Dom, who you may know from my Video Gear tutorial posts, recently directed a commercial for a mobile app called "Lomics." Having recently purchased a color grading monitor, I did the color correction for his spot. With his default FCP-X LUTs it looked like something out of a Tim Burton film at first. My sense is more toward naturalism, so that's what I tried to create here.
I saw an American Masters documentary about Mike Nichols today and he said one of his early mentors, Elia Kazan, had a saying, "the purpose of a director is to turn psychology into behavior." That really resonated with me because it also translates to, "the purpose of a cinematographer is to turn psychology into photographs." It's something for me to remember on each shoot and take to heart.
I ran into some difficulty when using S-log3 on the Sony FS-7 because my images always seemed to be under exposed, even though I was using a light meter. Part 1 of this tutorial series was mainly about looking at why lenses don't behave ideally and how to use an 18% gray card to get "correct" exposure. In part 2 I wanted to eliminate the need for a gray card - which I knew would just get lost or destroyed on set - and start to be able to trust my light meter again.
So I read the Sekonic manual and it gave me a way to compensate the exposure reading. Then I went into Video Gear and lit up a gray chart and found the correct compensation offset for each of my lenses. The end result is that I can use my light meter again with more trust - because it's been verified! That's what this video is about.
Most of my lenses are about half a stop darker than ideal. What really surprised me was that the optically excellent 21mm lens is almost a stop darker than ideal! I ran into this problem in post on the car mount video and the results of the color correction was frustratingly as good as I could get it. I won't have that problem anymore.
So here's part 2 of 2 of our log gamma exposure tutorials - shot in S-log3 on the Sony FS-7.
With my recent interest in log gamma curves I've been having to do some self-education on the meaning of 18% gray cards. One of the issues I've run into is that I can't justify the cost of true cinema lenses, so I'm using Zeiss ZF.2 lenses. I like the sharpness of the glass, but the lack of aperture measurement in T-stops makes getting accurate exposure with S-log a real chore. Those lenses also make my life difficult because I can't just use a factory calibrated light meter. See my previous post and tutorial video about why this is.
The first thing I learned is that 18% gray became a reference standard by accident. The history of it goes something like this... Kodak originally produced R-27 reference cards in the 1970's with the instructions that you were supposed to take an exposure reading off the card, then open up the aperture 1/2 stop more. Somehow Kodak's manufacturing organization lost those last instructions for 20-some odd years, so a whole generation grew up believing that the exposure reading was exactly what an R-27 card read in your exposure meter - making everyone's readings 1/2 stop off! Then it seems video cameras came along and adopted a "standard" that didn't exist.
The "official" ANSI reference is really 12.5% gray, which is 1/2 stop brighter than an 18% gray card. One thing I forgot to mention is that 18% gray is a rounded number. An 18% gray card is really 17.7% gray. The difference between the ANSI/Kodak standard and the vendor specific adoption of a "18% gray standard" is where the confusion lies. There are much more notable people on the web who have written longer diatribes on why this mythology is confusing...here, here, and here. I recommend reading all three links to get a more in-depth understanding - after you finish my article!
So as a personal educational experiment I took my Sekonic L-758DR light meter and measured both an incident meter reading of sunlight as well as a reflected reading off an 18% gray card. They matched exactly. So my light meter appears to be calibrated to 18% gray. If it was calibrated to 12.5% gray then the reflected reading would be 1/2 stop more closed than the incident reading.
The good news is that a Kodak R-27 card (18% gray) is still very useful if you use it correctly. An 18% gray card is also perfect for calibrating your light meter to the particular camera system you own, as vendors shift around the meaning of "middle gray" quite significantly. Middle gray can mean anywhere from 32-50 IRE in my recent experience, whether you're using rec.709 or a log gamma. It's camera manufacturer specific. So while the strict mathematics works out to 44 IRE for 18% gray (assuming 0-108.5 IRE range), the camera manufacturer may choose to put 18% gray at 50 IRE, for instance. I see a part 2 and 3 of our camera exposure tutorials already!
So here's the other experiment I did. I took both the Lastolite and R-27 card and measured their actual reflectance using a Gretag Macbeth SpectroEye. This color science instrument accurately measures color and is used in the print and textile industries. First, let's look at the R-27 18% gray card.
I know the instrument is measuring in LAB color space, but just stick with me here. I created a spreadsheet to calculate the reflectance from the L*a*b* numbers. If a* and b* are 0 this would be an absolutely neutral reference since those are the color coordinates. So we can see here that the R-27 card is pretty much perfectly neutral. The L* value of 49 translates to a reflectance of 17.6%, which is almost perfectly the 17.7% reflectance target for an "18% gray" card. So this inexpensive purchase from Amazon is actually quite good quality, despite being painted on cardboard. This is a reference chart I can trust to get any future measurements right.
This 18% gray "card" is nice to have since it folds up for travel. Here we can see that the gray side measures an L* value of 45, which translates to a reflectance of about 14.5%. That's a difference of about 0.3 stop between it and 17.7% gray. So any reflected light meter exposure measurements will be about 0.3 stop too bright - which is nothing really unforgivable when using a log gamma curve such as S-log3. It might even be a good thing considering that some of the new Sony cameras have dark noise that could benefit from a bit of over exposure. If you're really worried about it you can either subtract 0.3 stop from your exposure measurement, or set an offset in your light meter (Sekonic lets you set an offset value). One other minor issue is that the fabric material appears to be just slightly blue from the a* and b* measurements - but this is just slight. Probably nothing significant to fret about. Still, this isn't as dead-on perfect as the R-27 cards.
There's also the option of using the white side of the chart as a reference since camera manufacturers specify both the 18% gray and 90% white targets (at least Sony and Canon do). We can see here that the white side has an L* value around 90.4. This translates to a reflectance of 77%, so while we could use this side for white balancing, it appears to be too dark to use as a 90% reflectance chart.
This is all I have for now, but I'm sure we'll be producing some more camera exposure tutorials in the next few months. I know these items were confusing to me at some point, so there's a need for clear in-depth information.
Last weekend Dominique and I made this short tutorial on how to exposure your camera when using a log gamma picture profile, such as Sony S-log or Canon C-log. After we finished filming I took the Lastolite "18%" gray target home and measured its reflectance versus the gray cards I bought. It turns out that the Lastolite product is actually about 15% gray (a bit darker) and the R-27 gray cards I bought are almost exactly 17.7% (what "18% gray" is really supposed to be). So the cards are the more accurate way to go. The Lastolite target will cause an over exposure of 0.2 stop, which is sort of in the noise when it comes to exposure with log gamma picture profiles.
From a practical point of view you're perfectly OK using either. In fact, S-log usually likes to be a bit over exposed from a camera noise standpoint. You're going to color correct the picture either way. Since S-log3 is so linear all you need to do it make a slight gain adjustment and you're there - easy.
After doing this tutorial I found out a lot more about the history of 18% gray cards. I'll share that in the next post.
I've recently been doing more color correction with Sony's S-log3 and learning as I go. A problem that keeps creeping up is that Dominique's skin tones always look too dark. He has very dark skin and tends to wear typical set work black clothes during our blog videos. I was wondering why in the world does it look so dark when I get the footage into post on my computer(?) I use a rec.709 (800%) LUT on the monitor and it looks perfect there. Is there something wrong with my computer monitor?
As with most technical endeavors, I first set out to make a model of the problem in Excel; but first a brief explanation: Computer monitors use a standard called sRGB. This color space is pretty much what the web uses today. Rec.709 is a television and video monitor standard. Although both standard have the same color primaries, they differ in the encoded gamma curve. ...and that's where my problem lies. When I was viewing the footage for rec.709 it wasn't correct for display on an sRGB display, nor web videos.
...so how far off was my video? That's where Excel comes in. I started by plotting percentage reflectance of an object being recorded versus the output IRE value. So if you had a test target that gradually increased reflectance from 0% (absolute black) to 100% (absolute white) then this is what the luma waveform monitor would show for each standard. The biggest take away here is that sRGB produces higher IRE values for any given reflectance (blue curve). So if you have an 18% gray card in the scene then the standard says it would produce a 44 IRE level with rec.709 and a 50 IRE level with sRGB (assuming data levels 0-255, not broadcast or "legal" levels).
To further describe the relationship of sRGB to rec.709 I plotted the two against one another. The take away here is that sRGB produces higher values, especially in the dark tones - i.e. where Dominique's skin and clothing are. The process to fix this in Adobe Premiere is pretty easy. All we have to do is apply a "luma curve" effect that emulates the blue line in the graph below and we'll have a correction from rec.709 to sRGB - and a much better looking web video.
The question now is how does this look in the real world? I added the luma curve effect to the project below and set it to use approximately the same curve as the blue line in the previous graph. Notice how Dominique's skin now has much more detail and his clothes do too? Much better, I think.
So my learning here is that if you want to output the video for television, use a rec.709 monitor. If you want to output to the web and you shot in rec.709, you might want to use the luma curve above to correct your footage. Next time I know. Don't assume one equals the other.
Dominique and my friend Kevin helped make this tutorial yesterday on how to mount a video camera on a car. Yes...I know the camera is making a shadow on the car hood in the end shot, but this was a tutorial NOT a film. I had a hell of a time doing color grading. The contrast of the outdoor sun with the inner garage bay was a challenge, but the FS7 was up for it. Also, I had an "adventure" when I found out my ancient CS5 version of Premiere wouldn't import the Prores codec files. I spent many hours trying to find a work around using free conversion software, which added time and complexity, as well as lowered the video quality. -- If anyone wants to support the future of these tutorials please consider sending a few bucks my way to pay for Adobe Creative Cloud software!