Chip size effects on pixels

[Steve Phillipps]

For all the engineers out there! Apart from light gathering ability, what is the difference between a pixel on say a 1/3" vs 2/3" chip. What I'm wondering is say you have 1920x1080 on a 2/3" and a 1/3" chip - when these are displayed on a 50" HD screen is there a difference in resolution/sharpness? In other words is the 1/3" pixel being blown up twice as much or is the pixel from the chip just translated into the pixel on the screen? The latter is how I guess it to be, and therefore there will be no difference between the 1/3" and 2/3" pixels.
This then would be very different from film, where if you were to blow up a 35mm negative it would be much worse than blowing up a medium format 6x7 one.
Steve

[Brian Drysdale]

You'd need a higher resolution lens to get the full information from the 1/3" chip compared to the 2/3".

[Steve Phillipps]

Good point Brian. Assuming that was equal, is 1 pixel 1 pixel?
Steve

[Brian Drysdale]

Yes, although 1/3" cameras tend not to have the full number of sensor pixels.

[Dale Guthormsen]

Good afternoon,


Isn't this much like the 1/3 inch SD as compared to the 1/3 inch HD. the Hd has to have smaller pixels and hence less light collection. An xl2 does better in low light than an xlh1.

I would reckon that if both resolve 1440 x 1080 or uncompressed, then the larger ccd will give better light collection as the pixels would be larger??????


Interested to hear what aothers have to say ont his.


dale

[Bruce Watson]

Originally Posted by Steve Phillipps

For all the engineers out there! Apart from light gathering ability, what is the difference between a pixel on say a 1/3" vs 2/3" chip. What I'm wondering is say you have 1920x1080 on a 2/3" and a 1/3" chip - when these are displayed on a 50" HD screen is there a difference in resolution/sharpness?

To answer your question directly -- the difference between a pixel on a 1/3" sensor and one on a 2/3" sensor is that the pixel from the bigger sensor may be a little less noisy, and may be able to record a bigger subject brightness range (SBR).

Resolution and perceived sharpness (acutance) involve way more than just a single pixel -- everything in the capture to display chain (that is, from the lens to the anti-aliasing filter in front of the chip to the sensor chip itself to the camera's compression algorithms to the NLE suite's algorithms to the way the operator uses the NLE to the electronics reading the data from the blu-ray disk to the HDMI electronics to the display electronics and algorithms to the state of calibration of the display and the individual preference settings) effect resolution and perceived sharpness. So one can't easily talk about resolution the way you seem to want to.

On the chip it's mostly a trade off on the size of the individual photo receptor sites. Smaller sites, pushed closer together, demand a higher resolution capability from the lens. Bigger sites on the other hand tend to be somewhat less noisy, and tend to exhibit a greater dynamic range.

Then you get into DOF -- where a bigger chip requires a longer lens for the same angle of view. And a longer lens has shallower DOF for a given aperture. This is the "film look" argument and the popularity of the DSLRs for making video.

And of course money. There's a sweet spot in any process -- from there in either direction (bigger or smaller) the chips cost more. Typically the sweet is for fairly small sensors.

[Steve Phillipps]

Thanks Bruce, I understand all that. Just trying to get confirmation of something in my mind, which is that a pixel translates from the chip to the screen on a 1 for 1 basis, so the initial size of that pixel is irrelevant.
Are you saying that and individual pixel can have more noise than another? I kind of assumed that the noise in an image was a collective of all the pixels involved.
Steve

[Jim Andrada]

And whether it's CMOS or CCD - CCD sensors have an accumulator site in addition to the photo receptor site which is why they don't exhibit rolling shutter issues. On the other hand, if you leave off the accumulator sites you can use larger photo receptors, or put on more of them or...

And don't forget the role that the sampling algorithms in the camera play - a pixel is not a little square dot in reality. There's more to life than megapixels. Sometimes you get better results with fewer magapixels as counter intuitive as this may seem. In fact, the number of pixels in the output image doesn't necessarily bear a one to one relationship to the number of receptor sites.

And the comments about the effect of resolving power of the lenses is right on. The lens on my 5 X 7 view camera has nowhere near the resolving power of the lenses on my 35mm or video camera because it doesn't have to - the "sensor" (ie film is so much bigger that edge to edge covering capability is way more important than resolving power. And since nobody uses a 5 X 7 camera for casual handheld shots, super wide apertures aren't needed either.

These things are all rather complex systems and the balance of component characteristics is very tricky. I don't think there is a single figure of merit that tells the whole story.

[Steve Phillipps]

Yes Jim, I realise there are plenty of variables involved.
The only reason I asked is that I was thinking about it the other day and with film in mind I thought that enlarging a smaller pixel would produce poorer quality than a larger one, just as it does with a film negative, but then I suddenly thought hang on, it's not the same is it - a pixel contains number information on the luminosity and chroma of the scene and this is transposed exactly onto the screen, so the size makes no difference as it's just data.
Steve

[Jim Andrada]

Yes, I forgot to add that having said all the techie stuff, at some reasonable level of accuracy what you were thinking is right on! X numbers is x numbers regardless of whether they're numbers stored in 2 inch square boxes or 4 inch square boxes.

Sort of like an excel spreadsheet - doesn't matter how big you make the cells, you get the same number of cells in a X by Y range of cells - except that you could enter a longer number in each cell if it's bigger (ie a larger dynamic range)

Oh well, best not to stretch this analogy too far.

[Bruce Watson]

Originally Posted by Steve Phillipps

Thanks Bruce, I understand all that. Just trying to get confirmation of something in my mind, which is that a pixel translates from the chip to the screen on a 1 for 1 basis, so the initial size of that pixel is irrelevant.

Yes, if you capture a 1080x1920 image and display it on a 1080x1920 display, you get a one-to-one mapping between captured pixels and displayed pixels. The size of the individual sensors, or the sensor chip as a whole, doesn't effect mapping. But since sensor size does effects other things I wouldn't make the blanket statement that it's irrelevant. Just sayin'. ;-)