Re: Sony A7RII - FIRST video sample!!!
- Thread starter ectobuilder
- Start date
I think I watched the same clips. Do they indicate which frames were shot in the s35 mode and which in full frame? To the best of my concentration, that info is missing.
Not being an electronics engineer, I would assume that the FF 42 MPX readout has a ton of data tossed away immediately, which could be what creates these artifacts. The s35 has about 1/3 (1/2.5?) the data of the FF and thus the image may be "truer" to the original.
Yes, a huge auto-focus improvement and an oversampled 4K, similar to F65 (albeit using a different Bayer mask and fewer pixels).
Samuel H
Major Contributor
The pulsations I talk about are caused by the codec (either the camera's codec, or the youtube codec, or something else that may have been used in the middle). It has nothing to do with the sensor, or how much info is thrown away.
The advantages I see are the IBIS an the in-sensor PDAF. The oversampled real 4K in crop mode comes at the price of line-skipping in full frame mode, so at best I'd call it a tie.
The advantages I see are the IBIS an the in-sensor PDAF. The oversampled real 4K in crop mode comes at the price of line-skipping in full frame mode, so at best I'd call it a tie.
Well, using the process of elimination - this is the same bit rate as GH4, roughly the same codec but vastly more data (and, presumably more processing power), so when compressing from 42 MPX down to 8.3/8.8, a lot of more data needs to be discarded and/or processed compared to a native 8.8/8.3, so I can see artifacts introduced during that process too.
But, like I said, I am not an electrical engineer or a coder and I understand that online delivery will continue being the weakest link in the chain until the superfast broadband is introduced and streaming bit rates will cross into the 50-100 Mbps HEVC neighborhood.
Samuel H
Major Contributor
Sorry but I think You got an important detail pretty wrong, which serves my argument quite well: the GH4, AFAIK, records at 200 Mbps, whereas the new Sonys top at 100 Mbps.
(edit: it seems it does 100 Mpps IPB or 200 Mbps all-I, so, you're partly right, it would be similar to the sony's if you shoot IPB)
Also, on FF mode, it seems the a7R II does lineskipping, so it doesn't compress 42 Mpix to 8 Mpix, it reads 8 Mpix and saves 8 Mpix. In any case, discarding those other 36 Mpix wouldn't create this kind of artifacts. This is clearly a long-GoP kind of artifact.
(edit: it seems it does 100 Mpps IPB or 200 Mbps all-I, so, you're partly right, it would be similar to the sony's if you shoot IPB)
Also, on FF mode, it seems the a7R II does lineskipping, so it doesn't compress 42 Mpix to 8 Mpix, it reads 8 Mpix and saves 8 Mpix. In any case, discarding those other 36 Mpix wouldn't create this kind of artifacts. This is clearly a long-GoP kind of artifact.
Do we know it skips every 4th line?
Faymus Media
Well-known member
Base ISO for A7R II is 800 for Slog-2. I used it all day today. Its amazing. S35 crop means 0 rolling shutter.
That would suggest that A7R II is two stops less sensitive than the A7s, which has ISO 3200 for Slog-2.
By any chance did you test the 1080 60P? I would be interested in knowing what the picture quality is in that mode.
Samuel H
Major Contributor
I believe it only reads half the rows and half the columns, so, a fourth of the photosites. Because that leaves you a matrix on which you can do a 1-to-1 debayer to get a 4K image. But really, no idea. All I really know is that Sony says it doesn't line skip on APS-C mode, and says nothing about FF mode. From there on it's my speculation.
@FaymusMedia:
Please tell!!!
Do you see any codec issues? How is DR? How good are the high fps modes? Really, anything you can share, I'd love to hear
@FaymusMedia:
Please tell!!!
Do you see any codec issues? How is DR? How good are the high fps modes? Really, anything you can share, I'd love to hear
Have you heard of a camera skipping/binning specific photosites - i.e, fewer reds than blue or something like that - rather than a simple geometric binning? I am wondering that because I saw recently a Image Sensor World post that brought up a new Aptina patent where the red filter is larger than others.
You've probably seen it already -
http://image-sensors-world.blogspot.com/2015/06/on-semi-aptina-proposes-variable-size.html
Emanuel
Cross-Examiner
You Samuel and everyone here... Let's spill out the beans!@FaymusMedia:
Please tell!!!
Do you see any codec issues? How is DR? How good are the high fps modes? Really, anything you can share, I'd love to hear
E
ectobuilder
Veteran
Because there is no mention of a global shutter, Sony may be using this design they patented last summer:
http://www.thephoblographer.com/201...looks-eliminate-rolling-shutter/#.VX-H2M_BzRY
If true then they are reading 4 lines at a time.
http://www.thephoblographer.com/201...looks-eliminate-rolling-shutter/#.VX-H2M_BzRY
If true then they are reading 4 lines at a time.
ectobuilder
Veteran
So, if true - the sensor is read 4 lines at a time and then that information is averaged out and, if it changes several times per frame, that's what produces the artifacts.
See, I knew there had to be some info I failed to glance at. S35, however, shouldn't have the problems associated with the 42 MPX readout/pixel binning.
Let's see what the early reviews say. Bloom seems to not like the FF.
Samuel H
Major Contributor
No, it's usually just "read two rows, skip two rows, read two rows, skip two rows, etc" (what I believe the a7R II does) or "read two rows, skip four rows, read two rows, skip four rows, etc" (what I believe the 5D2 does), and things like that. The line-skipping is usually necessary because the sensor can't work fast enough for the intended fps. If the processor can't cope with the data either, instead of averaging the same-color photosites from each row, it can discard them (I believe the 5D2 was discarding, the 5D3 is averaging; no idea about the a7R II but I'm guessing it will discard; if it averages, there will be basically no moire to talk about, and aliasing will be very mild too).
About that patent, yes, I'd seen it (I read image-sensors-world almost daily), and it's very clever. They say it will help with diffraction at smaller photisite sizes, but it also helps by equalizing the amount of light each color channel gets (the red filter lets through a lot less light than the green one, that's why there's usually more noise in the red channel). But it's just a patent so far, it usually takes years for products to come out using them.
With A6000 getting a XAVC S upgrade, I wonder if a better codec or a higher recording bitrate is in the cards for A7R MK II somewhere down the road. They've obviously crippled it to protect FS7 with the 8-bit internal/external and 100 Mbps only. The rest depends on its sales figures, I suppose.
As to the asymmetrical sensor design (not that the original Bayer is symmetrical to begin with but, as a point of reference, this suffices), it seems like a fairly simple task to undertake, since it only concerns the change in the Bayer layout with everything else being rather standard. In that respect, it shouldn't be all that different from the F65 slanted Bayer mask in concept or execution.
PS. Fuji has a different Bayer filter scheme as well.