Sumix 2/3" 1920x1080 CMOS

[Régine Weinberg]

Well done
The APS format sensor is THE idea.
But it is a mass market product for the photo industry.
Let face this
The Nikon D3 is the only,
since the D1, full format sensor they have.
Why.
Less pixels, bigger pixel an a full format chip is the only way to have less digital noise. You can take pictures up to crazy 12000 ISO, you need this for taking pictures of live acts, or in door shoots in small bars etc.
The sister product the d300 has quite the same features but an APS chip and ISO can be pushed only up to 3200 for a asking price a fraction of the D3.

Aps is a billion mass market product, we even don't need 12 Mpix.
Our dream chip would be 5 K big pixels to have low noise and 30 fps. Could be done, sure.
But we are simply no multi million market and nobody ever will do such a silly crazy product. In the sense of marketing.

The money behind red is OAKLEY no kidding.
Same story as Ubuntu,
not perfect but very nice to use.
Not exact what we are crying for
but the best we can get.

[John Wyatt]

In the struggle to come up with a low cost high quality camera system, for people who would rather be filming than building portable computers for their cameras, is it possible to make some brief lists? To decide what we would like (would need some discussion), what we have now (would need some research), and what exactly needs work (would need some community!). If we can agree on what we reasonably want, we can quickly find out what's already available, and then concentrate on what are the problems left to be solved in order to bring it into reality. I know Jose and Daniel have been trying to achieve this with the help of the Sumix Corporation on this very thread, but I'm still not clear in my own mind what the problems are and what we should be looking at as a priority to bringing it closer. Not helped by the sheer variety of problems, such as turnkey or open source software, limitations of frame grabbers, how to power a micro computer with camera batteries, best lens mount, ever changing technology and standards for motherboards and sensors, preferred onboard storage strategies, RAW or compressed, 2k or aim for the future with 4k, LCD monitors you can't see in daylight, minimum orders for sensors and non disclosure agreements. It's endless. The fact that SI have battled the same issues and come up with a camera I'd love to own but will probably never be able to afford, makes me wonder if it's ever going to be possible before this decade is out (it really does seem like we're trying to get to the Moon sometimes!), to finally get the camera we need to make no budget movies. It is such a deep problem, you can easily get caught up in the building and almost forget why you wanted a camera in the first place. But perhaps a big problem like this can be broken up into smaller parts and solved by different people. Maybe a community mission (like Linux) can solve it; I just wonder how long it will take before we are over taken by events, and non film makers are shooting their kids on the lawn with that 3k camera...

[Régine Weinberg]

http://www.toshiba.com/taisisd/indme...tail_ikhd1.jsp
look at this

[Daniel Lipats]

There is much more to getting a complete camera than just hunting for the perfect sensor. The Sumix camera project is a very good example of this. The hardware, software, and optical aspects need to be addressed. A weakness in any of the areas make it all just an expensive paperweight.

There have been a variety of problems with getting a Sumix based camera running. Obstacles were in communication, technical, and implementation.

I have reached the limits of resources to invest into this project. I will take my losses and quit. Tonight I will be requesting a return of the camera to Sumix.

I regret having to do this but I don't see any other options. This is just not going in the direction I had in mind.

[Jose A. Garcia]

Règine, I've seen that camera many times. It's quite expensive, has 3CCD (I'd rather have just one sensor) and does just 1080i.

[Paul Curtis]

Originally Posted by Jose A. Garcia

What about this one?

http://download.cypress.com.edgesuit...upa_4000_8.pdf

It's a square sensor but its width is close to APS-C size. It says 15fps at full res. Could be 25-30fps at 1152 or 858 pixels of vertical res.

Anyway we would need a way to control the sensor.

Jose,

Yes it's quite interesting but the QE (Quantum Efficiency) and the FF (Fill Factor) together seem very low. And that's to do with how good the sensor is at converting photons to charge. So despite 12 micron pixels there must be a lot of on pixel circuitry to cut that result down. CCD on the other hand has the whole pixel so QE tends to be much larger.

You must remember by the time you've whacked a bayer mask over it the sensitiviy nose dives again.

That's one benefit of 3 chip systems, not only more sensitive but also the ability to apply analogue gain on the channels individually which makes for better colour balancing (tungsten vs daylight)

Even so, it's an interesting sensor heading in the right direction...

Im going to read the datasheet in more detail once i've had more coffee.

cheers
paul

[Paul Curtis]

Originally Posted by John Wyatt

In the struggle to come up with a low cost high quality camera system, for people who would rather be filming than building portable computers for their cameras, is it possible to make some brief lists? To decide what we would like (would need some discussion), what we have now (would need some research), and what exactly needs work (would need some community!).

John,

It's a nice idea but im not sure how practical is it. There are so many components and each component is evolving at a different rate (PC hardware changes very quickly, solid state is become more practical etc,.). Or new third party hardware makes something previously impossible, possible (thinking about the hard drive recorders about to appear or SxS as a recordig format)

It seems that every now and again some intrepid folk start off on a camera project, learn a lot and give up. One day im sure someone will crack it but i suspect the personal cost would be higher than just buying a camera, it's a work of passion more than finance.

This forum is a great resource and anyone starting out should go back and read old threads and then 'stand on the shoulders of giants' and learn from past mistakes.

SI have sunk an enormous amount of time (money) into software development. I doubt an indivdual has that resource. The fact that this development cost is split amongst many camera sales. I would hope that Jason et al are in it for the long term and their software evolves along with the sensors that are available.

On the software side leveraging existing solutions is the only way forward, use of After Effects workflow (doing log to linear) or even the adobe raw convertor (which has been used for vfx work anyway). The new adobe DNG format could be the one to watch because it includes the kind of RAW metadata that is required for moving images. Or perhaps cineform RAW will open up and become a public standard.

But these solutions are an individual system, you couldn't build a busines selling a camera to 3rd parties in this way. These solutions might be a single production job, come the next one and the playing field has moved again.

I hope to document my findings when i had a go at all this, but im in the middle of a production right now so time is limited!

cheers
paul

[Farhad Towfiq]

Hello all,

We are understanding what film makers need a little better. We will try to make the software much easier to use before encourage another DIY integrator to touch the camera head. Extra software features, like automatic look up table is already added. At the end the purpose of this camera head is to be used by artists who want to set their own setting and compose their own Look up tables. The purpose of this camera head is to allow talented and technically capable film makers to differentiate their work by controlling the sensor and post processing themselves. The extra effort can pay back for them not only in producing unique result, but also offering their integration work to other film makers for a price. Michal Dell was putting together computers for himself when his friends asking to purchase them from him and started his Dell computer business.

As far as sensor is concerned, Altasens is still by far superior to anything else in latitude and high signal to noise ratio. The reason Red is pushing higher pixel count is because it is easier technology for them. They are playing Houdini on you as he upped the challenge by suggesting that he would open the safe from inside. The real challenge is larger and more light adsorbing and less noisy pixels and not their plurality

[Gottfried Hofmann]

Originally Posted by Daniel Lipats

There is much more to getting a complete camera than just hunting for the perfect sensor. The Sumix camera project is a very good example of this. The hardware, software, and optical aspects need to be addressed. A weakness in any of the areas make it all just an expensive paperweight.

There have been a variety of problems with getting a Sumix based camera running. Obstacles were in communication, technical, and implementation.

I have reached the limits of resources to invest into this project. I will take my losses and quit. Tonight I will be requesting a return of the camera to Sumix.

I regret having to do this but I don't see any other options. This is just not going in the direction I had in mind.

This is sad. I'd like to know more about the current problems as I am very close to buying one. Same goes to Jose...

[Biel Bestue]

Originally Posted by Gottfried Hofmann

This is sad. I'd like to know more about the current problems as I am very close to buying one. Same goes to Jose...

yeah i agree, i'd like too

[Daniel Lipats]

Well in hopes of coming up with solutions, I will demonstrate...

One of the problems I faced when shooting with this camera is controlling light. It seems to be very sensitive to slight changes in light intensity. The picture below will do a good job of demonstrating:

http://www.dreamstonestudios.com/per...es/compare.jpg

Shot was intentionally composed to demonstrate the topic. Lit with a small light kit, and a par outside the window aimed at the vase.

The HV20 image copes with the scene just fine, there are no extreme highlights, detail is very well preserved on the corners of the table as well as under the vase. But the image is more or less flat. There is not a lot of difference between the brightest and darkest part of the picture on the table.

The Sumix image on the other hand is very different. There is a huge contrast difference on the table. Big highlight under the vase, and the corners of the table are dark as light intensity falls. The Altasens sensor is able to pick up a big difference between the darkest and lightest part of the table. It can detect slight changes in light intensity, but can not view a high range of brightness.

Unfortunately, this makes shooting anything more complicated, requiring more complex light setups and a higher degree of control. This is true especially outdoors. You have to stay in a small range of light intensity.

Strange enough, the images posted on the Silicon Imaging website for the SI-2K contradict my results. Without knowing what kind of light setup used in the shots its difficult to get a good idea of exactly how different the camera behaves, but it seems to have much less trouble with a range of light intensity.

[Jose A. Garcia]

I had the same problem too. I don't know why our tests are so different from SI2K tests. Both cameras have different Altasens sensors though, SI2K sensor is of course 2K and it's listed as "Professional Broadcast" but the Sumix one is listed as "Videoconferencing". I'm quite sure that's where the main differences are.

[Solomon Chase]

This is mainly a software issue. It is not mapping the values correctly. I have calibrated SMX12AC footage and it looks to have similar dynamic range to HV20 if not more, after tweaking.

The other part of the problem IS hardware related, but Farhad already mentioned it is because of a bad IR cut filter, that has been fixed.

[Daniel Lipats]

Solomon,

Do you mean tweaking as a post process to a RGB file?

[Solomon Chase]

Originally Posted by Daniel Lipats

Solomon,

Do you mean tweaking as a post process to a RGB file?

Remapping the levels with curves, and fixing the red/green color spectrum bleeding. I used After Effects.

Send me a raw HD size grab from the Sumix and HV20 and I can match them. (of course the color skew from the bad IR Cut filter intrinsically messes up the color matrix, but It can still look decent)

You have to expose for the top end (highlights) to get a perfect match though. It looks like the exposure on the Sumix is set a little too high on your example JPG.


Here's an example:
http://solomonchase.com/sumix/cc1.jpg