Hello,

Just to throw this up for discussion... I am noticing that with the release of the Panasonic AF-100 and the Sony F3, that the camera makers can make excellent images from large sensor cameras for affordable prices.

But wait, the 2/3" chip cameras were always the "big sensors" on the block and they can be quite expensive. It seems odd that the AF-100 could cost ~$5000 and offer more sensor area than an HPX3100 which costs $20,000. If Panasonic made a 2/3" version of the AF-100 would it cost $3,500? Sure one could argue about the quality differences but how far apart are they?

Strange days. The AF-100 and the upcoming Scarlet are going to make the 2/3" chip large format cameras seem un-invited to the party. I see people adapting 2/3" zoom lenses for the AF-100. Just seems like tings are going upside down. I do not think Panasonic will make a 2/3" version of the AF-100 but it does seem like some kind of handwriting is on the wall.

Your thoughts on the future?

For a lot of productions the 2/3" cameras, which allow relatively compact, large zoom range zooms, are ideal. I suspect some of the current ENG/EFP type video zoom lenses will be made in single sensor versions for the Scarlet, just as Canon did for 16mm.

It's the desire for a shallower DOF in sections of the video market, which is driving for the larger sensor camera.

Manufacturing the large sensors is probably no more difficult that registering 3 sensors on an optical block and I suspect it's now actaully easier, once you've got a large enough scale of production.

Checking the horizontal resolution specs on a couple 2/3" cams vs a 4/3 cam:

Panasonic AG-HPX500: 1080 Lines
Sony PMW-350L: 1000 Lines or more
Panasonic AG-AF100: 800 TV Lines

So maybe there is significant expense in getting RESOLUTION, regardless of chip size.

Ken

Brian, interesting point on creating eng zooms for a single chip application. I wonder if a conversion could be done for exisiting 2/3" lenses if needed?

Ken, I think good stuff always seems to be expensive but the AF-100 is showing to be very good and inexpensive...and have larger chips at the same time. BTW, the HPX-500 gets ~800 lines of resolution with it's pixel shifted chipset. Its bigger brother the HPX-2000 might get more.

Cameras like the EX-1 and Canon's 300 series have 1000 lines of resolution for under $7,000. With this state of the market right now, Scarlet seems to be in a perfect position if RED can deliver. 2/3" chip, 120 fps in 720p, at least 1000 lines of resolution & RAW processing for under well $10,000.

On the other hand, if a company like JVC could release a single chip, 2/3" version of the mini shoulder mount with the 50mbps codec I bet it would be quite a hit. And I wonder if the margins for a camera like this would be less, equal for more than the traditional 3-chip $20,000+ bodies?

I'd be very interested in a JVC with a 2/3" chip in an HM700 body. But I doubt that'll happen. Or how about a Canon XF300 with a single 2/3" chip? It already has the codec, and a 1/3" single chip cam.
I wonder why there aren't any single 2/3" cameras. Anybody with a technical background who can answer that?

The Scarlet has a 2/3"single sensor, as does the SI 2K.
In the past, Sony made 2/3" single tube colour Betacam cameras.

I'd be very interested in a JVC with a 2/3" chip in an HM700 body. But I doubt that'll happen. Or how about a Canon XF300 with a single 2/3" chip? It already has the codec, and a 1/3" single chip cam.
I wonder why there aren't any single 2/3" cameras. Anybody with a technical background who can answer that?

There is. The Sony F23 is a single sensor 2/3" camera. It was used by Michael Mann on "Public Enemies".

There is. The Sony F23 is a single sensor 2/3" camera. It was used by Michael Mann on "Public Enemies".

According to the Sony website, the F23 has 3 2/3" imagers.
I knew about Scarlet, but who knows when that'll see the light of day - if ever.
I'm just wondering - since one chip should be cheaper than 3 - why there aren't a lot more one-chip 2/3" cameras. Would a B4-mount lens not be suitable with a one chip arrangment?

Most B4 mount lenses are designed to work with three chips. So an adapter would need to be put in place. Scarlet will need this adapter when it comes to market in order to use B4 eng lenses.

I knew about Scarlet, but who knows when that'll see the light of day - if ever.


The Scarlet looks like it'll be out later this year, possibly just after NAB at the earliest. Certainly working examples are being demoed. How quickly you'll get one is another question given the potential order back log.

Unfortunately, announcing cameras just as the manufacturer come up with the product idea doesn't really help when customers just need the camera and want it yesterday. I guess it does show the man hours involved in developing these products. Although feature growth also adds to the development time scale.

It's important to distinguish between big chips that are designed specifically for video (such as the F3) and cameras which use big chips primarily designed for high res stills.

The key difference is pixel count on the chip. Too few and the resolution won't be adequate, too many, and compromises have to be made for video. The optimum for 1080 output is somewhere around the 3.5-4 megapixel range.

Think about it. If we're looking at the 15 megapixel range, it's extremely difficult to read all the pixels off the chip at up to 60 times a second, then process them, then do a decent quality downconvert. Not impossible - but will be extremely expensive and make the camera power hungry.

Except for the very top end, the way high pixel count chips get used for video is by either ignoring a large number of the photosites totally, or by "binning" - simply combining the charges from adjacent sites BEFORE processing. Or a combination of such techniques. Do it such a way, and it becomes a very good idea to use an optical filter to prevent excessive aliasing.

A common technique is to pool a 4x4 block of photosites to give a single "super-pixel". With a Bayer sensor, it will have a green value made from 8 green photosites, a red value from 4 red photosites, and blue also from 4 blue photosites. It has the advantage of being cheap to implement, and uses low power. Resolution wise, a typical such chip may be about 4,600 photosites horizontally, so this technique will give a resolution a quarter that, or about 1,150, so suiting 720p system resolution well enough.

As far as the future goes, then it is lenses that will be the limiting factor. It's expensive enough to make a f1.8 servo 16x zoom with a decent wide angle coverage for a 2/3" camera, let alone size/weight issues. For a sensor of four times the area, it all becomes prohibitive. The large format video cameras may establish a niche for work where the use of fixed lenses is acceptable, for other types of work they are a non-starter.

Thanks for your replies.

My main thought is for chips made for video. I do not know the econonies of scale between using chips for dual purposes etc... I am not a chip maker. But It would seem that making a 4 megapixel 2/3" single sensor would be affordable enough to offer a 2/3" video camera for under or well under $10,000

If the AF-100 has a chip much larger than a 2/3" chip for $5,000 one would think a smaller chip would cost less. What makes this a difficult situation is that the 2/3" size is kind of "off limits" because there are existing $20-$35,000 models on the market that are trying to be sold.

As a consumer, I am willing to accept the little bit less image quality from the $25,000 models to have the functionality of a 2/3" chip in a lower cost of a single CMOS implementation. Will a manufacturer meet me in this place? So far Scarlet seems to be the only one.

If the AF-100 has a chip much larger than a 2/3" chip for $5,000 one would think a smaller chip would cost less.
I believe the big cost is in the intial design - so the more you make and sell, the cheaper they become per unit. So the likelihood is that those initial costs for the AF100 chip are largely met by the ones produced for the stills cameras, the main extra work for the AF100 is the addition of the optical low pass filter to get rid of the worst of the aliasing problems.

Making use of existing multi-mega pixel chips for stills use is sensible from a point of view of keeping cost down, but it's not the best solution purely technically. Not, at least, with present technology.

As far as a single sensor 2/3" chip goes, it isn't likely to be as good as a conventional 2/3" 3-chip arrangement. It can't be because of the light lost in the single chip filtration, let alone the Bayer pattern being more complicated to process than 3 1920x1080 chips. Hence there is far less impetus to design something like a designed for video 3.5 megapixel sensor, hence little likelihood of good economies of scale. Equally, a 2/3" chip falls a bit between stools for stills use - not up to the DSLR market for quality, but too big for compacts.

So that's why you see quite a lot of smaller than 2/3" single chip cameras (they keep the size/weight/price low) and you see bigger than 2/3" single chip (a 3-chip design at that size starts to become unmanageable). 2/3" falls between those stools - you could use single chip, but it's more satisfactory to just use a 3-chip design.