2/3" vs 1/3"

It doesn't necessarily. His point is, I believe, that making a lens which can resolve all that detail for a 1/3" camera is much harder than a 2/3" camera. That and the pixels are physically smaller and thus have a lower signal to noise ratio.

Yep, bigger is better :D
 
Maybe an illustration would help?

You've got half-dollar sized pixels in a 2/3" chip.
You've got penny sized pixels in a 1/3" chip.

If you put penny sized pixels in a 2/3" chip, you would have space for a lot more pixels.
Conversely if you put half-dollar sized pixels in a 1/3" chip, you would have less space for them.

So, bigger chips have bigger pixels than a smaller chip with the same number of pixels.

As a generalisation, smaller pixels are less sensitive to light (don't do as well in low light as bigger
pixels) and that's a reason to buy a camera with larger pixels and larger chips, among others.
 
People get really hung up on pixel count, thinking that it has to be 1920x1080 or else it's "no good", or less than it can be, etc. Unfortunately, that's just not the way things work.

Bigger pixels are better than smaller pixels-- they're more light sensitive, they deliver better low light performance, better signal-to-noise ratio, cleaner video, less smear, they're just overall better.

However, if the pixels are too big, you don't have enough resolution. You need to have enough pixels to get adequate resolution.

And if you have a small sensor to start with, packing lots of pixels on a small sensor = lousy video. Whether it has "resolution" or not, it'll look awful.

Further muddling the issue is that you have three CCDs, but you're not recording 4:4:4 color with them. The recording formats available in these HD cameras records either 4:2:2 or 4:2:0. That means that in a 4:4:4 system, you'd need all three CCDs to have the same resolution and the same pixel density, and lined up in exactly the same position. However, in 4:2:2 color, that's pretty much just wasteful. The camera will only resolve half as much resolution in color, so why bother making 1920-pixel CCDs (and endure the sacrifices inherent with doing so) when the camera will immediately throw away half that resolution?

That's where "pixel shift" comes in -- offsetting the green ccd by 1/2 pixel from the red & blue. Using sophisticated processing the camera is able to deliver 1920x1080 resolution from a 960x1080 CCD. What do you lose? Color information that you're going to lose anyway due to the codec. So a pixel-shifted 960x1080 CCD could theoretically (THEORETICALLY) deliver all the practical, usable resolution of a 1920x1080 CCD, but gets to have pixels that are twice as big -- which gives better low-light and dynamic range than you would have gotten from the 1920x1080 CCD.

Furthermore, as Graeme has been saying, the lens limits the amount of information that gets through the glass anyway. There comes a point where the pixels are going to be so small, measuring down to maybe 4 or 5 microns... and the lens won't be able to resolve that much information anyway. So why bother going for overkill, making super-dense-pixel chipsets, if the lens can't deliver enough information to let the chips resolve it anyway?

In bigger cameras it's a no-brainer -- the bigger chip size means you can pack 1920x1080 on a CCD and still get big-enough pixels to get decent performance. Even so, Panasonic is using 1280x720 pixels on their new 2/3" HDX400 (which delivers 1920x1080 resolution, by the way) because it employs pixel shift in both the X and Y.

What I'm saying is, you can't just count pixels and then say "ah... more = better". It doesn't work that way. There are many, many things that go into the imaging chain. In general, you want pixels that are as big as you can get them, arranged in such density (and even pixel-shift) such that they can capture the maximum resolution that the lens can deliver to them. If you're recording 4:4:4 you want native resolution. But for 4:2:0 or 4:2:2, pixel shifting bigger pixels may result in better overall video.

As always, the proof will be in the testing.
 
As a point of reference the terms 1" video , 2/3" video date to the use of tubes in the cameras and represent the approximate outside diameter of the tube, not the active area used in forming an image. The image diagonal of 1" video is 15.8mm if memory serves me, 2/3" is of course 11mm.

For anyone who may not know, 16mm film is 12.8mm diagonal, super16 is 14.5mm, 35mm Academy aperture is 27.3 mm -- it is smaller than the old silent or full aperture (30mm) to make room for the sound track. There are several Super 35 formats but they tend to have a diagonal as small as the academy format or as large as full aperture but an aspect ratio wider than 4:3 ( 1.85 for example)

There were other larger tubes used going clear back to the Black and White Image Orthicon tubes that had a 40mm image diameter.

Bill Turner
Century Division
Schneider Optics
 
Also, all this talk of shallow depth of field is incorrect. 2/3" cameras have shallower depth of field because they usually have longer lenses, NOT because of the size of the CCDs...just thought I should clarify that. But this really was a great discussion! Peace.
 
Depth of field is a combination of film/sensor back size, lens focal length, aperture,
shutter angle/shutter speed and even the emulsion chosen if shooting film.

There are more factors than simply focal length.. This ought to be another thread entirely.

Brian
 
With the size of the pixels bigger in a bigger chip, does that not translate to it holding up a lot better on a large screen thus making a 2/3" chip better for use for theatrical release?

What kind of ballpark price differencial is a 2/3 chip camera to a 1/3 camera?(hypothetically pretending one could get the exact same camera just with different chips)
 
Here's what I don't get: how is it that a pocket-sized digital camera can have a 6 megapixel resolution (much larger than HD), but then, we can't have a cheap lens that resolves HD resolutions for video? What is the difference?
 
I get what you're saying thisiswells, but it still doesn't make sense to me... A Canon Digital Rebel (SLR) for example has a very high quality imaging system, which much less abberation than pocket cameras. Also, the color rendition on these SLRs is pretty good too, so why do we need three CCDs? To me it sounds like 1 big CMOS is the answer, but I'm sure I'm missing something.
 
Daniel--Your comment about 1 BIG CMOS is what everyone is the video community has been
shouting for the last couple of years since we knew about such a thing. Give it a couple of more.

I've concluded if it were possible they would be doing it. Things like refresh rate and ability to handle
high frame rates (like 60Fps in HD) come to mind as possible problems. I am terribly undereducated
on the subject and frankly I'm not a camera or lens designer. I'd recommend not making an
issue about the type of sensors, lenses, and cameras available now because it doesn't really
do much help... Just know that what they have and what we've got is all there is. There are
a couple of experimental 1CMOS cameras and a couple of mega-buck CMOS cameras but I
don't think you're missing anything. There aren't any real field usuable 1CMOS cameras.
 
I seem to recall seeing some software a few months ago that faked shallow depth of field. I think it analyzes your foreground vs. background elements and throws the background out of focus. Does this sound familiar to anyone? Does anyone know the name of it?

Thanks,
Ben
 
There's a lot of software like that but none of it works so well for video. The amount of motion tracking and masking would just be too intense.

Constantine, we all know that DOF is related to aperture setting and focal length rather than sensor size, but the cropping due to sensor size necesitates shorter focal lengths for small sensors. So while you may technically be right, it's simpler to think of it in terms of effect rather than cause. The cause may be the lens, but the effect is a shallowed DOF with the same image.

The canon's CMOS would overheat and die if it had to shoot that fast, and it can't even shoot over 5fps as it is. 2/3'' chips require lots of cooling in video cameras and they are much smaller than an APS-C sized chip. And point and shoot digital camera lenses may capture tons of resolution, but they have pretty major abberation issues and are VERY TINY and don't let much light in.

Rockwell makes a pretty sweet Bayer pattern CMOS for 700 dollars that's 2/3'' and offers full HD resolution. (1920X1080+). JVC will be using 3 of them for their new HDV camera, but what would be coolest would be a small single chip camera with one (sure, you'd lose some resolution...whatever) that doesn't use HDV and has a c-mount for 16mm lenses. That would be ideal so far as I'm concerned.

As for dynamic range/resolution/whatever...just compare 16mm film to 8mm film. Same difference. (Well, same idea. Sensors vary in quality just as film stock does and that's another independent factor.)
 
Back
Top