do ccd's "make stuff up"?

[Sean]

Not to keep posting digi vs traditional discussion, but the implications and differences between the two always interest me. I was reading an article on how the ccd translates an image, and came across something I hadn't noticed before. The article described the 'lattice' structure of the ccd, then said the gaps in between the lattice which contained no sensor activity were dead zones -and for those areas, the computer would guess the information that would most likely have been there and fill in the blanks. If that's the case, then a percentage of the ccd image would be somewhat artificial wouldn't it? It's already converting light to electricity which changes things, then it fills in blanks for the information it could not gather. So does this mean film will always be a representation of a scene, and digital will only be a numerical representation of what the scene looked like?

 

[bmac]

Yes.

 

[Tom Duffy]

But Sean, what about the light from the scene that impacts on the film, between the silver particles?

This would mean that 8x10 provides a truer representation of a scene than is possible with a 35mm, given that you are enlarging the "missing light" less.

I can live with this conclusion... what disturbs me is that t-grain film then would provide a truer representation of the scene than is possible with conventional emulsions. very disturbing...

 

[Robert]

</span><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td>QUOTE (Tom Duffy @ May 13 2003, 04:42 PM)</td></tr><tr><td id='QUOTE'> But Sean, what about the light from the scene that impacts on the film, between the silver particles?

This would mean that 8x10 provides a truer representation of a scene than is possible with a 35mm, given that you are enlarging the &quot;missing light&quot; less.

I can live with this conclusion... what disturbs me is that t-grain film then would provide a truer representation of the scene than is possible with conventional emulsions. very disturbing... </td></tr></table><span class='postcolor'>
I thought the older random films are actually truer. Something about the randomness. It was explained to me once and all I remember was it made sense-))

Audio CD players make things up to. I've seen CDs that look horrible play just fine. The player guessing on what it should be seeing.

 

[Sean]

I think gaps in the silver are ok because they are left as is and are not changing anything. The gaps in the ccd probably can not be left as is because it would have a major impact on the image -so it adds information that did not exist to smooth things out. Maybe it is silly to have an issue with this if our brains can't even perceive it, but the thought of artificial information added by a computer is unappealing to me. I am ready for new fims based on nano-technology which will make tmax 100 look like tmax 3200. Currently it looks like 16megapixels is required to compete with fine grained 35mm film. If kodak could evolve their film technology to widen that gap to say 100 megapixels it would rock the industry. Imagine an 8x10 sheet of film with that kind of technology -if it's even possible- I like to dream...

 

[Robert Kennedy]

But isn't it true that lenses can only resolve so much information? I remember hearing that there is a limit to the LPMs or a lens and that we have already hit it.

 

[Sean]

Well, several companies have been growing diamonds with some success, so maybe we'll have diamond lenses one day. I wouldn't want to open the wallet for it though. Then we'll have the problem of how much resolution our brains can see

"As methods for growing diamond, both at high pressure and by chemical vapor deposition, improve, and as science finds ways to take advantage of diamond's properties, the potential applications of diamond's superlative properties appear boundless. From super electronics, to indomitable optical windows, to unscratchable surfaces - maybe the next watch bezel - diamond is an obvious choice."

 

[Robert]

For 35mm and MF I think the film is still the hold up. LF you run into diffraction limits.

Doesn't Kodak claim 35mm consumer colour film would need something like 30mega camera.

 

[glbeas]

What you've heard about Sean is the way a standard CCD will extrapolate colors using the RGB array that is laid out in a checkerboard pattern. In order for each pixel to have it's own color the processor will extrapolate from the surrounding pixels what it thinks the "right" color for that spot will be. You can get some wierd effects with fine patterns that fall in a harmonic range of this extrapolation.
There is one digi camera out now that doesn't do that, the Sigma SD-9. It uses the Foveon X3 image sensor in which the photodetectors are layered in a similar fashion to color film. The results are much sharper and truer colors. In my opinion this is the only digital imaging system that has a chance in hell of really competing with film for image quality. Keep your eye on any system using layered photodetectors, they may be the ones that come out on top of all the other digi-junk.

 

[Sean]

ah yeah, I've heard of the foveon, didn't realise it reacted that way.

Robert Kennedy had a good point about lens resolution. Will consumers pay for a 100megapixel slr if the lens will not resolve that much resolution? So far consumer digitals are following the path of pc processors, where increasing resolution is their main selling point. Surely they are about to hit a brick wall with lens resolution,, then how will they continue? It would be like having a pc that could run any application thrown at it, then asking if you want to upgrade it... why would you? Then again they say the lifespan of these digitals is very short so...

 

[fhovie]

Besides the fact that ALL image capturing makes stuff up - Imagine putting film in a digital camera. I have some of these - they are 16mm cameras. I had to re-learn photography. Sure - the lens will stop down to f22 but the image will have so much diffraction from the 25mm lens that it will look very bad. So you have a film like CCD with a 25mm lens that has 35mm like resolution, you will still be no better off than 16mm in projective geometry. I would rather see the density of the CCD become less important and the SIZE of it increase. Then you could better manipulate your DOF. Then - use Zeiss lenses ... is that Sony?

Frank

 

[Robert]

Look up aerial lens resolution. Film can't keep up at wider apertures. Some lenses can reportly hit 400 lp/mm. None of this matters for digital. Digital is over loaded at much lower levels. That why the low pass filter is stuck on them. There is no use putting on a high resolution lens on a digital camera.
A crappy consumer zoom is likely better then the sensors can handle.


Second issue. I thought digital was hitting the wall in terms of size. Soon basic physics won't let them put more pixels into that small sensor. That means if they do make a 64M sensor it will need to be a bigger sensor. Meaning the number of pixels per sensor area will be the same. So no increase in quality other then bigger.

 

[Jorge]

</span><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td>QUOTE (Robert @ May 14 2003, 05:36 AM)</td></tr><tr><td id='QUOTE'> Look up aerial lens resolution. Film can't keep up at wider apertures. Some lenses can reportly hit 400 lp/mm. None of this matters for digital. Digital is over loaded at much lower levels. That why the low pass filter is stuck on them. There is no use putting on a high resolution lens on a digital camera.
A crappy consumer zoom is likely better then the sensors can handle.


Second issue. I thought digital was hitting the wall in terms of size. Soon basic physics won't let them put more pixels into that small sensor. That means if they do make a 64M sensor it will need to be a bigger sensor. Meaning the number of pixels per sensor area will be the same. So no increase in quality other then bigger. </td></tr></table><span class='postcolor'>
Question Robert. If what you say is correct, why are lensmakers making lenses specially designed for digital, such as the digitars? They market these lenses as having better quality than a "normal" lens. Do you know if this lenses actually perform better? or is this anothe digital world con?

 

[Robert]

Go to:

http://www.schneideroptics.com/info/white_papers/

And download the white paper on digital lenses.

Better that somebody who understands optical engineering explain it but the way I read it the better the lens the more false info. There for the "digital" lenses are actually dumbed down.

 

[haris]

Just one of the may CCD possibilities. Now, Big brother is child game... See http://www.kaya-optics.com/products/overview.shtml

 

[fparnold]

glbeas,

I remember seeing a review a few months ago where they compared one of the 6MP digilars to scanned MF. The upshot was they thought it was higher res, until they looked closely and found that the corrugated metal building was just the right size in the frame to set up a 'beat', and create a sharp (but non-physical) corrugation pattern.

I like the randomness of film, even the grain (though I do go to some lengths to minimize it). Thankfully, most digis are only interpolating between pixels, and not inserting new scenic elements, changing the colors, or correcting the composition.

 

[Robert Kennedy]

Those opaque IR filters are nothing new. A few years back Sony came out with a near-IR sensitive CCD on some of its' camcorders. It was learned that if you slammed an IR filter on it and turned it on during the DAY that you could, in certain circumstances, see through (sort of) clothing.

Christmas for peeping toms....

Anyway, Sony has since fixed the "problem"....

 

[Sean]

I've been trying to find some more info about lens resolution and ccd's. For example if there is a 100megapixel ccd, what good does that do if the lens only resolves 40 megapixels? Is this a roadblock that digital is about to hit? Anyone know how many 'pixels' a lens can resolve on a ccd the size of a MF film back? Does film already exceed the maximum resolution of the lens?

 

[Robert Kennedy]

A bit of this gets into the weird apples and oranges thing....

Traditonally (at least as I understand it) lens resolution was measured by how many lines per mm could be resolved on film.

But digital doesn't use film. It uses little dots.

Now, from what I understand, there is an absolute limit on len resolution due to physics. You simply won't be able to resolve past a certain point.

Now here is where I see it getting tricky.

Say you "hit the wall" resolution wise and have a CCD that is say twice as good as the lens. In other words each pixel is 1/2 the width of the smallest resovable line.

You just market it. Nobody can really tell the difference right? I mean sure only 1/2 the pixels will be "used" (o.k. will be needed....), but why stop there?

 

[Robert]

Physics limits the size of the pixel also and I think the limit is a bigger pixel then current lenses.

The diffraction limit with white normal light is something like

1600lp/mm / the F stop

So at F/1 1600 would be a perfect lens. I doubt this exists in anyplace.

At F/4 it would be 400 lp/mm. Supposedly some of the best 35mm lenses are already close to this limit.

A lab could test with monochromatic light and get higher numbers but that's not exactly usefull for most of us.

Now check your films. How many come close to even the F/8 number of 200 lp/mm?

 

[Robert]

I think pixels used in digital cameras are already close to the limits and can't improve much more.