Page 1 of 3 123 LastLast
Results 1 to 10 of 21
  1. Collapse Details
    Am I understanding Exposure Correctly? (LOG, 8-bit, 10-bit, HDR)
    #1
    Senior Member
    Join Date
    Jan 2007
    Location
    Portland, OR
    Posts
    4,525
    Default
    Please correct me where I am wrong, feel free to explain/flesh out where I am under-understanding

    Most screens - including computer monitors and tvs - display 10 stops DR at best, usually closer to 8-9 under ideal conditions and using 10-bit. Many down to around 6 stops displayed with 8-bit and generic lighting conditions. This is a factor of color depth (8-bit and 10-bit), mastered content, and brightness output. Many screens and monitors simply cannot get bright enough.

    HDR screens potentially display up to 17 stops of DR, if they can get bright enough, but in practice they can't yet get bright enough. If an HDR TV has 1000 nits brightness, it can display about 10 stops of dynamic range. (That is 2^10... because every stop (^10) doubles (2) the brightness and hence how many stops of DR can practically be shown).

    At Atomos HDR monitor with 1500 nits of brightness can display 10.5 stops of DR. (2^10.5). A SmallHD Focus with 800 nits could display somewhere around 9.5 stops of DR (but does it? Does the Focus HD support this? It is 10-bit, but will it show this many stops?)

    The Panasonic EVA1 can capture 14-stops of DR. Thus, any viewing display will show crushed blacks or blown highlights when in fact there is more data that is being captured. Even your monitor for editing will show crushed blacks or blown highlights. When color grading and correcting, these can be recovered because in fact, the data is there, you are just bringing it into the viewable area for your screen to now see and thus your eye to access.

    When recording on a screen, this is why you can't trust the LCD. You can't trust an 800-nit device that can only display at best almost 10 stops of DR under ideal conditions to show you a 14-stop image. Thus, you enter the tools and the waveform monitor, 0-100 IRE.

    Except, now things get a bit confusing. In LOG, you only show up to 80 IRE. If you treat 80 IRE as your blown out whites, you seem to be fine. But why? Why 80 and not 100, or 110?

    Essentially, I am looking to understand the pipeline and why I expose the way I do in LOG, grade the way I do on my screen, and see the content the way I do on an SDR and/or HDR display.

    I feel like I'm sort of stumbling through the dark here - like a chef who can fully a recipe, but doesn't really know why they are doing what they are doing...


    Reply With Quote
     

  2. Collapse Details
    #2
    Senior Member Cary Knoop's Avatar
    Join Date
    Jan 2017
    Location
    Newark CA, USA
    Posts
    1,492
    Default
    Quote Originally Posted by filmguy123 View Post
    If an HDR TV has 1000 nits brightness, it can display about 10 stops of dynamic range. (That is 2^10... because every stop (^10) doubles (2) the brightness and hence how many stops of DR can practically be shown).
    The Nits scale does not stop at one. You can have a screen that can show 0 nits (OLED screens for instance).


    1 out of 1 members found this post helpful.
    Reply With Quote
     

  3. Collapse Details
    #3
    Senior Member
    Join Date
    Oct 2008
    Location
    Bristol, UK
    Posts
    9,493
    Default
    WHen you describe a screen as able to show '6 six stops' it is worth maybe considering this.

    A screen can display six stops with a proper contrast ratio.

    A proper contrast ratio is stop two being twice as bright as stop one, stop three being twice as bright at stop two etc.

    The washed out log image might be 12 stops but if displayed with an improper contrast ratio (eg stop two 50% brighter than stop one) you can see all the stops on a siz stop monitor.

    When you apply a LUT (which is a conversion from the improper contrast ratio to the correct contrast ratio) you might then push the log image outside the range of the monitor.


    1 out of 1 members found this post helpful.
    Reply With Quote
     

  4. Collapse Details
    #4
    Senior Member
    Join Date
    Jan 2007
    Location
    Portland, OR
    Posts
    4,525
    Default
    I don't understand how display limitations effect footage monitoring during capture, as well as IRE value mapping?


    Reply With Quote
     

  5. Collapse Details
    #5
    Senior Member
    Join Date
    Oct 2008
    Location
    Bristol, UK
    Posts
    9,493
    Default
    The Panasonic EVA1 can capture 14-stops of DR. Thus, any viewing display will show crushed blacks or blown highlights when in fact there is more data that is being captured.

    not true - because log doesnt have a proper contrast ratio - a typical monitor can show all you capture - just not with a proper contrast ratio


    this is why you can't trust the LCD. You can't trust an 800-nit device that can only display at best almost 10 stops of DR under ideal conditions to show you a 14-stop image. Thus, you enter the tools and the waveform monitor, 0-100 IRE.


    not true - because log doesnt have a proper contrast ratio - a typical monitor can show all you capture - just not with a proper contrast ratio

    Except, now things get a bit confusing. In LOG, you only show up to 80 IRE. If you treat 80 IRE as your blown out whites, you seem to be fine. But why? Why 80 and not 100, or 110?


    Because Log doesnt record a proper contrast ratio (but a LOGarithmic interpreation of it) - it sounds like the panny spreads the 14 stops over 0-80

    Overall your on set monitor should be reasonably trustable and show everything you are recording.

    When you appy LUTs you start potentially pushing data into the 'invisible' regions.


    1 out of 1 members found this post helpful.
    Reply With Quote
     

  6. Collapse Details
    #6
    Senior Member
    Join Date
    Dec 2007
    Posts
    2,552
    Default
    To understand LOG, you need to know that the real work is done in the middle of the curve. It is the often the way that the low to mid tones are compressed which allows for proper distribution after the curve is evened out in post and highlights as well as shadows are repositioned to their proper place. The price for that is increased noise in the low to mids. However, what you are also not taking into account is superwhites. Information beyond 100 IRE. This means there is more information in the highlights than is distributed when normal contrast is reapplied through correction, LUT's, etc.. Essentially superwhites also allow a selection to be made with a mask and guarantee information to be there when it comes to windows, skies, etc. The superwhite maximum value for the EVA1 is 109 IRE I believe. I do not trust HDR monitors because until relatively recently there was no set standard as a conventional benchmark, so who knows what you are getting objectively speaking? And if you are just playing with a slider as on the Atomos monitors, you can subjectively put yourself in a place where your image looks good on set, but actually has less information to play with later because you are not exposing for a maximum of information in the mids. I have seen many folks underexpose for this very reason.

    End of the day, you really need to look at your scopes/RGB parades to be sure that you have the info you need and are not clipping.

    Personally, I don't give a rat's ass about seeing my image corrected on a camera monitor. I hate LUT's. I have NEVER seen one that came even remotely close to our more nuanced results in post. That includes ARRI as well. However, I am uses to looking at LOG and RAW images. Many clients are not.
    Last edited by yoclay; 11-04-2018 at 05:25 AM.


    1 out of 1 members found this post helpful.
    Reply With Quote
     

  7. Collapse Details
    #7
    Senior Member
    Join Date
    Jan 2007
    Location
    Portland, OR
    Posts
    4,525
    Default
    Quote Originally Posted by morgan_moore View Post
    The Panasonic EVA1 can capture 14-stops of DR. Thus, any viewing display will show crushed blacks or blown highlights when in fact there is more data that is being captured.

    not true - because log doesnt have a proper contrast ratio - a typical monitor can show all you capture - just not with a proper contrast ratio


    this is why you can't trust the LCD. You can't trust an 800-nit device that can only display at best almost 10 stops of DR under ideal conditions to show you a 14-stop image. Thus, you enter the tools and the waveform monitor, 0-100 IRE.


    not true - because log doesnt have a proper contrast ratio - a typical monitor can show all you capture - just not with a proper contrast ratio
    But, both true if I was monitoring with a LUT - yeah?


    Quote Originally Posted by yoclay View Post
    It is the often the way that the low to mid tones are compressed which allows for proper distribution after the curve is evened out in post and highlights as well as shadows are repositioned to their proper place. The price for that is increased noise in the low to mids. However, what you are also not taking into account is superwhites. Information beyond 100 IRE. This means there is more information in the highlights than is distributed when normal contrast is reapplied through correction, LUT's, etc..
    So I think I need help understanding this. Let me see if I can explain what I *think* is happening. 14 stops are being recorded into the camera on the SD, or onto the SSD external recorder. In order to "fit" a special format is being recorded (and the exact same thing which is being recorded is being displayed) - namely, a super low contrast "log" signal - which essentially needs a "key" to unlock and remap all the tone values. This key is the LUT, or any proper LUT that knows how to map the tone values it finds on this particular footage. There is no metadata in the footage. What you see is what you get.

    On a monitor, you can see all 14 stops in LOG because it's so low contrast. Put on a LUT that is intended to be viewed on standard SDR monitors, and of course you are limited to DR, making extra DR/Data that is in fact there "invisible". You can shift this visible spectrum in post, of course, but hence why monitoring with a LUT or without using tools misleading - because in that case, what you is what you DONT get.

    The only reason the display and tools show up at 80 IRE is because the camera is "tricking" them. To fit such a large range, its only use 80% ish of the *display's* Dynamic Range, because it's so low contrast. So your monitoring tools all show things a little wonky - hence 7%-80% IRE or so. And the signal clips after that. So your WFM shows accordingly, and your zebras must be set to match (otherwise zebras set to 100% instead of 80% would be essentially doing nothing). But it's all a trick, until the footage gets re-expanded in post (or via an on-monitor LUT, but be careful, it's just to get a feel for the scene - it's not actually showing you the full spectrum of data that you are recording!).

    Now, theoretically you could expose and monitor just with the LUT by eye balling it and using your tools as normal. And if you didn't know you had "extra data", it would all be a joyous surprise in post when you realized that you had extra range on either end of the signal. Woohoo! No harm no foul. It's only trouble if you monitor with a LUT and just sort of assume "my clipped highlights will totally be there in the computer!" only to discover later they are not. Correct? Well, correct with one other caveat I am guessing: the optimal middle grey values and skin tone values on LOG don't align with the conversion to the LUT neccesarily, so if you shoot and expose from the LUT you may be placing skin tone values at a non optimal range.

    Am I kinda-sorta tracking this whole thing correctly?

    Believe it or not I've been shooting LOG for a couple years now and it's always worked out fine, like I said, I just realized I've been following the included instructions without really getting WHY under the hood - and I'm fed up with that!

    Detailed nerdy responses welcome ;)


    Reply With Quote
     

  8. Collapse Details
    #8
    Senior Member
    Join Date
    Dec 2007
    Posts
    2,552
    Default
    You don't need a "key". I commonly grade LOG footage without a LUT. LUT's are just shortcuts. I simply bring the blacks down to between 0-3 IRE, equalize their individual channels to have a neutral black (unless I want to get creative), then I push the whites up to with 97% and work on the tricky part of midtone contrast and then, only then color. A LUT will never be as flexible in it's curve in the middle of the spectrum as doing it by hand, and when you work in a variety of situations with varying contrasts you will quickly find that LUT's can actually create unneeded extra work as you try to correct back from them.

    Let's also be clear here, there is nothing that says all LOG images will display at 80% before clipping. White is displayed as beyond 68% on the original SLOG for the F3 for instance. It is also still entirely possible to clip (go beyond the dynamic range of your camera) while in LOG if you overexpose too much. All log is, is a kind of tweaked curve which compresses the midtones to the point where it pulls back the highlights and shadows of the spectrum within a determined range. This varies from camera to camera.

    You don't need to look so much under the hood. What you do need to know is where middle grey should fall, as well as your white point. On a conventional grey card for the F3 in SLOG, 50% grey should be exposed at 38%, whites will automatically fall at 68% then. These two value change when in SLOG 2 or 3 on other cameras for instance. You need to know the same thing for your camera. The rest is gravy.


    1 out of 1 members found this post helpful.
    Reply With Quote
     

  9. Collapse Details
    #9
    Senior Member
    Join Date
    Oct 2008
    Location
    Bristol, UK
    Posts
    9,493
    Default
    But, both true if I was monitoring with a LUT - yeah?


    In the simplest form yes - true.

    Using a LUT you get proper contrast in exchange for on monitor clipped highs/lows that can be 'magically' recovered in post.

    On set Im toggling between lut on/off all the time so I can see my true clip points and also present a 'pretty' image to the director (and focus puller is IMO easier on a punchy image)

    --

    The curve balls and specifics.

    Whats the deal with 80% ?

    Panasonic log is designed to present maybe 15 stops from 0-100%, but the GH4/5 only does 12 or so stops thats why it clips at 80%.. get a Varicam Pure or some more complex log panny camera and you might well find you can go all the way to 100%
    This means one log curve works for all log panny cameras now and future - good - but you are not completely maximising the data use on the GH4/5 - bad.


    A lut clips all data beyond 6 or so stops ?

    Well actually they dont because a typical LUT is 'rolled' at the toe and knee - not a linear function but an S curve. so given 8 stops displayable you might map stop 9 and 10 into the last stop.. the contrast is rolled off. If you check the arri LUT against a greyscale you will see this Scurve.

    You don't need a "key".

    Indeed you do not. But any manufacturer with half a brain (probably none of them) would be using a camera specific LUT to both de-log the footage and also account for some of its known weaknesses - imagine you built a sensor with a bit of a magenta swing in your mid tones you would put a green swing in the mid-tones in the LUT - bingo.. perfect sensor.

    If you check various camera specific luts against a grey scale you can see that there are swings in hue across the data range - surely put there to bounce out weakness in the sensor performance.

    For this and other reasons I consider it to be good practice to start with a manufacturer issued LUT in post.


    1 out of 1 members found this post helpful.
    Reply With Quote
     

  10. Collapse Details
    #10
    Senior Member
    Join Date
    Jan 2007
    Location
    Portland, OR
    Posts
    4,525
    Default
    Thanks Morgan.

    Why do the GH5 and the EVA1 (with a 2 stop difference) both cap out around 80% IRE? (I think the EVA1 at 85%... but it's only .5 stops less than the Varicam...)


    Reply With Quote
     

Page 1 of 3 123 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •