Silent picture mode incorrect exposure

Started by tof, February 02, 2014, 08:35:22 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

tof

Hi,

I'm fairly new to ML and I'm currently experimenting with some of the features on my 5D Mark 3.

I like the idea of silent picture mode for timelapses but while testing the workflow, I encountered a problem when viewing the dng files on my PC. I took a silent picture, which I exposed to the right using the waveform monitors. When I looked at it on my PC, the exposure was lowered to a "correctly" exposed image, which didn't match what I saw in camera.

I also tried the bracketing feature. I saw that it worked correctly in camera, under and over exposing 2 additional shots. However, when bringing these images back to my PC, all 3 were exposed exactly the same.

(I'm using UFRaw to view the DNG images)

I feel like I'm missing something obvious here...

Any help would be appreciated :)

a1ex

1) You have used a YUV exposure meter while saving a RAW image.
2) If you took the image in photo mode, the histograms try to approximate what you are going to get in a CR2, not what you see in LiveView.

tof

Thanks for the reply, I have been this in both photo and video mode, using both the camera histogram and the ML waveform as a reference.

I now also have the problem that the images are extremely noisy, of tiny resolution and with a purple cast, nothing like what I'm seeing in camera.

I researched this issue, watched a few video tutorials on using this feature and read the user guide... no idea what I'm doing wrong. Could it be my sd card? or do I need to load another module besides silent picture for this to work?


a1ex

Yes, your SD card is definitely the cause for exposure and noise issues. You need to use a Lexar f1.4 or faster.

Expose to the right and make sure you follow my suggestion #1.

tof

Sorry I cant find the lexar f1.4 anywhere...would a 30mb/s sandisk ultra 16 gb do the job?

Walter Schulz

A1ex, there are indeed people around believing black magic happening after shutter being pressed. They have no clue how things work in technical terms and therefore got totally lost telling irony from advice.

@tof: Ignore a1ex's first and second sentence and rely on the third.
No offense but you should be aware that ML is not well suited for people without some background knowledge. And progress goes on and on ... making things more complicated at times.

Ciao
Walter

tof

@walter I'm trying to learn by doing. I followed his first advice to no avail, so pardon me for expecting a serious response when I'm genuinely struggling.

I know that the sd card has nothing to do with exposure...however I was referring to my other issue with extremely low resolution and nasty looking dng files, which I thought could be related to a corrupt sd card or something.

I followed the suggestion and exposed to the right, however no suggestions were made regarding my other issue...

a1ex


Audionut

Live view doesn't show the true exposure.

So for instance, the live view image could look really well exposed, however, the YUV data is not a correct representation of the true exposure.
So, when you use YUV based exposure meters (waveform for instance), these are also incorrect.

So, as a1ex said, you should use raw based exposure meters (histogram/zebras), which display a correct representation of the actual exposure, not some digitally modified brightness level.

With raw based exposure meters you have an accurate report on the clipping point of the captured data, and you can use this to ETTR.

Or you could try a new SD card.  Your call!

dmilligan

 
Quote from: tof on February 03, 2014, 10:59:08 AM
I know that the sd card has nothing to do with exposure...however I was referring to my other issue with extremely low resolution and nasty looking dng files, which I thought could be related to a corrupt sd card or something.

If the card was corrupt, you wouldn't get anything at all.

Quote
So for instance, the live view image could look really well exposed, however, the YUV data is not a correct representation of the true exposure
I think the problem is that he doesn't understand why the YUV data is incorrect, or the exposure appears brighter on camera screen and darker on the computer. There is a very good reason, that is not completely obvious to newbies, so I'll take the time to explain it to him:

Silent picture is just a frame grab of the LV data, basically a single still frame of a video feed. You should expect the low resolution, that is what you get for not actually using the shutter. LV doesn't use the whole sensor (until you take a 'normal' photo, which actually stops LV, takes a normal full res photo and then goes back to LV). LV data is only really for display on screen or hdmi monitor so the resolution isn't all that great compared to the native resolution of the sensor. Also, the actual exposure length is limited to 1/fps (theoretically, in reality it's actually even a little shorter -> additional time is spent to readout the sensor, reset the sensels, etc.), which might be shorter than what you're doing with the real shutter. In other words, if you use a shutter speed longer than 1/fps, the camera 'fakes' what the image 'would' look like by digitally boosting the exposure, making the image nosier. You can't take 1/8s exposures 30 times in 1s can you? The workaround is to use fps override, slow down the LV framerate and use the 'low light' mode.

EDIT:
since silent picture + intervalometer and raw_rec + fps override are essentially the same thing anyway, you might consider using raw_rec + fps override, rather than silent picture. It should actually make things a little easier on you.