CINE with SL2(s) L-Log FW3.0/2.1 = Zebras gone! Waveform flattened above 90IRE ???

[Slender]

Advertisement (gone after registration)

Dear video enthusiasts with SL2/s camera: Have you noticed how Zebras became unreliable or even totaly absent when using 10bits L-Log video/cine mode? I can't get them to show up anymore not matter how I set them or how overexposed the shot is.

In a similar fashion the waveform monitor now seems to "flat-out" all data above the 90/95IRE bar.... same as for the histogram, nothing shows in the far right part, not matter how high you overexpose your image.

When Gamma is turned off in video settings (no L-Log, no HLG) or using MP4 (8bits) modes, all work as intended: Zebras show up reliably, Histogram and Waveform display highlights properly again up to 105%.
@hansvons@Aaron Daniel@Photoworks@Steven @Leon_B. have you noticed anything of the sort? If so that will make highlights monitoring more tricky if not totally unreliable.... Especially so with L-Log you want to go as far to the right as feasible - and so you can get wonderful imagery out of the SL2....

[Photoworks]

I was trying to create some LUT for Rec.2020 and test different exposure compensation settings in order to maximize DR and noise.

I have created a LUT look that I like and loaded it on camera and Monitor.  I don't find "classic" useful for monitoring, it looks too stylized. Leica "Neutral" Rec.2020 LUT tends to underexpose from the meeter and a little too crushed on the blacks. 

My finding are similar to yours, on the wave form I can't control the exposure to well in C4K L-Log 10bit. Most of the time the wights don't reach the 100% on the camera.
The camera wave form monitor looks the same on my external monitor over HDMI when no LUTS are applied . It seams that the camera waveform monitor only and always shoes the LOG data and does not change with preview LUT activated.

On the external monitor the LUT will update the waveform monitor and peeking. You can preview clipping. that requires to bring a good monitor with you and all shoots. otherwise it is hard to keep consistent exposure monitoring from shot to shot, or have pickup shot from preview shot.

I think it is something that we should ask Leica if that is the intended waveform response.

 

on the other hand I love what comes out of this camera in color and DR. If you operate it like a cinema camera and control exposure and WB if is fantastic how much I am getting out the 10Bit files. I Stopped using the 6k Blackmagic and the sony 8bit for now.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[hansvons]

Hi Slender, I think the underlying issue is somewhat complicated.

• First of all, the Leica's colour science is Rec 2020 that has a much larger gamut than Rec709, the EVF's and screen's colour space.
• Secondly, it depends on how the LUT is applied. In a colour managed grading application, the viewing LUT is applied AFTER the grading process to retain the camera footage's full gamut when transforming the footage to a smaller colour space (here Rec 709).
• Thirdly, the transform from Rec2020 to Rec 709 is done (at least in Resolve and ADSK Flame) in such a fashion that the whites compress and don't cut at 100%. That means 100% IRE can't be reached.

In my experience, full white at 100% in colour managed systems, and when colour spaces are transformed (which must be happening here), can only be reached with white sRGB graphics or Rec709 video applied AFTER the transform. With a de-clicked lens, you can see the smooth compression process in the whites nicely on the Leica's waveform.

If Leica were doing the process BEFORE the transform, the LUT would do the transform, as you were doing it in a non-colour managed system and the whites would cut at 100%.

Ok. Please take all this with some pinches of salt. I know a thing or two on colour science but only from a user perspective. I might be completely wrong. 

Attached, my settings (apologies for the German naming, but the Leica is a German chap).

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[hansvons]

11 hours ago, Photoworks said:

I think it is something that we should ask Leica if that is the intended waveform response.

 

on the other hand I love what comes out of this camera in color and DR. If you operate it like a cinema camera and control exposure and WB if is fantastic how much I am getting out the 10Bit files. I Stopped using the 6k Blackmagic and the sony 8bit for now.

If Leica were to apply the LUT before the transform, or to put it more correctly, use the LUT as a transform to the sceeen's/EVF's Rec709, the camera's full gamut wouldn't be retained, the process of compressing the whites when opening the aperture wouldn't be visible and the user wouldn't get the same idea of the camera's gamut in the whites. 

Again, I could be wrong.

On the latter sentence I can't agree more. The Leica SL2 series cameras are a gem in video. I've done recently some hybrid work on a doc project. I don't look back to a cinema camera (used to own a Red Epic Dragon) for such jobs. The SL2(-S) does it all on a high level. Video and Photos. 

 

 

[Slender]

Thank you both very much for your first answers @Photoworks and @hansvons.

 

The issue, from my perspective, is new in the sense I managed to have a view of highlights rising above the 90/95IRE waveform when monitoring L-Log (WITH OR WITHOUT LUT Preview) when I was Beta Testing FW 2.7.3. And Zebras were a nice tool to warm me about how much I was going to loose/clip in the high key parts of my image when using the camera w/o an Atomos/monitor with FW2.0 and 2.7.3. I can't get them to show up anymore when shooting log, regardless of outputing a LUT preview or not in EVF or rear screen. This was normal for HLG before, but now it does the same thing when using L-Log...
 

It shouldn't change how the files look, but it hampers how reliably you can "expose to the right" before loosing parts of your image to overexposure. I will do some screen shots as well when possible and e-mail Leica about this.

[Slender]

@hansvons a very smart feature of the Leica is that the waveform remains unchanged whether or not you choose to display a preview LUT in the EVF/rear Screen, it always shows you what how the Log image is going to be exposed.

On the Atomos Ninja, your waveform will go mental when using a LUT preview, and you can't rely on it, you have to go back to display the Log image if you want to expose just using the waveform.

What I noticed to my great shock is that ZEBRAS do not function as they used too since FW3.0 when using 10bits Log, and so does the waveform. Is that the same on your SL2s on FW2.1 ?

Zebras/waveform should remain unaffected regardless what your screen is capable of, it should only show you what part of your image is exposed at whichever setting you set them to get triggered at... same for other monitoring tools, they just analyze whatever the signal, regardless what the attached display can actually show you. That's why you can reliably monitor some 4444 ProRes on any old SmallHD and reliably expose with their embedded tools. I need to test those on my SL2 see the difference in how waveforms behave and report back.

[frame-it]

Advertisement (gone after registration)

18 hours ago, hansvons said:

If Leica were to apply the LUT before the transform, or to put it more correctly, use the LUT as a transform to the sceeen's/EVF's Rec709, the camera's full gamut wouldn't be retained, the process of compressing the whites when opening the aperture wouldn't be visible and the user wouldn't get the same idea of the camera's gamut in the whites. 

Again, I could be wrong.

 

 

 

nope, you're correct, the LUT should be applied after the color grade is done as the last step.

[BernardC]

13 hours ago, frame-it said:

nope, you're correct, the LUT should be applied after the color grade is done as the last step.

This is an area where a Leica-supplied ACES IDT (input device transform) would be very welcome.

(skip this paragraph if you already know the terminology) A LUT converts from a camera file directly to a display colour space. That's why Leica has REC-2020 and REC-709 LUTs. An IDT converts from camera files to a much wider intermediate colour space. An ODT (output device transform) then converts that to a narrower display space.

One standard ACES workflow in Resolve is:

camera files ->> camera-specific ACES IDT ->> ACES colour space (ACEScc or ACEScct) ->> ACES ODT ->> Output colour space (Rec709, or Rec 2020, DCI-P3, or whatever your project requires).

You apply your IDT first, for instance in a group pre-clip node (where the group contains all files from the same camera). Your actual grading is in ACEScc/cct, but you never "see" that colour space because you apply an ODT in a timeline node (an output node that applies to everything on your timeline).

The beauty of this approach is that you can mix different cameras in the same project with minimal fuss, and that you are doing your grading in the widest colour space.

The problem with camera LUTs is that they don't preserve all of your original information, unlike an IDT (I am simplifying here), so you shouldn't apply them right away.

 

Short answer: Camera LUTs should be applied last in the grade, but it would be better if we didn't have to grade with LUTs at all. Camera LUTs should be confined to on-set monitoring, and the real transformation should be done with an IDT. Unfortunately, Leica doesn't provide an IDT for L-Log.

[hansvons]

1 hour ago, BernardC said:

This is an area where a Leica-supplied ACES IDT (input device transform) would be very welcome.

I've requested that in the SL2-S beta group. Perhaps someone is listening.

[Slender]

@hansvons@BernardC In the meantime my conclusion is that L-Log will clip at the 95% IRE point... Atomos/Swit/Small HD seems to concord on this... but I am surprised I haven't notices it before, I am pretty sure the WFM would display something different before FW3.0 when I was beta testing it in the form of FW2.7.3 (SL2).

Annoyingly for the Zebras, they now indeed barely show up if at all, at any setting, Annoyingly too, Leica only proposes the 0 - 255 scale (luminances in 10bits) to set them up, and so I am unsure what 95% IRE would correspond to: 245, 250, 253????

 

[hansvons]

If 256 equals to 100%, 95% equals to 243,2. Nothing exiting, of course. But since you've been asking ...

[hansvons]

@Slender I see the 95% max value my Flame's waveform too, when working colour managed. Let's assume that all of that is somewhat ok-ish.

[Slender]

21 hours ago, hansvons said:

@Slender I see the 95% max value my Flame's waveform too, when working colour managed. Let's assume that all of that is somewhat ok-ish.

Hi Hans, thanks again as always for your precious insight. I have done some shades of grey charts tests and I concluded that, if you want to use Zebras to warn you about clipping (about 95% IRE on the waveform as we both observed) while monitoring your L-Log, you have to use the (utterly weird choice) Luma value of 220.

That's because, for some reason I can't fathom, Leica seems to have switched to the 8 bits scale (8-225) for the monitoring tools?????? Whilst shooting using L-Log. Whereas when you shoot w/o Log or HLG, the 10 bits scale (0-255) becomes relevant again... irrespective of shooting 10 bits .Mov or 8 Bits .MP4.

This also means the lowest value you can choose for Zebras (200) only alows for an IRE equivalent of 80%.... Meaning you can't really have two different zebras setting you can use to expose... WHY????. For those who may not know the trick, you can use zebras at two distincts settings (one to avoid highlight clipping, and another one to properly expose 18% grey or whatever IRE you  prefer for skintones). Lumix cameras allows you to show two at the time, and use the more useful IRE (0-105%) scale. The Luma 8bits (8-225) or 10bits (0-255) are only useful for stills in Raw to push your exposure to the right as high as you can before clipping. Not really useful for monitoring Log, especially if Leica doesn't tell us how high L-Log can be exposed before clipping, we are left in the dark and concluded by reverse engeneering that that Value is around 95IRE.

Sorry to summon you @Leon_B., but is it possible to clarify this with Leica? SL2(s) is a wonderful platform for stills and video, but I have a feeling we are a bit left in the dark and we need to do an awful amount of DIY to extract the best out of it (L-Log). I am okay-ish with the waveform not displaying data above 95%IRE, but the Luma scale chosen for clipping in Stills is mostly irrelevant and hard to use in Video.

@Photoworks@bpLeica@BernardC@frame-it@Aaron Daniel if you care about my 2 cents conclusions:

• When shooting L-Log, set your camera (and external monitor/recorder if you have) to warn you around 90-95%IRE for clipping highlights. Onboard the camera, the useful value (in Luma 8bits) is now around 220.
• If you shoot without L-Log, in either 8bits or 10bits (both behave better for a ready-out-of-camera-image now that we can fine tune contrasts/shadow/highlights/IDR) the Waveform scale will look different and display values above 100%, the clipping point is theortically 100IRE.... and The zebra Luma scale is again 10bits (0-255 equiv 0-100%IRE).
• A smart feature Leica kept is that the waveform monitor and the zebras remain unaffected regardless you use a lut preview in the camera rear-screen or EVF. So you can expose with confidence whilst enjoying framing with a Lut.
• For HLG, I don't have much experience with it but you can go ahead and enjoy @Stevenincredible videos I believe are mostly shot wih his SL2s using HLG 😍. It seems the clipping point is 100IRE on the waveform.

 

Edited October 14, 2021 by Slender

[Steven]

7 minutes ago, Slender said:

For HLG, I don't have much experience with it but you can go ahead and enjoy @Stevenincredible videos I believe are mostly shot wih his SL2s using HLG 😍. It seems the clipping point is 100IRE on the waveform.

Too kind, thank you brother. I confirm, all shot on the SL2S in HLG, saturation -1. 

[Photoworks]

On 10/14/2021 at 2:42 PM, Slender said:

When shooting L-Log, set your camera (and external monitor/recorder if you have) to warn you around 90-95%IRE for clipping highlights. Onboard the camera, the useful value (in Luma 8bits) is now around 220.

Thanks for sharing. I did some test with the zebra, looks like clipping is 3 stops over from the in camera meeter.  I have created a LUT for L-LOG Rec.2020 +2 stops to see exposure and color correctly. It is amazing how good quality is in shadow and highlights and the colors.

Just keep an eye for hotspot that come into frame, you only have a stop before clipping.

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Slender]

@PhotoworksThanks! You are using the SL2, correct? Could you share this Lut with us do you think? Which is your prefer setting for the internal light meter? (Spot, MidField, MultiField, Highlight Priority?)

[Photoworks]

I have used Multifield on SL2

This LUT's where created to preview in camera not to edit , but can be used for quick edit. I am still working on it so they are not final.

https://drive.google.com/drive/folders/1O7y4IPzbRBR7G4XL1X-n_IglhrP5ExUL?usp=sharing

 

Edited October 17, 2021 by Photoworks

[Slender]

Much appreciated, I will try! Thank you.

[Speeding]

On 10/14/2021 at 11:50 AM, Steven said:

Too kind, thank you brother. I confirm, all shot on the SL2S in HLG, saturation -1. 

Hey Steven,

How do you handle ISO when shooting HLG?  

I tend to start at 800 as that fits most scenarios I encounter but any thoughts about 400, 1600, 3200 and do you tweak exposure compensation?

Thanks in advance.

[Advertisement]

Hallo,

Look here: Leica L-Mount

[Steven]

26 minutes ago, Speeding said:

Hey Steven,

How do you handle ISO when shooting HLG?  

I tend to start at 800 as that fits most scenarios I encounter but any thoughts about 400, 1600, 3200 and do you tweak exposure compensation?

Thanks in advance.

I set it to auto, and I let it go up to 12500, or more manually if needed.