Sony A7R III announced with 4K HDR

[jonpais]

This is getting way off topic, but how is it that for all those years Panasonic was saying OIS was good enough or even better for video, when as it turns out, the m4/3 sensor was expressly designed with room for IBIS all along. So confusing!

[Emanuel]

On 1/21/2018 at 11:00 PM, Matthew Hartman said:

Wha?

I'm not "looking down" per se on anything. I just don't prefer Panasonic when it comes to the character of their image, it looks "flat" or "thin" to me regardless of resolution or bit rate I view it at. Obviously, there are others who like or even love it, which is super for them. No hate here. Each to their own. 

My guess about YT's compression algorithm is a doubt, not anything prescriptive. I don't work for Google so of course I couldnt tell you their future intentions. I based my doubt off of what we've seen in the past and today.  As of today, 8bit is still the distribution standard. If that changes, great! Right? 

My "arguement" was actually in support of not looking down on 8bit rather than turning up my nose to it. Also, to get away from at size perceptible banding you really need to be at least 10bit, but more realistically in a 16bit + colorspace. 

I think there's a buzzword when it comes to 10bit right now and a misconception that 10bit means no banding. I've been a designer for over 25 years, that's simply wrong. If we're being rententively honest, you'd have to be in a 32bit/float colorspace to get zero artifacts/stepping, as far as pixel peeping in concerned. But ppl don't watch content at a 500% crop, unless they're anally retentive, so does it even matter? 

Maybe I didn't articulate this well, but basically dont worry about it. Banding, it's there, even if you cant see it at 100%, and a reality for most of us, just concentrate on making good content instead, because that's what will determine the success of it the most. 

The video you linked is very impressive on my 10bit BenQ 2.5k monitor.

Bit rate is more crucial than anything else...

sample extracted from (minute 28:30):
https://vimeo.com/114978513

[jonpais]

But isn’t that example demonstrating the relationship between banding and bit depth, not bit rate?

[markr041]

2 hours ago, jonpais said:

But isn’t that example demonstrating the relationship between banding and bit depth, not bit rate?

It seems to me that most of the banding disappears when the lightly-compressed, high bitrate codec at 8bit (probably ProRes 422 - that is what the Shogun uses) is used compared to the highly-compressed low bitrate internal codec (middle panel compared to the first - both 8bit). There is a gain from going to 10bit from 8bit, using the exact same compression (bitrate), but it is relatively minor. The poster should have indicated the bitrates, not just the bit depth.

[darrellcraig]

6 hours ago, Matthew Hartman said:

Curious, which image to you prefer? 

Matthew, on GH5 vs. A7RIII image...

For stills: the A7RIII is vastly superior to me. But A7RIII does weird white balance stuff (same problem with A7/A7II/A7RII) - inconsistent and often wrong. GH5 stills quality is vastly improved over the GH4 and the ghastly GH3/2. The jpeg engine on the GH5 is really good and I often don't bother messing with RAW. And I found DR is much better on GH5 over GH4, files are much more malleable, especially on highlight retention. But shadows are cleaner too. But A7RIII stills just have a lot more latitude and I actually prefer Sony color (feels more natural, accurate) when white balance is right. I shot with Fuji X-Pro 2 and X-T2 and X100T for a while and those cameras are awesome at auto white balance. Colors are really pleasing, if not accurate.

For video:  I haven't shot the A7RIII much, just got it and I've had a bunch of opportunities to shoot stills but not video, so far. So nothing meaningful to add on video image quality. But clearly IBIS is not in the same league, that is immediately clear. 

The main reason I bailed on the A7RII was usability as a hybrid cam rather than stills or video image quality. The A7RIII is much, much better as a hybrid camera. Not as seamless as the GH5 in switching between roles as stills and video camera, but much closer to the GH5 than the A7RII. The other reason I prefer Sony is the access to a wonderful range of glass, especially the ability to use wider angle legacy and manual focus glass without crop factor. I recently picked up the Voigtlander 40/1.2 and am really enjoying that lens. I think the Oly 25/1.2 Pro is one of the best 50mm equiv lenses I've ever used, for any platform (it really is all about the glass), but that is still like a 50/2.4 from DoF perspective, so somewhat limited in what you can do.

[jonpais]

1 hour ago, markr041 said:

It seems to me that most of the banding disappears when the lightly-compressed, high bitrate codec at 8bit (probably ProRes 422 - that is what the Shogun uses) is used compared to the highly-compressed low bitrate internal codec (middle panel compared to the first - both 8bit). There is a gain from going to 10bit from 8bit, using the exact same compression (bitrate), but it is relatively minor. The poster should have indicated the bitrates, not just the bit depth.

The way I understood it is that bit depth, or the number of 'shades' there are for each color channel, is what determines whether we see banding in an image. In 8 bit, there are 256 levels, in 10 bit, 1024 levels, which is why, when we push the image around in post, often (if not always!), the 10 bit image should hold up better. 

source

[markr041]

10 minutes ago, jonpais said:

The way I understood it is that bit depth, or the number of 'shades' there are for each color channel, is what determines whether we see banding in an image. In 8 bit, there are 256 levels, in 10 bit, 1024 levels, which is why, when we push the image around in post, often (if not always!), the 10 bit image should hold up better. 

source

You are right ceteris paribus, but the bit depth statement and the visual comparison ignores the role of bitrate (that is, you have compared different bit depths for highly compressed clips (I assume)). The test posted above shows us that bitrate matters also, and that it matters more. Remove the high compression, and the difference by bit depth shrinks almost to invisibility. So you are correct, it is just that it turns out something is even more important than bit depth (they interact). That is, when you use 8 bits but do not highly compress you get much less banding. I have heard this before; I am not claiming the bitrate effect is correct other than interpreting what the poster showed (and I do not see anything wrong with his test).

[Don Kotlos]

I just got my hands on an A7rIII and the greatest drawback of A7rII has been fixed: Shutter sounds great! 

[Renaud]

Is there anyone disappointed with A7r3 video af ? 

I found it way slower than on my a6500 with the same setting... 

[Don Kotlos]

But see comments:

"There was monitor attached to 6500 and as i did fund out later face detection do not work, or not work correctly with monitor attached!"

 

[Robert Collins]

22 hours ago, Renaud said:

Is there anyone disappointed with A7r3 video af ? 

I found it way slower than on my a6500 with the same setting... 

Have you tried setting 'af tracking sensitivity' - menu 2 - 2 - to responsive. By default it is set to standard.

[kidzrevil]

On 1/23/2018 at 1:04 PM, jonpais said:

The way I understood it is that bit depth, or the number of 'shades' there are for each color channel, is what determines whether we see banding in an image. In 8 bit, there are 256 levels, in 10 bit, 1024 levels, which is why, when we push the image around in post, often (if not always!), the 10 bit image should hold up better. 

source

you are right. It is bit depth not bitrate that causes the banding. The other poster is wrong on this one

[Renaud]

11 hours ago, Robert Collins said:

Have you tried setting 'af tracking sensitivity' - menu 2 - 2 - to responsive. By default it is set to standard.

Yes I have this setting since the beginning but it’s still slower than my a6500..

[Emanuel]

20 hours ago, kidzrevil said:

you are right. It is bit depth not bitrate that causes the banding. The other poster is wrong on this one

Let's give 2nd look on that test a chance...

People generally have that idea. Banding depends on the bit depth.

That test leads to a bit different path though.

Actually, it is even beyond color sampling, also bit rate.

That is, not only bit depth.

 

Practice redefines theory much often to keep going to ignore the evidence : ) Mark's @markr041 posts hinted it all :-)

[kidzrevil]

@Emanuel I may have misread @markr041 post. Although it is bitdepth or lack thereof that generally causes banding I guess in this case the image in question is so compressed and of such low quality bitrate is the limiting factor here. 

[tomsemiterrific]

Shot march 3, 2018