New Sensor Tech in EOS 7D Mark II [CR2]

[jrista]

That is the exact OPPOSITE of a shared pixel. Shared pixels SHARE readouts. Canon's DPAF use INDEPENDENT readouts.

Heh. You share readout circuitry between photodiodes ... to read their output independently.
What a paradox. And yet, that's exactly what the industry has been doing for a decade now (or more?).
Fascinating stuff. LOL.

Your still misunderstanding. Pixels are activated row-by-row, and all columns are read out SIMULTANEOUSLY. Every column of an activated row of pixels has the charge stored in the photodiode read, amplified, and shipped down the column line AT ONCE. Because the photodiode is split per-pixel, that they occupy the same row. Therefor, you cannot share the readout transistors, because both are read out simultaneously. Therefor, DPAF does NOT use a shared pixel design. There is, literally, two independent sets of transistors to read out each half of the pixel when the row is activated...and twice as many columns. During an image read, additional binning transistors combine the charge of each photodiode half, that total charge is amplified, and only one half of the columns are used to move the charge down to the CDS units and column outputs.

Shared pixel designs usually share DIAGONALLY (I already said this, but apparently the reason did not sink in.) By sharing diagonally, you avoid the concurrent row problem. The first row is activated, the first set of pixels that are sharing readout logic are read. The next row is activated, and this second set of pixels uses the same set of transistors to read out as their DIAGONAL counterparts in the row above. I've also read about patents that share pixels vertically, which achieves the same result, but ends up resulting in mixed color output for every set of transistors...green/blue, red/green, etc.

It isn't possible to share anything in the same row, though...because once a row is activated, everything in it has to be read out...and by nature, DPAF photodiodes for any given color filter share the same row.

 

[x-vision]

As for the rest, your making a LOT of assumptions, and piling assumption on top of assumption, then making bold claims about how you've discovered Canon has QPAF technology ...

Whenever I said anything on the topic, I've always started with 'by the look of things...', 'theoretically...', etc..
And I've been usually finishing my posts with 'it's all speculation', 'it's fun to speculate', or something like that.
I wouldn't say that I've been making bold claims. Mostly speculation and fun 8).

 

[x-vision]

Therefor, you cannot share the readout transistors, because both are read out simultaneously. Therefor, DPAF does NOT use a shared pixel design. There is, literally, two independent sets of transistors to read out each half of the pixel when the row is activated...and twice as many columns.

In other words, the two halves are read as ... independent pixels.
I feel that we are getting somewhere.
Oh, wait! That's what I've been saying all along.

The shared-transistor part is a different story.
I've never said explicitly which transistors are shared - and that's the key here.

 

[Lightmaster]

i hope it´s not just dual pixel v2.... :


ps: some of you guys should really try to get a life and spend less time on forums..

 

[Loswr]

Also, your claim that there are exactly TWO PHOTODIODES (and that's it!) is not based on fact.
We don't know for sure if Canon's design is a dual-pixel design (your assumption) or a quad-pixel design
(my assumption).

...

My assumption for a quad-pixel design is based on simple geometry.
If there are just two photodiodes per pixel, these photodiodes need to be rectangular.
This would be uncommon - if not even a first in the industry.
But with a quad-pixel design, the photodiodes are square just like in any other sensor.

...

Just a speculation, of course - but based on some informed assumptions.

Your claim of a quad-pixel design is based on a flawed assumption you're making after totally misinterpreting a single piece of data (a picture of a part of the sensor that's not the active imaging area). The claim of a dual-pixel design is not based on assumptions, it's based on actual published patents issued to Canon for the technology. That's established fact vs. ill-informed speculation.

If you want your claim to be in any way believable, you need actual evidence. Read the Canon patents on dual-pixel AF - show us where a quad pixel design is mentioned. Show us a verifiable image of the actual photodiodes of the 70D sensor (not the dead area at the very edge of the sensor in the Chipworks teaser). Until you can provide some evidence, your ASSumptions will just continue to make you look like a...foolish person.

 

[wickidwombat]

i hope it´s not just dual pixel v2.... :


ps: some of you guys should really try to get a life and spend less time on forums..

best post ever!

 

[Orangutan]

I feel that we are getting somewhere.
Oh, wait! That's what I've been saying all along.

The shared-transistor part is a different story.
I've never said explicitly which transistors are shared - and that's the key here.

Doubling-down on a weak hand is not wise. I suggest you fold, and cut your losses.

Oh, wait! That's what I've been saying all along.

Whatever you've been saying, you haven't said it very well. If you want to be taken seriously by some very well-educated folks on this forum, you might want to:

• do your research before you make assumptions
• be very clear about your definitions
• cite sources
• write clearly and concisely
• not write antagonistically to people who may know more than you do

 

[Orangutan]

I believe this is the original version: http://xkcd.com/386/

i hope it´s not just dual pixel v2.... :


ps: some of you guys should really try to get a life and spend less time on forums..

 

[scyrene]

I feel that we are getting somewhere.
Oh, wait! That's what I've been saying all along.

The shared-transistor part is a different story.
I've never said explicitly which transistors are shared - and that's the key here.

Doubling-down on a weak hand is not wise. I suggest you fold, and cut your losses.

Oh, wait! That's what I've been saying all along.

Whatever you've been saying, you haven't said it very well. If you want to be taken seriously by some very well-educated folks on this forum, you might want to:

• do your research before you make assumptions
• be very clear about your definitions
• cite sources
• write clearly and concisely
• not write antagonistically to people who may know more than you do

Very well put

I might add to that 'not retreat into "it's all fun speculation" when shown to be wrong'. Seriously, jrista and Neuro must have incredible patience.

 

[tayassu]

Interesting... This cartoon also showed up in the last sensor technology discussion. How did the wise man say once? "De-ja-vus aren't the errors in the matrix, they are the matrix." ;D

 

[x-vision]

Your claim of a quad-pixel design is based on a flawed assumption you're making after totally misinterpreting a single piece of data (a picture of a part of the sensor that's not the active imaging area).

My claim is not based solely on that.

I now admit, though, that the partial picture of the sensor die from Chipworks is not sufficient/reliable evidence to support any claims.
Therefore, I'm officially retracting it as evidence. I was wrong to quote it and I won't do it again.

The claim of a dual-pixel design is not based on assumptions, it's based on actual published patents issued to Canon for the technology. That's established fact vs. ill-informed speculation.

Of all those published patents, so far we've only seen a link to a schematic diagram of the 'dual-pixel' tech.
That's all. Everything else has been based on Canon's officially published information.

I'm not claiming that Canon's information is incorrect (hell, no). All I'm saying is that it's not guaranteed to be 100% revealing.

I asked for a link to the all-important patent that shows rectangular photodiodes.
As I said in a previous post, this patent would settle the discussion - but Jrista hasn't provided it.

Btw, this is the CanonRumors site.
By definition, most of the info published on the front page of this site is ... ill-informed speculation.
If you have such high standards for fact vs fiction, how come you are a regular on this forum ??
If you were holding CanonRumors accountable for the info that they are publishing, they should have closed shop by now.
???

If you want your claim to be in any way believable, you need actual evidence. Read the Canon patents on dual-pixel AF - show us where a quad pixel design is mentioned. Show us a verifiable image of the actual photodiodes of the 70D sensor (not the dead area at the very edge of the sensor in the Chipworks teaser).

Fair enough.

I've never presented my speculations as facts.
But if one day I decide to do that, I'll make sure that I have very solid evidence.

 

[jrista]

Therefor, you cannot share the readout transistors, because both are read out simultaneously. Therefor, DPAF does NOT use a shared pixel design. There is, literally, two independent sets of transistors to read out each half of the pixel when the row is activated...and twice as many columns.

In other words, the two halves are read as ... independent pixels.
I feel that we are getting somewhere.
Oh, wait! That's what I've been saying all along.

The shared-transistor part is a different story.
I've never said explicitly which transistors are shared - and that's the key here.

You seem to think your getting somewhere...convincing me, or anyone, that you know what your talking about. I don't really have the heart to continue the conversation, because the clueless one here is not me...and this is just becoming disheartening. Go educate yourself. Please. Once you have actually seen the actual electrical diagram for at least one of Canon's DPAF patents, AND preferably read the accompanying description of how it works, maybe then we can have a coherent discussion.

If you want your claim to be in any way believable, you need actual evidence. Read the Canon patents on dual-pixel AF - show us where a quad pixel design is mentioned. Show us a verifiable image of the actual photodiodes of the 70D sensor (not the dead area at the very edge of the sensor in the Chipworks teaser).

Fair enough.

I've never presented my speculations as facts.
But if one day I decide to do that, I'll make sure that I have very solid evidence.

The problem is that you keep "speculating" about the same thing, even though it's been proven wrong on multiple occasions. I DID link you several detailed pages that had the full patent information the last time we had this debate. It took a while to find a working link as well, but apparently you never read the link. I'm not going to go digging through the internet, spend all that time, to find something that your likely never going to read (most of the time, these patents are translated from Japanese, as they can be notoriously hard to find in the US Patent office's database...and therefor common search terms don't necessarily bring up what your looking for.) I'm not repeating that effort for someone who seems more interested in literally and intentionally ignoring the FACTS he's been presented with, and "speculating" about a falsehood that he himself initiated even though it's demonstrably false, invalid, incorrect, not real, not happening, and otherwise moot.

It's fine to speculate...when you ACTUALLY DON'T KNOW anything about whatever it is your speculating/rumormongering about. When it comes to DPAF...there is nothing to speculate about. Canon has files patents. That's the end of the story. WE KNOW. Your notion that "it's not guaranteed to be 100% revealing" is 100% absolutely wrong...patents MUST, BY DESIGN, INTENT AND FUNCTION, be 100% revealing. Otherwise they wouldn't be granted in the first place...specificity in a patent is key. Your ignorance in that only demonstrates you don't know much about patents, nor the technologies they can potentially describe. I'm no electrical engineer with a Ph.D, but that doesn't mean people who don't have degrees in electrical engineering are incapable of understanding an electrical circuit diagram or the terminology that describes them. I do my fair share of dabbling in electronics. (Right now I'm building a pelitier cooled DSLR cold box for my 5D III, which is involving a growing amount of electrical gadgetry and know-how to get it working the way I want.) Go educate yourself, read the actual patents, or in lie, read anything you can find about DPAF that isn't on Canon's site (so you can stop worrying that it's "just Canon marketing material"...although I'd offer that Canon is very up front about their technology, they have no reason to be misleading about DPAF)...and stop digging yourself into your hole. Your half-way to China right now.

Anyway, I think the myth of QPAF in the 70D or any other current Canon camera has been successfully debunked. The myth that DPAF or QPAF could be used to improve resolution in and of themselves, simply because they have independent readout, probably still needs to be revisited, however that's a debate for another day.

 

[cnardo]

Looks like the folks over at Canonwatch.com have a different take on this whole 7D2, new sensor info going around as well as both announcement dates and shipping dates! ???

 

[x-vision]

I don't really have the heart to continue the conversation, because the clueless one here is not me...

Here we go. Resorting to personal attacks and bullying again.

Admit it, you don't have a technical background.
That's why you run in circles, contradicting yourself, when you try to explain a technical concept.

You are the clueless one, not understanding the basic underlying principles.
I pointed out the flaws in your arguments about pixels vs photodiodes - but you are clueless to understand it.

Let's leave it at that.

 

[x-vision]

I'm no electrical engineer with a Ph.D ...

Yup, that was pretty obvious ... and I was right about that (so, I was definitely right at least about one thing 8) ).

Kudos for having the courage to admit it, though. Seriously.

Of course now all the people on your side of this argument should have second thoughts on
why they should be listening to you.

And btw, I happen to be an electrical engineering by training. Not a Ph.D., though.

 

[jrista]

I'm no electrical engineer with a Ph.D ...

Yup, that was pretty obvious ... and I was right about that (so, I was definitely right at least about one thing 8) ).

Kudos for having the courage to admit it, though. Seriously.

Of course now all the people on your side of this argument should have second thoughts on
why they should be listening to you.

And btw, I happen to be an electrical engineering by training. Not a Ph.D., though.

LOL. I have not admitted anything that diminishes my position, here. It doesn't matter that I don't have a Ph.D. You missed the point entirely. You don't need a Ph.D. to understand this stuff, if your otherwise well educated enough to understand the technology involved.

Even if you are an electrical engineer, you are not educated as to how DPAF works. Your lacking a basic knowledge set that has allowed you to wildly speculate, without any sound basis for your speculation. Here is what Canon's ACTUAL patent, from the US Patent office, actually states:

United States
Patent Application Publication
Yoshimura et al.

Pub. No.: US 2013/0147998 A1
Pub. Date: Jun. 13, 2013

IMAGE CAPTURING APARATUS AND FOCUS DETECTION METHOD
Applicant: Canon Kabushiki Kaisha, Tokyo (JP)
Inventors: Yuki Yoshimura, Tokyo (JP); Koichi Fukuda, Tokyo (JP)
Assignee: CANON KABUSHIKI KAISHA, Tokyo (JP)
Appl. No.: 13/692,173
Filed: Dec. 3, 2012

Foreign Application Priority Data
Dec. 13, 2011 (JP) ...................... 2011-272746

Publication Classification
Int. Cl.
H04N 5/232

ABSTRACT

An image capturing apparatus performs focus detection based on a pair of image signals obtained from an image sensor including pixels each having a pair of photoelectric conversion units capable of outputting the pair of image signals obtained by independently receiving a pair of light beams that have passed through different exit pupil regions of an imaging optical system. In the focus detection, an f-number of the imaging optical system is acquired, the pair of image signals undergo filtering using a first filter formed from an summation filter When the f-number is less than a predetermined threshold, or using a second filter formed from the summation filter and a differential filter When the f-number is not less than the threshold, and focus detection is performed by a phase difference method based on the pair of filtered image signals.

Note the words used here. This is DIRECTLY from the patent:

"An image capturing apparatus performs focus detection based on a pair of image signals from an image sensor including pixels each having a pair of photoelectric conversion units (photodiodes) capable of outputting the pair of image signals obtained by independently receiving a pair of light beams...

Performs focus detection. The intent of DPAF is to perform focus detection. There is no mention anywhere in this abstract that describes an increase in imaging resolution.

Pair. Pair is used repeatedly in this abstract, indicating that there are only TWO photodiodes.

The abstract quite explicitly describes a "pixel" as having a "pair" of photoelectric conversion units. In other words, a pair of photodiodes.

Everything I've said before was not based on assumption, or my own personal ignorance because I'm not a Ph.D. wielding electrical engineer. Everything I said was based directly on Canon's own patent, invented by Yoshimura and Fukuda. Invented, actually, quite well before the 70D ever actually used the technology in an actual commercial product...the earliest date of foreign application (application for patent in Japan, I assume) was Dec. 13, 2011!!! The patent in the US office was filed Dec. 3, 2012. This technology is not particularly new, and it has been well described for years. We aren't lacking knowledge about it.

Here is some more of the patent. The summary of the invention (just the first part...this section of the patent is a couple pages long):

SUMMARY OF THE INVENTION

[0007] The present invention has been made in consideration of the above situation, and improves defocus direction detection performance in focus detection of a pupil division phase difference method even When the pupil division performance is insufficient, and the influence of vignetting is large.

[0008] According to a first aspect of the present invention, there is provided an image capturing apparatus comprising: an image sensor including a plurality of two-dimensionally arranged pixels including pixels each having a pair of photo electric conversion units arranged to output a pair of image signals obtained by independently receiving a pair of light beams that have passed through different exit pupil regions of an imaging optical system; a first filter formed from a summation filter; a second filter formed from the summation filter and a differential filter; an acquisition unit arranged to acquire an f-number of the imaging optical system; a filtering unit arranged to perform filtering of the pair of image signals by selecting the first filter When the f-number is less than a predetermined threshold and selecting the second filter When the f-number is not less than the threshold; and a focus detection unit arranged to perform focus detection by a phase difference method based on the pair of image signals that have undergone the filtering by the filtering unit.

I don't assume when I have the option of actually knowing. I like to actually KNOW...know everything I can about as many things as I can. I have based my assertions in this debate on actual facts, derived from a patent that clearly and explicitly specifies exactly what DPAF is: What it's purpose is, how it works, why it works that way.

You now have the patent number. You have what you need to go look the patent up yourself and read it.

 

[Loswr]

Your claim of a quad-pixel design is based on a flawed assumption you're making after totally misinterpreting a single piece of data (a picture of a part of the sensor that's not the active imaging area).

My claim is not based solely on that.

I now admit, though, that the partial picture of the sensor die from Chipworks is not sufficient/reliable evidence to support any claims.
Therefore, I'm officially retracting it as evidence. I was wrong to quote it and I won't do it again.

Well, that's magnanimous of you considering that you've been told at least half a dozen times that the image was irrelevant and proof of nothing.

If your claim isn't based solely on that image, then upon what VERIFIABLE FACTS (you know...things like patents) is it based? One would think after the number of posts in which you insisted your opinion was plausible that you could have produced at least some additional documentation.

Btw, this is the CanonRumors site.
By definition, most of the info published on the front page of this site is ... ill-informed speculation.
If you have such high standards for fact vs fiction, how come you are a regular on this forum ??
If you were holding CanonRumors accountable for the info that they are publishing, they should have closed shop by now.

Personally, I have no difficulty in distinguishing rumor from fact from logical opinion from incorrect opinion based on bogus assumptions. I think most of know into which of those buckets your 'the 70D uses quad-pixels' idea falls.

 

[Sabaki]

Anybody think it's possible Canon could drop a dedicated processor into the 7D2 to handle AF subject recognition, like the 1DX does?

Better noise handling is my biggest want but a superb AF system is a close second.

 

[jrista]

Anybody think it's possible Canon could drop a dedicated processor into the 7D2 to handle AF subject recognition, like the 1DX does?

Better noise handling is my biggest want but a superb AF system is a close second.

It depends. If the 7D II hits with a new DIGIC processor, they may not need to resort to a dedicated AF processor. Each generation of DIGIC chips gets considerably more powerful than their predecessors. I'd imagine a DIGIC 7 would be quite considerably more capable than the DIGIC 6 used in some of the more compact cameras. DIGIC 6 does a LOT of image processing (It's more like Sony's Bionz X chip than the DIGIC 5), with high quality noise reduction, high quality video processing, etc. If Canon created a DIGIC 7 with some 7x or so more processing power than the DIGIC 6, they could easily handle high frame rates as well as high end AF capabilities all in the one chip (or two chips, as it's probably likely to be.)

 

[x-vision]

Here is what Canon's ACTUAL patent, from the US Patent office, actually states:

Oh my goodness! You haven't read the patent past the summary section.

Because if you had, you would have come across section [0038].
Here's it is on Page-13 of the patent's PDF file (which you can get from here):

[0038] FIG. 2A is a view for explaining the pixel arrangement
of the image sensor... FIG. 2B is an enlarged view of
the pixel 210G which includes a plurality of photoelectric
conversion units (to be referred to as “sub-pixels 201a and
201b” hereinafter) for pupil division. Each of the pixels 210R
and 210B also includes two sub-pixels 201a and 201b. Each
pixel can output an image signal obtained by receiving light
independently from each of the sub-pixels 201a and 201b.
The independently obtained image signals can be used for
focus detection, or added for each pixel and used for image
capturing.

So, you quoted this patent in support of your TWO PHOTODIODES claims - but the patent says otherwise.

Dude, you have to work not only on your technical skills but on your comprehension skills in general.
When arguing about something, it's plain retarded to argue with yourself.

Canon themselves are talking about sub-pixels.
Not only that. Here's the last part of Section [0038]:

The pixel group 210 having the above-described
structure is repetitively arranged. Note that in the arrange-
ment shown in FIGS. 2A and 2B, all pixels include the sub
pixels 201a and 201b. Instead, the pixels including the sub
pixels may be discretely arranged in the image sensor 107.

This is (maybe deliberately?) vague but one way to interpret it is that Canon is saying that the
arrangement illustrated in Fig. 2A doesn't need to be strictly followed.
That is, they are opening themselves to a different implementations of the same concept - without
actually specifying the implementation.

Anyway, I don't know what you are arguing about.

This patent is about a methid of focus detection on a sensor.
That is, they are patenting the method itself; nothing is mentioned about how exactly
they are going to etch the pixels and sub-pixels on the wafer.

They say that each pixel has two sub-pixels - but at the same time they are also leaving the door
open for alternative implementations as well.

That would be very clear to a technical person (or someone with a legal background as well - because of
how the last part of Section [0038] is worded).

But since you are not a technical person these things are escaping you.
(You could have at least read the patent in its entirety, btw - but you haven't done even that.)

I'm done.