Panasonic to restart development of image sensors to fast-track 8K video

joe1946

Veteran
According to a report from the Nikkei Asian Review, Panasonic has decided to resume image sensor development in order to speed up the drive for 8K video. The company suspended its development program a few years ago, the report says, but now Panasonic is preparing to invest ¥10bn/$80.8m to get its R&D back up to speed.

The renewed effort is said to be aimed at producing sensors for 8K video output for its consumer and broadcast cameras, with production models expected on the market by 2018.


Recently, Panasonic has used its 4K video functions for a range of applications in addition to video, such as 4K Photo, where a still is extracted from a 30fps burst, and the Post Focus feature that allows users to change what is in focus after the picture has been taken. Presently, the stills extracted from clips made using these modes are only 8MP, but if that clip were recorded at 8K video (around 7680 x 4320) the stills would be in the region of 33MP, depending on the image proportions chosen. This is very much in line with Panasonic's previous promise that we will be extracting 33MP still images from its cameras by the Tokyo Olympics in 2020.


The report says that the introduction of 8K video in cameras will eliminate the boundary between videos and stills. It will certainly change the way images are created in quite a dramatic way.
http://www.dpreview.com/articles/30...pment-of-image-sensors-to-fast-track-8k-video

2018 is only two years away. 8K GH6 ?
 
2018 is only two years away. 8K GH6 ?
I wonder if that article is putting two and two together and making seven......?

The concept of a 33Mp sensor as a "universal" video and stills sensor has been raised before - but with the 8K aspect applying to STILLS, the video being a very easily derived 4K.

The idea is to have a sensor exactly (2x3840)x(2x2160) - 7680x4320. (In other words, 8K) For stills you deBayer in the normal way, for video you derive simply - reading 2x2 Bayer blocks to directly get R,G,B values for high quality 4K via simple processing. The technique is exactly as used in the C300 to get high quality HD with simple processing from a 3840x2160 chip - but here scaled up by a factor of two on each axis. (If used in a hybrid video/stills camera, the C300 chip could only give 8Mp stills.)

Yes, with a full deBayer at frame rate, that 7680x4320 sensor could give 8K video - but you're then talking about truly massive processing loads in all aspects. I just don't see that in a "GH6" in just two years time - but very high quality 4K video and 33Mp stills, yes. (And maybe a very short duration burst of 8K "video"?)

There are a couple of things to think about. Firstly, still camera sensors tend to be 3:2 aspect ratio, with the 16:9 video mode being a crop of such. So if we assume the 7680x4320 is such a crop, it makes the full active sensor area about 7680x5120 - so about 40Mp total. Secondly, what about photosite size? If a 4/3 sensor, then 40Mp is going to make for individual photosites being very small - so I'd suspect to see a movement to at least s35 size and preferably full frame 35mm.

Is such possible in two years time? Well, the Sony A7R ii has now been out a while with a 42Mp sensor - and yes, full frame to keep the photosite size up. The dpreview test shows that 4K off a sensor crop is actually better than off the whole sensor, so it's not maybe yet performing as well as such a sensor may be theoretically capable of. (See http://www.dpreview.com/reviews/sony-alpha-7r-ii/9 ) and in stills burst mode it's nothing like 30fps even for a very short period. So it shows what may be possible soon - with 42Mp sensors already here - but I really don't see full 8K video in a camera of this price range for a lot longer than two years....... Bear in mind there's little point in 8K at framerates less than 50/60fps, ideally double that.
 
A 33-42 MPX sensor would deliver a an RGB (444) 4K off 26.4 MPX. It'd be in a crop mode, a la several current models.

With A7RII, Sony decided to make it an s35 size in order to make it more compatible with all s35/APS-C pro lenses out there. With the full frame coverage and a faster readout, the crop could be larger and the same 4K RGB in all aspect ratios (1.33, 1.66, 1.79, 2.0, 2.4).

Of course, I expect to see this with Nikon D5 (with Samsung sensors and processors inside) in a month's time, not in 2018.

As to Panasonic, they're presenting further development of an organic sensor (the public conference in February, I believe). On immediate MFT's lineup, they could also go the Samsung route - unless Samsung has a total non-compete clause with their Nikon deal - or with any custom designed chip from a variety of third party manufacturers. A key question here is that, much like Samsung, Panasonic is so huge, a single digit market share of photo-video industry is a spec on their ledger. I figure, they'll probably experiment with a full frame/4K model based on Leica SL design but, if they can't translate this into a full fledged consumer and pro ecosystem, they may have to consider dropping this portion of their portfolio altogether. In the M&A territory, there are likely to be more that a few interesting parties in China. Xiaomi, Huawei, Heier, Hisense, Tsinghua/Seiki, etc. can all be dipping their toes in the camera biz.
 
A 33-42 MPX sensor would deliver a an RGB (444) 4K off 26.4 MPX. It'd be in a crop mode, a la several current models.
Don't follow your maths (not if a Bayer sensor, anyway) - surely to get 444 4K you are going to need at least 7680x4320? You will only get it off about 25Mp if there are an equal number of R,G,B photosites on the sensor - but I was assuming we're talking about a Bayer sensor?
 
Don't follow your maths (not if a Bayer sensor, anyway) - surely to get 444 4K you are going to need at least 7680x4320?...
F65 and Alexa65 do it off 17.6 MPX in 4:2:2 (8.8 + 4.4 + 4.4), albeit with slightly different masks. One pixel per color would deliver 8.8 x 3 = 26.4.

Of course, that's presuming no additional sampling of the same data, which would require more processing power, or that the Bayer mask isn't more unique a la F35 (which was ... eh, a 3 x 2 matrix).
 
4K/8K will not soon eliminate photography equipment or lift the line between video and photography, for these hard reasons:


1 -Mechanical Shutter. A professional photographer will never take a fully electronic shutter that has rolling artefacts, light bands, and flash bands.

2 -Global shutter is the way to solve this, but it hugely affects image quality, taking us to the next point:

3 -Lowlight performance. To make an 8K FF Global shutter camera shooting video, lowlight performance will be far worse than an 8K FF Mechanical shutter stills camera.

4 -Compression. Being a video camera, the images are going to be degraded enormously by compression and wrapping, therefore unless 4K/8K cameras shooy uncompressed RAW video sequences, then it's unacceptable. And if it does, it takes us to:

5 -Data rates. 4K/8K Uncompressed RAW FF 24p is HUGE amount of data, HUGE. Almost no photographer is going to cope with such data rate ever, due to cost and simple storage and media size, which takes us to the next point:

6 -Camera size and price and weight: an 8K Camera shooting RAW off a FF sensor (an equivalent to a D810), is at least ten times as large, and is not handholdable in anyway. The only 8K camera we've seen, From Canon is a C300II body, with enormous fans, four stacked 7" CD Odyssey Q7+ recorders carrying SSDs on a full rail system, and that's even only an APS-C 8K camera. A Red Weapon Vistavision is also unusable huge for stills photography applications. It's simple physics, a D810 shooting at 24p is going to be a huge camera.

7 -Autofocus systems, the lack of mechanical shutter and mirror assembly forces full reliance on sensor based AF, which is to this point, is far lower quality AF systems for stills photography (even Canon PDAF) vs DSLRs high end AF systems with a separate sensor. They are faster, momentary, accurate, and track better at extreme speeds.

8 -The lack of optical viewfinders is a huge, huge deal, in itself a dealbreaker to many

9 -The lack of Flash systems, that work and shoot at 24p, and even if they were to be invented they'd be humongous and require enormous power, and still don't work because of rolling shutter. Continuous light is going to be their only choice. A dealbreaker.

10 -The power the camera system needs, at least a huge V-Lock Battery mounted to a plate on the back, vs a small LP Canon or Nikon battery that hides in the grip.

11 -The ergonomics and size of an 8K cinema camera to shoot stills vs the ergonomics of a DSLR for shooting stills.

12 -The huge cost increase in camera price and storage and power and processing

13 -Lastly, the simple lack of need. The vast majority who shoot all types of photography find 5-8fps more than they need, and only one small genre of photography needs that concept which is extreme sports: and those have sports cameras like the D4s/1DX and more coming. Stills cameras are extremely fast now that we don't need to shoot video to get that spray and pray style that video offers, sports cameras that are much less expensive, smaller, better to use for stills, shoot 14fps at raw, not far from 24p and certainly more than enough. Even a tiny 7D shoots 10fps RAW with all the advantages of a stills camera to a stills shooter. That's only going to to get even better and stills cameras are going to improve each year reaching 15ish fps speeds. No need for the extra 24p when it comes with all those huge downsides.

***So does it mean there is no place for high resolution photo-based video aquisition with 4K/8K cameras? Of course there is!
They will be used for many great applications. Mostly:
-The simultanuous shooting of photography-quality stills whilst shooting video/film projects and ability to offer both at the end. This is an advantage to videographers/filmmakers and not really photographers.
It's an enormous bonus. I deliver 8mp stills from video and clients are simply shocked, they adore the extra photo book.
Therefore, I expect this technique and usage of video for stills aquisition is only going to be for videogeaphers rather than photographers, who will continue using dedicated photography cameras and will never give up their convenience, unless a company makes:


A video camera, that shoots 24p 4K/8K FF uncompressed RAW, using a Global shutter sensor yet deliver the same IQ and lowlight performance as highest end stills cameras, having a top notch EVF system that compares to OVFs, and an AF system much better than Canon PDAF, all to small Card-sized media and batteries, with a flash system that can coup with it and still have the same size and power, and a body that has the same stills ergonomics.


This camera is not going to happen anytime soon.


And the shooting of video by photographers is not going to happen on more than an experimental stage, until and if technology of video cameras really out run the technological advances of stills cameras and the previously mentioned camera is built, and I doubt it, as by then, they'll be even greater stills cameras, shooting 20fps raw with mechanical shutters and OVFs and amazing image quality, rendering the use of video useless aside from the simultaneous acquisition of both media types.
 
By the way, to correct some of my thinking - 33 MPX 4:2:2 Bayer will be 17.6 green MPX and 8.8 of red and blue each, the later being "full" 4K but the effect being closer to 6K 4:2:2. To go 444 at 26.4 would require a different color filter. The other option is to go Bayer 4:2:2 with 17.6 MPX in the s35 mode (prorating to full frame, it'd be a ~ 50 MPX sensor), which would deliver 4K similar to F65 and Alexa65, as mentioned previously.
 
4K/8K will not soon eliminate photography equipment or lift the line between video and photography, for these hard reasons ...
The high end photography and video will have their own small niches in what is becoming a very small niche industry. With that there will be "cameras for every occasion". You want low light? Here's a medium format camera with 16-20 MPX BSI sensor. High frame rate? High resolution? A tailored Bayer mask or lack thereof? Everything off the shelf.

The consumer will be taking selfies with their smartphones and the enthusiasts/pros will have a handful of jack-of-all trades that will be suitable for a vast portion of their adventures/assignments in one body and arguably one lens.
 
The other option is to go Bayer 4:2:2 with 17.6 MPX in the s35 mode (prorating to full frame, it'd be a ~ 50 MPX sensor), which would deliver 4K similar to F65 and Alexa65, as mentioned previously.
I don't follow that - "4:2:2" defines colour resolution as equal to luminance vertically, but half horizontally. Fundamentally asymmetrical.

But that's not true of either Bayer 17.6Mp sensors, or the 17.6 Mp Q67 design as used in the F65.

http://content.provideocoalition.com/uploads/Q67_rev_2_thumb.jpg

In each case chrominance resolution is less than luminance - but symmetrically. The difference is the same horizontally and vertically.
 
Back
Top