Sony a7s III 48mpx Quad Bayer sensor

well, you'd only get a B/W 48MP picture. the color info would be the same

the more disappointing thing is they didn't implement DGO. seems like it would be easy by alternating high- and low-gain pixels
 
ahalpert said:
well, you'd only get a B/W 48MP picture. the color info would be the same ...
Click to expand...

Most smartphones have a Quad Bayer sensor, where can easily switch the readout to get a sharper image. It may not be the exact 4-for-1 but it's somewhere 80% in terms of the resolution.
 
DLD said:
Most smartphones have a Quad Bayer sensor, where can easily switch the readout to get a sharper image. It may not be the exact 4-for-1 but it's somewhere 80% in terms of the resolution.
Click to expand...
Your comments are nonsensical. Thr a7s III is not crippled at all. Please discontinue spreading misinformation.
 
DLD does have the highest blather / talking completely out of his *ss post density on this website, empirically. But...it probably would be within Sony's abilities to implement circuitry into their sensors to either boot in high res or low res (improved readout, video optimized) mode with these types of sensor designs. In this case they appear to be able to make small alterations to create either a 50 or 12 mpx (slightly cropped) sensor and maximize their return on the design and it's production. Accompanying components obviously contribute to part of the large price difference for two very differently marketed cameras.

He is correct about some (I doubt it is a majority) smartphone sensors already having this functionality. Their actual resolution however is so much lower than all the nonsense marketing specs. I don't doubt the sensors have all those pixels, but it sure as hell isn't resolving that much actual detail for a variety of reasons. Even the pixel level detail on these *large* new iphone sensors largely look like mush to me - unless the photo was shot in broad daylight (base iso).

Regardless, I would imagine this built-in functionality will make it's way to hybrid mirrorless cameras in a couple years
 
Last edited:
DLD said:
Most smartphones have a Quad Bayer sensor, where can easily switch the readout to get a sharper image. It may not be the exact 4-for-1 but it's somewhere 80% in terms of the resolution.
Click to expand...

Well, I think the detail and lines should be better resolved. But the colors will be effectively splotchier or something? But obviously no worse than the (effectively) binned capture to begin with. You're only adding capabilities by enabling 48MP capture, not that I need that
 
ahalpert said:
Well, I think the detail and lines should be better resolved. But the colors will be effectively splotchier or something? But obviously no worse than the (effectively) binned capture to begin with. You're only adding capabilities by enabling 48MP capture, not that I need that
Click to expand...
There's need for some pretty powerful processing to demosaic the full count (x4), so the image quality on cheaper smartphones - especially if used for 8K video - can be smudgy and noisy. Or both. But a $3,500 camera should have enough oomph behind it to deliver something pretty close to its full specs. After all, R5, Z9, A7RIV can all do the high pixel count to the various degrees of performance already. A7SIII should at least have that option.

But Sony didn't want to cannibalize its photo-centric models and no one in the outside world sought to challenge the company - or even tried finding the correct technical specs of the sensor - and so it whacked it with a cripple hammer and no one has been wiser. Until now.
 
DLD said:
There's need for some pretty powerful processing to demosaic the full count (x4), so the image quality on cheaper smartphones - especially if used for 8K video - can be smudgy and noisy. Or both. But a $3,500 camera should have enough oomph behind it to deliver something pretty close to its full specs. After all, R5, Z9, A7RIV can all do the high pixel count to the various degrees of performance already. A7SIII should at least have that option.

But Sony didn't want to cannibalize its photo-centric models and no one in the outside world sought to challenge the company - or even tried finding the correct technical specs of the sensor - and so it whacked it with a cripple hammer and no one has been wiser. Until now.
Click to expand...

The a7s III is not whacked with any cripple hammer and virtually every single reputable objective reviewer on the Internet disagrees with your assessment.

The a7s III remains among the bestselling video-centric mirrorless cameras at N. America's leading independent retailer.

The a7s III has the least rolling shutter measured of any mirrorless to date and is the only one to shoot full frame 4K 60p and 120p and on and on...

It is much more popular among videographers than either the R5 or the 8-bit a7R IV, both of which have a 30-minute record limit.

Some of the best cinema cameras in the world have from 7.5 - 9 megapixel sensors, from Canon's C70 to the C300 Mark III to the Arri Alexa Mini (20 out of 52 feature films screened at Cannes 2021 were shot with the Mini) and professionals like Steve Yedlin scoff at pixel counters. There is nothing smudgy about anything.

You'll have to do better than that.
 
Last edited:
That's not what he meant by smudgy. He meant bad demosaicing of 48MP with a 2x2 cfa is smudgy. I don't think you get his line of reasoning. Its not that the a7siii is bad, it's that its hardware is capable of additional functionality that they chose not to enable. Doesn't matter much to me though.... except dgo
 
No, I get it. He said the a7s III is crippled, then suggested that the R IV and R5 can do un-cropped full frame no line skipping, no pixel binning 4K 60p and 120p, which they are emphatically unable to do. The a7R IV uses line skipping in full frame mode. From what I've seen, the Nikon does line skipping or pixel binning, too! Rolling shutter was definitely visible as well, whereas the RS of the a7s III is the lowest of any full frame mirrorless camera on the market.

Not sure what added functionality you're referring to, but THE a7s III IS NOT CRIPPLED.

If someone were to go up on stage and make outrageous and unfounded accusations to that effect at SMPTE, they would be ridiculed by all the engineers and representatives of all the manufacturers.

The same sensor is in the FX6, which is Netflix approved, while the R5 never will be. The R5's got a meager 10.4 stops of DR, among the lowest of any full frame mirrorless camera ever tested by CineD, alongside the equally bad R6. The R5 sensor will never be used in a Canon cinema camera. Countless filmmakers have said how much they deplore editing Canon's impossible codec on their machines, whereas the files of the Sony are easy to edit. When it comes to professional filmmaking, workflow is a massive consideration.

"In order to achieve certification from Netflix, cameras must meet minimum resolution and capture specifications, as well as other ‘attributes like dynamic range, form factor, stability and workflow compatibility were also taken into consideration,’ according to Sony. While this isn’t huge news for your average photographer or cinematographer, it’s worth noting the FX6 is effectively a cine version of the a7S III that features more production-level inputs, outputs and adjustments. This effectively means the core components that are approved for shooting Netflix Originals are also inside the much smaller — and more affordable — a7S III". -DPReview, writing about the FX6 Netflix approval

The bestselling FX3 also uses the same sensor as the a7s III.

I wouldn't hesitate to choose the 9 megapixel C70 over the R5 any day of the week - and neither did Netflix!

On the other hand, if you like artificial 30-minute record limitations, micro HDMI ports, hard to edit files, low dynamic range, diving into the menus each time you need to custom WB, rolling shutter, line skipping and pixel binning, 8-bit and overheating, then yeah, one of the cameras mentioned by DLD might fit the bill.
 
Last edited:
OK, folks; here's the deal.

The personal insults are getting out of hand lately. You all have been around here long enough to know that it doesn't fly.

Yes, posts in this thread have personal insults, and no, that's not up for debate.

So, if the insults continue, bans are coming. This is your warning.
 
ahalpert said:
That's not what he meant by smudgy. He meant bad demosaicing of 48MP with a 2x2 cfa is smudgy. I don't think you get his line of reasoning. Its not that the a7siii is bad, it's that its hardware is capable of additional functionality that they chose not to enable. Doesn't matter much to me though.... except dgo
Click to expand...

Hmm! Interesting!

The Landingfield. A group of diehard Astrophotography followers have done a strip down on the A7Siii and had a good look at its sensor and layout. In a nutshell, they have titled their observations as:

"Sony a7S III has a 2×2 pixel binning IMX510 BSI sensor". They go on to say amongst other things.

"Let’s heat remove the cover glass and inspect the pixels under a 50x microscope objective. It turns out this sensor was a 2×2 binning design. This means IMX510 actually has a 48MP native resolution. The RGGB Bayer pattern is spread across a 4×4 grid. After sensor readout, the four pixels in each pf the same color are then combined digitally to give one pixel before sending out on the SLVS-EC interface. This could explain the increase in read noise. From my knowledge, none of Sony DSLR CIS supports charge binning due to limitation in its pixel architecture. By combining four pixels digitally, you would increase the noise variance by four and hence read noise almost doubles (sqrt to RMS). The bright green pixels are phase detection pixel for hybrid AF system."

Photos are included with their strip-down assessment.

All can be seen and read here:

https://landingfield.wordpress.com/2021/09/03/sony-a7s-iii-has-a-2x2-pixel-binning-imx510-bsi-sensor/

Chris Young

A7Siii sensor grab.jpg
 
DXOMark found a 1/3 stop difference between 2nd and 3rd generation but there is a massive jump in readout speed with the a7s III, which is excellent for video, since (a) AF speeds can be faster, (b) the entire sensor can be read more quickly, meaning no crop in 60p and 120p (c) we get no line skipped 1080p and (d) it results in the least rolling shutter of any full frame mirrorless camera.

As the article points out, all these engineering choices involve tradeoffs. Would I rather have a slightly cleaner image but massively worse rolling shutter, cropped 60p, no 120p and line skipped 1080? No to all the above! Which is also why I won't be picking up another camera that shall remain unnamed...

A little noise can be cleaned up in post, but rolling shutter cannot. Or I could upload all my videos to YT without the very tiny bit of noise reduction I apply and no one would be the wiser!

Fast sensor readout increases every facet of camera performance, especially in video (and also in sports photography).
 
Last edited:
jonpais said:
Not sure what added functionality you're referring to, but THE a7s III IS NOT CRIPPLED.
Click to expand...

Added functionality would be shooting high-resolution stills that record an independent luma value from each photosite, which would have made the a7siii a much more imposing hybrid camera
 
jonpais said:
As the article points out, all these engineering choices involve tradeoffs. Would I rather have a slightly cleaner image but massively worse rolling shutter, cropped 60p, no 120p and line skipped 1080? No to all the above!
Click to expand...

The idea is that the camera would employ two readout methods. One is the pixel-binned method it currently uses, resulting in fast readout of a 12.1 MP sensor (or, rather, the 16x9 crop of that sensor). The other would be a slower readout of a 48 MP sensor for the purpose of shooting high-resolution stills. (Or maybe high-resolution video with worse rolling shutter).

This concept was bandied about on sonyalpharumors before the release of the A7SIII. When they released it, I assumed that they had gone with a different sensor design. There had also been talk of DGO via alternating high- and low-gain photosites
 
ahalpert said:
The idea is that the camera would employ two readout methods. One is the pixel-binned method it currently uses, resulting in fast readout of a 12.1 MP sensor (or, rather, the 16x9 crop of that sensor). The other would be a slower readout of a 48 MP sensor for the purpose of shooting high-resolution stills. (Or maybe high-resolution video with worse rolling shutter).

This concept was bandied about on sonyalpharumors before the release of the A7SIII. When they released it, I assumed that they had gone with a different sensor design. There had also been talk of DGO via alternating high- and low-gain photosites
Click to expand...

"The third gen has a huge improvement in readout speed due to its BSI architecture. After all, this camera is mainly aimed for cinematographers. Its all-pixel scan rate has drastically increased from 30FPS to 90FPS. And 1080P60 no longer needs subsampling like in IMX235".

Fast sensor readout speed boosts every facet of camera performance, as we have already shown.

Sonyalpharumors is a troll magnet and their comments section is the very last place on earth anyone should be looking for reliable information.

Your comments fail to demonstrate how the video-centric a7s III - which, along with the FX3, is among the most successful video-centric mirrorless cameras over at N. America's leading independent retailer and elsewhere, and which has earned the praise of virtually every objective, reputable reviewer on the planet - has been crippled. BECAUSE IT HAS NOT BEEN WHACKED WITH A CRIPPLE HAMMER - any more than the Canon C70, the Sony FX3, the Sony FX6, the ARRI Alexa Mini (7.5M), the Canon EOS C300 Mark III, the Panasonic Varicam, the LUMIX BGH1, or any other 10 megapixel or less cinema camera has been crippled. In fact, it absolutely destroys the competition in several ways, which I won't waste your time or mine by repeating. The FX3 was exciting to filmmakers not because it shoots high resolution stills - which it doesn't - but because it was even more video focused than the a7s III. As a matter of fact, the three best selling digital cinema cameras at B&H are all 10 megapixel (or less) cameras. Concerning DGO, its implementation in the $5,500.00 Canon C70 yields a dynamic range equal to that of the Sony, so in no universe can the lack of DGO be called cripple hammer - and Canon does not deign to put DGO in any of their mirrorless bodies either, for that matter. Let's face it - once someone resorts to holding up the a7R IV - an 8-bit line-skipping camera with an artificial 30-minute recording limitation, no full HDMI out, no articulating LCD, no RAW video, etc. - or the Canon R5, whose artificial record limit and overheating due to intentional lack of proper thermal management were the object of derision around the world, or the Z9, a camera that has neither been released nor tested that will sell for $2,000 more than the Sony and weighs as much as two Sony a7s III bodies - as examples of how Sony might have improved the a7s III, the conversation has already ended.
 
Last edited:
You must log in or register to reply here.
Back
Top