Sony A7sII Initial Impressions

For A7S II 4K XAVC-S 100Mbit/s footage on a GTX980ti and 12 3GHz 12-Core MacPro with Samsung 850Pro SSD, Premiere Pro CC 2015 can't render in real-time at full resolution. Must go to 1/2 Playback resolution, then real-time is possible. This is on Windows 10 and dual 4K monitors, will try booting into OSX to compare performance.

1/2 res i can live with. I assume if you put a single color effect it really slows down. Thanks for the info
 
2) Adapted lenses with electronic contacts that transmit the distance information and focal length data through the adaptor (like Sony A-mount lenses), also work with 5-axes stabilization by the camera. I think Canon lenses with Metabones fall into this category, but I am not sure about that. Commlite or other inexpensive adaptors don't work very well for IBIS, and you have to set your focal length manually...

This is exactly the piece of information that would be helpful for lens strategy. If using an electronic adaptor such as Metabones preserves the 5-axis stabilization for EF lenses, I'd go that route. Not only Canon, but Zeiss lenses would benefit from this.

On the other hand, if even using a metabones adapter gives just 3-axis stabilization, then the Sony IBIS is not much more stable than a stabilized lens, and canon has a plethora of those. A stabilized lens provides yaw + pitch anyways, so what's the allure of Sony's IBIS if not using Sony lenses?
 
1/2 res i can live with. I assume if you put a single color effect it really slows down. Thanks for the info

Lots of Lumetri effects as well as Mercury accelerated effects can be added and still run in real-time. It appears that 4K performance is mostly a Premiere CC software architecture issue vs. a codec issue as even ProRes 4K (very fast to decode) won't run real-time unless using 1/2 Playback resolution.
 
OSX 10.9.5, Quad core mac pro tower, NVIDIA Quadro K5000, 35G ram

So sorry, I stand corrected. Did not look before I kept (posted). I am pulling 4K into a 1080 timeline almost always. THAT is what plays back real time with no issue. You are right that 1/2 rez is the only way I get real time playback @ 4K.

Sorry about that.
 
Take a look at this page regarding Sony A7x IBIS -- it should help to clarify things:

http://briansmith.com/sony-a7ii-hands-on-review/

I already did, and it's still ambiguous:

"If the lens adapter can transmit the focal length and focus distance electronically to the A7II, the camera apply all 5-Axis sensor shift stabilization. But if the lens adapter can’t transmit this information to the camera, you must set the focal length manually for the particular focal length used."

There are two things that haven't been clarified:

1) Which lens adapters transmit focal length and focus distance electronically?
Does Metabones IV transmit focus distance accurately?
2) If you set the focal length manually, what exactly does that give you?
All five, even without focus distance? Or just three?

As of today, B&H's description:

"All 5 axes of stabilization will function at all times, even when used with third-party lenses and adapters or lenses with built-in optical stabilization."

There's a lot of misinformation there.
 
Last edited:
Thanks for that explanation Paul, and for confirming that the HDMI output is still full range -- as opposed to the legal range internal files.

1. Just wanted to confirm: The LUT you generously supplied in this post is for legal range internal files ("This is assuming you're importing the XAVC directly") and not full range HDMI output. Is that correct?

2. As you continue to work with this LUT, are you having any nagging issues -- or are you largely satisfied with the results?

3. Curious, have you created a similarly performing LUT for recordings made using the full range HDMI output?

Thanks very much.

Yes the LUT was for the XAVC and legal. I will produce some full range ones too as soon as i can grab a moment!

I was using the LUT for quick and dirty proxies but have been happy with the results. I would generally shape the data going into the LUT, sky grads and so on.

You know this prompted me to look closer and i'm going to need to do these tests again. Broadly speaking it's all okay but i was looking at values through a display gamma (sRGB) so those absolute values are off a little. Because of this i think i'm going to go back to the source again. However HDMI is certainly full range compared to XVAC.

But here's the thing i noticed. I've got several ways to decode XVAC and they are producing quite different results. This one floored me. I see different colour shifts and even more importantly different levels of noise too. So now there's an added complication to all this. For example in comparing a real world scene XAVC with ProRes the XAVC is skewed towards green in the reds. Noticeably. Then XAVC decoding in PPro vs Nuke - the PPro is producing a better image or it is doing something to the image. I think PPro is dithering the UV channels, a form of chroma noise reduction.

So now i have another variable to take into account a quick step back is needed. Most of what i've posted has been based on ProRes recordings from the camera.

cheers
Paul
 
... but i was looking at values through a display gamma (sRGB) so those absolute values are off a little.
I see. You formulated a LUT to work with a legal signal but were basing your 'look' decisions on a full range display. I had wrongly assumed you were using a 709 calibrated display.

But here's the thing i noticed. I've got several ways to decode XVAC and they are producing quite different results.

This suggestion won't address the monitoring issue referenced above, but maybe the XAVC-S decoding discrepancies can be negated by using Sony's own Catalyst Browser or Catalyst Prepare. It would seem to make sense to rely on one of those applications to at least get a handle on how the footage is supposed to look -- and I seem to remember one or both of them supporting LUTs.

http://www.sonycreativesoftware.com/catalystbrowse
http://www.sonycreativesoftware.com/catalystprepare
 
A stabilized lens provides yaw + pitch anyways, so what's the allure of Sony's IBIS if not using Sony lenses?
For me, that it doesn't require a stabilized lens. In particular, that it works with my vintage Leica lenses. In this sense, I don't really care if it works better or worse than a Canon stabilized lens, it's enough for me that it works.
 
Yes the allure is that instead of the choice of a few IS lenses mostly slower zooms and only very few medium speed primes all of which have IS that looks and feels a bit different from each other, you can use any lens at all from vintage to fast primes to most standard fast zooms as well as cine lenses, none of which would otherwise be stabilized.
 
Thanks for that explanation Paul, and for confirming that the HDMI output is still full range -- as opposed to the legal range internal files.

1. Just wanted to confirm: The LUT you generously supplied in this post is for legal range internal files ("This is assuming you're importing the XAVC directly") and not full range HDMI output. Is that correct?

2. As you continue to work with this LUT, are you having any nagging issues -- or are you largely satisfied with the results?

3. Curious, have you created a similarly performing LUT for recordings made using the full range HDMI output?

Thanks very much.

I see. You formulated a LUT to work with a legal signal but were basing your 'look' decisions on a full range display. I had wrongly assumed you were using a 709 calibrated display.



This suggestion won't address the monitoring issue referenced above, but maybe the XAVC-S decoding discrepancies can be negated by using Sony's own Catalyst Browser or Catalyst Prepare. It would seem to make sense to rely on one of those applications to at least get a handle on how the footage is supposed to look -- and I seem to remember one or both of them supporting LUTs.

http://www.sonycreativesoftware.com/catalystbrowse
http://www.sonycreativesoftware.com/catalystprepare

Sorry i should have been more clear. The values i was quoting for Black and White point would have been done through an sRGB transform and i should have quoted via a linear one. You see if you take Slog footage into PPro for example, you are looking at Slog with a sRGB transform, there's no way to remove the viewing transform in PPro. Whereas in Nuke i can take the footage in, have no curve applied to it and turn off viewer transforms with the eye dropper and then when i find a value that is the actual code value in the file. It's easy to forget that most of what we see has an automagic sRGB gamma transform applied for the monitor.

The XAVC decoding differences appear to be different decoders, simple as that. Different ways to do the same thing. IMHO i think PPro is 'fixing' the footage a bit (which is a good thing in some ways). The way the decoder handles the shadows appears to show a large difference in the visibility of blobby noise.

This is a comparison, showing two decodes, on the left you see finer detail in the red channel and more controlled noise, on the right you see dodgy looking decodes and emphasised blobby noise.

https://www.dropbox.com/s/43936zr0104p3ti/xavcdecoding.png?dl=0

I'm going back to the developers to find out what decoding methods are being used. I believe it is through the QT subsystem.

cheers
Paul
 
Rather than edit the original.

I realised what the noise difference was. Both were slightly different frames and the PPro frame must have been a fully compressed frame rather than intraframe. Whereas the Nuke frame was intraframe (between key frames). Therefore the intraframes show the difference compression noise. When i match frame for frame the noise is identical. There is still a colour difference though but it's more subtle now.

So the XVAC intra frames can show more blobby noise.

Sorry, this was a bit of thinking out aloud going one

cheers
Paul
 
Rather than edit the original.

I realised what the noise difference was. Both were slightly different frames and the PPro frame must have been a fully compressed frame rather than intraframe. Whereas the Nuke frame was intraframe (between key frames). Therefore the intraframes show the difference compression noise. When i match frame for frame the noise is identical. There is still a colour difference though but it's more subtle now.

So the XVAC intra frames can show more blobby noise.

Sorry, this was a bit of thinking out aloud going one

cheers
Paul

Thanks Paul -- for this, and the previous explanation.
 
My understanding is that
1) All native E-mount lenses work with 5-axis stabilization.

Just a small correction. The Loxia lenses do not provide distance information so it is only 3-axis there (mildy annoying considering there price tags or that if you go for a Sony FE lens there is no proper manual focus)
Also not all A-mount lenses have the "D" function to support 5-axis (but most of them do)
 
Hi All. Just wanted to chime in with some more anecdotal data on overheating. I had not been following chatter on the A7RII before picking up the A7SII, and was not aware that overheating was a "thing". But I soon discovered this when the A7SII mysteriously shut down on an interview shoot. I was not watching the screen so I didn't see the hi temp message. In fact it wasn't until I ran a few tests at home, filming the screen with an iPhone (which is rock solid for really long clips, btw), when I caught it. In my case, the camera succeeds in recording an entire 29 minute clip, then cuts out anywhere from 10-15 minutes into the next clip. This is in a nice, cool 70-degree room with no lights hitting the camera.

As a test, I ran the same setup but allowed for a two-minute break between clips. In that case, I was able to get three full 29-minute recordings and nearly fill up a 64GB card. If your shoots allow for some flexibility, where you might be able to spare a brief coffee break every 29 minutes, this may work for you. As usual your mileage may vary, so I'd test this workaround yourself, preferably under a variety of representative temperature conditions.

A note on battery life during the tests: I have the battery grip, and on two fully-charged batteries, the first two 29-minute recordings left 2 or 3 percent on the initial battery, and a full charge on the remaining one. After the third recording I believe I still had 80+% or so on the remaining battery.

Now, while I have you, a newbie question. I'm upgrading from a hacked 5D2 here (finally!) and numerous C100/300 rentals, so I'm just getting used to Sony operation. I've scanned the manual for this and can't find an answer, and I've done the requisite Googling and can't find it either. Here goes: In movie mode, after you're done filming a clip, the LCD will black out after a certain amount of time. This much I expect. But when you want to film another clip, how do you turn "live view" back on? I know that sounds silly but... I just can't figure out how to do it. Simply hitting the movie record button won't do it. Nothing happens. Hitting the scroll wheel where it says "Disp" doesn't do anything either. In the Canon world obv we've got a dedicated live view button. On the A7SII, I've resorted to hitting the "menu" button, waiting for the menus to display, then hitting "menu" again to dismiss them. That seems to activate live view. This can't be the only way?

Thank you!
 
That is the auto shut down to save the battery power, you can set it to stay on longer in the menu, or you just have to turn it off and on again or you can shoot in M mode instead of Video mode and all you have to do is hit the shutter to turn it back on again without having to turn the camera off and on again.
 
Thanks for the heads up. I have to say, that is just bizarre. The camera isn't totally "off" when the screen goes dark - the menu button will wake the screen back up, but only after a trip through an irrelevant menu. I still feel like I'm missing something. But if this is the normal operation of the camera... Is it also the case that there's no way to turn off live view either? You just have to let it time out?
 
Back
Top