Always Underexposed :( Any help?

[redaber]

Hello everyone,

I have just shot a vacation video on the ''Landscape'' picture profile, now since its RAW i heard it does not matter in what picture style i shoot since it's going to adjusted anyway.
Now here is the problem for me whenever i choose the colorspace ''BMD Film'' Setting my footage always seems to be underexposed by up to 2 points. Is there any way to fix this problem? Is there any picture profile setting that maybe matched the ''BMD Film'' color space so we can get correct exposure.

Thanks !

Here is a grab of what's coming soon!

[Audionut]

If you're shooting raw, then you should probably expose based on raw based exposure feedback.

[redaber]

i will look into that! thank you very much

[Levas]

Also using DaVinci Resolve (and Raw Therapee)
And as far as I know, you never have a right exposure with "BMD Film".
I think it's designed that way, it compresses the highlights and lifts the shadows, so you can adjust everything to your specific taste.

You probably know this already, but you can adjust a lot of raw exposure stuff on the tab with the "camera" icon on the left in DaVinci Resolve.
If you open the video scope window and look at the wave parade or histogram, you can adjust BMD film colorspace to fill the whole histogram.

If you choose rec.709 colorspace it should be more like the landscape profile

[redaber]

Thank you for your comment,

Yes indeed know that but the thing is i am afraid that it will introduce more noise then by just exposing ''right''. because diffrent picture style's give diffrent liveview feedback, some clip faster than t he other.

[dyfid]

Try using the raw exposure features as Audionut suggested.  The raw exposure tools aren't affected by picture style. Then just use a picture style to help with focus if using manual lenses, such as Landscape or maybe like me you swap between h264 & raw, so Landscape for viewing, VisionTech for recording (h264).

I have no problems with ML raw in Resolve, when using the raw exposure tools in ML , the raw interpretation looks just fine with rec709, no underexposure unless I was struggling with light.

Using BMD Film is pointless imho for ML raw same as Highlight Recovery which is unnecessary if exposing 'accurately' using the raw exposure tools in ML , ie: minor clipping in spec highlights.

If memory serves me right ML raw recording was a discovery by the devs whilst creating the raw exposure tools in preference to UniWB.

[Levas]

I use raw histogram.
Nice tool:
You can even choose if you want it linear displayed(as raw data is) or logarithmic scale (as your final output histogram is).
It also shows which channels are clipped. If only the blue channel is clipping, it shows.

[redaber]

I use BMD film to grade with ImpulZ LUT's from Visioncolor and also i think i get it! RAW Zebra's will come up with if am underexposed and those zebra's are based off RAW feedback right :D? thank you guys so much.

i swear i have told my mates that if i am ever going to make some big cash with movies i will give the main devs some good money man

[dyfid]

Going back to your original post and rereading your issue is with BMD color space applied to ML raw, as Levas says BMD on ML raw will give the appearance of underexposure whether you use the raw exposure tools or not.

But with the raw tools, exposure is more accurate from the point of view that we see if clipping is occurring in the raw data which ultimately will affect how well it can be post white balanced, how much DR is captured after WB and demosaic and whether or not as a last resort highlight recovery of some sort has to be used to try and remedy bad exposure control from the point of view of raw data clipping. Blindly applying HR can screw up badly exposed shots even more than help.

But going back to BMD as a starting point for lut f**kery, I understand that's what Visoncolor recommend, but with Resolve 11's camera raw adjustments that happen under the rec709 curve and gamut is there really any need to push your raw data into a logish curve like BMD, personally I don't think so. It's 16bit linear raw data filtered through a rec709 curve, which at that bit depth will provide more than enough levels to work within and you are in under the curve to adjust raw data before going to more typical 1D tools like RGB curves and LGG, even with the rec709 2.2 ish gamma applied without resorting to log, if going end to end in Resolve 11.

But you have the Repulsz luts so your own tests will help you decide, would be interesting to hear your findings.

[redaber]

Dyfid,

Thank you contributing your time and effort into helping me so much, i really appreaciate this (the guys over at EOSHD dont even bother helping you)

Reason why i am using BMD Film is beceause i wanna get rid of the ''video'' look and try to emulate a film stock, Kodak's legendary Vision3 500T 5219 to be exact, and and rec709 simply is to ''baked'' for me. so to my understanding i really only should worry about getting it right IN camera since i can get that information back after apply BMD Film because it simply compresses the highlights and shadows right?

[redaber]

here is a video done with BMD film and a kodak film lut :

i always had to add up to 2 points on exposure in these clips

[dyfid]

Dyfid,

Thank you contributing your time and effort into helping me so much, i really appreaciate this (the guys over at EOSHD dont even bother helping you)

That's kind of you to say but am I really helping.

Reason why i am using BMD Film is beceause i wanna get rid of the ''video'' look

But what do you mean by video look? Is that really your goal?  Applying a film emulation lut, some recorded film grain loop and a logish gamma curve is not going to imho remove the video look if the underlying shots motion isn't right, so much is to do with motion, framing, movement through the frame, transition from one scene to the next, how one scene relates to another, what you're trying to say, the feeling and emotion you're trying to evoke. Otherwise you could end up with a video that looks like it was shot on an mobile phone obfuscated with layers of cr*p to try and make it look filmic.

and try to emulate a film stock, Kodak's legendary Vision3 500T 5219 to be exact,

Sure and that's cool, we do what we like, it's supposed to be fun right. I personally couldn't tell one film stock from another.

and rec709 simply is to ''baked'' for me.

Except it's not in the case of raw, you can adjust under the curve?

so to my understanding i really only should worry about getting it right IN camera since i can get that information back after

Using the ML raw exposure tools to avoid to much raw channel clipping will give you the most out of the camera. Clipping a raw channel too much will be detrimental to what you can get out of the raw data, even if it is raw because ultimately when applying WB to get a neutral starting point, the remaining channels levels will be scaled to the one that is most exposed depending on the light and colour in the scene, possibly dragging up noise in the weaker channels.

BMD Film because it simply compresses the highlights and shadows right?

Imho the purpose of log is at capture time, to compress high performing cameras with wide DR into a economic format that maximises capture in minimal acceptable bit depth and compromised codec at the time of capture. When using linear would demand high bit depth storage and result in excessive file sizes making excessive impact on capture media and therefore cost, management and handling.

And that applying a BMD log curve to 16bit raw linear data when working end to end in the same application such as Resolve 11 appears to be pointless and nothing more than a one shot 'fix all' before the lut f**kery.

[redaber]

Wow thank you so much again,

One more question and than i think im done with asking . i feel like the the overexposure zebra's isnt as accurate, reason is, when i clip 90% i can get that info back with BMD Film .
Oh and btw, its true film stock luts indeed almost look the same

[a1ex]

Zebras are by default JPEG-based (for speed reasons). Make sure you choose RAW zebras in menu.

More info here: www.magiclantern.fm/forum/index.php?topic=12096

[redaber]

Thank you sir! can i also ask you, once overexposed is possible to still get data back if yes, up to how many stops ?

[dmilligan]

If you only over expose a single channel it may be possible to reconstruct highlights from data in the other channels, but if you overexposed all channels, it is impossible by definition => overexposed means clipped to white and unrecoverable.

[Levas]

it's 14-bit raw data. every pixel has a "brightness/saturation" value between 0 and 16383 (14bit).
There is no secret or hidden data stored beyond the value 16383...otherwise it would be 15 bit or more 

[redaber]

Understood! So once overexposed based on raw feedback, you cant get it back:)

[Audionut]

Under certain conditions you can recover detail, as explained by dmilligan.  Depending on the scene, you main gain close to a full stop.

[redaber]

Alright thanks mate sorry for asking so much, guess everyone started out as a newbie also how many stops of DR can the 5D III capture in raw mode? and how much in h264 + audionut i sent you a PM

[Audionut]

http://www.magiclantern.fm/forum/index.php?topic=10111.0

About 11 stops @ ISO 100 in photo mode.  Some movie mode numbers are near the end of that thread.  H.264 is limited to 8bits, so it will be less then 11 EV.

Will respond to your pm later tonight.