A7s sample footage

macgregor said:
I have no idea where and how they got that number. My tests don't show anything close to 14 stops.
Click to expand...

I'm sure it greatly depends on your methods, but just rolling through the iris/shutter exposing a fixed white spot on a wall from just under clipping white down to where you can no longer tell the difference between stopping down, you can quickly see 13-14 stops pretty easily. Obviously any test that actually quantifies noise and accounts for noise levels or simply uses some 709 (or other non-log) gamma curve without digging into shadows may find differently. But I do find it a bit more objective, if not always as useful, to measure from white clip down to the bottom of the noise floor where signal is no longer distinguishable from noise at all.
 
CharlesPapert said:
For what it's worth, I do feel that the 8-bit debate is overstated by many. I shot the a season of TV that way a couple years ago (F3 with s-log recording to external XDCAM) and we did heavy grading and never saw any issues. And as cyork says, it's not yesterday's 8 bit to boot.
Click to expand...
Well said. And it is 4:2:2 recorded externally, 8-bit but 4:2:2... Don't forget.

Moreover, 8-bit can be critical but when there is a sky plenty of layers of blue.

Other than that, you can even feel you can beat any Ferrari with your Honda if, after all, the highway has speed limit or you are driving in downtown.

E :-)
 
macgregor said:
I have no idea where and how they got that number. My tests don't show anything close to 14 stops.
Click to expand...
Looking forward to your findings Mac. If there is anybody on this site to get to the bottom of things it be you. All anyone has to do for my point to be valid is look at Mac's work. Mind blowing!
 
macgregor said:
I have no idea where and how they got that number. My tests don't show anything close to 14 stops.
Click to expand...

There's a massive difference between technical range quoted by people and actual useable range. I've been doing some extensive testing and would be inclined to agree with you there.

But that's true of most real world cameras out there and relative to the others the A7s is very very capable.

Soon i will publish the results

cheers
Paul
 
Erik Naso said:
Wow. Nobody has called me a troll before. Congrats on post #5. Welcome to DVXuser.
Click to expand...

Erik, I apologize for doubling back with my own negativity. It was whiplash from having experienced senseless negativity on Vimeo and having met James IRL (he's a really nice guy!). I assumed it was blasting/flaming of James on another forum where he could not defend himself, and that was where I took a wrong turn...ASSuming.

I was out of line for saying your activity was trolling when I don't know the context. I hope you accept this apology as I enjoy everyone's contribution to this forum. Of course as you are a long-time user, I should have shown more grace & respect.

Thanks!
 
Now this looks great. Colors look natural and skin tones are looking excellent while still looking filmic. Im in on this lut.
 
paulcurtis said:
There's a massive difference between technical range quoted by people and actual useable range.
Click to expand...

Yes, with the primary difference being that the technical range from white clip to noise floor is substantially more objective than "useable" range which is highly subjective and usually has to rely on a particular workflow, look, and preference regarding the appearance of noise. That said, I think a more standardized quantification of noise may be called for to distinguish between a camera that has 14 stops of range with a lot of color noise in the lower 5 stops and a camera that has 14 stops of range but much less noise and less colorful noise in the lower stops. Then again it can be quite difficult to judge such things since compression and gamma and resolution can have a huge role in the quantification of noise in shadows so a lot of normalization would have to happen to make meaningful comparisons in "useable" DR.
 
nyvz said:
Yes, with the primary difference being that the technical range from white clip to noise floor is substantially more objective than "useable" range which is highly subjective and usually has to rely on a particular workflow, look, and preference regarding the appearance of noise. That said, I think a more standardized quantification of noise may be called for to distinguish between a camera that has 14 stops of range with a lot of color noise in the lower 5 stops and a camera that has 14 stops of range but much less noise and less colorful noise in the lower stops. Then again it can be quite difficult to judge such things since compression and gamma and resolution can have a huge role in the quantification of noise in shadows so a lot of normalization would have to happen to make meaningful comparisons in "useable" DR.
Click to expand...

Measuring DR is always a tricky business, that's why I only care about direct side-by-side comparisons using exactly the same scene and methodology. The cinema5D test is using the same method (expensive xyla-21 chart) for the a7s and GH4, that's why I think the difference must be real.

A year ago I favored measuring "absolute DR", since, as you said, it is more objective: if the waveform shows a difference, there's information there, so it counts, simple enough.
Then I did this "usable DR" test here, where I defined the noise floor in a very strict way (being able to see the fine detail in the dark shadow here), and got incredibly smaller numbers. Like: for the 5D2 in H.264 mode and using Prolost Flat (dialed down Neutral, it was a loaner and it didn't have Flaat installed), I measured 10 stops of absolute DR, but just 6 stops of usable DR. The reason? A poor codec that eats up all the small detail in the shadows: you can recognize big shapes but no texture at all.

In any case, what's clear to me is that most people are pretty bad at gauging DR in non-technical tests. Get both cameras on a tripod, and severely overexpose and underexpose, to see what you can get back and where the clipping point is, anything else will end up in "they are about the same", even if they aren't.


I will be running a7s vs 5N vs pocket vs D5200 sharpness and DR tests when I get my camera. I'll try to borrow a GH4 too...
 
I wish C5D would have measured absolute DR instead of cropping at acceptable noise and without saying what that number is. I really find it hard to believe the C300 has less DR than 5D3 h.264. Having used those cameras its simply not true. I also wish they would have given exact picture settings. It can make a big difference on some cameras like the GH4. Although no where near the A7s it still had a pretty healthy number as measured by Adam Wilit.
 
So, I searched for that Adam Wilt test, it's here. He finds 10 stops of DR in the regular image, 12 if you push the shadows pretty hard in post and don't mind the noise (13 discernible patches means 12 stops of DR, and the two brightest patches in his chart are always equally clipped). So, not really that different from the cinema5D test (here; I see 10 stops in the unmodified image, but pushing it in post reveals another noisy stop, which their software did count as available DR), and the C5D test has the virtue of comparing two different cameras with the same methodology. Looking at the images (here and here), I would say the a7s has around 12 stops of usable DR, the GH4 just 9.

BTW, C5D used dxomark to take the measurement, which means their noise floor is defined as a specific SNR level, which I can't seem to find right now.

Things like different ºK of the background lights, different lenses, light reflection and spill, etc, can easily create deviations of around 2 stops in these kind of measurements, that's why I insist on comparing side-by-side numbers where everything is common to both cameras. Whenever I do my a7s DR test, I will have to test the pocket and the other cameras again as well: even if I try to match everything as well as I can to what I did the last time, it will never be good enough, these measurements are very tricky.
 
Last edited:
Still working on my review. But here's a quick snapshot. High contrast scene pushed to the limits of the raw files. A7s in RAW at ISO100 vs a very popular DSLR. Same exposure. Same Camera raw settings. No comments...

Untitled-1.jpg
 
Last edited:
macgregor said:
Still working on my review. But here's a quick snapshot. High contrast scene pushed to the limits of the raw files. A7s in RAW at ISO100 vs a very popular DSLR. Same exposure. Same Camera raw settings. No comments...
Click to expand...

I always hated that red Canon noise. An A7s lens cap test is very telling, there's very little noise up into very high ISOs in movie mode as long as you ETTR.
 
squig said:
I always hated that red Canon noise. An A7s lens cap test is very telling, there's very little noise up into very high ISOs in movie mode as long as you ETTR.
Click to expand...

Hold on to that comment Squig, you might be surprised which camera is which... A7s is the one on the left.
 
Last edited:
A happy coincidence (tho not for my wallet, who needs food anyway right?), the A7S came in the same day as the DJI Ronin for me.

Instead of a straight-forward test, I decided to really push the A7S in particular. I wanted to see if I could get away with usable footage at f/7.1 so that everything is in focus. With some noise-reduction, it's not turribbbllle (see that in Cleveland Brown's voice).


The second one, I decided to not go above f/4 and ISO 16,000.... and honestly, even though the compression hurts it, the noise coming out at that place isn't that bad.


If you'd like to read some more: http://www.llirik.com/2014/08/dji-ronin-sony-a7s/
 
You must log in or register to reply here.
Back
Top