5D3 + ETTR + Lightroom

Thought I would post up a few shots I did and try and get some feedback from the ETTR experts out there in terms of Canon and Blinking Highlights and the amount of available room to still pull back details.

I began shooting weddings last year and basically adopted the approach of metering and setting the exposure based on the 'white' dress such that the exposure was generally as high as possible (assuming someone had a dark suit on) without the dress Blinking with highlight alerts ON - i.e. exposing to the right (ETTR) without getting into trouble - and I have been generally very happy that I was able to do whatever I needed to in post without getting into trouble - I don't think I will change this but since the camera (5D3) does NOT have a RAW meter, and we do not know exactly how much safety canon has built into the meter, I thought I would run an experiment, post up the pictures, and see if anyone has developed any rules-of-thumb that might dictate how much into the Blinking Zone we can go and still be able to recover the Blinking Highlights

What I did was hang a white wedding dress in a window, place camera on tripod (35mm L), first meter using Evaluative Metering - take picture, then use Spot Metering - take picture (spot was 1/3rd of a stop more light), and then jumped forward 1 full stop more light when I started seeing Blinking (ON THE DRESS). Even with the initial Evaluative Metering there was blinking in the top right corner window - so Blinking showed up on the dress +1 1/3 stops more light on Evaluative and 1 Full Stop above Spot (but still this blinking was really the light shining through the dress).

What I then did in Lightroom was adjust Highlights down as required, then moved to Whites down in addition as required and then Exposure down in addition as required so that I had NO Blinking Highlights - I also, starting when the dress started blinking, increased the Shadows by +97 in every picture to bring life to the bottom of the dress but more importantly to analyze the Black Calculator and Book that was sitting in the Window to see if there was more detail / less noise by using ETTR - I should have put a much bigger black thingy there to replicate a Black Suit but live and learn - on this issue there was some improvement for sure but not as much as I was maybe hoping for - and thus, to jump to the answer, I think you have to be careful when you are using ETTR once you get into the land of the Blinking Highlights as there isn't that much to be gained but there is some for sure.


SOOC with Evaluative Metering (but Highlights reduced -13 to remove Blinking Highlights in top left window - iso 100, f2, 1/200)

IMAGE LINK: https://flic.kr/p/qJWL​C6

IMAGE LINK: https://flic.kr/p/qsxQ​xS

IMAGE LINK: https://flic.kr/p/qsEk​oK

IMAGE LINK: https://flic.kr/p/qsEk​dp

IMAGE LINK: https://flic.kr/p/pNkg​Wp

IMAGE LINK: https://flic.kr/p/qK31​2m

IMAGE LINK: https://flic.kr/p/qsEj​BV

IMAGE LINK: https://flic.kr/p/qK2Z​Go

What do you have your picture style set to?

I used Faithful

For what you're trying to do I would suggest neutral with everything turned to the lowest settings. The histogram is generated from a jpeg created from your picture style, therefore it won't be accurate for what is captured. In other words you have more leeway to expose higher because the bump in contrast and color in the picture style will increase your highlights prematurely.

May I suggest a couple tools that will take the guess work oUt of ETTR? First, the Magic Lantern hack for Canon cameras provides a Raw histogram and Raw highlight clipping warnings by color channel. I find it invaluable. Second, the applications Raw Digger and Fast Raw Viewer, from the same authors, display a Raw histogram that will enable you to see precisely at what point each channel clips and to know how far above the point where the camera starts blinking that is. A Raw histogram is always precise and accurate, but the camera blinkies are inconsistent because not only are they based on a jpg, they are based on a white balanced jpg and white balancing will push the jpg into clipping at different points depending on the color of the light.

I was going to give you a lengthy spiel about how I do it with my 1Ds3...
until I read Eli's reply...his is better!

tzalman wrote in post #17374462
May I suggest a couple tools that will take the guess work oUt of ETTR? First, the Magic Lantern hack for Canon cameras provides a Raw histogram and Raw highlight clipping warnings by color channel. I find it invaluable. Second, the applications Raw Digger and Fast Raw Viewer, from the same authors, display a Raw histogram that will enable you to see precisely at what point each channel clips and to know how far above the point where the camera starts blinking that is. A Raw histogram is always precise and accurate, but the camera blinkies are inconsistent because not only are they based on a jpg, they are based on a white balanced jpg and white balancing will push the jpg into clipping at different points depending on the color of the light.

Thanks - appreciate the note.

I know about ML and the RAW histogram but have so far chosen not to install it as I also have a C100 that I use for video and I don't really want to deal with ML.

With your experience with a RAW histogram, do you have a best guess, taking an average of WBs, what the average amount of highlights in Stops you can pull back? Is my guess of 1 to 1 1/3 close? I'm a stats person, and given what I would image the Canon engineers built in for safety margin, I would guess it is likely a good guess.

It is hard to say exactly what the difference in headroom between a SOOC jpg and the Raw file is, because the jpg processing can vary so much. Consider the following factors:
1. WB. White balancing is done by pushing up the red and blue channels to first of all bring them up to parity with the green (the sensor is most sensitive to green) and second neutralize the color cast caused by the light. So a sunlight (blue light) WB might be R x 2.0 and B x 1.3 while a tungsten (orange light) WB could be R x 1.4 and B x 2.3. Outdoors its easy to make a jpg with blown reds and indoors blown blues - either of which would set the blinkies going, but because the multipliers are always changing according to the light, it's hard to determine exactly where the clipping point is.
2. Color space. The camera can capture a very wide gamut, but the jpg has to be either Adobe RGB (about 50% of the native gamut) or sRGB (about 33%), so if colors have been captured that are out-of-gamut they will be rendered as clipped. Obviously there is a greater chance of gamut-clipping with sRGB- mostly in the greens.
3. Picture Style. Some are more contrasty or more saturated than others, both in the basic tone curve embedded in the profile they apply and in the additional user-setable parameters, Contrast, Hue and Saturation.
All the above will determine the shape of the jpg histogram and the activation of the blinking warning. Many Raw shooters will try to prevent them from expanding the histogram by setting the camera to Adobe RGB, Neutral P.S. and Contrast and Saturation at the minimum. Some will use a special custom WB setting (Uni-WB) that cancels the effect of WB.

To answer your question, on my 5D2 with it set as above the jpg clips 1.5 stops before the Raw. To be safe, I would consider it to be 1.3 - so your estimate sounds right.

Here is a jpg extracted from a Raw that was ETTRed to put the white flowers on the edge of clipping:

Screen cap from LR:

Just want to say a big thanks for the last 2 posts. In most things I do I generally go completely off the deep end to get into the techical stuff so I truly get what is happening behind the covers but I have been resisting for a while doing this in terms of colour space and everything RAW because you have to choose your battles in life and I figured, being a math stats guy, I could probably guess the answer statistically and as long as I left a little room for error I would be good - but much appreciated.

Prior to my experiment I would push to the right just before blinking but now I'm pretty confident that if I really need it, I can push another 2/3rd of a stop and still have 2/3rds protection.

Your techical response was very much appreciated - thank you.

Paul

Just be aware that LR PV2012 compresses the crap out of your highlights to avoid blinkers showing up. Switch to PV2010 to see the difference. If highlight detail is what you are after, ETTR and pull down exposure in PV2010 versus 2012 and you will see a noticeable difference.

Folks like PV2012 because they feel like they can do all sorts of crazy pushing and ETTR Without the highlight clipping warning showing up. It is possible because it it compressing all that highlight detail into tepid nothingness.

Check your data with Raw Digger, not LR.

Kirk

Although this thread has passed its expiration date, I have pulled it out of the freezer in order to append this article by Iliah Borg, author of Fast Raw Viewer and Raw Digger to illustrate the advantages of using those products. It says what I was trying to convey, but far better and with illustrations (one picture is worth ...). Keyboard shortcuts mentioned in the article refer to FRV.
'Why bother shooting RAW if you are culling JPEGs?'
http://s3.amazonaws.co​m …_culling_RAW_vs​_JPEG.docx

Unsolicited advertisement: FRV is on sale at an introductory price until the end of the month. Highly recommended for all Raw shooters and especially recommended for users of LR/ACR because it can generate an .xmp of both its ranking tags and image adjustments that can be applied in those converters.

Thanks very much - I very much appreciate the follow up.

Subscribed to follow up on the utilities mentioned.
Thanks
Myron