Sony announced new A7r mk2, RX10mk2 and RX100mk4 all with internal 4k

[David Heath]

That sounds as if you're equating "very sharp" with "video look"?

I know where you're coming from, but I don't think it's true any more.

It's almost the opposite. Go back about 20 years - when HD (let alone 4K! :-) ) was a dream for most - and the basic problem with video images was LACK of real definition, and it became the norm to wind in large amounts of detail enhancement to try to give an illusion of sharpness. Think obvious edges round objects. And THAT is what I'd say is the real objectionable side to "video look".

But the sharper the image really is, the less need for electronic sharpening, and indeed many cameras now have an ability for negative enhancement. You can always make a picture softer, but not always the opposite!

[Larry Secrest]

True, we can always make something softer. I think for me there is something I simply can't define that tells me something is video or at least not film.
I used to be biased against the Black Magic Cams and now I realize it's simply because there is very few nicely graded footage out there.
Since them I've stumbled on a few decently graded footage, from all 3 current smaller BMC and I must say there is something amazing, for the price, in all the BMC footage I've seen that I don't see in the NX1. The more I think about it and the more I'm convinced that it must have something to do with the color sampling. Shooting raw or ProRes 4:4:4 is going to allow more latitude in post and I think this is what I see lacking in all the mirrorless cam such as NX1, GH4 etc...
I'm slowly confirming something I had always felt in my gut, a very good color space, provided some decent res is there, is much much better than just razor sharp definition. It's personal and very subjective.

[Betsy Moore]

Originally Posted by David Heath

I think you may be being unnecessarily pessimistic, but I'll confess I shall have to make some assumptions about what is going on.

I've thought for a while that about 42Mp is a "magic" figure for sensor resolution and the reason why is an extension of the technique Canon used for sensor read with the C300. Simply, in FF mode it can theoretically do the same trick as the C300 used to derive HD - read the sensor in 2x2 Bayer blocks, and derive each output pixels R,G,B values directly - no conventional deBayering.

For the C300 to do this for HD it needed 4x1920x1080 photosites (about 8.3 million). To do the same for 4K (or QHD at any rate) would mean 4x3840x2160 - about 33.2 million. That's for 16:9 - adjust for 3:2 (cropping to get 16:9) and it brings it to a total of about 39.3 million effective. Add in extra pixels for the surround and hey presto we're getting very close to the quoted count.

Are these assumptions correct? I don't know for sure, but they certainly make me optimistic! If - IF - it is true, then far from "massive line skipping" it could open the way to a highly clean 4K output achieved with fairly simple processing.

And more - by reading 4x4 blocks - it should give very good HD output, again with fairly simple processing (here averaging R,G,B values within each block).

And it's a similar thought regarding sensitivity - although each photosite may be relatively small, the averaging process will counteract that and you can expect to get back to square one. That should also hold true in s35 mode, here the averaging happening via the scaling process.

Yes, there are a lot of assumptions there - but I certainly see enough to make me very, very optimistic. Only proper chart tests will definitively show if it is indeed using the 2x2 read technique as seen (for HD) in the C300, but the numbers look right......

I certainly hope you're right and the RGB approach has precedent in HD cameras like I believe the Genesis. That said, Sony's own press releases seem to downplay the full frame quality--really pushing the super 35 aspect and shooting the first promotional 4k video exclusively in that mode and bragging on charts that only the crop mode had no "pixel binning", as does Philip Bloom's first hands on impressions. Maybe the tiny body and economic restrictions didn't allow for a more elegant use of all those great pixels. Now, for the crop mode, they're obviously doing something clever to combine those pixels from 1.8 to 1 and that's exciting.

I read today that the native iso in s-log mode is 800 so I'm sure it can be happily stretched a few more iso's at least. 1600 should be easy--wonder if 3200 is asking too much. And I wonder what the native iso is for the crop mode.

As to that, does the metabones adapter in crop mode increase the available light the way it does on other cameras? That could be very exciting in and of its self.

[Charles W. Hull]

Originally Posted by Betsy Moore

As to that, does the metabones adapter in crop mode increase the available light the way it does on other cameras? That could be very exciting in and of its self.

Metabones makes 1:1 adapters, and then the "Speed Boosters" you are referring to. All of the excitement about the a7RII focusing so well with Canon lenses is with the 1:1 adapters (the excitement is because earlier Sony Alpha cameras did not focus very fast with Metabones.) In principle the Speed Boosters should work and be great for Super 35 video, but I don't think anyone has demonstrated that yet.

[David Heath]

Originally Posted by Betsy Moore

Sony's own press releases seem to downplay the full frame quality--really pushing the super 35 aspect and shooting the first promotional 4k video exclusively in that mode and bragging on charts that only the crop mode had no "pixel binning", as does Philip Bloom's first hands on impressions. Maybe the tiny body and economic restrictions didn't allow for a more elegant use of all those great pixels.

I think you may be looking at that the wrong way. If they are doing a "2x2" readout ("direct read") of a 7680x4320 photosite matrix then it makes it EASIER for the "tiny body and economic restrictions" to do the processing - you're "just" reading out values directly for the R,G,B of each output pixel.

That is much, much easier than the more conventional way of doing a conventional deBayer, and then having to do complicated scaling - as easy as line skipping methods have been, but without the quality issues they have been associated with. That's the beauty - theoretically, if it is what it's doing! :-)

But it's only possible with exactly the right number of photosites - not too many, not too few, that's the point. And why I've been wondering when we were going to get exactly such a 40Mp sensor (3:2 - cropping to 33Mp for 16:9)

The point about super35 is that it's the basic video large format standard, so an ability to do it well is important as well as the full frame abilities. But such a crop from such a sensor means you are now nowhere near any of these "magic" numbers, so the only way to get quality now is to do the conventional deBayer and scaling. Well, only from about half the total sensor count, but likely still a lot of processing and that's why I think Sony are flagging up that it's not pixel binning - not for anything it implies about the FF mode.

It needs proper tests to be certain - but well, from the number count I say again I'm optimistic!

Originally Posted by Betsy Moore

Now, for the crop mode, they're obviously doing something clever to combine those pixels from 1.8 to 1 and that's exciting.

Yes, but "complicated" may describe it better! :-) If they are doing 2x2 read in FF it's nowhere near as clever in terms of processing, but actually is more exciting!

Originally Posted by Betsy Moore

I read today that the native iso in s-log mode is 800 so I'm sure it can be happily stretched a few more iso's at least. 1600 should be easy--wonder if 3200 is asking too much. And I wonder what the native iso is for the crop mode.

It's often said that sensitivity is dependent on photosite size - which is true to a point, but needs a lot of caveats. In particular of the "all else equal" type, and that is what often gets misunderstood.

If you keep sensor size the same, but reduce the size of the photosites, it must follow that the resolution will increase.

It further follows that whilst each photosite may then be noisier at a given illumination, it's effect on a given output image will be smaller in area, so the effects tend to cancel out for a given size of output picture.

It's a bit like comparing two film cameras with differing negative sizes. Which will be the best in low light? DIRECTLY speaking, it may seem a silly question - as the answer may seem to be as simple as "it's irrelevant - it only depends on the film ISO".

But think a bit deeper. If we're interested in a given level of grain for acceptability, then it follows that the larger format can tolerate a higher ISO film stock than the smaller format before grain becomes a problem. So negative size DOES have a bearing on low light capability, yes? :-)

Originally Posted by Betsy Moore

As to that, does the metabones adapter in crop mode increase the available light the way it does on other cameras? That could be very exciting in and of its self.

The adaptor I think you are referring to does NOT "increase the available light" - that would be impossible according to the laws of conservation of energy! What it DOES do is make for a smaller - but brighter - image. Same AMOUNT of light, but differently distributed, which is a very different matter indeed.

(And practically, if we assume twice the brightness, but half the size, then for corresponding sensors you can expect very comparable performance. Halving sensor size - with all else equal - is likely to reduce the native ISO rating by about a stop. But since the image is twice as bright it brings things back to square one - same performance.

The oft made mistake is to focus solely on the way one variable changes, and ignore other variables. "Amount of light" and "brightness" are related, but are *NOT* the same thing.)

[Betsy Moore]

Originally Posted by Charles W. Hull

Metabones makes 1:1 adapters, and then the "Speed Boosters" you are referring to. All of the excitement about the a7RII focusing so well with Canon lenses is with the 1:1 adapters (the excitement is because earlier Sony Alpha cameras did not focus very fast with Metabones.) In principle the Speed Boosters should work and be great for Super 35 video, but I don't think anyone has demonstrated that yet.

Charles, do you imagine the AF will work with the Speedbooster?

[Joe Ogiba]

For 4K UHD in Super35 mode the Sigma 18-35mm f/1.8 DC HSM Art Lens for Sony A mount with A to E mount adapter looks interesting. Maybe Metabones will come out with a Speedbooster A-E mount adapter for lenses like the new full frame Tamron SP 15-30mm f/2.8 Di USD Lens (Sony A).

[David Heath]

But a Speedbooster is designed to use a full-frame lens with a s35 sensor - that's fair enough if you have a camera with that size sensor and full frame lenses. It'll keep the lens angle as when used on a a full frame camera, and won't waste any light.

But if your camera has a full-frame 4K mode, then what's the point of the Speedbooster!? For a given full-frame lens, then why not simply just put it on the camera and use that in full-frame mode?

To answer my own question, then it may be valid if the s35 crop mode was noticeably better than the full-frame mode, but for all the reasons above I strongly suspect that will not be the case. But OK, let's wait for proper tests.

[Michael Kraus]

Originally Posted by David Heath

But a Speedbooster is designed to use a full-frame lens with a s35 sensor - that's fair enough if you have a camera with that size sensor and full frame lenses. It'll keep the lens angle as when used on a a full frame camera, and won't waste any light.

But if your camera has a full-frame 4K mode, then what's the point of the Speedbooster!? For a given full-frame lens, then why not simply just put it on the camera and use that in full-frame mode?

To answer my own question, then it may be valid if the s35 crop mode was noticeably better than the full-frame mode, but for all the reasons above I strongly suspect that will not be the case. But OK, let's wait for proper tests.

Sony is advertising the S35 mode on this camera as advantageous. No pixel-binning in that mode means less weird stuff happening during processing that produces the moire, aliasing, and softness that plagues most of our DSLRs. I can see why someone might want to use a speedbooster rather than a standard adapter for this camera. I'm pretty sure all of the video they have released thus far was shot in S35 mode.

That being said I'll still probably go with a standard adapter. I just love the full frame aesthetic too much to abandon it all together.

[David Heath]

Originally Posted by Michael Kraus

Sony is advertising the S35 mode on this camera as advantageous. No pixel-binning in that mode means .......

Unless you've seen anything else, the only thing I've seen from Sony on the subject is :

The impressive video credentials of Sony’s new α7R II camera include the ability to record movies in 4K quality (QFHD 3840x2160) in either Super 35mm crop mode or full-frame mode.

In Super 35mm mode, the camera collects a wealth of information from approximately 1.8x as many pixels as 4K by using full pixel readout without pixel binning and oversamples the information to produce 4K movies with minimal moire and ‘jaggies’.

Which praises the s35 mode, yes, but is certainly not saying it's advantageous over the full-frame mode. It's making a comment about the s35 mode but not saying if the final quality is worse, better or the same as full-frame.

It's being INTERPRETED as meaning "better" because we've got used to the idea of a very large number of photosites will mean it will have to pixel skip or bin. By and large that may be true, but once you get up to a count of 4x3840x2160 that theory all goes out the window for the reasons I gave earlier. ("2x2" block read, or direct read etc.)

We'll have to wait to see what actual results show, but I think dismissing the full-frame 4k mode and assuming speed-boosters will be necessary is very premature. But let me know if I've missed anything Sony have said.

[Michael Kraus]

Originally Posted by David Heath

We'll have to wait to see what actual results show, but I think dismissing the full-frame 4k mode and assuming speed-boosters will be necessary is very premature. But let me know if I've missed anything Sony have said.

To be clear, I certainly was not dismissive of the full-frame mode and even indicated that I'd still likely prefer it over S35. I still think Sony is making it sound like S35 mode will have less moire and processing artifacts. But yes agreed time will tell. I wish they'd hurry up and release some full frame images.

[Charles W. Hull]

Originally Posted by Betsy Moore

Charles, do you imagine the AF will work with the Speedbooster?

I really don't know. I've never used a speedbooster on a Sony camera. I've used the full frame Metabones adapter with Canon lenses, and the autofocus was very slow on a A7II. But the A7RII is reported to have solved this. Whether this would apply to a speedbooster has not been discussed by Sony. More will be revealed over time. I'm personally not very interested in this, I'm not a fan of lens adapters, and I'll just use my wide angle lenses in the Super 35 mode with the A7RII. But the idea is intriguing if it comes to be.

I'm more interested in whether Sony and third party APS-C lenses will work in the Super 35 mode. There is a great selection of these lenses. I'm pretty sure this is going to work, but again, Sony hasn't discussed it.

[Betsy Moore]

Originally Posted by David Heath

Unless you've seen anything else, the only thing I've seen from Sony on the subject is :

Which praises the s35 mode, yes, but is certainly not saying it's advantageous over the full-frame mode. It's making a comment about the s35 mode but not saying if the final quality is worse, better or the same as full-frame.

It's being INTERPRETED as meaning "better" because we've got used to the idea of a very large number of photosites will mean it will have to pixel skip or bin. By and large that may be true, but once you get up to a count of 4x3840x2160 that theory all goes out the window for the reasons I gave earlier. ("2x2" block read, or direct read etc.)

We'll have to wait to see what actual results show, but I think dismissing the full-frame 4k mode and assuming speed-boosters will be necessary is very premature. But let me know if I've missed anything Sony have said.

I think what I and others have been trying to say, and it's my fault perhaps for not being clear enough, is that there are several early articles, videos, etc. that are touting the super 35 abilities and downplaying the full frame 4k abilities. Sony, bless 'em, is like most other camera manufacturers in that it allows its marketing department to sometimes painfully oversell a product and under deliver (auto focus in video mode on the a99 comes painfully to mind). If Sony has included an incredible full frame, no compromises 4k functionality, and chosen not to play up that fact at all, and hide videos of how great this mode will be and only show videos of the super 35 mode, then the execs at Sony must have been visited by 3 ghosts last Christmas and changed their ways:)

[David Heath]

Originally Posted by Betsy Moore

I think what I and others have been trying to say, and it's my fault perhaps for not being clear enough, is that there are several early articles, videos, etc. that are touting the super 35 abilities and downplaying the full frame 4k abilities.

Yes, but AFAIK those have been by people who haven't had the facilities or access to the camera to perform any real quantitative tests. Until somebody does exactly that, it's dangerous to draw too many conclusions from slight inferences in marketing press releases - which unfortunately is exactly what has been happening. On top of that, conclusions have been reached by extrapolating from other camera tech - in a fashion which is sensible on the face of it, but when you think about what the implications of a sensor with twice the desired resolution *MAY* be, those certainties go out the window.

If it is the case, the rule book has to be rewritten.

The first test that could be done to try to prove or disprove whether what I'm saying is true is to first lock the camera off, ideally looking at something like a ruler laid horizontally. Then take a still frame using full frame max resolution mode, then shoot some 4K full frame 4K video without moving the camera, and compare the angle of views of the two results.

The max resolution for stills is 7952x5304. If my theory is correct, it must use 7680x4320 if it's doing the 2x2 read for video in this mode. The implication is that if I'm right, we'll see a slight reduction in the angle of view between the still and video modes - video mode will be a bit tighter than stills. It's possible to predict how much it be which is the difference between the horizontal counts, divided by the total, and multiplied by 100 to get a percentage. That's about a 3.5% crop.

So - does that happen?

Originally Posted by Betsy Moore

Sony, bless 'em, is like most other camera manufacturers in that it allows its marketing department to sometimes painfully oversell a product and under deliver (auto focus in video mode on the a99 comes painfully to mind). If Sony has included an incredible full frame, no compromises 4k functionality, and chosen not to play up that fact at all, and hide videos of how great this mode will be and only show videos of the super 35 mode, then the execs at Sony must have been visited by 3 ghosts last Christmas and changed their ways:)

Well, no need for supernatural explanations! I can think of two other possibilities, always assuming I'm right!

Conspiracy would be that they want to keep the exact details from their competitors for as long as possible, in which case they are going to appreciate what I've said like a hole in the head!

Cock-up would be that in many organisations there is a big gap between the designers and engineers and the marketing and sales people - and it's the latter who tend to write the public facing literature. Believe me on this one! And all too often the marketing people don't fully understand what the technical people tell them, let alone there being issues with needing to condense it into a public friendly form anyway.

And in general, I normally favour cock-up over conspiracy...... And either of those are more likely than 3 ghosts coming a-visiting! :-)

[Rakesh Malik]

Originally Posted by Betsy Moore

1) Sony (almost) always Best Buy showrooms up the color and sharpness and contrast of their promotional videos so I'm not worried about the color just yet.

I'm not worried about Sony color at all: Rakesh Malik

2) Say you what we will about its video potential, this looks like a world-class stills camera--and that's its primary customer base.

That I think is the most sensible statement so far on this thread. :)

3) I'm not very technically oriented so forgive the extremely crude estimate but I'm not sure how great the low light capability in video mode will be--at all. In full frame still mode, using every pixel on the sensor, the camera is already 2 ISO stops down from the a7s. Okay, not a big problem since the a7s is a low light monster, though that already means you can't deep focus in moderate light like you can with the a7s.

I have no problems focusing with my a7r, so it wouldn't worry me.

But remember, just as the 409,600 mode on the a7s is not aesthetically usable in most situations, neither will the 102,400 iso mode on the a7rii. For most folks the a7s stops putting out a magnificent picture if it goes beyond 51,200 iso. If the same ratio applies to the a7rii, then in full frame still mode one will probably start to lose great quality beyond iso 12,800. Still impressive--but already no a7s.

I find it amusing that in most discussions about the a7s, people seem to assume that its defining characteristic is its high ISO support, which IMO should be a tertiary consideration, at most. Far more useful is its very wide dynamic range (comparable to what you get with an Alexa or a CineAlta), its very wide color gamut, and as a fringe benefit, its high sensitivity.

It gets worse though, in full frame 4k video mode the a7rii has to engage in massive line skipping. Forgetting about moire for the moment, even in 16 x 9 mode the a7rii in still mode has roughly the resolution of an 8k camera. So, crudely guessing, in video full frame 4k mode you're using 1/4th the pixels--wouldn't that reduce the light capturing ability by another 3 or 4 iso's? In full frame video wouldn't we be looking to max out at iso 1600 before quality starts to go down?

Worse? Not likely, based on the quality of what the a7s can record internally.