AF sensors -technical issues

Status
Not open for further replies.
Does anyone on this forum know how phase detect AF sensors actually work and perhaps, how expensive they are to manufacture. I'm not talking about the general principals of splitting light beams and measuring the difference in phase, but the 'nitty-gritty' details. For example how does this pattern of silicon photodetectors:

http://a.img-dpreview.com/reviews/CanonEOS5DMarkII/Images/features/af02.jpg

relate to the pattern seen in the viewfinder? And how does it relate to this:

http://a.img-dpreview.com/reviews/CanonEOS7D/Images/whatsnew_afsensor.jpg

i.e. how exactly does one pattern lead to 9 AF points (1 cross type, 6 AF assist points) and the other lead to 19 AF points,all cross type?
 
The dark bars, which come in pairs, are the actual AF sensors used to analyze the light from the splitter lenses.

Each pair of dark bars maps to a single non-cross AF point.
For cross-type AF points, there should be two pairs of bars (that is, 4 bars).
Note that long bars are actually pairs of small bars, not single long bars.

The center AF points in both the 5D and 7D are double-precision points, so the thicker bars are presumably the ones for the center AF point.
Also, the center AF point in the 7D is cross type in X-formation, so the 45-degree bars seen on the 7D AF chip are obviously also for the center AF point.

Other than that, it's a guessing game which bars map to which AF points.

The general principle is that a) the longer the bars and b) the further apart they are from each other, the more accurate the AF point is.

I don't have an idea about how expensive AF chips are but the same principle applies as with normal sensors – the larger the size, the more expensive the chip.

And for lots of high-precision AF points, a large chip is needed.
 
Upvote 0
Aug 11, 2010
827
5
I think the "insufficient room" claim has been thoroughly shouted down and disproved as a lame, last-ditch, cover-my-rear excuse from canon. while a powerful AF chip may need some more space, I don't think anyone would have complained about the weight or form factor had Canon decided to beef up the 5DII to accomodate better AF.

it certainly would have impacted the price, and I think Canon decided to forego the AF module R+D in favor of keeping the final price tag low, guessing it would sell well despite the fossil-like AF. and they were right. I'm one of those people who was willing to overlook the 5DII's ancient AF system due to all its other perks, but it doesn't mean I wouldn't be a happy guy if they stuck a 7D AF system into the future 5D Mark III
 
Upvote 0
Digging a bit further, I found this page:

http://www.digitaljournalist.org/issue0512/westfall.html

To quote:

"One of the differences between the 9-point AF sensor used in the 20D versus the 45-point Area AF sensor used in the EOS-1D class cameras is the size of the individual pixels on each. The pixels on the 45-point AF sensor are much smaller than those on the 9-point AF sensor, and one inevitable result of this difference is superior low-light sensitivity for the 9-point sensor. This is reflected in the specifications for each camera, which state the 20D's AF sensitivity at EV -0.5 compared to EV 0 for the 1D Mark II N. " (Chuck Westfall)

Hmm... Interesting.
 
Upvote 0
E

Edwin Herdman

Guest
And from that quote, which agrees with my reading elsewhere (I've said before that I forgot whether smaller or bigger AF sensor sites are better, but think about it - smaller sites equals more precision), it's clear that you can make a better sensor in the same space if only you are willing to put some money into making them on a smaller process (and, of course, assuming that the lithographic production process allows this). The same economic and logistical challenges exist in making a good AF chip as in making the main sensor or the Digic / EXPEED processing chips.
 
Upvote 0
Status
Not open for further replies.