Sony A7SIII = $3500, 10-bit 4K120p internal, 16-bit 4K60p RAW out

Silenkiller said:
anyone have any links to some off camera files of people from the a7s iii?
Click to expand...

I think maybe some of DPReview's A7SIII samples are out of camera, but I'm not sure:

https://www.dpreview.com/videos/4655202205/sony-a7r-iii-uhd-4k-sample-video-footage

https://www.dpreview.com/videos/8582701288/sony-a7s-iii-4k-sample-reel

Philip Bloom has his original uploads on Vimeo which can be downloaded but they are not camera files (might require Vimeo Pro or Business account to download, I can't remember):

https://vimeo.com/441676150

https://vimeo.com/441675945
 
joema said:
I think maybe some of DPReview's A7SIII samples are out of camera, but I'm not sure:

https://www.dpreview.com/videos/4655202205/sony-a7r-iii-uhd-4k-sample-video-footage

https://www.dpreview.com/videos/8582701288/sony-a7s-iii-4k-sample-reel

Philip Bloom has his original uploads on Vimeo which can be downloaded but they are not camera files (might require Vimeo Pro or Business account to download, I can't remember):

https://vimeo.com/441676150

https://vimeo.com/441675945
Click to expand...

@joema - Thanks a million! Will definitely throw them in Resolve tonight!

@ahalpert - As regards Philip Bloom, he stated openly that he was commissioned by Sony to film a promo short with the A7SIII, adding he'd still be impartial in his review and evaluation. Besides his 'heat test' looked fairly authentic to me. So did his low light (or rather NO LIGHT) samples - giving similar results to John Behiri's tests... Aside from his irrational dislike of BlackMagic cameras, his reviews are spot-on, particularly his long term ones - and reflect what a day to day user earning his living from said camera would experience.... and remember he got his break with Canon 5D MkII video revolution, and used Canon cameras most of the time - so if anything he'd be biased towards Canon.

No camera is perfect - but this looks like a great workhorse if you can justify spending three and a half grand - currently I can't and will be sticking with my BMPCC - but will jump ship to Sony once business picks up again.....
 
tull_power said:
Rather then look to 'see' in the highlights, I think a lot of shooters look for a smooth, film like roloff. To this day many cameras have harsh highlights roloff unless you carefully underexpose and then adjust exposure in post - so it's a bit the Holy Grail of large sensor cameras. The low range is equally important though, particularly if you're going to make up for bad / poor lighting in post.

Thanks for the chart you've posted earlier BTW!
Click to expand...

You're welcome!

There is *no* rolloff with SLog-3. It is linear past middle gray all the way to the clipping point. DR doesn't give any hints about rolloff either. All we're doing is counting stops from the clipping point.
 
tull_power said:
No camera is perfect - but this looks like a great workhorse if you can justify spending three and a half grand -
Click to expand...

Yeah it is a workhorse, that's why I plan on getting one.

Where is Bloom's overheating test? I can't find it

And yeah I think dan Watson is a canon shill. He tested the R5 against the A7SIII in the only mode in which the R5 has an advantage and he wasn't terribly clear about it

That being said, I like to know all the problems with a camera before I buy it so I am on the lookout for issues. So far so good

Regarding color - I see nice skin tones in the A7SIII stills I've seen. I was also just reviewing some video interview I edited recently that someone else shot on a7riii. Skin tones looked great. Not sure why but I feel like a lot of the early youtube reviews of the A7SIII are not doing it justice
 
ahalpert said:
Well, I think the A7S3 can keep a lot of highlight detail and produce a nice roll-off. I just mean that a lot of cameras lose chroma info even when they keep luma gradient at the top of their exposure range. I found this neat tool at the digital picture recently which has a series of color charts at different exposures, and you can see how the color from green/yellow just vanishes on the R5 at 3 stops overexposure: https://www.the-digital-picture.com...&ISO=0&CameraComp=1221&TestComp=0&ISOComp=100
View attachment 139885

I think that's what happened here - that the camera was well-overexposed and the sky comes out looking gray:
View attachment 139883

But as far as highlight handling, especially for a properly exposed shot, I'm feeling very good about this:
View attachment 139884
Click to expand...

You are seeing the sun low on the horizon, the polarizing effect of a variable ND filter is very strong at 90 degrees, thereby causing the strong vignetting evident in the corners. The sun and some of the reflections are clipped; the areas that are not clipped were recorded at lower IRE. The DR improvement is evident in the dark details of the shirt and swimsuit.
 
Tom Roper said:
You are seeing the sun low on the horizon, the polarizing effect of a variable ND filter is very strong at 90 degrees, thereby causing the strong vignetting evident in the corners. The sun and some of the reflections are clipped; the areas that are not clipped were recorded at lower IRE. The DR improvement is evident in the dark details of the shirt and swimsuit.
Click to expand...

Yeah I think the DR looks great. The sun will always clip. People use 14-stop ND to shoot the sun. I'm more interested in the area around it, which looks nice to me. That was why I showed the shot - the sky is holding and the shadow details as well.

I'm not sure the polarizing effect is causing the vignetting because it's happening in the exact same spot in both shots and he's facing a different angle. I think something is physically obscuring his lens or else he added it in post
 
ahalpert said:
The sun will always clip. People use 14-stop ND to shoot the sun. I'm more interested in the area around it, which looks nice to me.
Click to expand...

The gradient around the sun is diffraction, the bending of waves around the corners of an obstacle or through an aperture.
 
Tom Roper said:
Here's the running DR compilation, all @ SNR 2.0:

Alexa 14.0
C500MKII 13.1
A7SIII 13.0
C300MIII 12.8
FS7 12.7
S1H 12.7
UMP4.6 12.6
C300MKII 12.3
BMPCC6K 11.9
BMPCC4K 11.6
XT3 11.2
R5 10.9
GH5s 10.7
A7SII 10.6

Updated to include R5
Click to expand...

Tom, where do these numbers come from? Forgive my ignorance - but is there math involved with the SNR 2.0 statement? And would be interested to know where BM new sensor comes in on the 12k cam...
 
Every sensor captures light linearly and clips the same way but yes you can have built-in roll off in your captured gamma if you want, no? With s-log 3 just roll it off in post

There is math involved in SNR 2 in that the signal is 2x the average noise. Choosing that ratio is arbitrary under the belief that at that ratio we can discern the signal from the noise. Of course it does nothing to take different types of color noise into effect. There are a lot of complications that one can subjectively evaluate.

The ursa 12k has less DR than the UMP G2
 
@markfpv

The DR numbers are a compilation from Gunther's Imatest DR tests at Cinema5d, with a couple more added recently by Gerald Undone who used the exact same model. The 2.0 SNR is an arbitrary but agreed upon noise level of where to stop counting DR stops in the low key noise region. DR stops are counted *down* from the clipping point, and where they blend in or disappear into the noise zone has always been fuzzy, so 2.0 is an agreement to stop counting once the signal is less than twice the noise. It's the best of the imperfect methods that we have, because a camera that performs in-camera temporal noise reduction gets preferential SNR treatment and thus more stops of DR, when it would be most desirable if we had a way (which we don't) of directly measuring the shot noise of the sensor.
 
ahalpert said:
Every sensor captures light linearly and clips the same way but yes you can have built-in roll off in your captured gamma if you want, no? With s-log 3 just roll it off in post

There is math involved in SNR 2 in that the signal is 2x the average noise. Choosing that ratio is arbitrary under the belief that at that ratio we can discern the signal from the noise. Of course it does nothing to take different types of color noise into effect. There are a lot of complications that one can subjectively evaluate.

The ursa 12k has less DR than the UMP G2
Click to expand...

Beat me to it!

Where did you find a DR of the Ursa 12k?
 
ahalpert said:
BM themselves claim 15 stops for the G2 but only 14 stops for the 12k.

And if you look at the released footage, that seems to bear out.

The G2 does DGO and the 12K does not
Click to expand...

It's too speculative to assume BMD will apply the same level of exaggeration to its promotion. We have to have a test. It's a brand new sensor design with equal number of RGB pixels plus white pixels now; many ways it can be binned.
 
I'm hearing that the A7S3 only does a 12-bit adc raw readout and then dithers it up to 16-bit before outputting it. Also:

"4K 24-60P and 4K 120P have almost the same level of sharpness and detail. And 1080P is not supersampled, but pixel-binned from 24-60P and line-skipped from 100-240P."

None of this info impacts my preorder. Just passing it on
 
Tom Roper said:
It is an observation, not a complaint. There is no highlight rolloff in SLog-3. That's not a complaint either. It's a fact.
Click to expand...

And a very good/valid one at that. The assessment is appreciated and I didn't mean it as critique, just a playful observation on how far we've come since I joined the forum!
 
ahalpert said:
I'm hearing that the A7S3 only does a 12-bit adc raw readout and then dithers it up to 16-bit before outputting it.
Click to expand...

That could not possibly be true. You would only have the potential for 12 true stops of DR. The specs seem to infer 14 bit ADC, 16 bit data. This makes sense because HDMI doesn't support 14 bit.
 
You must log in or register to reply here.
Back
Top