{ illuma·blogspot·com }

Saturday, February 28, 2015

••◊ Color Correcting Footage From The Blackmagic Design Pocket Camera

My friend David recently purchased a Blackmagic Design Pocket Camera.  It's a rather minimalist design with a number of significant quirks, but cheap at $420 on the used market.  He essentially just needed something to record himself and fellow actors practicing lines for his acting class.

What we soon discovered is that the BMPC doesn't have a very complete feature set when it comes to exposure metering.  It has zebras and a histogram, but that's it.  No IRE waveform, no exposure meter...none of the usual fare of a professional video camera.  This made for a difficult situation when filming outdoors with my variable ND filter.  The image from the camera is "flat" in either video and especially film mode.  It also lacks saturation, which makes approximating exposure from the back LCD rather difficult.  (Blackmagic - if you read this PLEASE update your firmware to include a viewing LUT!)

In all fairness, I do own a very nice Sekonic exposure meter.  However most variable ND filters are not marked in stops, much less third of stops.  You get generic 1,2,3...markings that are completely without measure.  So when you're outdoors at ISO 800 (recommended sensitivity) then you're forced to use ND to get a reasonable stop on the lens.  I don't know anyone who carries around a set of screw on fixed ND filters.  I have a nice of set ND's that go in a matte box, but the matte box is easily ten times the size of the camera plus lens!

So my best estimate was to get something that looked kind of close on the back LCD and use a GretagMacbeth color checker chart to set the final exposure in post.  I used the histogram on the back LCD to make sure that my exposure wasn't clipping black or white and the distribution was approximately centered.  This is the same way that you would exposure a Red camera when shooting in raw.  Then I made a final tweak to the variable ND filter to bias the exposure to something that looked a little more correct to my eye.  I knew this would get me in the ballpark for adjustment in post. 

One really great thing about this camera is that it includes two very robust codecs, 12-bit Prores HQ and CinemaDNG raw.  Unlike 8-bit highly compressed footage from a DLSR, adjustments to exposure and color in post are very easy and robust on the BMPC.  These same adjustments would likely destroy footage from my 5D mark II.

So we're starting with something like this...

If you look left to right on the chart, the fourth neutral patch (bottom row, just below yellow) is just just less than middle gray.  I know this because I recently tested it with the Canon gammas in my last blog post.  So adjusting the mid tones to have this particular patch just below middle gray should get my exposure approximately corrected.

The first thing I'm going to do is put a four point garbage matte around the neutral patches and look at the luminance waveform in Premiere.


If the picture was properly exposed we would see the fourth patch just below 50 IRE.  In the graph above we see it around 45 IRE, so this tells me the shot was a bit underexposed.  That's pretty easy to fix by bringing up the mid tones with a luma curve.


For this particular test shot we were using the "video" gamma on the BMPC, which requires a bit less correction than the "film" gamma.  To get a more finalized color I used the fast color corrector effect in Premiere.  The first step is to use the eye dropper in the effect and choose the middle gray patch so that we can establish a good white balance.  Note that I chose 5600K white balance on the camera when recording this shot, so it was already pretty close...just not perfect.  This step also removes the green bias from the variable ND filter.  Then I added about 40% more saturation to the image to get the skin tones where they look natural to me.  Depending on your taste, this may be anywhere from 30-60%.  In "film" gamma it was more like 210-230%, depending on the look you're after. 

The great thing is that with such robust codecs it's not a destructive adjustment.  It's like Blackmagic Design, developers of Davinci Resolve color grading software, want you to do color grading.  Hmmmm...  The downside of this is that you MUST do color correction on all your footage to get it to look decent.  There is no easy-to-use baked-in "look" on this camera.  David is an engineer by training, so he probably likes that characteristic of his toy.

One final step I do is take a look at the highlights and shadows and make small adjustments to them using the luma curve.  This is just for personal taste and not a step that's required.  Often these are very, very minor adjustments.  In this case I want to back off the highlights just a bit to give the pictures a more "filmic" highlight roll off.  The only place you'll probably see the difference here is in the sky reflection in the window.  The shadows look normal to me so there's no need to correct them.


Before I conclude this post I just want to point out that what I was trying to do here was color correct the footage, NOT color grade.  That's a whole advanced topic for which I'm not an expert.  I think of color correction as more of a technical process and color grading as an intentional artistic process that requires experience and technical know-how.

If you have any questions feel free to leave a comment.

Sunday, February 22, 2015

••◊ C-log and S-log: Why On 8-bit Cameras?

A lot of the people I congregate with in my small film maker world would easily be described as "low end."  They can afford 8-bit beginner or mid-range cameras and dream of shoots with the Alexa.  One issue I constantly run into is that the camera manufacturers put log gamma curves into their cameras - on 8-bit cameras - and low end folks I work with think this makes their camera just as good as the high end cameras.  Reality is a little different and I'll explain with some simple math.

My two most recent examples are C-log on the Canon C100 and S-log on the Sony A7S.  Let's start with Canon.

In the best case C-log has 5.3 stops above middle gray and 6.7 stops below middle gray, as shown on the picture above.  I stole the picture above from Canon's white paper site on C-log.  From my previous article I measured middle gray on C-log somewhere close to 32IRE.

Now this is going to involve some basic math - and I know videographers generally don't like math.  Just stay with me.  An 8-bit camera produces 256 levels of gray, measured as 0-255.  That same 8-bit level can be measured as somewhere between 0-109IRE.  Doing the algebra here, 32 IRE equals 76 8-bit counts.  That means that the 6.7 stops below middle gray in C-log is represented by 76 counts (i.e. 0-75).  The 5.3 stops above middle gray is represented by the remaining 180 counts.

Doesn't that seem backward that the majority of the dynamic range is represented by 76 counts, while the minority of the dynamic range is represented by 180 counts?  This only gets worse when you decrease the ISO.  For instance, looking at the picture above you can see that at ISO 320 or ISO 400 it's nearly 2/3rd's of the dynamic range that is represented by 76 counts, whereas the remaining third is represented by 180 8-bit counts!

The same is true of S-log on the Sony 8-bit cameras, such as the A7S.  In their case they have Slog-2, which places middle gray at 32 IRE again, with 6 stops above middle gray and 8 stops below middle gray.  So the shadows get 76 counts to represent 8 stops - that's almost the entire dynamic range of a Canon DSLR!

The thing is that these log curves were really designed to work on 10-bit cameras.  In that case you have 300 counts (i.e. greater than an 8-bit range) below 32 IRE.  Stretching the shadow range in post doesn't seem like such a big concern when you start with 10-bits (0-1023).  It's the camera manufacturers' marketing folks that listened to all the low end folks and probably thought they would sell more cameras if they had a box spec that the camera supported a high end feature, like log gamma curves.

The other two issues I usually warn people about is that you need a monitor that supports a LUT in order to see what the final image might look like after color grading.  That's usually not cheap ($2k+).  Plus you need someone that knows how to properly expose (experienced DP) and color correct (experienced DIT) footage shot with a log gamma curve.  If you're eye glaze over while reading about IRE levels, you shouldn't use log gamma curves!  Don't understand vectorscopes or luminance waveforms?...DO NOT use log gamma curves!...Don't use a light meter or know how to measure contrast while lighting - DO NOT USE LOG GAMMA CURVES.  There's a reason log gamma curves are for higher end professionals. 

That said, it's nice to believe for a little while that a low-end 8-bit camera has the same features as a higher end professional camera.  It's almost like those kids who put fancy wheels and a loud muffler on their Honda Civic car and think it's now a sports car.

I color graded some footage off the Blackmagic Pocket Camera yesterday.  Let me tell you; 12-bits makes a huge difference!  What we were doing with that footage would have destroyed an 8-bit capture - and it looked great in the end.  More on that later.

Sunday, February 15, 2015

••◊ Exploring Canon WDR and C-log gamma curves

I was helping Dominique, aka "El Gun Legro" record his Kickstarter video using the Canon C100 with the WDR gamma and looking at the monitor, saying to myself "I sure hope this is right."  The image looked a bit dark, de-saturated, and lacked contrast.  So this weekend I went into the shop and borrowed the use of the camera for a few minutes to gather some test data to see if I was actually using the correct camera settings, or this was another case of a stupid ignorant DP believing he was being "smart."

In the first test I recorded a few seconds of footage using the Gretag Macbeth chart.  This told me about color rending, as well as contrast.  In the screen shots below you can see that the standard setting is pretty mild mannered by default.  The EOS standard gamma is extremely contrast-y and saturated (yuck).  It reminds me of the bad old days of fighting a DSLR to work for filmmaking.  The WDR custom setting shows that the gamma reduces contrast quite a bit (just look at the white chip) and de-saturates the colors.  C-log does the same, but to a much more severe degree.  In each case I chose the corresponding color matrix for each gamma so I wasn't mixing the color setting with the wrong gamma setting.

C100 standard settings

EOS Std. gamma custom setting

WDR gamma custom setting

C-log gamma custom setting

Now if we just look at the neutral patches at the bottom of the chart this tells us about the contrast of each setting.  The third patch from the bottom right is just a little darker than middle gray by design of the chart.  So it should show up somewhere less than 55 IRE on the graph.  In the case of the standard gamma and the WDR gamma it does just that, however the highlights are more protected with the WDR gamma.  The EOS Std. gamma raises the almost-middle gray patch to nearly 60 IRE, artificially brightening the image, as well as adding eye popping contrast.  C-log does the opposite and darkens the patch to less than 30 IRE, as well as reduces the contrast.  With an 8-bit camera I would be concerned about the loss of shadow detail given that there aren't a whole lot of 8-bit counts below 30 IRE.  Bringing back shadow detail might be a bit difficult in certain situations.  C-log seems make more sense for a 10-bit camera like the C500; at least to me.  It's the same thing with Sony's S-log - it starts to make sense when you have a 10-bit camera.

Neutral patches with standard settings

Neutral patches with EOS Std gamma

Neutral patches with WDR gamma

Neutral patches with C-log gamma

I took the camera out to the parking lot and used my light meter to determine proper exposure for the sun so I wouldn't bias this real world test.  I also double verified with the C100's exposure meter - which was dead-on when I adjusted the iris.  As before, the standard settings are fairly good and generic by default.  The EOS Std. setting is horribly contrast-y and already over-saturated.  Now if we compare the standard settings to the WDR settings you can see that the highlights are a bit more protected (just look at the sky).  The WDR settings also seem to be lacking a bit of contrast and saturation, which is generally a good thing for protecting any clipped highlights - *IF* you plan on color grading the footage.  If not, then the standard settings seem decent by default.  The C-log image is very dark and needs quite a bit of correction to bring it back to life.

 standard settings
EOS Std gamma

WDR gamma

C-log gamma

Here are two quick color grades I did with the WDR and C-log images.  I found that typically the WDR footage needed a boost in saturation of 20-30%.  The C-log footage typically required about a 70-80% boost in saturation, which can be a delicate operation on 8-bit highly compressed AVCHD footage.  It also required a large adjustment with a luma curve to bring the exposure and highlights back to where they should be.  If the production allows for color correction, I would personally still stick with the WDR gamma on the C100 because less post production adjustment is required.  In my opinion this holds with the Ninja Blade recorder as well.  Prores HQ is better than AVCHD, but the camera only outputs 8-bits either way.

WDR gamma, luma adjustment and saturation boost

C-log gamma, luma adjustment and saturation boost

Dominique gave me a small clip of the Kickstarter video.  After adjusting the contrast and increasing the saturation about 25% the footage went from what you see above to what you see below.  So now I can sleep a bit better knowing that everything was OK with the footage.  I just wish I had a monitor that allowed for a programmable viewing LUT so I could know that my final footage would look OK while we're recording the video.

Sunday, February 1, 2015

••◊ Perhaps the Most Expensive Kickstarter Video Ever Produced

I spent most of yesterday helping Dominique, a.k.a "El Gun Legro" make a fundraising video for Kickstarter.  In my previous blog entry was a description of why we didn't use the A7S/Shogun combo.  In these pictures you can see we replaced that camera with a C100/Ninja combo plus a whole lot more.  If it wasn't for the fact that we both work with, or in his case for a rental company, this could have possibly been the most expensive Kickstarter video ever produced.  Luckily, it just cost him dinner at a local Chinese restaurant afterwards.

We used both the Wasp and Bee plasma lights from Hive.  The Kino-flo gaffer kit with daylight tubes was used to light up the green screen.  We had a teleprompter running so El Gun Legro could look into the camera and do his thing.  There's also a P180 LED light from Fiilex providing the shoulder and hair light.  Lenses used were mostly Zeiss CP.2's with my 50 f/2 macro for one extreme close up. 

As for diffusion panels we had a 4x4 of 216 and another roll of half soft diffusion from Lee.

We doubled up on the audio with my large diaphragm condenser mic from Pearlman as well as a Sennheiser lav well hidden under his shirt collar.  The problem is that the shop isn't designed to be a production studio so it echoes quite badly.  I figured we could use one or the other mic in post since there wasn't a good way to A/B the audio options while in the echoing room.  Maybe he'll mix the two?

Another problem we had to solve was the very loud laptop.  I put a sound blanket right in front of the laptop to deaden the fan noise in the direction of the mics.  It seemed to work from what I heard on the headphones.

El Gun Legro has one more scene to shoot, which he can probably do without me.  The target date for him to start begging for money on Kickstarter is the third week of February.  It should be fun video if he pulls off what we discussed.  Good luck El.