Sony A6300 announced!! 4K internal! S-log3!! 1080 120fps! 4K for the masses :-P ;-)

[Wayne Morellini]

how are they getting 4k from 24 mpix sensor with no pixel binning? Line skipping haha? This is nonsense...
I see they claim real time down conversion... impressive if true...
This should be the sharpest camera on the market then...

Sensor 101, they just down scale which is probably better than native Bayer and binning. In binnng there are issues that short out the opportunity to debayer for better color first. So they probably debsyer and down scale. If you think it should be the sharpest camera on.the market thren you dont understand the consumer camera business enough, the business takes opportunity not to.ptesent tje brst quality possible, hunce no 4:2:2/4:4:4/bayer 10 bit plus at 200+ mb/s.

Maybe you were trying to say you proferred pixel ninninh instead, but as above, this is inferior and of limitrd low light advantage, as you cannonly really bin the same colored pixels togrlether fir increasef accurscy or some complementary or hybrid scheme. Then with binning you can get pixel fill issues to we're details can make the binned pixel.crawl in movement. But there is another vital issue,except some odd irregular bin/averaging, 4k does not fit evenly into this resolution, so you can't group 4 pixels say to produce a 4k pixel, but would have to stripple yjdm or something. A nightmare. So 32mp+ with 8k+ horizontal resolution is needed. There is one more problem, but I have forgotten it.

 

[Wayne Morellini]

Now for everybody else. Sure most of the really mentally talented individuals have sharp edges, they often get bent out of shape by constslant ill treatment by knuckle draggers. We should be appreciative of them, because without them we are more likely to be in caves picking berries. I know which life I prefer. Supporting them is a small price to pay for the truely great benefits they bring to most levels of society.

 

[Karlis Korbens]

Thanks Tim !
It's a pretty big mix of Cameras, lots of Epic and Epic Dragon in there -

Most recent Walkman - Walkman w410 https://vimeo.com/151730293 was on the A7R2 using Cooke Anamorphics. Pretty silly having the tiny camera attached to those huge lenses, but using 16X9 FF I got a 4k sensor that was 20.3mm x 36mm wide - Crop that a bit at the edges and it was basically a 3k 4:3 ... worked nicely with the Cookes.

Bruce Allen helped me figure it all out and loaned me his Odyssey 7Q to monitor with.

Matt Uhry
www.mattuhry.com

Liked your video. I'm interested about your prefered picture profile settings.

 

[Matt Uhry]

Liked your video. I'm interested about your prefered picture profile settings.

I'm pretty much always using PP4 with Cine Color and Cine Gamma, It's a bit flat and desaturated, but gives you a good place to color towards your end result. S-Log was just a bit to flat for me to be happy looking through the EVF and you can get some banding on gradients if you grade a bit on the hard side. We should get Bruce Allen to weigh in on this - He has a A7R2 now and did some legit testing with it.

Matt Uhry
www.mattuhry.com

 

[Jacek Zakowicz]

Sensor 101, they just down scale which is probably better than native Bayer and binning.

I'm not sure it's sensor 101 actually- their images would be pristine- better than Red.
which they are obviously not.
The fact that they are not gives two options
a- they are lying (unlikely)
b- they are doing really sh*&ty job with the downscale.
In either case the statement about downscaling from the whole sensor is irrelevant for all practical purposes because it does not produce results to be expected from that statement.
May as well be binning or line skipping since it would produce eqal results.
But it makes a catchy marketing line as this thread shows...
I hope I'm making myself clear now....
And please no emotional outbursts and personal attacks- this is strictly technical....

 

[Bruce Allen]

I'm not sure it's sensor 101 actually- their images would be pristine- better than Red.
which they are obviously not.
The fact that they are not gives two options
a- they are lying (unlikely)
b- they are doing really sh*&ty job with the downscale.
In either case the statement about downscaling from the whole sensor is irrelevant for all practical purposes because it does not produce results to be expected from that statement.
May as well be binning or line skipping since it would produce eqal results.
But it makes a catchy marketing line as this thread shows...
I hope I'm making myself clear now....
And please no emotional outbursts and personal attacks- this is strictly technical....

Jacek, you're starting from the wrong conclusion - that the downscale is bad - without testing.

I actually tested the 6K -> 4K downscale modes on the a7r2 and NX1, and tested the a7r2 against Matt Uhry's Epic Dragon. The a7r2 was sharper than Epic Dragon in 4K mode.

Put the Dragon up to 5K mode and run at 4:1 compression and you have a closer fight. But then of course you can also run the DSLRs into an external recorder!

Colors on the Dragon with Skin Tone Highlight OLPF were better of course. Latitude was... surprisingly similar, actually!

Finishing up a project. After I'm done I'll put up some results!

Bruce Allen
www.boacinema.com

 

[Jacek Zakowicz]

Bruce please do post some.
However- Sony is downscaling from 6K 16:9 sensor to 4K output on HDMI (?)
It should be actually sharper on the live out than Red in 6K recorded and outputted in 4K because all things being equal you are not compressing with sony and you are with Red.
I think that's logical...

 

[Wayne Morellini]

I'm not sure it's sensor 101 actually- their images would be pristine- better than Red.
which they are obviously not.
The fact that they are not gives two options
a- they are lying (unlikely)
b- they are doing really sh*&ty job with the downscale.
In either case the statement about downscaling from the whole sensor is irrelevant for all practical purposes because it does not produce results to be expected from that statement.
May as well be binning or line skipping since it would produce eqal results.
But it makes a catchy marketing line as this thread shows...
I hope I'm making myself clear now....
And please no emotional outbursts and personal attacks- this is strictly technical....

No it's not strictly technical. What I said is. It is a basic thing and completely relevanthat ou are accessing something as wrong onna faulty basis. Binning Bayer is difficult and unless somebody has a way to combine debayering in the binnjng, low quality, and the little performance advantage of binning gets compromised by suchba circuit unless the circuit is something more that (which I'm not getting into). Binning is only tjenbrst on multiple chip/layered such as in 4:4:4, or in.monochrome sensors.

There are so many processes of downgrading the image you vsnt just blame.the downscaling, it is like rpulet,you might be right one time out x. But I can tell you what to do, test a 4k wide and a native resolution shots in.still mode. If the native resolution is good then we can stop.blaming the low pass filter, if the 4k native still is good we can stop blaming downscaling, as it obviously can work. We then can start looking at the video chain itself. Maybe there is a low grade hardware debayering algorithm being used in the video downscaler (such as you might get if you use a cheap third part chipset design) maybe it some other part of the chain. But what does get people is that images that aren't processed to minimize the diffusion of the Bayer olpf, or shsprned may appear soft, but we are used to seeing it sharpened up, and contrasted up. You might be seeing a gstch job, or you might be seeing a more authentic image, you ha e to have a look. Another thing done is crafting the olpf to get better resolution, which Red did, and Graeme cleaned up the rest. A normal.olpg cam.be so bad as to effect the image sharpness on a 2:1 downscale. I.also forgot to mention that dowscsles that are not even suffer image pixel boundaries falling accross recorded pixel.boundaries, so that a colored pixel is hated accris multiple target pixels, a reason I eas not in favor of diwnscslling so much years ago, so that has to be compensated for.

 

[David Peterson]

Two questions:

Better than the 5Dmk2 for stills and video? I still love the image out of the 5D2 but need to upgrade as mine is a little worse for wear. Seems like a nice cheap option, but want to keep my L lenses and maybe use some Leicas.

Undoubtedly better than a 5Dmk2 for stills, and *HUGELY* better than a 5Dmk2 for video. (heck even a cheap and older Nikon D5200 is better than a 5Dmk2 for video!)

 

[Mike P.]

Did no one post this A6300 video? They clearly say in the description that there's no pixel binning and that they oversample the entire sensor ('2.4x the amount of pixels required for 3840x2160')... Videoy as shit, but that's probably because it's 30p and a straight-up video-styled colour correction (similar to most japanese tv I've seen).

Still don't know about rolling shutter though.

Undoubtedly better than a 5Dmk2 for stills, and *HUGELY* better than a 5Dmk2 for video. (heck even a cheap and older Nikon D5200 is better than a 5Dmk2 for video!)

Bare in mind the A6300 be ~s35, not FF35, which in my opinion, makes the 5Dmk2 better for stills... FF35 always looks better spatially (not just FOV, but FOV and DOF and how objects look within those values.)

 

[Wayne Morellini]

Yeah, that's why I wanted to design a medium format multi sensor camera. Apparently Apple is patenting an Iphone with multiple sensors for this, but mathematically you need more than two tiny plain lens to do this. This type is not as good performance as a single large sensor at the moment, but that is this type. It is not so much a race to the bottom as Jim said, but a race to the top. Eventually 90%+ of the market could be covered by this sort of thing.

 

[Karim D. Ghantous]

Yeah, that's why I wanted to design a medium format multi sensor camera.

Can you be a bit more specific? Is this sort of like what Light did with the L16 (not released yet)?

I had an idea which I suggested a few years ago for the RED. It was basically a different (but more costly) way to do HDR. You have two sensors, and one receives the images via a semi-silvered prism. But its sensitivity is several stops lower than the main capture sensor, so that you could blow out the highlights on the main sensor and simply replace them with the information from the secondary sensor.

Another idea was to place a transparent LCD (with the same number of pixels as the sensor) right on top of the sensor. The sensor would relay the highlights to the transparent LCD, which would become darker in the areas where the highlights were strong.

All of these ideas are nice but I think manufacturers want to keep things as simple as possible. Sony has a novel way to increase DR by simply having every other group of two lines on the sensor at a lower sensitivity. Simple? Yes. But possibly quite effective.

 

[Wayne Morellini]

This multi lens/sensor computational imaging configuration is an old concept. The Light concept is rather more elaborate. If you look at my question sometime back to Red on which firmware technique update they maybe using to get single lens 3D, I list many techniques I had been working on, including similar to what Lighf is now trying to do. Mine was regular spaced, until I could get to.develop the technology I really wanted.

The ideas you mentioned and the issue is older. I suggested the multiple exposure hdr as a convenient fill in technology to Red, until they could make a hdr sensor. Back last decade I can up with many ideas to induce hdr, including a sensor LCD overlay, electrical and photochromatic, which I think somebody has done now. But I realized that withinh a few years sensor performance would get so good it couldn't really compete, so abandoned it. With in a few years I could buy 16stop+ cameras for under $100, including HD.

With a computational camera it becomes simpler/cheaper than a 70mm medium.format camera, you just loose some light and some sensor performance. Sensors are becoming better to.the extent that 90% or so of filming situations could be handled good enough, filling people around in.the pitch black, maybe maybe not.

 

[Karim D. Ghantous]

You know, the only digital technology that can kill film is the Lytro. I have looked at the Illum with fresh eyes recently and it's better than I used to believe. The problem is not resolution but light field artefacts which occur under certain conditions. No need for focus stacking anymore!

 

[Daniel Kelly Brown]

In either event, the result looks pretty darn good. Not an amazing jump (probably their algorithms aren't the highest quality, seeing as they have to scale 6K->4K up to 30x per second without killing the tiny battery / overheating!) but certainly a good-looking image.

Bruce Allen
www.boacinema.com

Mildly interesting- I have noticed that the A7sII does not allow for recording with screen at sunlight level while shooting 120fps HD (drops screen brightness down even when sunlight mode is selected). I figure this is another way they avoid overheating issues.

 

[Wayne Morellini]

You know, the only digital technology that can kill film is the Lytro. I have looked at the Illum with fresh eyes recently and it's better than I used to believe. The problem is not resolution but light field artefacts which occur under certain conditions. No need for focus stacking anymore!

Lytro is a real issue, you have to make it big to.get the light into.those tiny pixels, and read them out at video. It is also far more than needed. On my original design I considered 3 sensors as a minimum, because 2 you can't normally see under or over a plane to make a correct unidirectional image and image blur. Very simple, lost on some ideas. I settle for 5 for a very simple reason, but I had something else in preference. 9 lens seems to the better number (stacked on one sensor) but I don't want to go that far. I'm actually been very long term plan single lens/sensor system, based on theoretical work decades ago. But that sort of custom.design costs money.

 

[Jim Skinner]

Just wanted to give an update for the curious. My company does a lot of non-profit work, and we were hoping the A6300 would be the perfect camera to put in the hands of the young filmmakers we work with. The slog footage, however, is pretty horrendous. It cannot stand much of a grade at all. It is significantly worse than the same slog profile used with A7Rii or A7Sii. The sensor also has a much cheaper coating than the A7 line, and of the 3 cameras we tested each had some kind of white goo on the sensor. It was only visible when shooting slog (which forces a high ISO, and thus a very sensitive image plane). I've never seen anything like this with the A7 line.

The A6300 is essentially trying to win the market on specs (4K!!!) with no attention to quality. This is why we LOVE Red and look forward to adding to our Red collection this year.

 

[Samir Patel]

I have one. I primarily use it for fast action stills with adapted EF lenses (using an old adaptor so manual focus at the moment). It is simply a camera that is small enough to go some places where you don't want to take a bulky DSLR. Rolling shutter is horrendous on video but I would never use it to shoot a feature. Just need it for occasional video clips (in between shooting stills). So in that regards. I'm good. I'm not shooting in slog. Just trying to use a baked in profile I find the colors pleasing enough on.

 

[Patrick Tresch]

Are you using SLog-2 or 3?

 

[Emir Aksay]

Its a great gimbal B-cam..