FS7: Shooting Custom Slog3 can we see a Look on a monitor out of camera?

Donny, there's already a ton of quite heated threads on that subject, hence me saying leave it alone :)

And for the rest, we all get wayyy too caught up in being determined to capture the sensor's entire DR in every scene. In reality there's really not that may times when one needs to make use of everything available above and below middle gray in a single shot.
 
Yeah but if Sam typically is having to go up to 4000 or sometimes 8000 ISO.. in even day light..maybe because of a slow lens. then your better doing that in cine EI and avoiding all the noise and lack of DR.. or in any case rating the camera lower and getting a cleaner image..
 
starcentral said:
Alister Chapman has argued implicitly that there is no difference between adding gain in camera or in post. I have never set out to test this but I also thought that in-camera is better than in-post, and I still don't know the true answer to this.

My theory is that if Sony is going to add gain to go from 2000 ISO to 4000 ISO you can bet they may introduce some variation on their processing, algorithm, or NR -- again no one knows for sure but this is sort of what I believe Sam has shown with his tests.
Click to expand...

Wasn't he saying that adding NR in post was better than in camera.. ?..
 
Like I said earlier, there is not a situation outside undercover filming where I would consider shooting at 8000ISO, even on my A7s which is absurdly good in low light. From what Sam says there's only really a couple of scenarios where he would too, and he happened to use one as an example which then became a multi-page exercise in surmising precisely just what the details of that example are.

He's a crazy guy with some crazy ideas (I've still not got over the grip destruction/re-purposing) but he makes nice pictures that look how he wants them to look and they are the result of him testing and discovering what is best for him.

I don't really think that idea warranted the complete derailing of the thread about a different subject for several pages, even if I was complicit in that as well.

Apologies OP. I'll shush now.
 
Ben Scott said:
I don't really think that idea warranted the complete derailing of the thread about a different subject for several pages, even if I was complicit in that as well.

Apologies OP. I'll shush now.
Click to expand...

The question was answered in the first three posts.. get an external monitor that supports luts.

S
 
morgan_moore said:
Im referring to any scene where at maybe 2.8 (I don't like shooting wider) the WFM would be hitting 30-40 max at 2000 and getting up towards the ideal 70-80 at 8000, the first image would need a huge bump in post, the second would not. I suggest the second image would be cleaner once the two images were 'equalised' in post
Click to expand...
In such an example, "cleaner" is not necessarily the issue as much as bit-depth. The darker the signal the fewer bits are assigned to use to describe it. Fewer bits means less information, less information means various issues, including increased banding.
 
Mitch Gross said:
The darker the signal the fewer bits are assigned to use to describe it. Fewer bits means less information, less information means various issues, including increased banding.
Click to expand...

There we go! In that sense you could say the neg is "thicker" (even though usually something else is meant by that).
 
I've had my FS7 less than a week, yesterday I shot some back to back Custom SLOG and CineEI with the Alexa LUT baked in for comparison. The latter looked utterly stunning. My partner came into my office while I fart arsing about and I showed her the Alexa LUT, she liked it but when I showed her the Custom SLOG delogged in FCPX with its overly saturated colours she almost had kittens, it was a big win.

To my jaded media hack eyes the Alexa LUT looked completely flawless and everything I've ever wanted a camera to look like but to a lawyer the technically inferior over saturated Custom SLog was the winner by a big margin.

It's a funny old game...
 
If it makes you feel any better every single TV showroom around the world operates on a similar principle. It's not an indication of quality :)
 
This is what Mitch is referring to in terms of data allocation specifically of course for slog:

ScreenShot2014-11-11at95059AM_zps5b93d478.png
 
With a 14-stop sensor, we don't use all the bits with S-Log3, which makes me wonder if S-Log2 might give the smoother gradations of greys and gradients.

Lately, I've been testing my FS7 with the baked in 709 LUT and rating the ISO at 500. I did some tests with a static outdoor shot that is very busy (testing CODEC quality) and I found that the noise was lowest and the artifacts most minimal at 500 ISO. 2000 ISO looks pretty noisy in comparison, under magnification. 3200 ISO, forget it. I could not tell any loss of dynamic range at the 500 ISO setting. In fact, the banding in the shadows that I saw at 2000 and above, was gone and more detail revealed in the darker areas.

I did some further testing indoors, and with the MPEG 422 CODEC at 1080, just to see how clean it would be as compared to my EX3, shooting to the Samurai Prores 422. All I can say is WOW. In a direct comparison, the MPEG2 out of the FS7 is clean, free of any noise, free of banding, except near total black, and very crisp, while the EX3-->Samurai footage looked almost like a consumer cam shot it, in a direct comparison. Surprisingly, I saw mosquito noise in the EX3-->Samurai footage, and some noise, but more suprisingly, since this footage was supposedly 10-bit, the shadow areas had banding, which exaggerated the noise. There was also no detail in the shadows, just flat black. The MPEG encoder in the FS7 is amazing. It actually looks better than the Prores 422 off the external recorder on the EX3. And I had forgotten what CA looks like, after using decent lenses with the FS7, something that jumped out at me when I scrubbed over the EX3 footage. My god, how did I ever tolerate that Fujinon lens for so many years?

Finally, the FS7 seems not to care much about white balancing. Shooting in mixed light, the rendered images look like what my own eyes see in person. The EX3, too pink/orange where the tungsten lighting is, and just right where the daylight lighting is. My hunch is that the wider color gamut of the FS7 gives more realistic rendering of mixed lighting, while the REC709 camera can only look right in the lighting it was balanced for.
 
basspig said:
With a 14-stop sensor, we don't use all the bits with S-Log3, which makes me wonder if S-Log2 might give the smoother gradations of greys and gradients.
Click to expand...

The way I read the graph that starcentral has posted is that SLog3 will give better blacks/greys and SLog2 with be better for brighter low contrast shots.
 
True, but also, the bit values are higher in S-Log2, so more grey steps across the 14-stop range. In S-Log3, the levels above 14 aren't used, and the upper bits not used.
 
starcentral said:
This is what Mitch is referring to in terms of data allocation specifically of course for slog:

ScreenShot2014-11-11at95059AM_zps5b93d478.png
Click to expand...

What's important to understand about this chart is that not all exposure levels are treated equally (or in Animal Farm terms, some animals are more equal than others). In S-Log3, the top 8 stops have about the same amount of data space allocated to them, then it curves gradually down for a few stops and then drops off considerably for the bottom three stops. S-Log2 data allocation is more extreme, with a huge allocation for the top five stops, with a quick drop off for the next five stops until there is little data room left for the bottom few stops.

What this means in practice is that there is a lot more room for adjustment and functional use of the image in the upper parts of the exposure than there are in the lower parts. When Sony first developed S-Log2, its sensors could technically reach into the lower parts of an exposure range with many stops of dynamic range, but they got fairly noisy quite quickly, so there was no reason to devote a lot of data space to the part of an image that would be eliminated in post anyway. As their sensors improved as well as changes in market forces and influence by client requests, Sony improved performance throughout the range and allowed more information in the midrange as this held a great deal of info that could be utilized well in an image.

My general taste is that I don't really need 14 stops of dynamic range, but I do want the part of the image that I care about to be stuff that I can use and manipulate properly. So I like to rerate the sensor and overexpose some, then bring it down in post. If I overexpose the image by two stops and then push it back down by the same amount in post, I'm effectively eliminating the bottom two stops on the chart above. Not only have I taken a bunch of the noise in the image and pushed it down into blackness and that's all that most people think about. But the other benefit is that the darkest parts of my image have three times as much data allocation, so I have a far greater ability to adjust the image in post without it falling apart or creating banding. No one is ever going to really use 14 stops, in fact hardly anyone will ever actively use 12 stops. What's important is to know how much you can push and pull the sensor and the recorded signal about to get the data you need in order to achieve the image you want.
 
Mitch - thanks for explaining this * edit * (and bonus points for using Animal Farm as a frame of reference).
 
Last edited:
One thing.. and Im not clear on it but this is my 'feeling'

Slog3 allocates very similar data levels to many stops.

To me this says 'it grades, it has lattitude' - the two things I want.

S
 
You must log in or register to reply here.
Back
Top