A7s sample footage
- Thread starter squig
- Start date
Korean Rally with a mix of frame rates:
http://www.youtube.com/watch?v=hAFjdLD_QSM
http://www.youtube.com/watch?v=hAFjdLD_QSM
vincent thomas
Well-known member
When the Shogun or similar 4k recorder will be available, i hope the 4k to 2k conversion should be able to solve the problem
The 60p aliasing is killing me. It's all over the stairs on this shot. A drop in anti-aliasing filter should fix it but it's going to be a long wait for one.
squig
Veteran
It doesn't shoot 60p @ 4k
Uwe Lansing
Veteran
squig
Veteran
1080p definition looks great even on youtube, no DSLR softness.
cyork
Well-known member
Thanks Kholi and tell me about it. I put half this thing on plastic and now I'm looking around at my current gear to see "what can go and what can stay"
I've got a BMCC and BMPCC and the A7s will fill the low light void and make for a better event cam. Image wise, after playing with the last batch of footage I've found, I'm not seeing anything that couldn't be matched pretty well with the other two cameras. Plus, the overcrank will be nice to have again, even with the aliasing and moire. I imagine the successor to the FS100 is going to be full on beast mode so I guess it's time to start socking away the pennies.
cyork
Well-known member
Good IR and I'm wondering if exposure isn't causing the color shift along with the lenses used. Can't remember what he said he shoots with but I want to see some Zeiss paired with this thing. Zeiss and the FS100 with the right profile is solid gold, IMO so I can't imagine it will be much different once we figure out how to dial in the right settings and for SLOG2, learn to expose for LOG. Somebody had posted a great article on that a few pages back and it was new to me. I usually ETTR and call it a day but with 8bit, sounds like you have to nail it.
The Irish TRex
Well-known member
Damn. This bites. Makes me not want to use any of my APS-C lenses now. Is this really how big of a difference there is?
...
strancali
Well-known member
Kholi, you got it man. It should arrive on Saturday. I'll let you know and you test the s*%t out of it next week. I also have the metabones v4 adapter so throw that 70-200mm of yours on it
vincent thomas
Well-known member
Any word about how the Slog2 shown here will perform compared to the futur Log gama on the gh4?
Philip , James... What are you though about color and dynamic range difference?
On a less important note, Which camera version is it James? So is it the world first camera or not? Is it best to wait the european release even if i could order one from USA or Asia?
Flickering issue? Could i hope it will be fixed with a 4k recording?
Thanks, can't wait to see more test!
Philip , James... What are you though about color and dynamic range difference?
On a less important note, Which camera version is it James? So is it the world first camera or not? Is it best to wait the european release even if i could order one from USA or Asia?
Flickering issue? Could i hope it will be fixed with a 4k recording?
Thanks, can't wait to see more test!
Amen. He just posted this a bit ago. https://vimeo.com/99676933
The "natural reference" had my jaw hitting my keyboard a bit, I'll admit
Loving the potential here. View attachment 88302 Also found some other clips that were decently shot. (graded)View attachment 88303
paulcurtis
Veteran
I'm not sure everyone is really getting Slog2 here, it's not about dynamic range it's about the distribution of light values in a quantifiable way. Slog2 offers a better way to record and to manipulate the image afterwards.
The cinecurves may contain the same lowest black value and the same highlight pixel of white but the distribution of values between differs wildly, most cine curves will very quickly roll off the very top highlights so perhaps the top 2 or 3 stops might be described in a handful of actual code values, it looks like clipping basically and it practically is.
Remember we have a paltry 256 steps to describe this scene. a 12 stop scene in RAW minimum needs 4096 steps but really needs a 16 bit container, 65k steps. So you can see that an 8 bit output is a massive difference.
Light is a doubling of value, so for 12 stops that very last stop occupies code values from 2048 to 4096, a massive waste. Log is an attempt to distribute stops so that each stop takes a similar number of values. In our 8 bit case describing each stop with 20 values would be possible, Slog2 sort of does this but changes the curve in the shadows to make the best of it. The black is also elevated so black is not 0.
When we talk about stops over and under it's difficult to quantify with cinecurves because they are not officially defined - there's no mapping between real world stops and what's in the file.
The point of Slog2 is that it is a mathematically defined curve that an NLE or Grading app can unfold back into scene linear space, so that the light is photometrically accurate. If you reduce exposure via footage that has been correctly interpreted as Slog2 then your scene will look 'right'. Much like RAW. RAW is also a known photometrically accurate way of storing light, although it doesn't use a log curve (usually), it's simple linear by nature.
Each Slog container sony has done has tweaked the curve a little. The original Slog was created for the F35 i believe and was designed to hold the range of that camera. Slog2 was more for the F5/F55 generation and has a range of 13.5 to 14 stops. Slog3 is a tweak of this that opens up the shadows and makes the actual log part of the curve more like LogC which is the basis of AlexaLogC and is a slightly better, truer log curve.
The thing with range is that it's difficult to quantify. You can peg mid grey at whatever you want and then you can tell very easily where clipping is because, well, it's clipped. However where do you place black? Because the darker you get the noiser and it's up to you to decide when has the noise taken over the signal. I think the FS700 has maybe 4 stops under grey on a practical level based around the 12 bit output.
It would be useful for Sony to issue a white paper explaining how they arrived at a base iso of 3200. Like cpc said i feel there is some definite messing around happening between the sensor and output - much the like FS700 which requires an exposure gain for the RAW footage to work correctly. It's not wrong, but it would be useful in understanding how the sensor works best.
cheers
Paul
The cinecurves may contain the same lowest black value and the same highlight pixel of white but the distribution of values between differs wildly, most cine curves will very quickly roll off the very top highlights so perhaps the top 2 or 3 stops might be described in a handful of actual code values, it looks like clipping basically and it practically is.
Remember we have a paltry 256 steps to describe this scene. a 12 stop scene in RAW minimum needs 4096 steps but really needs a 16 bit container, 65k steps. So you can see that an 8 bit output is a massive difference.
Light is a doubling of value, so for 12 stops that very last stop occupies code values from 2048 to 4096, a massive waste. Log is an attempt to distribute stops so that each stop takes a similar number of values. In our 8 bit case describing each stop with 20 values would be possible, Slog2 sort of does this but changes the curve in the shadows to make the best of it. The black is also elevated so black is not 0.
When we talk about stops over and under it's difficult to quantify with cinecurves because they are not officially defined - there's no mapping between real world stops and what's in the file.
The point of Slog2 is that it is a mathematically defined curve that an NLE or Grading app can unfold back into scene linear space, so that the light is photometrically accurate. If you reduce exposure via footage that has been correctly interpreted as Slog2 then your scene will look 'right'. Much like RAW. RAW is also a known photometrically accurate way of storing light, although it doesn't use a log curve (usually), it's simple linear by nature.
Each Slog container sony has done has tweaked the curve a little. The original Slog was created for the F35 i believe and was designed to hold the range of that camera. Slog2 was more for the F5/F55 generation and has a range of 13.5 to 14 stops. Slog3 is a tweak of this that opens up the shadows and makes the actual log part of the curve more like LogC which is the basis of AlexaLogC and is a slightly better, truer log curve.
The thing with range is that it's difficult to quantify. You can peg mid grey at whatever you want and then you can tell very easily where clipping is because, well, it's clipped. However where do you place black? Because the darker you get the noiser and it's up to you to decide when has the noise taken over the signal. I think the FS700 has maybe 4 stops under grey on a practical level based around the 12 bit output.
It would be useful for Sony to issue a white paper explaining how they arrived at a base iso of 3200. Like cpc said i feel there is some definite messing around happening between the sensor and output - much the like FS700 which requires an exposure gain for the RAW footage to work correctly. It's not wrong, but it would be useful in understanding how the sensor works best.
cheers
Paul
Samuel H
Major Contributor
^ great post
My guess about how Sony arrived at the ISO3200 number: they shot a 18% grey card, with minimum analog gain and using their curve, and throwing at the card the appropriate amount of light so that it would hit the 32% mark in the recorded footage. Then they used a light meter to see how much light that was: with our aperture of f/X and shutter speed of 1/Ys, this is ISO3200.
At least that's how I would have done it. Except I would have put mid grey quite a bit higher.
My guess about how Sony arrived at the ISO3200 number: they shot a 18% grey card, with minimum analog gain and using their curve, and throwing at the card the appropriate amount of light so that it would hit the 32% mark in the recorded footage. Then they used a light meter to see how much light that was: with our aperture of f/X and shutter speed of 1/Ys, this is ISO3200.
At least that's how I would have done it. Except I would have put mid grey quite a bit higher.
cpc
Well-known member
It is possible Sony's s-log DR distribution choice is based on simulating film's stops under mid-grey (around 5-5.5 for 5218, and less than 4 stops from toe to midgray; possibly 1 more for), and then letting overexposure latitude fall wherever it will, and then rescaling midgray to accommodate that. If you follow this procedure, the bigger the total DR, the lower the midgray point: exactly what happens with s-log and s-log2.
The original log format Cineon has midgray at around 45% but it compresses highlights a bit, unlike s-log, and also provides for base variation.
The original log format Cineon has midgray at around 45% but it compresses highlights a bit, unlike s-log, and also provides for base variation.
Last edited:
dfreerider
Well-known member
Seems like Stu got his A7s. He sent his GH4 back, so I am curious what his reaction to the A7s will be once he's had a chance to test it out.
philiplipetz
Veteran
8 bit Slog2 may be usable but thin, however a 4k A7s image resized to HD will approximate 10bit chroma with 8 bit luma. Will that be enough to make a significant difference in gradability?