FS7: Overexposing on the FS7 for cleaner shadows

R

RASL

Member
From my experience, properly exposed Slog3 footage on the FS7 – Cine EI at base 2000 ISO – is clean and beautiful when there is plenty of light (for example, daytime exteriors). But when there are a lot of dark midtones and shadows (for example, darker interiors that are in fact supposed to look dark), even properly exposed Slog3 footage will be a bit noisy in those regions, albeit cleaner for me than on the FS5. I know some noise is a caveat from log formats, but I really prefer to stay in log because of the look and color I can get out of it.

Many overexpose shots to get cleaner shadows by rating the camera in CINE EI at 1000 ISO (one full stop). And some even rate the camera below that. I have shot many times at 1000 ISO myself. But I ask all of you: is it really worth it? In practice – not on paper, which I already understand – do you feel images are really much cleaner when overexposed? What is your experience? Especially, what is your experience with the final footage after grading it and exporting it in good quality, like ProRes?

After grading and exporting in ProRes from Resolve, I don’t feel it seems too beneficial to overexpose, at least in some of the recent tests I have made. And when dynamic range is very needed, due to uncontrolled lighting with a lot of contrast and strong highlights, it does hurt to lose a stop in the highlights. But since I will be delivering soon graded footage in ProRes, thus in high quality and not in compressed web formats, every detail is important. And noise is not an exception.

By the way, two more questions:
1. How do you feel about NR in CINE EI? In custom mode, the effect is very noticeable, but I honestly don’t see much difference in CINE EI. Besides, I rather get the footage right from the beginning without needing NR and risking banding issues.
2. Do you see a big difference between internal XAVC-I and recording in ProRes on an external recorder? I have not done any tests yet, but I am looking forward to your opinions.

Since I know most of you, if not all, are more experienced and knowledgeable, your feedback would be very appreciated! Thanks in advance for your time. And please forgive me this long post.
 
When I've shot with the FS7 I've rated it at 1000 ISO. But not to overexpose it, that's just to expose it correctly (as simple tests with a grey card and light meter show that you have to expose it at 1000 ISO for middle grey to sit where it should on a waveform).

So as far as I can tell, rating the camera at 2000 ISO would actually be underexposing the camera by a stop. Now the sensor handles a stop of underexposure pretty darn well, but it's never going to yeild optimal image quality for you.
 
RASL said:
From my experience, properly exposed Slog3 footage on the FS7 – Cine EI at base 2000 ISO – is clean and beautiful when there is plenty of light (for example, daytime exteriors). But when there are a lot of dark midtones and shadows (for example, darker interiors that are in fact supposed to look dark), even properly exposed Slog3 footage will be a bit noisy in those regions, albeit cleaner for me than on the FS5.
Click to expand...

You pretty much answered your own question with this statement. If an image with areas of the frame with higher IRE levels looks cleaner to your eye, then overexposing will do exactly that when dealing with low light - it will place everything higher on the curve hence giving higher IRE levels per given area of the frame. Another thing to remember too is how Slog gammas distribute bits of information - shadows get very few of them in comparison to highlights so the higher you push them up the curve the more information they will get.

Now it is debatable if rating the camera at 1000 ISO is really overexposing or if you are simply saying that this is what you consider the camera's true base sensitivity. If you go by Sony's spec which claims the camera's native / base ISO is 2000 then yes, it is a full stop of overexposure. However, even based on Sony's aimed numbers for 18% grey for Slog 3 gamma, it is relatively easy to figure out that the camera's native ISO is actually closer to 1250 - 1600 ISO. There was a thread about this here on dvxuser (started by Grug who just responded to your question): http://www.dvxuser.com/V6/showthread.php?331819-Beware-the-camera-s-not-really-2000-ISO
 
The way i figure it Sony has just called ISO 1250 EI2000. Which is still a lot better than most S35 cameras out there.

I monitor with a SmallHD702 with a 1-stop under compensated LUT on it for gauging stuff by eye and at the same time try to hit 75IRE for 90% whites instead of the 61% Sony specifies.

Then in circumstances where light is an issue I can't fix I have 1/3 stop and 2/3 stop incremental LUTs if I'm prepared to put up with a touch more noise, which in most cases I am.


I think in-camera noise reduction is useless for how I shoot and after testing it for a bit and always left it off.
I think if you can expose the image properly you don't need noise reduction as the images usually look nice and clean.
I think in the few instances where this is not possible then Neat Video or even Red Giant Denoiser do an exponentially better job for just clips where you need it.
 
Last edited:
Thanks a lot to all of you that replied, sharing their own experiences. I had forgotten completely, due to the fact that I had to stop shooting for about 8 months, that many indeed have showed that the true base ISO of the camera should be 1000 and not 2000. Thus shooting at 1000 ISO would not be overexposing 1 stop - or at least that is debatable, just like you guys said. And the other thread started by Grug is also very useful; I will take a close look at all the posts there, so thanks for the link! If anyone still would like to share his/her experience or make a suggestion, please feel free to do so, it will always be appreciated.

Scott, thanks for your comments on NR. They confirm what I was seeing. It helps a lot to have inputs from someone with a lot more practice, so thanks a lot. The doubts with this issue are clear and done for me.

Just to clear things up a bit more on other issues:

1. For those that usually rate the camera at 1000 ISO, do you do it even for sunny exteriors? Or just for interior shots with darker midtones and shadows? I don´t know, but it would not seem so necessary to rate the camera at 1000 ISO when there is plenty of light. But I understand it might be a better practice to always go with the same ISO.

2. How about internal XAVC-I vs. recording in ProRes on an external recorder? Any tests or observations? Personal opinions? Or do most of you just go with the XAVC-I? I imagine the latter might be the most used, just curious to know so I can get the best source footage to later edit, grade and deliver.

Thanks a lot to everyone! After working for many years in this area, I am starting to work on my biggest project to date on the FS7. So I really do appreciate you comments and I apologize in advance for any ignorance in my behalf.
 
One thing that I'll add is that, while there may be some norms and recommendations, most of the things discussed here are relative and very much a personal taste. What's acceptable will greatly depend on your personal taste, type of a project, its aesthetics, specific shots, delivery method, etc, etc. Advices you'll likely hear in this thread will greatly vary too because of that. If you look at the history of photography, when the first 35mm camera was invented by Oscar Barnack (you know, that German camera), many thought that the negative size was too small and that it produced images that were too grainy when compared to large format cameras. History shows that it was a tradeoff worth consideration, many accepted and took great advantage of a much smaller and user friendly tool. Believe it or not, Henri Cartier-Bresson was often criticized for grainy images yet he still simply did his thing. Was it worth it? On the other hand, try to imagine, i.e. Edward Weston's pictures taken with a 35mm camera...
 
Shadow latitude/noise vs. highlight latitude. Just find your sweetspot for various enviroments. I usually do this: EI1000 70% (interiors, dark exteriors) EI2000 25% (bright exteriors) and EI500 for 72p and up.
 
In EI mode the camera, as we all know, is always shooting 2000ISO, and underrating the camera just helps you ETTR.
Knowing that, I pretty much ignore ISO and use the WFM on my Odyssey to make sure I'm always capturing a 'thick' negative in all situations.
 
bill totolo said:
In EI mode the camera, as we all know, is always shooting 2000ISO, and underrating the camera just helps you ETTR.
Knowing that, I pretty much ignore ISO and use the WFM on my Odyssey to make sure I'm always capturing a 'thick' negative in all situations.
Click to expand...

Yep, that's pretty much what I do. EI ratings don't jive with ISO, just ignore and check your WFM and zebras.
 
The interesting behavior of this sensor and processing at low light levels has raised my curiosity.
It behaves as if there were automatic gain control at the very low levels below IRE 20. In fact, you can watch the noise floor rise on a scope as you close down the iris.
Even at 12,500 ISO, if you increase the light, the noise decreases drastically. That suggests there is some sort of AGC action going on with the sensor processing, apart from any gain settings the user may be implementing. I think it's more than an S-Log curve effect here. It happens with REC 709 gammas too.
 
Mitch Gross said:
At about three minutes in, we explain and illustrate how this reduces noise. It's quite clearly seen in a side by side comparison.

https://www.youtube.com/watch?v=QdOb7A0csyE
Click to expand...

That's a concept which I understand, however, what I'm observing suggests that something is turning up the gain on the sensor as light level diminishes. I can see the noise appear as the iris lets in less light, and I can see it rise on the scope. This is a dynamic thing, not the static noise that increases as a fixed function of s/n ratio.

Think of an audio amplifier that has a quiet noise floor when program is present, but then as the signal fades away, the amplifier brings up a rather loud hiss. Normally that doesn't happen, unless there is a compression amplifier that's got a lot of gain and is cranking its gain way up, hunting for a signal.

I would not expect noise levels to increase as I close down the amount of light hitting the sensor, if all of the analog electronics were of fixed gain. I would just expect it to go to black.

Even more interesting, is that if I cap the lens, the noise drops down again! It seems that some small amount of light hitting the sensor stimulates noise, but take away that light completely, and the noise goes to nearly invisible. For instance, if I close down to F16 with custom hyper gamma and ISO 12,500, the screen is filled with noise. As I open up to f/1.4, the noise completely disappears.
 
Thanks again to everyone that replied! Many interesting points were made. I agree with broughtonfilm and others that most things come to personal taste. What struck me the hardest is seeing that most, if not all, rate the camera at a lower ISO to get cleaner images. Even those like Bill and Liam, that use the base ISO of 2000, end up making some kind of compensation for exposure through the monitor's waveform. This is also clear in the other thread on the F5, started by Grug a few years back. So even though each one has his one methods and preferences, which is comprehensible, pretty much all agree that the base ISO of 2000 is actually underexposing the image.

From your advices and comments, it is pretty clear to me that going with 1000 ISO on the FS7 is definitely a better option for cleaner results in interiors and dark exteriors. It might hurt to lose a stop in the highlights, but it is worth it for overall image quality. I will make some more tests with bright exteriors, since my previous tests showed clean images in 2000 ISO out in the open with plenty of light; but I will have to pay a closer attention to shadowy regions. So I might be going with an approach similar to stokstad, or depending on the results of the tests only with 1000 ISO in every scenario. Another thing that might be interesting to take in mind is what basspig indicated.

Thanks for your comments and observations. Not meaning to be rude, but if anyone still has anything to say about internal XAVC-I vs. recording in ProRes on an external recorder, please go ahead. But if no one does it, I won't complain at all since your comments so far have been very useful. Thanks!
 
Recording to my Odyssey 7Q+ I find the image is marginally better in 4K but HD and 2K HFR suffer from additional noise. Given the extra hassle of using an external recorder, I wouldn't recommend it over shooting internal. It's just not that big a gain.
 
RASL said:
It might hurt to lose a stop in the highlights, but it is worth it for overall image quality.
Click to expand...

You may want to experiment with using S-LOG2 in outdoor, uncontrolled lighting situations because it allows more headroom on highlights. S-LOG3 clips at 94-IRE while S-LOG2 allows highlight up to 107-IRE for a little added safety. With my F55 and FS7 I generally use S-LOG3 in controlled situations and S-LOG2 outdoors for eactly that reason, but to be honest, unless you clip the highlights they can both be graded to look almost identical in Resolve.

I think Sony's ISO numbers are meaningless and I don't use them to determine exposure, but if you want to use that terminology, you could say that I'm rating my FS7 at about ISO800 and my F55 at around ISO400. That allows me plenty of headroom in most situations, and with any ISO higher than that and you will definitely be under-exposing. But just to be clear, this technique is different than ETTR. I don't just push every shot to the edge of clipping and then back off a little. There's no consistency between shots with that sytle of shooting and so it adds lot of work in post. So yes, I am over-exposing beyond Sony's recommendations, but I am doing it in a consistent manner so nearly every shot is offset by the same amount -- thus making post easy.

https://vimeo.com/ondemand/resolve
 
Liam, thanks a lot for answering my question about recording ProRes on an external monitor vs. the internal XAVC-I. What you said sums up very well what I have researched. I will to do some tests, of course. But, having feedback from others is very useful to me, because I know that at the end others have more experience and can outweigh things better. Even if the final decision is always personal, good feedback like the one everyone has given so far in this thread ends up being great food for thought.

And Doug, also to you, thanks a lot for your answer. It raises another good point about Slog2 being better for uncontrolled lighting situations, which I have read a lot about. The only problem I have with Slog2 is achieving more natural and pleasing colors; I just fill this is easier to do with Slog3, for which I have a lut/dpx of my own. If I could only get similiar colors in Slog2... Can you do that easily? (Doug or other reading this). Matching SGamut to SGamut3.cine is a whole different chapter for me. But I will do some new tests with Slog2, because what you pointed out is definitely worth trying more. I will also take a look at the guide you linked.
 
Last edited:
Liam Hall said:
Recording to my Odyssey 7Q+ I find the image is marginally better in 4K but HD and 2K HFR suffer from additional noise. Given the extra hassle of using an external recorder, I wouldn't recommend it over shooting internal. It's just not that big a gain.
Click to expand...

At 200 % magnification, quite dramatic differences in block artifacting can be seen. These crops of a UHD image illustrate the differences:

XAVC 200 pct crop.jpg
Prores 200 pct crop.jpg

Also, you'll notice the warmer color rendering of Prores. I don't know whether that's the CODEC or the Shogun's doing.


RE: ISO, in S-Log, all bets are off. ISO varies depending on how you grade, ie., what curve you apply in post. As such, you can have a higher effective ISO in the midtones and lower in the shadows, if you play with the curves.
 
Just another note:
If you're shooting a scene with plenty of light, shallow depth of field without a lot of detail in the shot, ISO 2000 will look quite clean because you're not stressing the CODEC.
However, if you're shooting 180fps in the woods, with a lot of fine detail, forget it. Get yourself a 'thick' negative or you're going to have A LOT of noise.
 
Scenes with lots of high frequency detail will break the high compression CODECs. With XAVC, mixing dense foliage with open patches of deep, blue skies will produce some particularly nasty artifacts.

The worst cases are the dense foliage, waterfalls and dense foliage, choppy ocean water with sun glare, and flames--zoomed up close. To be safe, I record in Prores HQ when equipment configuration allows. And Prores HQ is a tad sharper on fine detail, for those times when you're using a clinically-sharp lens.
 
You must log in or register to reply here.
Back
Top