,
midtones). If you have a tricky white object which is very easy to
overexpose, measure for highlights. If you don't, probably the best
thing is to measure for midtones. Keep in mind that highlights are
easily overexposed on digital cameras.
Then, the algorithm behind bulb ramping is a very simple one (at each
step do half of the exposure correction). If you want a mathematical
proof of why this works, and why I don't do full corection, just ask.
You will only have to learn a bit of control systems theory, PID
controllers, stability, robustness and such (the algorithm behind bulb
ramping is actually a P controller, the process is also considered a
simple gain K without dynamics, and in the first 3 pictures it does
system identification... it computes K).
There is no fixed recipe for bulb ramping, you can setup it in a lot
of ways (depending on what you need to achieve... the compromise is
basically less flicker vs. better highlights) and you need to
experiment. Default setting should be a good starting point (every
time I start the bulb ramping with the defaults, it works very well),
but you **MUST** expose properly.