This is something that keeps coming up in my workshops. It’s very important if shooting with S-Log2 or S-Log3 not to under expose and in most cases it can be highly beneficial to over expose a bit. Especially if you are using a camera like the A7s or FS5 in 4K when you only have 8 bit data.
Take a look at this chart. It plots the S-Log2 and S-Log3 gamma curves on a log scale of f-stops against the amount of 10 bit or code values used to record each stop. The center line of the chart is middle grey. Both S-log2 and S-log3 provide 8 stops below middle grey and 6 stops above. Take a look at the darkest stop, the one that is -7 to -8 and look at how much data is allocated to that stop. With 10 bit recording you have according to this chart about 10 code values for S-Log2 and about 20 for S-Log3. That’s if you have 10 bit, and it’s not a lot of data. Admittedly there isn’t going to be a great deal of scene information in that darkest stop, deep in the shadows and the noise. But there’s part of the issue, the noise. If you have under exposed and you take this in to post and have to stretch out the shadows, the noise in these darkest shadows is going to look pretty coarse because it hasn’t been recorded with many shades/steps so stretching it out will make even “rougher” for want of a better term. If you are recording with 8 bit the problems is even worse. With 8 bit, S-Log2 will only have around 2 or 3 code values for that bottom stop, in effect the noise will have two values – black or +1 stop. Imagine how nasty that will look if you need to raise or stretch you blacks because you are under exposed, it will become very blocky and grainy.
The solution is to over expose a bit. By over exposing your footage by a stop when you go in to post production you will in most cases be bringing your levels down. So instead of stretching the noise out and making it worse you will be shrinking it down and reducing the negative impact it has on it’s image. Because cameras like the FS5, A7s etc have 14 stops of dynamic range this small bit of over exposure is going to make very little difference to your highlights in the vast majority of situations. Any slight over exposure you may have will likely look quite natural anyway, after all our own eyesight does also over expose, we don’t have unlimited dynamic range. On top of that the display technology does not exist to show a 14 stop range shot in it’s entirety and with natural contrast.
20 thoughts on “Why It’s Helpful To Over Expose S-Log, Especially If You Only Have 8 Bit Recording.”
Sounds about right given my experience with 8-bit log in general.
I suppose this doesn’t necessarily apply in the 10bit 4:2:2 HD mode unless you want to overexpose a bit for creative reasons right?
Something irrelevant but I wanted to ask if the hd mode is always 10bit and 4:2:2 when you use the clear zoom function of the camera and or the slowmotion modes,something i didnt recall you mentioning in the Vocas seminar.
Yes it’s always 10bit 422 in HD as far as I am aware. Doesn’t hurt to overexpose log by a stop for most things.
Thank you thats good to hear.These run and gun style cameras -in general not only Sony- truly need a broadcast compliant format in hd,that’s the true future investment in my opinion,weird that this is not being mentioned a lot with all that 4K and log curves hype.
I know you are talking S-Log 2 and 3 but what about all the previous hype on UNDER exposing S-Log like this, “If you using S-Log then the normal exposure for middle grey is 38% of the way up the S-Log curve. Because the curve is not linear and compressed (squashed) a white card will be at 68% on the S-Log curve, not 90%. It doesn’t matter how bright or dark the scene that’s where the fixed reflection ratio of 500% or 2.5 stops between grey and white cause the levels to fall when you expose correctly using S-Log.”
It used to be all about undercorrecting S-Log a few years ago?
You bring up a good point, but it wasn’t about under exposing log it was about exposing it correctly. Little has changed regarding what we are recording. Log uses the same amount of data for each stop above middle grey, meanwhile the scene you are shooting has twice as much light in each stop you go up. The net result is that if you over expose and then simply reduce the levels in post the image looses contrast in a non-linear and un-natural way. When log was first appearing on cameras like the F3 people were exposing it as though it was Rec-709. This meant that their base exposure was already 2 stops over the recommended level. In addition we didn’t have the wonderful correction tools that exist today to deal with over exposed log footage, things like exposure compensated LUT’s or free grading software like Resolve. The net result was a lot of people really struggling to get good results from log because it was over exposed. The main issue being a lack of contrast and texture in skin tones etc that was very difficult to correct with the tools available. In a very short time we have gained great new tools that make dealing with over exposed log much easier and as a result it is now reasonably straight forward to shift the exposure down in post while keeping the correct contrast and get a good result. There is still a limit to how far you can over expose, I find 2 stops to be as far as I can go and get decent skin tones etc and it’s important to remember that this is a brightness shift of only 14% within a very flat looking image (the equivalent of less than 1 stop of 709) so it’s still very easy to end up over exposing if you don’t know what you are doing.
So, thanks to new tools we have the ability to much better work with over exposed log footage. Over exposing log in a very controlled way by between 1 and 2 stops brings benefits with the signal to noise ratio and color reproduction. But that exposure does still need to be tightly controlled.
Thanks alisterchapman, this was very informative. I have a few questions thoguht.
I’m working on a shortfilm where I’ll be using low-key and middle low-key lighting. I’ll use the Canon 6D for this. So:
1. I want to know how far I can over expose on 709, to then be able to restore skin tones and texture in the characters’ clothings and skins. And also to have a nice register of the shadows, wich I will later shrink down in post.
2. Is there a way to bring the over exposed highlights down to what I was seeing on set, by not having to isolate or apply a mask in, lest’ say, Davinci Resolve? Remember I’m working on 709.
Thanks in advance
You can’t over expose 709 without killing the image due to the highlight roll-off. S-Log does not have a highlight roll off so you can over expose it.
Anyone able to clarify what Sony mean on the A7sii when they say and I quote;
Clean HDMI output
This function supports 4K and Full HD and allows uncompressed movie output to an external recorder or monitor.
Im aware that it still only outputs 8bit to an external recorder but does recording externally on the a7sii actually help anything with respects to grading/banding/possible issues against the internal recording codec sony use?
Thanks in advance,
The internal recordings are of good quality, but highly compressed. By using an external recorder you could use a less highly compressed codec that will give a small improvement in image quality.
Thanks for the reply, and great information on here.
Do you know what the numbers are with respects to 100mbs internally, 420 8 bit (assuming 4k as2ii), what could one get in theory externally? Presumably the HDMI out has a ceiling on what it outputs? Id love to be able to test some footage under the same circumstances recording internally & externally to see what latitude recording externally gives as far as image quality and what you can do with it in comparison, id love to see that.
The HDMI output is uncompressed UHD 8 bit 422 which is the equivalent to approx 3.5Gb/s. So a lot more data than the 100Mb/s 35:1 compressed XAVC. Most external recorders will use ProRes which is about 800Mb/s. In either case there will be fewer compression artefacts so when grading very heavily there will be a benefit to an external recorder. The FS5 isn’t really designed for heavy grading applications as it lacks the all important 10 bit UHD and Exposure Indexing for log. If your going to spend money on both an FS5 and an external recorder (and don’t forget your media costs will go up significantly) really you should be considering the FS7 or F5 which record at 400Mb/s 10bit 422 internally which is comparable to ProRes (XAVC is a much more modern codec than ProRes).
Thanks for the info, seems my number pad wasnt working when I wrote asii I meant A7sii which is what I have. Presumably same comparisons apply with both cameras being 8bit. Sounds like you get a ton more data recording externally then. Logical but I read people saying theres not much difference which I find hard to believe if your doubling or even quadrupling the data rate
Alister thank you for taking time to educate the crowd! I’ve listened to the “Vocas”seminar couple times and its been extremely helpful and informative!
You mentioned that zebras for slog2/3 should be set to 70%. What setting should be for pp3/709?
70% Zebras for WHITE with S-log2/3
709 is the normal 70% Zebras for skin tones or 90% for white.
Hi Alister, thanks for all the info, really useful as always! I am working on an FS5 and I want to start shooting log. I have an external monitor but as you already said somewhere else, the Sony gamma assist is not great because does not take into account the fact that you should overexpose Log a bit. I was wondering, is there any LUT that you think works best when shooting in Log with the FS5? Something that I could put on my monitor and that would make exposing Log easier than with the gamma assist? Thank you!
My LUT sets always include luts for footage exposed brighter than base.
Hi i am so confused by all this – i used to use ETTR on my bmpcc now i have a sony FS5. What if i was in a studio and wanted to use either slog2 or slog3, what zebras would i expose my skin tone to get the most possible databits or code values (information) for each slog version . would slog 2 give me more skin information? ive seen charts that show it would. Some insight would be great. sorry im so confused. i.e. which zebras should i use for each slog version to give me the greatest amount of information for things like skin tones. for example you might say slog between 50% – 94% has the highest amount of information. hope this makes sense.
S-Log2 does give marginally more data in mid and high range which is why I prefer to use S-Log2 with an 8 bit camera. S-Log 2 has exactly the same amount of data per stop from 32% all the way to 105%, so you could expose a face anywhere in that range and have the same amount of data. Typically I put skintones at around 55-60% with S-Log2 which is about 1.5 stops above the base exposure level (approx 45-50% for skin tones). You could go a bit brighter f you want, but there is little to be gained going much beyond this.
Does this range (for skin tones) hold for S-Log3 as well, assuming one is wanting to also overexpose by 1.5 stops.
It’s beneficial to over expose S-Log3 by 1 to 1.5 stops too, but S-Log3 is a very poor match to an 8 bit camera, so the end result won’t be as good.