There's been a lot of speculation about what an a7S III would look like, perhaps this rumored sensor from Sony Semiconductor provides some clues.

We’ve been speculating for a while about what a Sony a7S III would need to offer to filmmakers to offer a compelling advantage over the a7R III and a7 III. Rumors of a new sensor from Sony Semiconductor could make even our wildest speculation look rather unambitious.

With the $2000 a7 III able to shoot 4K footage from a 6K region of its sensor (capturing more detail than the a7S II can), Sony would have to do something clever to make an a7S III look attractive. The office was split down the middle between 'more resolution,' perhaps using an a9-style stacked chip to give a7 III-esque resolution but perhaps at 60p and with minimal rolling shutter, or possibly a move to 10-bit recording, to give more flexible files, particularly for anyone shooting Log or HDR video.

In a recent interview, Sony's Kenji Tanaka told us the a7S II successor would be "more than [a7S II users] expect."

Rumored specification of sensor capable of 36MP stills and 8K Ultra HD capture
Pixel Size (µm) 4.88
Aspect ratios
  • 3:2
  • 16:9
Pixel Resolutions
  • 7380 x 4932 (3:2)
  • 7730 x 4352 (16:9)
Active sensor area
  • 36 x 24mm
  • 37.1 x 20.9mm*
Rumored readout rates
  • 10FPS @ 16-bit
  • 60FPS @ 10-bit**

*Although the 16:9 region is wider than the 3:2 area, its diagonal length is smaller, so would be guaranteed to fit within the standard image circle of full frame lenses.
**10-bit capture would limit the camera to around 10EV of dynamic range. Video is more likely to be made from 12 or 14-bit capture, the rates for which are unspecified.

However, Sony Alpha Rumors is reporting a fairly plausible chip spec that would blow such cameras out of the water: an oversized Full Frame sensor that could shoot 36MP stills or 8K video while maintaining the same diagonal field of view (so the scope of wide-angle lenses isn't reduced by cropping a 16:9 image out of the middle of a 3:2 sensor).

This could give a camera that makes no apparent compromise between stills shooting and video capture.

It's possible to use a 16:9 crop (yellow) that's wider than the standard 36 x 24mm dimensions of a full frame sensor (pink), so long as the diagonal is the same. This maintains the same diagonal angle of view: a trick Panasonic and Canon have made use of in the past.

This sensor could form the basis of one hell of a stills/video camera. Details of the chip are hazy to the point where they could be downright inaccurate, but a camera that can shoot both high-res stills and high-res video would make sense as an a7S III, since it would cost more than an a7 III but offer a lot more to videographers than the a7R III can.

Don't get side-tracked by talk of the a7S cameras being great for low light stills: it's no better than the a7R III

And let's not get side-tracked by talk of the a7S cameras being great for low light stills. I'd happily put money on that being a marketing angle decided as the camera was being launched: the a7S II is no better in low light than the a7R III if you scale files down to the same size (except in extreme low light). The only reason it ever looked good in low light was that it's the first time we'd seen 'Dual Gain' technology appear in a Sony sensor, and didn't initially recognize its significance.

Does 8K capture mean 8K output?

Just because a sensor can capture 8K doesn't mean that it would necessarily output 8K footage. Every manufacturer we've spoken to has told us how difficult it is to process and compress 4K footage in a camera with a small, stills-style body with no fans, because too much heat builds up. It seems unlikely that Sony could go straight from 4K stills/video cameras that have some heat buildup limitations straight to one that can process and compress four times as much data.

And that's before you question how many people have computers or software that can edit the footage.

Could we be looking at perfectly oversampled 4K output, rather than 8K

But the question is: would it have to? There's very little call for 8K footage yet, given that 4K is only starting to find its way into people's homes. Sony's current cameras can downsample 6K capture into 4K output, so could we be looking at perfectly oversampled 4K output, rather than 8K? After all, the math is likely to be easier.

The Sony a7S used 3840 x 2160 capture to produce 1920 x 1080 video with no aliasing, since this 2x oversampling let it perfectly capture all the frequencies that can be included in 1080 footage.

There's a precedent for this, of course. The original Sony a7S captured 3840 x 2160 pixels, not to create 4K footage but as the basis for excellent 1080 video. This has a number of advantages: 2160 is the fewest pixels you need to capture to accurately represent all the frequencies you can include in an image 1080 pixels tall (This article explains why). So the a7S captured 4K, filtered-out (blurred) everything that couldn't be conveyed in a 1920 x 1080 video, and then downsampled to produce some of the best 1080 footage we've ever seen.

An '8K' camera could output cleaner, more detailed 4K footage than anything we've yet seen

A camera sampling 8K could output cleaner, more detailed 4K footage than anything we've yet seen. Better still, creating one output pixel from four capture pixels means you're capturing all three primary colors at every location, so you could potentially output a 4:4:4 color signal over HDMI if you wanted. Though it's noticeable that the a7S didn't do this with its oversampled 1080, and most consumer recorders will only cope with 4:2:2 signals.

Does 10-bit capture mean 10-bit output?

And, just as 8K capture shouldn't be taken to mean 8K output, 60 frame-per-second readout shouldn't be assumed to mean 60p recording. The specs of the chip are rumored and hence hazy but if it's true that it can only shoot 60FPS with 10-bit readout, then I'd expect a camera that records 30p (or whatever rate the sensor can maintain in 12-bit mode). This is down to the difference between capture bit-depth (which is linear and can limit the system's dynamic range), and output bit-depth (which can contain any amount of DR). 10-bit output is pretty flexible, but 10-bit capture (with a maximum of around 10 EV of DR) is pretty limiting. So I'd expect video to come from 12-bit capture and then, ideally, be recorded and output as 10-bit files.

Venice-series/anamorphic

Of course this could all be wrong. The rumors might be completely incorrect or the idea of an a7S III (or a9S if componentry or the desire not to overshadow the a7R III saw it sold at a higher price point) could be wide-of-the-mark.

Maybe this chip could be destined for something in Sony's CineAlta range, with the 3:2 region used for anamorphic shooting. We doubt it, though.

For instance, the expensive, Stacked CMOS sensor used in the Sony a9 (or one very closely related to it) is almost certainly the one also used in Sony's pro-video Venice camera that costs tens of thousands of dollars. This new chip could be for a sister model for that range, with the 3:2 aspect ratio used to capture footage with anamorphic lenses, for instance.

Or maybe it'll be a different camera maker than ends up making use of the chip.

But, whether this 36MP still/8K video sensor turns up in a Sony a7/a9 series camera or not, it seems fair to say it could be the basis of one hell of a stills/video camera.