IMEC codevelops 4k/2k CMOS 12 bit sensor w/Panasonic for 2/3 inch

[Tom Roper]

Looks set to launch in 2014, Panasonic joins the 4k party


Imec news-imec

[Glen Vandermolen]

I read the article, but where does it say the chip will be a 2/3"?

[Philip Lipetz]

The pixel pitch is 2.5 microns, less than one quarter that of the BMCC and BMPCC.

[David Heath]

Originally Posted by Philip Lipetz

The pixel pitch is 2.5 microns, less than one quarter that of the BMCC and BMPCC.

Really what is likely to be crucial is the implementation - whether they get used in a 3chip fashion, or as a single chip, with something like a Bayer filter.

Smaller photosites compromises sensitivity, adding Bayer filtration to a single chip compromises it even more (relative to the same basic chip used as a 3 chip design with beamsplitter).

If used as the foundation of a 3 chip design this could be very interesting, if used with a Bayer filter as single chip I'd say much less so.

Compared to an s35 chip with similar photosite count, the resolution, chroma aliasing, etc will be almost identical - but at an eighth the area the native sensitivity will be down by at least three stops!

The other issue comes down to something Alister brought up in another thread - http://www.dvinfo.net/forum/digital-...ml#post1808242 - diffraction limiting. That already causes severe limitations in 1/3" HD cameras, and 2/3" at 4K is likely to be comparable to 1/3" at 1080.

[Tom Roper]

Originally Posted by Glen Vandermolen

I read the articles, but where does it say the chip will be a 2/3"?

At 4000 pixels and a pixel pitch of 2.5 microns, that's 10mm, which is basically a 2/3 inch sensor. 2/3 inch sensors are 9.587mm x 5.393mm

[Glen Vandermolen]

Originally Posted by Tom Roper

At 4000 pixels and a pixel pitch of 2.5 microns, that's 10mm, which is basically a 2/3 inch sensor. 2/3 inch sensors are 9.587mm x 5.393mm

Thanks. I have no idea what formula you used, math never being my strong suit.

So...4K, 2/3" 3-chip cameras? Perhaps a single chip camera, like the HPX600?
Go ahead, David, tear it a new one.

[Tom Roper]

Originally Posted by David Heath

Really what is likely to be crucial is the implementation - whether they get used in a 3chip fashion, or as a single chip, with something like a Bayer filter.

Smaller photosites compromises sensitivity, adding Bayer filtration to a single chip compromises it even more (relative to the same basic chip used as a 3 chip design with beamsplitter).

If used as the foundation of a 3 chip design this could be very interesting, if used with a Bayer filter as single chip I'd say much less so.

I agree.

The other issue comes down to something Alister brought up in another thread - http://www.dvinfo.net/forum/digital-...ml#post1808242 - diffraction limiting. That already causes severe limitations in 1/3" HD cameras, and 2/3" at 4K is likely to be comparable to 1/3" at 1080.

I think diffraction limiting is not affected by size of the photosites, it is the phenomenon of light bending as it passes around the edge of an object (iris). The amount of bending depends on the wavelength of light and the size of the opening. As the opening gets closer (i.e. smaller) relative to the size of the wavelength, the bending is more. It doesn't care how small or large the photosites are. So at 2/3 inch, diffraction limiting would be the same as for a regular 1080 panel of the same dimension.

[David Heath]

Originally Posted by Tom Roper

I think diffraction limiting is not affected by size of the photosites, it is the phenomenon of light bending as it passes around the edge of an object (iris). The amount of bending depends on the wavelength of light and the size of the opening. As the opening gets closer (i.e. smaller) relative to the size of the wavelength, the bending is more. It doesn't care how small or large the photosites are. So at 2/3 inch, diffraction limiting would be the same as for a regular 1080 panel of the same dimension.

I *THINK* that whilst the first part of your argument is correct, (that the actual amount of diffraction is a function of the size of the iris) then the finer the photosite array, the less is the level of diffraction that can be got away with before it becomes objectionable.

That's why 1/3" chips were never seen to have any large diffraction limiting problem when they were SD, but did come HD. In each case, at a given aperture, there's the same amount, but whilst it may have visibly softened an HD image, it may have been still better than fundamental SD resolution.

Hence for 1/3", diffraction may only have become a problem stopped down below about f8-f11, with HD it's more like f4-f5.6.

Glen - you seem to find the physics inconvenient. I'm sorry, I don't write the laws.

But with everything else equal, a single chip camera can't perform as well as one with three of the same chip, but beamsplitter instead of Bayer filtration. It will be over a stop down in native sensitivity. Doesn't matter who makes it, what the chips are, that's fundamental to camera design.

Same with the diffraction argument above. Tom has raised a very sensible point and I hope I've answered it. I still think that for 4K, and a 2/3" sensor, then diffraction limiting will start to become an issue at similar f stops to we currently see with 1/3" and HD, typically f4-5.6.

If you think the theory of my argument is flawed, then please explain why you think it to be the case.

[Glen Vandermolen]

Sorry, David, couldn't resist.

So, can we use the JVC HMQ10's performance as a barometer for a 2/3" chip 4K cam?
The HMQ10 has a single 8.3 million active pixel 1/2.3" CMOS chip. Smaller than a 2/3" chip, but I think it's the closest we have for a comparison. I understand the HMQ10 operates under a very tight F-stop range to get usable 4K (or quad 4K) video. Plus, I've heard it's not the best camera in low light conditions.

[Tom Roper]

Originally Posted by David Heath

I *THINK* that whilst the first part of your argument is correct, (that the actual amount of diffraction is a function of the size of the iris) then the finer the photosite array, the less is the level of diffraction that can be got away with before it becomes objectionable.

That makes sense, I see that now. In other words, the light is being bent a finite amount, but for the same amount of deflection means 2 pixel error on a 4k panel that would be only a 1 pixel error on a 1080 panel.

[David Heath]

Originally Posted by Glen Vandermolen

So, can we use the JVC HMQ10's performance as a barometer for a 2/3" chip 4K cam?

The HMQ10 chip is a bit less than half the area of 2/3", and if 8.3 million pixels that's about the same count - 3840x2160.

Hence a 2/3" 4K should be expected to perform something like a stop better.

Originally Posted by Glen Vandermolen

The HMQ10 has a single 8.3 million active pixel 1/2.3" CMOS chip. Smaller than a 2/3" chip, but I think it's the closest we have for a comparison. I understand the HMQ10 operates under a very tight F-stop range to get usable 4K (or quad 4K) video. Plus, I've heard it's not the best camera in low light conditions.

Well, I've always understood the limit for 1/3" at HD to be about f4-f5.6 before noticeable softening, and theory gives that to be the figure for 4K with 2/3". Hence theory also then predicts about f2.8 for the HMQ10. Since max aperture is only going to be around f2, you can see why it may be said to have a limited usable f-stop range! So theory backs up what you say.

Go the other way, and s35 is about 8x the area of 2/3". So by the same logic, expect diffraction limiting there around the f11-f16 point for 4K.

[Jack Zhang]

The one thing that caught my eye was native 60p right off the bat.

However... This will still come with the consequence of the rolling shutter. Remember the HPX300? Let's all hope that the incredibly high amounts of skew of that first gen CMOS cam does not carry over to this first gen sensor.

One thing we're guaranteed to not see coming out of anyone is affordable Global Shutter sensors anytime soon.

[David Heath]

Originally Posted by Jack Zhang

The one thing that caught my eye was native 60p right off the bat.

I'd have been surprised if it had been any less. Current thinking is that 4K needs at least 50/60fps, otherwise the static resolution is being largely wasted.

As far as global shutter goes, then the F55 gets it's global shutter by virtue of a drop in sensitivity (compared to the F5). That may be reasonable on larger sensors with much higher native ISOs to play with, but far less feasible the smaller the sensor.

[Jack Zhang]

And the current pricing on Global Shutter sensors is prohibitive. The big camera manufacturers will only take it seriously when it can be produced cheaply and in mass quantities, while still maintaining decent sensitivity. Who knows, this hurdle may not be overcome for the 1080 generation...

[Alister Chapman]

Might be an interesting camera for those needing to shoot 4K run and gun, but sensitivity is likely to be an issue, you really want a small aperture for run and gun for a deep DoF but diffraction effects will force you to use larger apertures. My bet would be for a single chip bayer design. I don't know what the practicalities of creating a high enough quality prism are? Might be too expensive.