Big sensor to replace three chippers cams?

[Don Miller]

Red is a single chip camera? Is that the next generation? Canon and Sony are by far the largest and most experienced big cmos producers. Are the three sensor ccd cameras going away? Perhaps thinking about 1/3 v. 2/3 ccd is soon a thing of the past. I have a Canon dslr with video to the LCD that has a chip about the area of the Red sensor. So is Red just talking about what all the big guys will have soon?
I would like my control of DOF back, so this would be good.
Anyone know how big the image circle is on the two lenses for the XL H1?

[Greg Hartzell]

I'm by far not an expert here, but I'll post a response to answer at least a few of your questions:

I don't know what the specs are for your dslr, but I can almost guarantee that it isn't capable of outpting HD level resolution at even 24fps. Large sensors that can do this are expensive to produce at this point, which is why we still see the 3-chip method. (Question for any one else: Is there any advatages that 3-chips offer over one?) Also, with a single chip, you are essentially using one chip for three colors, effectively reducing the resolution by 1/3.

Red isn't the only camera out (or not out) with a single sensor. Panavision developed a single super35mm cine cam with Sony (Called the Genisis and used on Superman Begins), Arri has released a camera called the D20 (I saw a demo last week on how much of a diference there was between a cinealta and the D20 in terms of flare with a direct light shined in the lens, pretty cool) and Dalsa has a camera called the Origin that can produce incredible resolution. Dalsa is also a large semiconductor producer that makes cmos and ccds. Are ther more?

Kodak is the largest manufacturer of still camera images, they were also the ones that started the game. In todays world of outsourced manufacturing, it wouldn't be that difficult to design and build and market a single chip video camera, but I doubt that we'll see one that is practicle enough for the low end video market, and I don't know that it would be that desirable over three chips.

[Glenn Chan]

1- In practice, a lot of the theory out there is not that good at predicting image quality. As well, technical quality isn't a pre-requisite for beautiful/interesting images (think Holga photography). Or you have situations certain film stocks have their own aesthetic which may be desirable.... and it's hard to put a number to aesthetic look.

2- In terms of luminance and color resolution...
Most single imager systems will implement a Bayer pattern. To avoid Bayer aliasing artifacts, you can implement a high degree of optical low pass filtering (blurry glass in front of the sensor). This will reduce the maximum resolution possible, and also seriously reduce luminance aliasing artifacts. With the Red system, I believe Graeme was saying resolution of about 70% of the sampling rate.

There are many other single chip systems which don't implement as much OLPF... and they achieve higher resolution at the expense of aliasing artifacts. This tends to happen a lot in consumer 1-sensor cameras.

Color resolution of course isn't that high. A lot of the performance depends on the de-Bayering algorithm used.

2b- With 3-sensor designs, most of them implement pixel shifting. It's kind of Bayer-ish in that you get better luminance performance than color performance.

In my Panasonic GS70 (3CCD consumer cam), luminance resolution seems to be 4X+ color resolution. (This figure isn't very good though.)

3- (might be wrong here) To use 35mm film lenses, you can't use a 3-sensor design since 3-sensor designs require a prism to split up the light. The prism sets a limitation is the distance between lens and sensor.

[Don Miller]

Has Red claimed proprietary sensor technology? I assume it's a bayer sensor. As far as the issue of single chip quality, it seems Red has proven that's not an issue.
Let me play "devils advocate". Is Red just trying to get a jump on the big guys? Is it really going to be special, or just the evolution of video technology?

[Glenn Chan]

Has Red claimed proprietary sensor technology?

I'm not sure what you mean by proprietary (it doesn't seem to be useful to describe it as that / the distinction is not important). From what I understand, they are keeping their technology a secret (as opposed to patenting it), to avoid other companies ripping off their technology. That's why they won't tell you who is making their sensor.

Let me play "devils advocate". Is Red just trying to get a jump on the big guys? Is it really going to be special, or just the evolution of video technology?

I'm not sure what you mean.

[Don Miller]

"Proprietary" is the key word. They have to have advantages they can protect to be successful. The leaders in big, complicated cmos chips is Canon and Sony. The market for a 4K camera may be too small to interest the large japanese companies at this point, however.
I can see that the price/performance Red has announced is unique. I can't see where they have any special advantage in producing a 4K product.

[Lawrence Bansbach]

Originally Posted by Don Miller

"Proprietary" is the key word.

If by "proprietary" you mean patented or patentable, I would guess that nearly all imager technology is proprietary. An idea isn't patentable -- only its implementation.

Originally Posted by Glenn Chan

From what I understand, they are keeping their technology a secret (as opposed to patenting it), to avoid other companies ripping off their technology.

Has Jim Jannard actually said that the Mysterium isn't patented?

[Glenn Chan]

1- Sony announced that they plan on making a 4k camera.

2- I will see if I can dig up Jim's post about Mysterium...

[Don Miller]

Anyway, my point wasn't really to evaluate Red. My point/question was to discuss a possible shift towards single sensor cameras. The relative market size for a 4K camera is small. But what about a ~28x16mm sensor that does 1080p? That would return control of DOF and be low noise, considering the fairly low noise of the Canon and Sony single chip HDV cameras.
The "bookends" of new affordable stuff - Red on the high end, and the $1000 1080i cameras on the low - suggests to me a switch away from the 3 CCD 1/3 2/3 paradigm. My point of bringing up Red was to suggest that single chip is the industry direction. As opposed to to the possibility that Red has created new technology.
The name "Mysterium" suggests to me that there using a standard part. Like $200 sun glasses.

[David Mullen]

There are already cameras with 35mm-sized single sensors limited to HD or 2K recording -- the Phantom HD camera, the Panavision Genesis, the Arri-D20 (though it has more of a 2.8K sensor). The Genesis doesn't use a Bayer-filter, unlike the others, but a RGB striping system.

And there are some three-sensor cameras with 35mm-sized sensors, mostly experimental prototypes.

What's "proprietary" about the RED is not particularly it's use of a 35mm-sized 4K sensor -- afterall, the Dalsa Origin camera has already been doing that for a few years.

In theory, a single-sensor design is cheaper simply because you don't have three-times as many sensors to put in the camera, nor a prism-block. But you do have to deal with processing a RAW Bayer-filtered signal in real time into RGB for display devices, plus processing it into RGB for post, whether in-camera or later in post. So the computer processing, with the de-Bayer algorithm, may add to the costs saved by not using three sensors and a prism block, for all I know. At least for high resolution cameras where you are talking about processing a lot of data in real time.

And in theory, a 3-sensor design does provide more resolution for each color record, than recreating RGB from a Bayer-filtered single sensor.

Of course, a 3-sensor design with a prism block takes up more space, makes the camera heavier, which becomes a problem when the sensors get bigger than 2/3"...

I asked Sony why the new F23 HDCAM-SR camera uses three 2/3" CCD's instead of a single 35mm-sized sensor like the Genesis, which is otherwise very similar, and Sony's response was that their customer base is heavily invested in 2/3" optics, having bought millions of dollars worth of HD lenses over the past few years, so would be very upset if their latest HD camera now suddenly used 35mm lenses instead.

I think you are going to see more single 35mm-sized Bayer-filtered sensor cameras limited to HD/2K coming out, a few attempting 4K like RED and Dalsa -- but from smaller companies than Sony at first.

[Lawrence Bansbach]

Originally Posted by Don Miller

Anyway, my point wasn't really to evaluate Red. My point/question was to discuss a possible shift towards single sensor cameras. The relative market size for a 4K camera is small. But what about a ~28x16mm sensor that does 1080p?

A single larger chip would be great, but then you've got to deal with bigger, significantly more expensive lenses. Sure, you could use a set of fixed still-camera lenses, but who wants to give up zooms? Putting aside permanently giving up deep depth of field, any zoom lens would have to cover a larger imaging area and maintain exposure through the zoom range (or most of it). Unless I'm remembering wrong, a 28- x 16-mm sensor is actually a little larger than the image area of 35-mm motion-picture film (even Super 35). You can only imagine how big (and expensive) a zoom with a reasonable speed and zoom ratio would be. This would put the camera beyond the reach of most "prosumers." The lens alone would cost thousands.

[Don Miller]

Originally Posted by Greg Hartzell

...............................

I don't know what the specs are for your dslr, but I can almost guarantee that it isn't capable of outpting HD level resolution at even 24fps. .

Canon's newest DSLR will do about 100 million pixels per second. It's obviously not designed as a video camera. It also is doing full resolution with each capture, so it has a load that a video camera doesn't.

[Daniel Browning]

Originally Posted by Don Miller

Canon's newest DSLR will do about 100 million pixels per second. It's obviously not designed as a video camera. It also is doing full resolution with each capture, so it has a load that a video camera doesn't.

Yes, the 1d2m3 can pull a lot of data off the sensor, but it overheats after a while, so it wouldn't be any good for long-form recording (I only used it for 4 days, but I've heard that noise begins to increase after 5 minutes). It's a step in the right direction, though. I'd love to see greater convergence between still and video sensors.