There is an ongoing and much heated debate on another forum about the practicalities of using the LUT’s or Looks built in to the PMW-F5 and PMW-F55 for setting the correct exposure of your SLog or Raw footage. In response to this I put together a very rough video demonstrating how this actually works.
Before watching the video, do please understand the following notes:
Correct exposure is normally determined by the level at which middle grey is recorded. This is true of both video and film production. Light meters are calibrated using middle grey. Expose with a light meter and you will find middle grey at the levels indicated below.
Different gamma curves may use different middle grey levels depending on the contrast required and the dynamic range of the gamma curve. Generally speaking, the greater the dynamic range, the lower middle grey must be set in order to leave room above middle grey for the extra dynamic range. This means that the relationship between middle grey and white will be different from curve to curve. Don’t always expect white to be some fixed value above middle grey. Some of the Sony looks for example LC709TypeA are very low contrast and while middle grey still sits at around 42% (The ITU standard for Rec-709 is 41.7%), because it is a low contrast, high dynamic range curve white is at a lower level, around 70%. The Hypergamma LUT grey points are given by the “G40″ or G33” number – G40 meaning middle grey at 40%.
When you take Slog or raw in to post production it is expected that the middle grey of the recordings will be at the correct nominal level (see chart below). If it is not, when you apply a post production Slog or raw LUT then the footage may appear incorrectly exposed. If you try to bring Slog or raw into an ACES workflow then ACES expects middle grey to be at the correct values. So it is important that your Slog or raw is exposed correctly if you want it to work as expected in post.
Having said all of the above… If you are using CineEI and lowering or raising the EI gain from the native ISO then your Slog or raw will be exposed brighter or darker than the levels above. But I must assume that this is what you want as you are probably looking to adjust the levels in post to reduce noise or cope with an over exposure issue. You may need to use a correction LUT to bring your Slog levels back to the nominal correct levels prior to adding a post production LUT.
Anyway, here’s the video.
Alister, I am doing a review on the Sony a7S and my contact at Sony directed me to your s-log video you had done for them. I learned a lot watching it, so thanks.
I know you said a couple of times that s-log2 does not make the best match with an 8-bit codec, but I had to test it anyway for my own curiosity.
In this test https://vimeo.com/100562117 you can see produces the cleanest image (2-stops overexposed), and the other is to expose for middle gray and reduce two stops (2-stops underexposed). You said in the video that 2-stops under is about 38% IRE (I am seeing that on my scopes), however there’s a ton of noise in the image.
Will I be giving up a ton of dynamic range if I do this?
Thanks,
Dave Dugdale
Every time you go up one stop you double the light in the scene. So for a gamma curve to be perform uniformly it must use twice as much data for each successive stop. Rec 709 is close to this. SLog however records every stop with roughly the same amount of data for every stop. This means that as you over expose Slog you are in effect halving the amount of data per stop. This has an impact on the quality of textures and subtle skin tones. With 10 bit and approx 75 code values per stop there’s some freedom to do this. But with 8 bit and only approx 24 code values per stop it’s very easy to end up with flat looking, low contrast faces or highlights that have no texture and look almost plastic like.
A further issue is how you handle the log material in post. If you use conventional grading tools designed for traditional linear video, then when you take those low SLog levels and boost them to normal 709 levels you will add gain. Adding gain increases noise. So instead of bringing the levels up by grading, you should use a LUT (Look Up Table). A LUT will transpose the data levels used for SLog2 to the correct data levels for 709 without adding gain, so does not add noise in the same way.
SLog2 is very different to conventional gamma and to get the very best from it, it needs different handling. If you choose to over expose it to deal with any noise issues, then you will need a good quality correction LUT to un-compress the mid-tones and highlights.
Middle grey for SLog-2 should be 32%, white should be 59%.
Thanks Alister for answering my question, that helps a lot.
I am using Speedgrade (sometimes Resolve). Do you know of a LUT in existence for the a7S yet?
Dave
Alister is it ok to quote you from your Sony video for my a7s review? My review will get watched 100k times in the first couple of months so I want to get it right.
I will be talking about s-log2 and 8bit recording on the a7s.
In the video you said:
45:05 “You really, really, really want to use 10bit recording, I can’t not emphasis this enough, it makes a huge difference when you grade, you can use 8bit, it is not impossible, it can be graded, it doesn’t look terrible but you will find 10bit preferable.”
12:00 “… because with 8bit you only 235 gray gradient of brightness, when you go to 12-13 stops of dynamic range when you grade it those steps will get pulled apart and you will see banding in the image, you want 1023 steps in 10bit”.