A7s sample footage

@ReelWorksMedia

That's why I was (still am) considering the Sony a7S, as a replacement for my Canon 5Dmk3.

Not to sidetrack the conversation too much, but the C100 is a really nice camera, I've always been super impressed with the video it captures.

https://vimeo.com/74847141

Real question: Would you rather shoot with the C100, or the 5Dmk3 in RAW?


definitely the c100 over 5d raw. I have shot lots of full productions on the 5d raw. Looks great! Huge pain in the arse too!

Im in the same boat though... Considering it a 5d replacer....and I imagine that is exactly what Sony intended this to be.

hmmmmmmm decisions decisions
 
3200 base in slog is not a deal breaker for me, just throw on a ND filter and you're good to go. Those magnetic xume filter rings are my next purchase to make swaps easy and not have to deal with vari-nd issues. To me this is a FF mini FS100 without the horrible moire and aliasing. Everything in this price range has warts, just find the ones you can live with.
 
Because there's relatively little content from the camera, I think it's easy to look at the tests that have popped up, which are mostly testing to see how the camera performs in worst case scenarios (i.e. extreme low light, APS-C mode, rolling shutter tests, etc.) with minimal or no grading. We're not seeing the best you're going to normally get out of the camera. Those first two films (the one in the temple and the one that Den Lennie shot in Scotland) looked pretty good to me.

To my eyes the colour response and highlight rolloff looks far more pleasing and natural than its nearest competitor (GH4). And the way people talk about 8bit vs 10bit makes me wonder how we graded ANYTHING this time last year. It *would* have been really great to have 10bit (Sony you missed a marketing trick vs the GH4 which doesn't even really need to record 10bit because its image is so noisy that it's got baked-in 8bit dithering! ;-) )... but I'm looking forward to putting it up against my C300 to see which one becomes my favourite. I think the 800% profile is going to be my go-to profile (just like wide-DR on the C300).
 
This base iso will be a deal breaker for a lot of folks(pre-order cancelled). Getting the best quality image while shooting even semi wide open with the sun still in the air will be a problematic for this camera. It seems like there are some big tradeoffs for getting this kind of sensitivity, which makes me think we are heading down the path of having separate cameras(or sensors) optimized for either shooting in daylight or night time conditions.
 
If any of these sample images are a good indication of best DR, I would say that many of you are waaaay too optimistic with your figures. Things might change as sample images progress but there is nothing compelling about this performance so far - save, perhaps, the low light potential.
 
... (Sony you missed a marketing trick vs the GH4 which doesn't even really need to record 10bit because its image is so noisy that it's got baked-in 8bit dithering! ;-) )...

Haha..., yup this basic noise in the image of the GH4 is very annoying for me too. But initially I thought that the a7s will be the much better choice in terms of DR, 120fps slomos, low light, colors... etc. I´m not so sure anymore from what I have seen so far (except low light abilities). Waiting for a a7s vs GH4 side-by-side comparison....

A nice picture:




out of this little video:
http://www.sony.jp/ichigan/introduction/a7/a7s_movie.html#r=s
 
ok, let's see...

About the ISO 3200 slog2 thing:

With my current camera (5N), I usually need 6 stops of ND to shoot in sunlights in a bright day at f/2.8, 1/50s and ISO 100. If my camera could see three stops more into the shadows, I would leave those 6 stops of ND in place, to get a non-clipped background; since it doesn't have that much DR, I take away 3 or 4 stops of ND, and get my highlights clipped.
Now, let's see how slog2 changes that...
With slog2 on the F55, I would have 6+ stops of DR above mid gray, instead of <3 on my 5N. So I could put just 3 stops of ND instead of 6 and the clipping point would still be somewhat higher, meaning I won't clip my highlights shooting in sunlight on a bright day at f/2.8, 1/50s and ISO 100. Now, if I can't shoot ISO 100, but only ISO 3200, that is 5 more stops of ND that I have to put in front of my lens, for a total of 8.
So: if the a7s matches the slog2 curve of the F55, then the clipping point with 8 stops of ND at ISO 3200 will be the same as I currently have with 6 stops of ND at ISO 100 on my 5N.
This is a very small change of plans for me: I will get the Hoya proND 500x (9 stops) instead of the 100x that I had planned (6.6 stops), and I'll have to add the 32x one (5 stops) to have a good filter that is not so strong.

About slog2 base ISO being 1600 then 3200:
I bet it's exactly the same as it was and only the name changed.

About slog2 vs 800%:
I'll have to test it but I think I want slog2

For guys who like riddles:


That is one very nice test. Pity that we won't get the which-is-which until next week.
My guess is that camera 2 is the a7s, because of the colors, the DoF, and how clean the shadows are. As long as it is not camera 4, which has more noise in the shadows than the others, I'll be a very, very, very happy camper. Hell, I'll even be hapy if the a7s is camera 4, it is pretty close to the big boys. Looking good.
(Camera 5 must be the F55 with slog3, it is head and shoulders above the others)

edit:
Looking at this again, I'll definitely be happy with camera 4 too, even if it has noisier shadows: it has more headroom into the highlights than cameras 1, 2 and even 5 (on 1,2,5, the drive on the left disappears into the window, on 3 and 4 it doesn't).

No matter which of these is the a7s, I think this test shows that it has a hell of a lot of DR, as much as its big brothers using slog2 and internal recording.
From what I see here, with internal recording (10:00 onwards), the F3 has two stops better DR than the C300. The a7s seems to match that. I don't need to know how many stops it is, it is awesome.

Samuel you should be able to measure the rolling shutter at around 15:00
I tried, but the numbers make no sense. It is handheld, so the camera tilts sideways as he pans. The camera has to be on a tripod for my measurements to have any kind of accuracy.
 
Last edited:
That is one very nice test. Pity that we won't get the which-is-which until next week.

Yes, that is just stupid. I'm sure it's fun for him, and I guess he's doing it because they are pretty close, but come on we are looking for information to help us decide wether to buy this camera or not. So I'd rather he give me all the information he's got, rather than the riddles.
 
Last edited:
I don't get it, what's the point of S-log2 3200 ISO limit? The whole point of using S-log is to get the most dynamic range out of the camera and the the dynamic range maxes out at 100 ISO.

The 3200 s-log2 is odd...Den Lennie lists all of his s-log2 shots as 1600 in the "Arbroath Smokies" short...

What's the surprise here? Doesn't most, if not all, Sony cameras that shoot S-Log2 have a rather high ISO for that? FS700 is 2000, F55 1250? Should be no surprise then that the A7S, which is more light sensitive have an even higher ISO...

If DR is good enough nothing should stop you from exposing s-log at 1600 even though it is nominally 3200.
You will probably want to do this for clean blacks anyway, and this is something people do with s-log on other cameras all the time.

It's possible that's just the gain setting with the lowest read noise and best DR and everything else is inferior. Wouldn't be that surprising since slog2 spec has about 7stops of DR over it's middle grey value as well, so rated ISO has to be high.

About slog2 base ISO being 1600 then 3200:
I bet it's exactly the same as it was and only the name changed.

Is it possible that it's not actually ISO 3200 (or even 1600) but actually a lower ISO that's rated/named higher to force you to under expose (giving greater highlight latitude) but with a digital shadow and mid boost that makes the perceived brightness inline with ISO 3200? So its basically a super "highlight tone priority" mode?
This seems more likely to me because if it was a genuine analogue gain ISO Boost it would only loose highlight latitude at a given exposure.
 
Last edited:
To my eyes the colour response and highlight rolloff looks far more pleasing and natural than its nearest competitor (GH4). And the way people talk about 8bit vs 10bit makes me wonder how we graded ANYTHING this time last year. It *would* have been really great to have 10bit (Sony you missed a marketing trick vs the GH4 which doesn't even really need to record 10bit because its image is so noisy that it's got baked-in 8bit dithering! ;-) )... but I'm looking forward to putting it up against my C300 to see which one becomes my favourite. I think the 800% profile is going to be my go-to profile (just like wide-DR on the C300).

I thought the A7S examples looked better than the GH4 examples, however after using the GH4 for a couple weeks it works really well and looks very good: http://youtu.be/4CVADWamp0k?t=17s (watch in 4K- something not possible with A7S (until around September with the Shogun). The GH4 noise quality is actually pretty nice- a fine grain and mostly monochromatic, and may indeed act like dither, which is a good thing. See any noise issues?)

When comparing A7S 4K to GH4 4K and recording in the same way, externally, the GH4 provides 10-bit, along with the 'dither' means '12-bit-like' performance in post ;). On a serious note, they are both excellent cameras, especially for the price. A note on price: for the GH4 I purchased 3 lenses in m43 so as not to have to deal with adapters and to also get fast autofocus (the Panasonic 12-35 F2.8 + 35-100 F2.8, and the Voigtlander 25 F.95 (manual, no IS)). For the A7S I ordered a Metabones (version 4) so I can use my Canon lenses (but no real autofocus for stills). I can also use the SELP18200 Sony lens from the FS700 on the A7S (in APS-C crop mode): fast AF + IS & power zoom (slow, but useful). Depending on what lenses you have, the A7S could be lower cost vs. the GH4. 14-bit 5D3 RAW is still the image king IMO, but with only 12mins per 64GB, it's use is limited, especially when doing many takes (solvable with on-set off-loading, but extra time & gear, etc.). Disk storage is really pretty cheap, so for a serious project the drawbacks of 5D3 RAW aren't a big deal (other than time). Converting 5D3 RAW to CinemaDNG is super fast, after which the files can be directly edited in PPro in real-time. With the latest PPro CC update, it's now possible to do basic RAW adjustments within PPro. Not as fancy as ACR in AE or PS, but a step in the right direction (ultimate quality is ACR+AE).

There is no all-around camera yet- each has strengths and issues. The new ARRI Amira looks excellent, however it's quite large & heavy (a good thing for stable image), and it's priced similarly to a RED Epic Dragon (which is actually pretty nice, color & image-wise: a move in the right direction from RED toward ARRI). All of these new cameras can produce quality good enough for just about everything seen on the Web and TV, even the big screen: GH4 4K and 5D3 RAW 2K looked great projected on the big screen. The A7S will work on the big screen too.
 
ok, let's see...

My guess is that camera 2 is the a7s, because of the colors, the DoF, and how clean the shadows are. As long as it is not camera 4, which has more noise in the shadows than the others, I'll be a very, very, very happy camper. Hell, I'll even be hapy if the a7s is camera 4, it is pretty close to the big boys. Looking good.
(Camera 5 must be the F55 with slog3, it is head and shoulders above the others)

edit:
Looking at this again, I'll definitely be happy with camera 4 too, even if it has noisier shadows: it has more headroom into the highlights than cameras 1, 2 and even 5 (on 1,2,5, the drive on the left disappears into the window, on 3 and 4 it doesn't).

No matter which of these is the a7s, I think this test shows that it has a hell of a lot of DR, as much as its big brothers using slog2 and internal recording.
From what I see here, with internal recording (10:00 onwards), the F3 has two stops better DR than the C300. The a7s seems to match that. I don't need to know how many stops it is, it is awesome.

Look at the wire on the robot's head (artifacts) and the left side of the robot (relative to camera view) for the highlight => black pixel issue. The black pixels are likely FS700 or A7S (someone mentioned they were present too on the A7S). If the A7S has the black pixel issue, curious if that's something there for business reasons (as it may be for the FS700, since with the 7Q the black pixel issue goes away- not a hardware design issue).
 
I think the bugaboo on all these DSLR's is rolling shutter.
Frankly it still is a dead give away. For many art directors it screams "low budget"
Until the DSLR world gets faster read times or global shutter, the playing field pretty much remains the same.
 
yoclay, completely disagree. For 99% of shoots, rolling shutter is a minor issue that only crops up occasionally, and rarer still to be at the point that it screams "low budget"
 
I think the bugaboo on all these DSLR's is rolling shutter.
Frankly it still is a dead give away. For many art directors it screams "low budget"
Until the DSLR world gets faster read times or global shutter, the playing field pretty much remains the same.

I haven't had any issues with the rolling shutter on the MK3 (I never shoot handheld, should rig or steadicam always). As long as the A7s rolling shutter isn't any worse than the MK3 I'll be happy with it. If the punters notice the rolling shutter the rolling shutter isn't the problem.

The only thing that really concerns me is this bloody 3200 ISO S-log thing. The APS-C aliasing sux but that's not something I really need but would have been a nice option.

@jcs On a typical narrative shoot day I fill up 5x 64Gb CF cards (15min per card 2.39:1 crop). 1000x cards have dropped in price a lot since the MK3 went raw and I figured safer to just pocket the cards and offload at the end of the day to a raid 5 and tape. I'm buying the A7s to determine whether the IQ (recorded to a pix 220 prores4444) is good enough to replace the MK3 raw so I can save time in post.

There's been a few occasions where I couldn't get the shot I wanted with the MK3 like an actor on a dimly lit subway platform in the foreground where I wanted the background only slightly out of focus. To get the shot I'd have to push the ISO too high on the MK3 so I could stop down enough to get the DOF right. The A7s should easily be able to nail those kind of shots where you can't light. I have to shoot a night scene in a quarry, the A7s will be worth it for that scene alone. Instead of a bunch of blondies and a generator I should be able to light it with battery powered LEDs.
 
Look, it's either you carry some ND filters or a bunch of lights. slog2 (3200) is kind of odd but hey if I can slap on an nd filter, no biggie. I rather have that than to carry around extra lights for other cams that are not great at low lights. I also think that Sony will update the firmware back to 1600 or less in due time. Maybe they didn't want to upset the higher end owners right off the bat?
 
NDs and lights aren't a problem. What I wanna know: is S-log @ 3200 ISO going to give me as much shadow detail as another profile @ 100, 400, or 800 ISO? And how much dynamic range are we losing with S-log @ 3200 ISO compared to S-log @ 1600 ISO? Den's stuff shot @ 1600 ISO S-log looked great and that's the camera I thought I was getting. Sony have some explaining to do.

Screen Shot 2014-06-22 at 3.59.31 pm.jpg
 
Look at the wire on the robot's head (artifacts) and the left side of the robot (relative to camera view) for the highlight => black pixel issue. The black pixels are likely FS700 or A7S (someone mentioned they were present too on the A7S). If the A7S has the black pixel issue, curious if that's something there for business reasons (as it may be for the FS700, since with the 7Q the black pixel issue goes away- not a hardware design issue).

OK, looking at those wires now...

They are fine on cameras 1, 3, 4. They show aliasing in cameras 2 and 5. Still not a big issue for me, I think.
But I believe this rules out my theory that camera 5 is F55 with slog3 (because it should show the same aliasing in slog2 and slog3). And if it can only be the FS700 or the a7s, then I need to know if the DR in the FS700 is clearly better than that in the F3 or F55 in terms of DR, because if not... Except, it just can't be, it would be too great. Plus I don't really see that kind of highlight issues in any other clip shot with the a7s in FF mode and with XAVC-S.
I think I'm already overthinking this.
I've already commented twice in the video asking him to give us the results sooner rather than later.
 
NDs and lights aren't a problem. What I wanna know: is S-log @ 3200 ISO going to give me as much shadow detail as another profile @ 100, 400, or 800 ISO? And how much dynamic range are we losing with S-log @ 3200 ISO compared to S-log @ 1600 ISO? Den's stuff shot @ 1600 ISO S-log looked great and that's the camera I thought I was getting. Sony have some explaining to do.

I don't think that is going to be an issue. I believe the ISO 3200 thing is just a rating: the sensor is working with minimum analog gain (that would be ISO 100 in the dxomark graph), the profile saves as much info from the sensor as possible, and Sony just tells us an ISO number so that we set our aperture+shutter+ND in a way that mid gray falls where they want us to put it.
With log, as with RAW, you can always overexpose or underexpose, and all you are doing is redistributing the DR across the image (so, if you have 6+ stops above and 6- below at ISO 3200, you'll have 3+ above and 9- below if you leave the camera as-is and just think of it as being set at ISO 400).

This is what I will do when I get the camera: I'll bring it out on a bright sunny day, and I will shoot things with slog2 and ISO 3200 and a fast shutter (probably 1/3200s, which is 6 stops faster than 1/50s). I will try to find the widest aperture that doesn't give me any unwanted clipping in any channel (so: a sun reflection in a car can clip, but faces can't, clouds can't, white or bright-red or bright-green wardrobe can't, etc). I expect that will be f/5.6. If that is the case, it means I need 9 stops of ND to shoot f/2.0 1/50s ISO 3200 slog2, and that's the widest I will ever shoot (I only have one f/1.4 lens and it is soft wide open), so I will buy the Hoya proND 500x.
I had planned to do this anyway (it's time for me to get better ND filters), it's just that I'm going to need a stronger filter than I anticipated, and some in-betweens to have some flexibility (I'm not going to be shooting f/2 all the time). This adds some small $$$ to the cost of the camera. It's already starting to build up, but I'm happy with the image.
 
Last edited:
In one of the interviews Atomos CEO mentioned that Sony contacted them because after 5-6 minutes of 4K recording A7S was overheating and they decided to go external 4K recording. I WISH they would leave that in and limit recordings, let's say, to 4 minutes. All that together would make a decision so easier to pick it up over GH4... Well, but maybe some hack? :)

Anyway...
XAVS S in A7S is 50mbs, but it's 4:2:0? So not broadcast approved? Ehhh.... if it would record broadcast ready format internally in full hd that for me would be probably much bigger functionality then all that 4k...
 
Back
Top