Musings on EOS autofocus sensor development

Started Sep 16, 2009 | Discussions
OP Karl Gnter Wnsch Forum Pro • Posts: 11,408
Re: Musings on EOS autofocus sensor development

number95 wrote:

I am so curious to know/learn whether precision even for f5.6 is similar to f2.8 or so!

If you are talking about the center AF point, there would be no sense in the complications brought by the longer baseline sensor for lenses faster than f/2.8 if there were no differende. For the rest of AF point, if you consider the metric of the AF precision (measured in DOF units), then yes they will be equally precise...
--
regards
Karl Günter Wünsch
Visit my gallery at
http://www.fotocommunity.de/pc/pc/mypics/461808

 Karl Gnter Wnsch's gear list:Karl Gnter Wnsch's gear list
Canon EOS 7D Mark II
OP Karl Gnter Wnsch Forum Pro • Posts: 11,408
Re: Musings on EOS autofocus sensor development

MM1 wrote:

Somehow I don't see this as an tremendous advantage over 1D - the f/5.6 lenses usually don't require such an accurate AF, unless you have a big focal length, since the AF error hides in the DOF.

Sorry to contradict you there, but in terms of AF the DOF is the metric of the precision measurement. A f/5.6 lens will only be precise within the DOF at that aperture, a f/2.8 will inherently be much more precise to focus anyway - for that lens focus errors will hide much more readly as you stop down.
--
regards
Karl Günter Wünsch
Visit my gallery at
http://www.fotocommunity.de/pc/pc/mypics/461808

 Karl Gnter Wnsch's gear list:Karl Gnter Wnsch's gear list
Canon EOS 7D Mark II
OP Karl Gnter Wnsch Forum Pro • Posts: 11,408
Re: Musings on EOS autofocus sensor development

PC Wheeler wrote:

Karl Gnter Wnsch wrote:

Yes they do, see page 5 of: http://cpn.canon-europe.com/content/technical/eos7d.do

Be sure to give page 6 a good look, too.

I must never show this page to my wife, there is a damning remark there which would mean that I either have to buy another 7D or that I am back to using the 50D:

"These enhancements have culminated in a fast, accurate and stable tracking AF system that will get you the shot no matter what the situation. It can even be used when shooting macro subjects – something that was not easily possible before."
--
regards
Karl Günter Wünsch
Visit my gallery at
http://www.fotocommunity.de/pc/pc/mypics/461808

 Karl Gnter Wnsch's gear list:Karl Gnter Wnsch's gear list
Canon EOS 7D Mark II
number95 Regular Member • Posts: 401
Re: Musings on EOS autofocus sensor development

Karl Gnter Wnsch wrote:

PC Wheeler wrote:

Karl Gnter Wnsch wrote:

Yes they do, see page 5 of: http://cpn.canon-europe.com/content/technical/eos7d.do

Be sure to give page 6 a good look, too.

I must never show this page to my wife, there is a damning remark there which would mean that I either have to buy another 7D or that I am back to using the 50D:

"These enhancements have culminated in a fast, accurate and stable tracking AF system that will get you the shot no matter what the situation. It can even be used when shooting macro subjects – something that was not easily possible before."
--
regards
Karl Günter Wünsch
Visit my gallery at
http://www.fotocommunity.de/pc/pc/mypics/461808

The more I read the Canon document for 7D the more I understand 7D has unique focusing characteristics. In addition to what u have written now, the first part which you didnt copy describes the warming up case for AF focus tracking for the all existing EOS models is eliminated for 7D! I guess new 1D series will use the same AF infrastructure and probably will add more of cost-no-object design! I wonder a HP dual zigzag for f8

BiscottiGelato New Member • Posts: 8
Re: Musings on EOS autofocus sensor development

So... then... In terms of the just the center AF point alone... What really is the difference between the 7D, 50D and 40D? Other than the 'Spot AF' option which the 7D allows for, there seems to be no difference? The dual cross-type seemed to have existed and implemented the same way since the 40D...? What is the 'Spot AF' actually doing to narrow down the focus area anyways? Can these sensors be turned off and shortened or something? or is it something done during post processing of the sensor read-outs?

40D picture from BobAtkins

50D picture from a blog written by this Jeffrey guy

7D picture from the-digital-picture.com

OP Karl Gnter Wnsch Forum Pro • Posts: 11,408
Re: Musings on EOS autofocus sensor development

BiscottiGelato wrote:

7D picture from the-digital-picture.com

That's not the 7D AF sensor - that's the module from the 40D/50D, it is missing 10 AF points.
--
regards
Karl Günter Wünsch
Visit my gallery at
http://www.fotocommunity.de/pc/pc/mypics/461808

 Karl Gnter Wnsch's gear list:Karl Gnter Wnsch's gear list
Canon EOS 7D Mark II
BiscottiGelato New Member • Posts: 8
Re: Musings on EOS autofocus sensor development

Karl Gnter Wnsch wrote:

BiscottiGelato wrote:

7D picture from the-digital-picture.com

That's not the 7D AF sensor - that's the module from the 40D/50D, it is missing 10 AF points.
--
regards
Karl Günter Wünsch
Visit my gallery at
http://www.fotocommunity.de/pc/pc/mypics/461808

Hmm... maybe you are right... but then, the question still remains... if 40D and 50D also have center dual crosstype AF, what is the difference between a 7D and the 40D/50D in terms of the center AF point?

OP Karl Gnter Wnsch Forum Pro • Posts: 11,408
Re: Musings on EOS autofocus sensor development

BiscottiGelato wrote:

Hmm... maybe you are right... but then, the question still remains... if 40D and 50D also have center dual crosstype AF, what is the difference between a 7D and the 40D/50D in terms of the center AF point?

More processing power and tight integration with the color sensitive light metering, enabling the focus system to compensate for light frequency dependent differences in refraction index of the phase separation optics... Also the pixel pitch of the respective sensors might be different but that is pure speculation unless someone is willing to sacrifice a 50D and 7D to do a comparison of the respective modules under a microscope.
--
regards
Karl Günter Wünsch
Visit my gallery at
http://www.fotocommunity.de/pc/pc/mypics/461808

 Karl Gnter Wnsch's gear list:Karl Gnter Wnsch's gear list
Canon EOS 7D Mark II
hpjfromdk Regular Member • Posts: 454
Re: Musings on EOS autofocus sensor development

Any phase type AF point uses two line sensors segments per direction of sensitivity. The two segments are refereed to by Canon as "normal" and "reference" respectively and are marked as a and b below.

A "+" type AF point needs two "directions" one horizontal and one vertical. In the table below the horizontal and vertical line sensors used by each AF point on the 7D is listed.

The vertical line sensors 4, 14 and 20 in the figure are actually divided into a number of zones each representing an AF point. This is how it possible to "pack" the large number of AF points within a fairly small area. It further has the advantage that the length of the zones can be changed on the fly, given the proper read circuitry, hence allowing for handling large defocus..

At the bottom of the figure I have tried to map the coverage of the line sensors onto the AF points. It should be taken with a grain of salt though, as the actual magnification of the SIR optical system is not common knowledge..

~ hans ~

 hpjfromdk's gear list:hpjfromdk's gear list
Sony RX100 IV
OP Karl Gnter Wnsch Forum Pro • Posts: 11,408
Re: Musings on EOS autofocus sensor development

hpjfromdk wrote:

Any phase type AF point uses two line sensors segments per direction of sensitivity. The two segments are refereed to by Canon as "normal" and "reference" respectively and are marked as a and b below.

A "+" type AF point needs two "directions" one horizontal and one vertical. In the table below the horizontal and vertical line sensors used by each AF point on the 7D is listed.

The vertical line sensors 4, 14 and 20 in the figure are actually divided into a number of zones each representing an AF point. This is how it possible to "pack" the large number of AF points within a fairly small area. It further has the advantage that the length of the zones can be changed on the fly, given the proper read circuitry, hence allowing for handling large defocus..

At the bottom of the figure I have tried to map the coverage of the line sensors onto the AF points. It should be taken with a grain of salt though, as the actual magnification of the SIR optical system is not common knowledge..

Very interesting read - in fact pointing out my bad that I overlooked a few line sensors that are there in the previously cited schematics of the 7D autofocus sensor...
--
regards
Karl Günter Wünsch
Visit my gallery at
http://www.fotocommunity.de/pc/pc/mypics/461808

 Karl Gnter Wnsch's gear list:Karl Gnter Wnsch's gear list
Canon EOS 7D Mark II
BiscottiGelato New Member • Posts: 8
Re: Musings on EOS autofocus sensor development

Thanks for the explanation and diagrams. Very informative and learned a lot on how these AF diagrams are supposed to be read.

In terms of sensors then, it seems the center point of 40D/50D/7D are all pretty much the same. In terms of processing tho, I agree that there is more advanced processing on the 7D (wavelength based, selective read-out of sensor for spot AF, more processing power for faster AF in general, etc).

For those differences, I'm most likely keeping my 40D for all the shoots the remainder of the year until I spot an awesome deal (of sort) on the 7D. Thx!

hpjfromdk Regular Member • Posts: 454
Re: Musings on EOS autofocus sensor development

BiscottiGelato wrote:

Thanks for the explanation and diagrams. Very informative and learned a lot on how these AF diagrams are supposed to be read.

In actual fact the mapping of the line sensor to AF points is a little more complicated than detailed in my diagram. For example the uppermost parts of the vertical line sensor pair 4 maps onto AF point 4 and not onto AF point 2 as one would assume, etc.. This "reversing" is caused by the optics of the SIR systems, but since the AF sensor is pretty symetric in its layout, IMO it is mostly relevant Canons HW & SW engineers as it does not change the basic understanding of how the phase type line sensors map to an AF point.

In terms of sensors then, it seems the center point of 40D/50D/7D are all pretty much the same. In terms of processing tho, I agree that there is more advanced processing on the 7D (wavelength based, selective read-out of sensor for spot AF, more processing power for faster AF in general, etc).

Even though this is partly speculative in my behalf (i.e. not officially published by Canon marketing) I believe the 40D thus 50D AF sensor chips to contain areas used to compensate for CA offsets (wavelength offsets). In total 4 sensor pairs are used with one sensor in the pair more sensitive to "normal" wavelengthes and the other sensitive to red and near IR. These sensor areas were taken out on the 7D AF chip to make room for the additonal AF points. Instead of having only 4 of these areas/zones covering the AF chip, the 7D uses the 63 zones of the AE sensor, which results in an resolution and hence CA compensation at AF point level.

So in my book the CA/wavelength compensation is not new, it was just dramatically improved now allowing it to cater AF wise for very localized illumination (small colored edges etc..).

One interesting aspect of the "new" 7D AF system is however also how much of its improved responsiveness and capabilities originate in the fact that it has a dedicated AF processor - something the 5DmkII lacks for instance

In the end what counts is not the amount of claimed new features, but rather that they work as advertised in real life..

~ hans ~

 hpjfromdk's gear list:hpjfromdk's gear list
Sony RX100 IV
hpjfromdk Regular Member • Posts: 454
Re: Musings on EOS autofocus sensor development

Karl Gnter Wnsch wrote:

Very interesting read - in fact pointing out my bad that I overlooked a few line sensors that are there in the previously cited schematics of the 7D autofocus sensor...

In fact the 1Dmklll series AF sensor is composed entirely arround the principle of splitting long line sensor horisontal rows and long vertical columns, "merged" into array type blocks, into AF point zones. Each pair of rows and columns in an array are further configured in "zig zag" (i.e. offset approx. 1/2 pixel) to increase sensitivity and large OOF capability.

The arrays are paired, with one holding all the "normal" line segments and the other holding all the "reference" segments. The vertical array pair has a baseline. The horisontal pair has a F2.8 baseline but the two central horizontal rows have larger and more sensitive pixels and are F4.0.

~ hans ~

 hpjfromdk's gear list:hpjfromdk's gear list
Sony RX100 IV
carlk Forum Pro • Posts: 15,940
Re: Musings on EOS autofocus sensor development

The fact is a lot of people still use the center AF point only and f/r. Not the best way for all situations but that's what many feel comfortable to do. For those who want to use all points, e.g. AI servo and focus with AF ON button, I agree this new module is really good and likely the best ever after I read your op.

Karl Gnter Wnsch wrote:

Tapani Tarvainen wrote:

pcunite wrote:

Contrary to popular option and even my own until recently... the 7D does not have Canon's best AF that they have had for 5 years or more.

Yes. Even EOS 3 had center sensor that was high-precision at f/4 and could AF at f/8 (as can current 1D series bodies), 7D cannot.

If you reduce the focus module to this single aspect, then you are correct. Taken as a whole (so looking beyond this very limited aspect), the 7D AF module wipes the floor with any previous focusing module...
--
regards
Karl Günter Wünsch
Visit my gallery at
http://www.fotocommunity.de/pc/pc/mypics/461808

 carlk's gear list:carlk's gear list
Canon EOS 50D Canon EOS 7D Nikon D800E Fujifilm X-E1 Canon EF 85mm F1.2L II USM +6 more
Keyboard shortcuts:
FForum MMy threads