CMOS inventor discusses pixel count and diffraction limit

Nikon Rumors Forum » Nikon DSLR

CMOS inventor discusses pixel count and diffraction limit

(8 posts) (5 voices)

Tags:

No tags yet.

donaldejose

preferred member
Joined: Mar '11
Posts: 1,043
offline

I recently found this YouTube video in which the inventor of the CMOS image sensor (Dr. Eric Fossum) admits current pixels are smaller than the diffraction limit of green light and therefore higher pixel count does not really allow you to see the image better. The higher pixel count is just a marketing advantage. Look at section 38:00 to 45:28.

http://www.youtube.com/watch?feature=player_embedded&v=JkBh71zZKrM

<iframe width="560" height="315" src="http://www.youtube.com/embed/JkBh71zZKrM" frameborder="0" allowfullscreen></iframe>

I should note that Dr. Fossum says the diffraction limit of green light at f 2.8 is 4 microns and we now commonly can make sensors with 1.1 to 2.2 micron pixels and sensors with less than a micron pixel are coming soon. So each point of light as best as we can focus it now covers more than one pixel. Thus, once you make pixels smaller than the diffraction limit of your lens you really do not gain any increased resolution.

Here is my question: If the upcoming D400 is really 24 megapixels and the upcoming D800 is really 36 megapixels, will the size of those sensor pixels be so much lower than the diffraction limit of Nikon glass that the increased pixels won't really produce increased resolution? I know we will have to wait for production samples to test the resolution of the new cameras and it will depend upon the software processing the data from the sensor but I thought it interesting to see the inventor of the CMOS image sensor admit increased megapixels are more a marketing ploy than a real improvement in resolution.

Interesting.

Posted 1 year ago #
kyoshinikon

preferred member
Joined: Jan '10
Posts: 1,200
offline

I saw this video and I thought the whole lecture was pretty cool. However he is a sensor maker and probably doesn't know the whole process behind optic making (I don't either) so with the technology we have now there is probably a way to make the glass at a finer grain. I would say profitability is the issue that would keep companies form doing it.

The question is what is the diffraction limit of the lenses and each lens? Canon really goofed on that one because so few of their lenses resolve nicely at 21mpx. Sony also seems to have goofed here, because even though more of their lenses resolve nicely wide open than canon, most cant seem to handle the 24mpx to it's fullest potential (the 70-200mm f/2.8usm IS2 vs the 70-200mm f/2.8 Sal G is a good example).

People make light of the issue by arguing that the bokeh is better because the lens is better, but it really is only better because the picture isn't sharp at all. While Canon and Sony seems to have a large amount of lenses that are consistently "soft" at high Mpxes, Nikon seems to jump around this one. Their 70-200mm resolves just as bad as the Canon if not worse, but their 85mm f/1.4 at f/2.8 has a higher level of detail than its sony and canon counterparts.

I honestly don't know the technical specs but as someone who does fine art photography and a camera buff, I study pics by quality, detail, and sharpness. If I see a consistent problem in a particular camera and brand I test it against its nearest comparison. If it is consistent (Ie Iv'e used 5 individual 5dmk2's with 5 different 70-200mm f/2.8 and they all produce the same result) I rule it a problem.

That is not to say the 5Dmk2 is a bad body or the 70-200mm f/2.8 USM is2 is a bad lens, because they are both great products, it is more that the results I've got from the combo tell me that the lens can't handle the camera. Most don't seem to have a problem with it as it isn't bad, it is just not great either...

I don't know what the microns diameter of each piece of glass is, but even if they are beating a dead dog, Nikon is doing a much better job at not trying to do it...

Posted 1 year ago #
satellites

member
Joined: Oct '11
Posts: 30
offline

kyoshi, i think you have confused what bokeh is or are not getting your thoughts across into words effectively :]

anyway, he is talking in terms of physical laws, not lens qualities...so sharpness of the lens isn't really an issue.

Posted 1 year ago #
kyoshinikon

preferred member
Joined: Jan '10
Posts: 1,200
offline

satellites said:
kyoshi, i think you have confused what bokeh is or are not getting your thoughts across into words effectively :]

anyway, he is talking in terms of physical laws, not lens qualities...so sharpness of the lens isn't really an issue.

I an not the best writer so forgive me on this one.

What I am trying to say that people will often compliment images from a canon for having smoother and creamier bokeh (blur in the out of focus area) but they use it to say that the lens performs better than it's Nikon or Sony counterpart.

The problem is that while the out of focus area is softer, so is the in-focus area (the subject or what is supposed to be in focus) which means the lens is less sharp than the Nikon or Sony, so in essence it is not a better lens. Using one copy doesn't prove this, but if using several different ones yields the same result it is safe to say it is a problem with the product itsself and not just a defect...

Also I believe you missed the whole point of my post which is to say that many lenses already cannot resolve at the MPxes that are being put in DSLR's. There is not a set law that all glass only can capicitate a certain resolution... Many older optics cannot resolve at 6mpx well but work fine on a 2 mpx sensor.

There is no set limit for optical quality (as not all lenses resolve at the same diffraction limit) except how much money a company chooses to invest in optics. The grain of the lenses in a 18-105mm may be 6microns wide while a 24-70mm may have grain at 3microns wide meaning it will resolve better because the structure is finer. I don't know the size of particle in a glass structure (I assume it is much smaller), this is just an example but it is to illustrate the differenced between lens resolution.

Posted 1 year ago #
TaoTeJared

preferred member
Joined: Apr '10
Posts: 2,422
offline

Interesting to watch - good find!

The one thing any debate about this is that everyone forgets one major issue, each pixel only records 1 RGB color and the average of adjacent pixels are calculated to create one pixel in an image. So you have multiple 4 microns pixels = creating 1 pixel of color (8-32 pixel microns recording color). So if one pixel happens to miss the "Green" it is picked up in the pixel next to it. Either way, they are averaged together and it is not a 1-1 matching.

Diffraction limits continuously get argued here and are usually in two camps - The Vacuum Bubble camp-raw definition and physics and the Holistic-don't underestimate software lens material & design. Both have valid points. No one here remotely has the background or knowledge base to explain how Nikon, Canon or the others use to compensate software and design for the raw physics.

I fall into the Holistic for what Kyoshinikon has just described, Nikon keeps piling on the pixels and the images keep getting sharper.

Should you worry about the diffraction limit? No - that is Nikon's job and they are not going to release a camera who's images look like garbage. Most what people point to are 1/16" pixels on an actual print size of 5 feet (1.6m) that are smooched. Obviously none of them have ever been exposed to a billboard up close.
----------------------------------
Very well put kyoshinikon!

kyoshinikon said:
There is no set limit for optical quality (as not all lenses resolve at the same diffraction limit) except how much money a company chooses to invest in optics. The grain of the lenses in a 18-105mm may be 6microns wide while a 24-70mm may have grain at 3microns wide meaning it will resolve better because the structure is finer.

I have an older Tokina that has an Angenieux design. It is sharp but not razor sharp and I always go to it for portraits. Digital has created the pixel peeper complex which I believe has overshadowed the wonderful properties how lens design "draws" the image in a gorgeous way. Zeiss & Leica seem to be the only companies who have not forgotten how to do this.

Posted 1 year ago #
donaldejose

preferred member
Joined: Mar '11
Posts: 1,043
offline

As to how all three colors are attributed to one pixel see 35:30 in the above video. This is an oversimplification but well illustrated.

Posted 1 year ago #
El_Pickerel

preferred member
Joined: Jul '11
Posts: 581
offline

TaoTeJared said:

The one thing any debate about this is that everyone forgets one major issue, each pixel only records 1 RGB color and the average of adjacent pixels are calculated to create one pixel in an image. So you have multiple 4 microns pixels = creating 1 pixel of color (8-32 pixel microns recording color). So if one pixel happens to miss the "Green" it is picked up in the pixel next to it. Either way, they are averaged together and it is not a 1-1 matching.

There's also the anti-alias filter blurring the final image somewhat.

Posted 1 year ago #
TaoTeJared

preferred member
Joined: Apr '10
Posts: 2,422
offline

That there is. I'm not sure what effect it also has on diffraction limits as well.

Posted 1 year ago #

RSS feed for this topic

Reply

You must log in to post.

The new Nikon Rumors Forum is now live at http://forum.nikonrumors.com/discussions. This forum is now in "read only" mode until I figure a proper way to import all data over to the new platform. Please register over at the new forum.

CMOS inventor discusses pixel count and diffraction limit

Reply