Sony A7S II fares worse in low light than A7R II and the original A7S

[ahsanford]

DXO -- the A7S II is "Low light performance redefined":

http://www.dxomark.com/Reviews/Sony-A7s-II-sensor-review-Low-light-performance-redefined/Comparison-1-Sony-A7S-II-versus-Sony-A7S-versus-Sony-A7R-II-Impressive-high-ISO-DR

Yet it's A7S predecessor and A7R II outperform it in their high ISO testing. Hmmm. Shouldn't the title be something more like "Low light performance nearly reaccomplished?"

Also, does anyone buy their conclusion on high ISO here? I thought the A7S II would handily beat the A7R II -- that (and video) is expressly what it's built for, isn't it?

- A

 

[ahsanford]

Sony have used the same sensor in the A7SII as the A7S. Any difference is likely in their BIONZ chip.

It would be great for the A7SII to have had a 12MP sensor using the same tech as in the A7RII's sensor.

Oh, I didn't know that. So there's no backside illumination hotness in that A7S II? That would certainly explain some of this.

- A

 

[9VIII]

I think it actually makes (financial) sense that they aren't using BSI on the A7S right now.
(I'm just linking to this article because they have a pixel pitch chart)
https://photographylife.com/sony-a7r-ii-use-considerations

The A7RII has a pixel pitch of 4.51 microns, or 20.34 square microns.
The A7S has a pixel pitch of 8.4 microns, that's 70.56 square microns.

If the amount of surface area occupied by circuitry is the same per pixel, just for a harsh estimate lets say 10 square microns, then the A7S would be losing approximately 14% surface area per pixel... But the A7RII would lose 50%.
Using an expensive fabrication method to get back 14% of your sensor's potential doesn't sound that appealing.

Sony probably loves that the A7S is so popular, the margins on a sensor with pixels that large are probably a lot higher just based on the lack of complexity.

I have a theory that IBIS is the reason the A7SII is worse than the original though.
The A7II (not the S or the R, the regular 24MP model) also had worse low light performance than the original A7, and the only reason I can think of for that to be the case is the In Body Image Stabilization, probably just because of thermal dissipation.

One thing I notice when shooting Macro with long exposures is that I get way more hot pixels when I'm using live view, I hate to say it but I actually prefer not to use live view if I really want a clean shot.
I assume that most cameras are designed with a good thermal management solution, but IBIS demands that the sensor be floating, you probably can't have a big heat sink stuck to the back of it at the same time. Thus, noisy pixels at high ISO.
Or so goes the idea in my head anyway.
How does the A7RII get away with it? Who knows. Maybe Backside Illumination also lets it offload heat better because the circuits are on the back? And maybe it also insulates the photon wells from heat generated by the circuits? This would indicate that BSI on the A7SII would be a good idea, but, margins. Just like Canon and their old sensor fabs, if you think the improvements are minor and you can sneak out old tech for a while longer, that's going to be extremely enticing.

What I know for sure is that I like high resolution cameras and I'm very glad that BSI is working so well on the A7RII.

 

[afonsoclj]

I think it's just because they used compressed (pre-cooked) RAW on a7S Mark I and uncompressed on Mark II.

 

[kaihp]

I think it's just because they used compressed (pre-cooked) RAW on a7S Mark I and uncompressed on Mark II.

That's a fallacy. Compression or not doesn't make a quality difference - only in size of the files.

Lossy compression, is what makes a difference. Most compression schemes are lossless (zip, lzh, arc, 7z, rar, gzip, bzip, and so on), but the JPEG, MPEG (H.264 and H.265) algorithms are specifically designed to be lossy with a low visual impact on still and moving images respectively

 

[3kramd5]

Lossy compression, is what makes a difference.

Which is exactly what Sony uses on the alpha series.

That being said, I've seen no appreciable quality difference between (lossy) compressed and uncompressed files from my Sony, just a file size difference. The scheme is designed to be lossy in the noise; maybe it's mostly successful.

 

[raptor3x]

If the amount of surface area occupied by circuitry is the same per pixel, just for a harsh estimate lets say 10 square microns, then the A7S would be losing approximately 14% surface area per pixel... But the A7RII would lose 50%.
Using an expensive fabrication method to get back 14% of your sensor's potential doesn't sound that appealing.

This ignores the effect of microlenses. Microlenses will act to lessen the effect of fill factor.