New CMOS technologies

[rosuna]

Advertisement (gone after registration)

Canon is incorporating many new technologies into its new sensors for the 50D and 5D successor. This article is written in Spanish:

 

DSLR Magazine

 

Valentín Sama renowned journalist, former editor of the FV magazine.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[farnz]

Thank goodness the diagrams are 'written' in english. My rusty Spanish isn't good enough to fully understand the article but the pictures sure help!

 

Thanks for posting, Ruben.

 

Pete.

[mjh]

What’s an old Fuji diagram of their SuperCCD SR doing in an article supposedly about Canon sensors, I wonder? (And why does it carry a “DSLR Magazine” watermark when it’s the intellectual property of Fuji?)

[sdai]

dpreview watermarks the camera companies product images all the time, nobody cares.

 

One thing really new with this 50D CMOS sensor is that Canon is also using these gapless microlenses now. Here's the one slide they showed to everyone at the 50D launch meetings.

 

http://dc.watch.impress.co.jp/cda/static/image/2008/08/26/shinagawa_004l.jpg

[gravastar]

If what the article says is true it would seem that competition from Nikon will be of great service to us Canon users .

 

For a better insight that's in english here is a Google translation.

 

 

Bob.

[ho_co]

Kommt mir spanisch vor.

 

My Spanish is rudimentary at best, but what I got is this:

 

21 MP

 

the closely-spaced microlenses are already in the 50D; purpose: improvements in S/N (article mentions M8 as also having special microlens pattern); reduced voltage with closer pixels means reduced sensor heat, reduced noise

 

new Digic IV processor (without which obviously some of these advances would have been impossible)

 

some noise reduction occurs onboard the CMOS sensor at the pixel level; automatic averaging of voltages of several (2, 4 or up to 8 currently) adjacent pixels to enable local noise reduction before sending data on from the sensor (not possible before the Digic IV processor)

 

thinner low-pass filter, closer to sensor for reduced effect; to be removed in next generation (article mentions DMR in this regard) (elimination of AA filter will require Digic V processor)

 

greatly improved dynamic range by specifically adjusting shadow pixel *sensitivity* (other brands manipulate the histogram as a whole)

 

14-bit; next generation will be 16-bit

 

planning is under way to make the 14- and 16-bit outputs convertible to 8 bits without loss of information

 

 

That's what I get from the Spanish. I hope this is helpful and I'm ready for any corrections.

[sdai]

Advertisement (gone after registration)

No one should doubt the Spanish of a Texan, Howard ... and thanks for the translation.

 

Basically, the DSLR author's information regarding the 50D is completely wrong.

 

As you can see from the picture I embedded above and Canon's news release from Tokyo, the gap between microlenses on the 50D's CMOS sensor has been eliminated and there's no space in between.

 

Not to mention about the 14 bit or 16 bit converted to 8 bit thing, eliminating AA with Digic 5, that's too funny ...

[ho_co]

Comment:

The article seems to me to be honest and above-board, but to a certain degree it is marketing-speak.

 

For example, it's clear that looking at charge levels of individual pixels to give them all similar anti-noise treatment and then adjust them to extend dynamic range requires a processor with that capability built in. The fact that "X and Y and Z weren't possible before the Digic IV processor" goes without saying, because the processor and the sensor were designed to accomplish these things.

 

The author acknowledges that the M8 was first to rearrange microlenses to get results not previously possible, and that the DMR was the first 35mm format camera to eliminate the AA filter. Canon will do the same in the next generation, not in this one. But by making it thinner and moving it closer to the pixel substrate, Canon will reduce its effect this time around.

 

Specifically manipulating less-illuminated pixels to increase dynamic range sounds new and interesting to me; and if I understand correctly, it's something that can be done only with CMOS because CMOS does some processing on the sensor itself, between pixel rows. The article states that this new methodology will do a better job than Nikon, Sony and Olympus in extending dynamic range.

 

It sounds to me as if Canon has taken a look at the M8's style of RAW compression, when the article remarks that as Canon moves from 14-bit to 16-bit output, it is looking at ways to maintain that "richness" of data in 8 bits. [but I'm less sure of my interpretation of that paragraph.]

 

Good to see such a full article on the "5D Mk II." Thanks, Rubén!

[sdai]

It sounds to me as if Canon has taken a look at the M8's style of RAW compression, when the article remarks that as Canon moves from 14-bit to 16-bit output, it is looking at ways to maintain that "richness" of data in 8 bits. [but I'm less sure of my interpretation of that paragraph.]

 

That's not the case.

 

Canon has already implemented raw compression with previous models and it is nothing new.

 

The author ripped off the Fuji Super CCD diagram, others from Olympus and the last one who knows where it is from but definitely not from Canon.

 

It has zero credibility because almost everything this guy has mentioned about the 50D is wrong.

[ho_co]

As you can see from the picture I embedded above and Canon's news release from Tokyo, the gap between microlenses on the 50D's CMOS sensor has been eliminated and there's no space in between.

Sorry, Simon, I misrepresented the article's author.

 

The diagrams he shows look as if he got them from the closet; so far as I can tell, they have nothing to do with the article. (Probably the editors said, 'we need something here to break up the tech-talk.)

 

You are correct: According to the article, Canon "sneaked" the new-style 100%-coverage microlens design into the 50D without major publicity. The word "reduction" was my choice instead of "elimination."

 

I thought about cross-referencing your comment, but figured that my rend(er)ing of the article would just be seen as repeating what you had said.

 

But I hadn't followed your link, so now that I have seen it, here it is:

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[ho_co]

Canon has already implemented raw compression with previous models and it is nothing new.

1) I clearly said this was MY reading.

2) I also said I am VERY unsure of the translation of this paragraph.

3) My take on it was that so far, there isn't an 8-bit lossless compression of 16-bit data.

 

It has zero credibility because almost everything this guy has mentioned about the 50D is wrong.

No, Simon, don't blame the author for my interpretation. I shouldn't be attempting to render Spanish to English, but I'm interested in the information, and no one else has stepped up to do the job. (They've got better sense than I, I guess.)

 

You're off base to say the author has no credibility, because you didn't read the article. Don't confuse my summary with the article.

 

I don't believe the article had much at all to say about the 50D, whether right or wrong.

 

 

Please, someone with Spanish competence, do a brief summary of the new technologies article. And accept my apology for attempting to do so. As I said in the summary post, please do post corrections and clarifications.

[sdai]

It's ok, Howard. I'll not put any doubt to your Spanish translation in any forum post because I know it has become many Texans' second language especially when close to the borders. We've travelled down there in the hunting seasons many times, English was only spoken on the ranches ... many people couldn't understand me at gas stations, restaurants.

 

I'll gladly withdraw my comments on the Spanish article and it's author.

[mjh]

Not to mention about the 14 bit or 16 bit converted to 8 bit thing, eliminating AA with Digic 5, that's too funny ...

Well, the bit about the 8 bit compression is wrong; what is meant (by Canon anyway; I cannot tell whether the Spanish author got that one right) is that they attempt to preserve a high dynamic range captured by the sensor in an 8 bit JPEG, probably by in-camera tone-mapping.

 

However, it is correct that Canon intends to move towards 16 bit A/D conversion and 50 MP, and that they want to do without an AA filter and remove moiré in-camera with the next generation of the DIGIC processor. Given the amount of additional resolution to be gained that way, that’s just a natural step. On the other hand, implementing a fast and effective algorithm for moiré-removal in-camera is far from trivial.

[rosuna]

Sama didn't say the next generation Digic processor is required for nothing. He is referring to the next sensor generation. The Fuji sensor is mentioned just because future Canon sensors will have variable sensivity in each pixel. The Digic will analyze the scene and the sensivity of the pixels will be set at different values. I cannot read anything wrong about the 50D... etc.

[ho_co]

Thanks, Rubén and Michael, for the corrections and clarifications.

 

The article is good promotion, well placed. Here we are discussing the technology of a camera not yet on the market instead of discussing the advantages of the new Nikons.

[like_no_other]

Well, the bit about the 8 bit compression is wrong; what is meant (by Canon anyway; I cannot tell whether the Spanish author got that one right) is that they attempt to preserve a high dynamic range captured by the sensor in an 8 bit JPEG, probably by in-camera tone-mapping.

 

However, it is correct that Canon intends to move towards 16 bit A/D conversion and 50 MP, and that they want to do without an AA filter and remove moiré in-camera with the next generation of the DIGIC processor. Given the amount of additional resolution to be gained that way, that’s just a natural step. On the other hand, implementing a fast and effective algorithm for moiré-removal in-camera is far from trivial.

 

50 mpix sounds like a big demand for lens resolution. Too much?

[jaapv]

It is way beyond the theoretical limit of lens resolution - but it looks great on a spec sheet - or in a advertorial...

[wildlightphoto]

I wonder why it took Canon so long to make the leap to 14-bit processing, while they acknowledge that 16-bit has the potential to produce better DR and gradation? It's not like it's unheard-of technology, or perhaps for CMOS it is.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[rosuna]

50 mpix sounds like a big demand for lens resolution. Too much?

 

It depends on lens design and manufacture, format size and aperture values:

 

Do Sensors Outresolve Lenses?

 

R.

[jaapv]

It depends on lens design and manufacture, format size and aperture values:

 

Do Sensors Outresolve Lenses?

 

R.

 

In which this is the conclusion:

 

Sensors for larger formats are approaching the diffraction limit of real lenses, and it is more difficult to get high levels of aberration suppression for them. The point is that you cannot fully exploit the resolution potential of high-resolution sensors with regular mass-produced lenses, particularly for larger formats.

 

Assuming a diffraction limited lens. Of which I know but one or two examples - by Leica. And we are talking about a doubling of the resolution which was regarded as the maximum when that article was written.....