Flaw in 10D's AF design

[mishkin]

Finally figured it out.

FACT:

10D's AF sensor is the same as Elan7's (introduced in 2000).

FACT:

D30/D60's AF sensor is the same as EOS650's (introduced in 1987 - first EOS camera).

THINKING:

Why would Canon's engineers stick an archaic, 13 yr. old AF sensor in modern and rather expensive (D30 - MSRP $3000?) camera? Why wouldn't they use Elan's sensor in the first place?

ANSWER:

There was something special about EOS650's AF sensor that Canon's engineers chose it. It was not accidental.

That special was focus precision. The first EOS camera tried to demonstrate that electronic AF works and gives sharp pictures.

"Canon's first EOS camera had the latest technologies including a super microcomputer and a Canon-developed BASIS sensor for *high-precision* (sic!) AF."
(http://www.canon.com/camera-museum/camera/1987-1991/data/1987_eos650.html)

FACT:

D60's AF was slow and didn't work in low light. Users complained a lot.

FACT:

Canon quickly (next camera upgrade) responded by incorporating quick and sensitive Elan's AF sensor into 10D.

THINKING:

Would this alone explain 10D's AF low precision/inconsistency? Probably not.

FACT:

10D's AF speed was additionally improved over D60 thanks to "a revamped AF sequence (AF processing at the early sequence)".

THINKING:

What that means, I guess, is that AF gathers data at the early stage of lens movement and thus has to calculate the focus position by extrapolating over the longer distance. This longer extrapolation makes for greater AF errors.

But would all this explain 10D's AF low precision/inconsistency? There's more.

FACT:

10D has 1.6x crop factor.

THINKING:

Could the crop factor affect AF accuracy? Absolute accuracy (offset of focal plane from the sensor plane) - no. But what about APPARENT - or call it effective - accuracy? By effective accuracy I mean how much image is blurred due to misfocus.

ANSWER:

Yes, 1.6x factor does affect effective AF accuracy compared to full-frame Elan.

First of all, 10D's image on the main sensor has to be magnified 1.6x more to get the same say 8x10 print, compared to FF camera. Thus, the circle of confusion created by misfocus is now 1.6x larger in print. Effectively, amount of detail in the print is thus reduced 2.56x (1.6x1.6) compared to Elan which misfocuses by the same absolute amount.

Example: say both Elan and 10D's AF sensor misfocuses (e.g. due to miscalibration) by 100 microns. Say we shoot at f/2.8. The CoC due to misfocus will be 100/2.8 = 35 micron, on both film in Elan and sensor in 10D. Now, Elan's frame is magnified 8.5x to get 8x10" print (8 inches divided by 24mm). So, that CoC becomes 0.035*8.5 = 0.3mm. Not bad at all. Remember, acceptable CoC for 8x10 print is 0.25-0.3mm. Now, 10D's frame is magnified 13.5x (8 inches divided by 15.1mm). The misfocus CoC becomes 0.035*13.5 = 0.5mm. That's much more visible! The print is now visibly blurry.

Secondly, due to 1.6x factor, lenses with 1.6x shorter focal lengths on average are used. For example, you shoot waterfall from the bridge. You can't move. You want to fill the frame with the waterfall. So if with FF camera you would zoom to say 160mm position, with 1.6x camera you zoom to 100mm position (e.g. if your lens is 70-200). Why does this matter for AF precision?

Remember the Main Equation of Measurbation (http://www.canondslr.com/articles/m-dream/theory.html). The misfocus amount (distance between your subject and the plane of sharpest focus) is given by the formula

da = dx * (a/f)^2

where dx is AF error (offset of the focal plane from sensor plane), a is subject distance, f focal length.

Now it's obvious that when you use 1.6x shorter lenses, the misfocus amount da will increase by 2.56x. That's an eye-opening descrease in AF precision!

Example: say you shoot a model with Elan and 10D, which both misfocus by 100 micron (that is, their AF sensors are both miscalibrated by 0.1mm). Subject distance is the same (say 5m) in both cases (you don't want to move too close due to unflattering perspective, but can't move too far due to space constraints). To get the same framing, you use 160mm focal length on your 70-200 with Elan and 100mm with 10D. How much both cameras will misfocus?

Elan: da = 0.1*(5000/160)^2 = 98mm
10D: da = 0.1*(5000/100)^2 = 250mm

That's a huge difference! Suppose the DOF in 8x10 print from Elan was +/-100mm. Then the misfocus is hidden just within the DOF boundary and thus is not visible in print. Now, due to 1.6x larger magnification (which reduces DOF 1.6x), but 1.6x smaller focal length (which increases DOF by 2.56x), the net effect is that 10D has 1.6x larger DOF in 8x10 print, that is, DOF is +/-160mm. But misfocus amount is 250mm! Well outside of DOF and annoingly visible!

Besides all the above factors, there's also, of course, "100% crop examination factor" - most people printed 4x6's and 8x10's from Elan and never noticed any focus issues. Viewing 10D's images at 100% on screen is equivalent to 20x30" print.


CONCLUSION:

Putting Elan's AF sensor into 1.6x crop, 7.4-micron digital camera, and speeding AF even more by making AF algorithm guess/extrapolate more - was a bad, ill-fated decision. It is a technical flaw. It's an engineering oversight, probably done under pressure from marketing department.

--
Mishkin

"He's out there measurbating without any decent restraint, totally beyond the pale of any acceptable human conduct."

[lightandlife]

Thanks a lot, Michkin. That is an impressive logic.

I did not comprehend all of your explanations. But over all, it seems to make a lot of sense. If the variance of focusing error of a full frame digital is 1, that of 10D (due to 1.6x factor) is 2.56, and your calculations are consistent with it.

I was wondering why then focus errors do not show up in shots taken in bright light, but become more conspicuous under low light conditions. The answer may be, that the forucing error is so small during the day, that multiplying it by 2.56 is still so small. However, under low light conditions, the variance of the focusing error may become more visible.
Fewer photons make it more difficult for 10D to collect distance information.

In any case, I think 10D is a stop gap during transition. We wil eventually move to full frame digital cameras, hopefully within a year or two.

In the meantime, I am going to enjoy 10D.

[dbarthel]

Mishkin, you've conned enough people over at rob galbraiths forum. Pseudoscience always sounds impressive to the gullable.

[Derek Smith]

Yo dbarthel,

If Mishkin is conning us all, please elaborate just how this con is working so that we feeble minds can understand the way in which we are being gulled.

Or is your explanation just too complicated for us to understand?

[CyberDyneSystems]

I don't see a con,.

But despite a well thought out theory (it is very clear that mishkin has put a good deal of thought to this) there are many leaps of faith that one needs to take to agree with mishkin's conclusion.

It remains a theory that we can take or leave,. the vast majority of digital cameras that have been built rely on autofucos technology that was developed for film cameras. That the adaption of the Elan's autofocus system is an inherently bad idea with no hope of succes is a bit of a leap of faith. Why do other digitals succeed if using a film cameras autofocus won't work on a crop factor.

[jimmyd]

mishkin wrote:

Why would Canon's engineers stick an archaic, 13 yr. old AF sensor in modern and rather expensive (D30 - MSRP $3000?) camera? Why wouldn't they use Elan's sensor in the first place?

ANSWER:

There was something special about EOS650's AF sensor that Canon's engineers chose it. It was not accidental.

That special was focus precision. The first EOS camera tried to demonstrate that electronic AF works and gives sharp pictures.



Now I'm just thinking.... if the EOF650's sensor was so good, why did they reinvent the wheel (and spend a bunch of R&D money) and not just put the EOF650 sensor into the Elans?

[Guillermo Freige]

And why they didn't keep the EOS 5 5-point AF and infrared assist light? It works fast, and almost in total darkness, and is 1993 technology!!! I don't think it's expensive to manufacture today a 10 years old tech.

[ron chappel]

I couldn't belive this rubbish!I joined this forum just to answer!

Let's start at the begining shall we?

The D60's focus sensor is from the 650...Really?!
That would be why it has three sensor points while the 650 has one.
What a joke

[mishkin]

The most important part is that, according to D60's specs, its "AF precision is the same as the EOS 650's".

And if you prove mathematically that my arguments about effects of 1.6x factor on prominence of misfocus are wrong, I'll paypal you 20 bucks ;)

[dbarthel]

Mishkins assertions about which Canon autofocus technology is deployed in the D60 and 10D were directly refuted by Chuck Westfall from Canon on Rob's site. Yet Mishkin has the balls to move to this forum and restate the same misinformation. Shame, Mishkin, shame.

[ron chappel]

"af prescision same as the 650's"
Ok,if so it's an interesting theory.
not interesting enough to prove true or false though.Maybe one day...

[bnpndxtr]

mishkin wrote:
The most important part is that, according to D60's specs, its "AF precision is the same as the EOS 650's".

And if you prove mathematically that my arguments about effects of 1.6x factor on prominence of misfocus are wrong, I'll paypal you 20 bucks ;)

I don’t think that’s what this formula implies. I think it just means that with the 1.6x factor that the focus adjustment must change to accommodate, not that there is some kind of “residual mis-focus”. In other words the autofocus servo sees this as a “bias” added to the focal length. But….so what? I think your 1.6x argument falls apart if you just apply the same logic to a zoom lens for example. Those lenses have focal lengths that vary by huge amounts, yet the autofocus “works”. The autofocus is a closed loop servo, probably even including feedforward or predictive drive to speed it up. One could imagine that at some extreme zoom level that the motor drive can’t be controlled well enough to keep the lens from constantly over shooting the correct focal point, essentially resulting in an unstable servo loop. But it would seem you’d need a big focal length to make that happen. I think that all the 1.6x factor does to the scenario is make that “unstable” point get reached a little sooner (the unstable focal length doesn’t change, but the longest lens that gets you there becomes reduced with the 1.6x factor accounted for). Just my $.02

[BobbyC]

dbarthel wrote:
Mishkins assertions about which Canon autofocus technology is deployed in the D60 and 10D were directly refuted by Chuck Westfall from Canon on Rob's site. Yet Mishkin has the balls to move to this forum and restate the same misinformation. Shame, Mishkin, shame.

This says it all.

[mishkin]

dbarthel wrote:
Mishkins assertions about which Canon autofocus technology is deployed in the D60 and 10D were directly refuted by Chuck Westfall from Canon on Rob's site. Yet Mishkin has the balls to move to this forum and restate the same misinformation. Shame, Mishkin, shame.

CW's argument is weak. Basically, he found that one of the several strong statements is incorrect and argued that then everything is incorrect. Although I backed up my statement (see above), it's not most important of the several. Most important are the ones about 1.6x factor.

[BobbyC]

Most important is that you state something as fact that isn't, nuff said.

[mishkin]

Elan misfocusing by 100mm (within its +/-100mm DOF) and 10D misfocusing by 250mm (well outside its +/-160mm DOF) for the same AF sensor offset (constant due to miscalibration or random due to variances in calculating focus in particular situation) and the same framing of the subject is a sound example Chuck haven't commented on. Same for 0.5mm blur vs. 0.3mm blur in 8x10 print example. Is he afraid to acknowledge?

[Orogeny]

I guess my question would be if there is a design flaw, why don't all 10Ds have autofocus problems? Even if you assume, for the sake of discussion, that most 10Ds have an autofocus problem, what about the cameras that don't? If there is a design flaw, it seems that ALL 10Ds should have a problem.

[lightandlife]

Most 10Ds, if not all, have autofocusing problem under low light conditions, I think. (D30 won't take pictures even).

The manual seems to acknowledge it and suggests manual focusing when AF fails.

[Orogeny]

How does the 10D autofocus problem in low light compare to that of film cameras?

[NickC]

I'm just wondering how much Nikon is paying mishkin to spout off here in a Canon-friendly forum ;)

[mishkin]

Orogeny wrote:
I guess my question would be if there is a design flaw, why don't all 10Ds have autofocus problems? Even if you assume, for the sake of discussion, that most 10Ds have an autofocus problem, what about the cameras that don't? If there is a design flaw, it seems that ALL 10Ds should have a problem.

Imagine a car with a steering column which has some play. Even if steering wheel is centered (perfect calibration), due to this play the car will sometimes steer away from the straight course. When it's not centered (miscalibrated), the car will always steer away from straight line in one direction, plus the "random walk".

Now imagine a car with the same steering column with the same amount of play, but now the transmission ratio from the steering wheel to the axle is 1.6x greater.

Even if the steering wheel is centered (perfect calibration), the car will sometimes steer away, but with 1.6x larger amplitude than that of the first car.

When the steering wheel is not centered (miscalibration), the car will always steer away in one direction, plus the random walk, but both will be 1.6x larger in magnitude than that of the first car.

It is a design flaw to put the same steering column with the same amount of play in the 1.6x more sensitive car.

Even if your car's steering column is perfectly centered, the random component will always be there. It will not always reveal itself, but oftentimes it will. And 1.6x car will have it 1.6x larger.

All cars with 1.6x "magnification factor" thus have this design flaw.

It exhibits itself most prominently at fast speeds. Most drivers drive slowly and never notice 1.6x larger miscalibrations or 1.6x larger random variations. But someday they drive at full speed and then the flaw becomes obvious to them. Some drivers very often drive fast and for them there's no question that the flaw exists.


Translation:
steering column = AF
1.6x transmission ratio = 1.6x crop factor
centered steering wheel = calibrated AF
play in the wheel = variations in AF from shot to shot
fast speed = fast lens

[mishkin]

NickC wrote:
I'm just wondering how much Nikon is paying mishkin to spout off here in a Canon-friendly forum ;)

1200/5.6L per post ;) You know, I'm Canon user, so they have to pay me in my currency.

In fact, I do own Nikon. It's a... Nikon strap on my 10D.

[Belmondo]

I'm probably mistaken, but the 1.6 magnification factor is not magnification at all---it's a narrowing of the field of vision thus creating the effect of magnification. In other words, the magnification of a lens on a 10D or a full-frame camera is the same; the difference is that the 10D records roughly 60% of the field-of-view that a full-frame camera does. That field-of-view is equivalent to a lens with a focal length 1.6 times greater than the actual lens used. Depth of field remains the same in either case.

I suppose it's analogous to taking a picture with a full-frame camera but discarding the outer 1/3 of the image.

I guess what I'm trying to say is that there is no 1.6X magnification at the sensor. It's actually a cropping of the image, so it doesn't make sense that focusing issues should be any greater on the 10D than on any other SLR or DSLR.

Or am I confused?

If so, please be gentle.

[mishkin]

belmondo wrote:
I'm probably mistaken, but the 1.6 magnification factor is not magnification at all---it's a narrowing of the field of vision thus creating the effect of magnification.

Or am I confused?
If so, please be gentle.


You are correct. But to make the same 8x10 print, you magnify 10D's image 1.6x more than Elan's. Thus, any imperfection (such as blur due to misfocus) is 1.6x more visible. Also, if misfocus was hidden in DOF on FF camera, it may now show on 1.6x camera.

[lightandlife]

orogeny wrote:
How does the 10D autofocus problem in low light compare to that of film cameras?

Oh, well. 10D has an autofocus problem in low light.

Film cameras have no autofocus problem. Pictures will be all black!

[lightandlife]

NickC wrote:
I'm just wondering how much Nikon is paying mishkin to spout off here in a Canon-friendly forum ;)

Actually, I think Mishkin likes 10D, and he is debating whether the grade is A- or B+. On the other hand, Nikon did not offer any digital camera that can compete with 10D. Not even Sigma with fovion chip can compete with 10D at the moment.

[sjprg]

Come on guys, get off Miskins case. Even if he should be wrong, he is at least thinking and not just accepting the marketing hype. His logic sounds reasonable and time will show whether he is right or wrong. Instead of launching a personal attact, read and try to learn as much as he is trying to learn and form your own conclusions. This is a discussion group where ideas should be expressed and debated without personal bias. You aren't being forced to particape.

[ron chappel]

Get off mishkas case?
-when he's not so confrontational we will.
His theories being refuted by chuck westfall,the head honcho at canon usa should say something.

By the way mishkin,the guy who said the 1.6 cropping factor should make zero difference to the AF acuracy was right and no amount of enlargement factors can change that...

IF THE THEORY IS CORRECT THEN IT WOULD SHOW UP AS A VERY SLIGHT FOCUS INACURACY ON THE PRINT *NOT* OVERALL SOFTNESS

A tempting theory for the 10D focus accuracy problem but it cannot in any way be associated with any image softness problems

[Belmondo]

In the words of that great American, Rodney King: "Can't we all just get along?"

I think you all should just respect each other's passion and call it a day. It's clear you're never going to agree---at least not on this subject.

[sjprg]

With all due respect to Chuck, he is part of the Canon technical infomation group which if I read the title correctly is part of marketing, and as such can only disclose a limited amount of information. This article found by Michael explains a lot

http://www.sciam.com/article.cfm?articleID=000E3DD0-F20E-1C73-9B81809EC588EF21

After reading this I think that the whole film hertiage should be scrapped for digital cameras. If you think about how the image sensor works the whole focus, exposure, and shutter speed can be programmed digitaly. This would eliminate the mechanical shutter, the mirror, the diafram, the separate focus sensors with the mechalical alignment problems. The only reason for these items is the fixed component of film. This would leave only the lens motors as mechanical components.

Thinking out of the box if you programed the image sensor as the focus sensor and than changed its function at image time (shutter button press) it solves a lot of alignment problems, and would sure make the camera cheaper with all those machanical items removed. This in turn would allow a larger sensor with the money saved.

Paul

[RichardtheSane]

A bunch of interesting theories, some good, a couple seem flawed to me. Makes good reading.
Nothing more than good reading for me though, and anyone else who is happy with their 10D. At the end of the day if you are happy with your 10D and you are getting good sharp images then there is nothing to worry about. I get the same AF performance in low light with my 10D as I do with my EOS 33 (elan 7) - which is nice.

[ron chappel]

sjprg wrote:
After reading this I think that the whole film hertiage should be scrapped for digital cameras. If you think about how the image sensor works the whole focus, exposure, and shutter speed can be programmed digitaly. This would eliminate the mechanical shutter, the mirror, the diafram, the separate focus sensors with the mechalical alignment problems. The only reason for these items is the fixed component of film. This would leave only the lens motors as mechanical components.

Paul
They allready do this with digicams but it's harder or impossible with CMOS type image sensors.there may be a shift towards this in time.I for one would like preview and movie ability in my DSLR

[bnpndxtr]

mishkin wrote:

Imagine a car with a steering column which has some play. Even if steering wheel is centered (perfect calibration), due to this play the car will sometimes.............

You are being very thoughtful about this, but I am afraid the analogy still doesn’t fit well. I think you can boil it all down to this (a reduction of my earlier post):

1. The 1.6X factor manifests itself simply as a focal length bias applied to all lenses used with this camera, nothing more.

A bit more-

2. To an AF system, the 1.6x bias represents nothing but a gain parameter in the feedback loop (how fast the focus changes with a given amount of lens barrel drive).
3. The AF system already deals a huge variation in lens types and focal lengths, so the 1.6X factor is nothing different from this.
4. The part of the focus error that is not resulting from the servo loop action (i.e. the rest position slop) will penalize any positive magnification zoom lens, so the 1.6x bias only has meaning as it is factored into a particular lens stacking- but there is no mystery here- see item (1).

The effect of the scale factor, whatever it is, is indistinguishable from just plopping a positive magnification zoom lens on the camera. So I think your picking on a moot parameter in the scheme of things.

[Belmondo]

RichardtheSane wrote:
A bunch of interesting theories, some good, a couple seem flawed to me. Makes good reading.
Nothing more than good reading for me though, and anyone else who is happy with their 10D. At the end of the day if you are happy with your 10D and you are getting good sharp images then there is nothing to worry about. I get the same AF performance in low light with my 10D as I do with my EOS 33 (elan 7) - which is nice.

Once again RichardtheSane has 'hit it out of the park' (a popular baseball metaphor often used by Colonists).

We can spend endless hours speculating about why some cameras don't focus well, or for that matter, speculating why some people spend untold hours fretting about such things. For those of us whose cameras work properly and don't feel the need to ponder such imponderables, our sacred duty is to go forth and snap pictures.

Say 'cheese.'

[mishkin]

bnpndxtr wrote:
The effect of the scale factor, whatever it is, is indistinguishable from just plopping a positive magnification zoom lens on the camera.

1.6x crop factor requires 1.6x larger magnification of the images to final print size. Thus, any imperfection is magnified 1.6x more. Including misfocus blur.

Besides, use of 1.6x shorter lenses to achieve the same framing of the subject leads to 2.56x larger misfocus amount and 1.6x greater ratio of the misfocus amount to DOF.

[Belmondo]

mishkin wrote:1.6x crop factor requires 1.6x larger magnification of the images to final print size. Thus, any imperfection is magnified 1.6x more. Including misfocus blur.

Besides, use of 1.6x shorter lenses to achieve the same framing of the subject leads to 2.56x larger misfocus amount and 1.6x greater ratio of the misfocus amount to DOF.



Okay, Mishkin, let me see if I've got this.

I have to use a 50mm lens on my 10D to get the same image as an 80mm lens on a full frame camera. Right? And you contend that the 50mm lens is 1.6X shorter. Right?

First, my lens is not 1.6X shorter; a number higher than one, and the term 'shorter' are mutually exclusive. if you want to express it in those terms, it's .375X shorter. (The 50mm lens is .625X as long as an 80mm lens----the reciprocal of 1.6.) The amount by which it is shorter is thus the complement of .625, or .375.

Correcting your equation of 1.6 crop factor multiplied by 1.6 image magnification equals 2.56 greater misfocus, I get the following:

.625 misfocus (because of the shorter focal length lens) multiplied by the 1.6X magnification of image equals 1.0 times the amount of misfocus. In other words, it's a wash. Check the math.

Okay, this is a little tongue-in-cheek; I realize I'm making certain conclusions that may or may not be valid. I suspect they're just as valid as yours, however, and on the issue of 1.6X shorter, I know I'm right.

Unlike some of the others in the group who have heaped scorn and derision on you, I feel the need to compliment you on your tenacity. On the other hand, I really think that before you continue to espouse this line of reasoning, you should sit down and really think it through. Your crusade only serves to confuse people and worse. If one person decides not to purchase a 10D because of your 'theory,' then you've done great disservice to Canon and the person who otherwise would have owned what is arguably the finest camera on the market for the price.

Suggestion: run for governor of California. Right or wrong, at least you try. That's more than can be said for the bulk of our politicians.

[lightandlife]

The proof is in the pudding. The market will do what it will. If 10D produces good pictures, people will buy them, regardless of what experts say. There is no need for Mishkin to feel guilty. He has a right to expression his views.

Scientific matters are not determined by a mavority vote. Only one persons needs to prove or refute an argument.

[mishkin]

First, my lens is not 1.6X shorter; a number higher than one, and the term 'shorter' are mutually exclusive. if you want to express it in those terms, it's .375X shorter. (The 50mm lens is .625X as long as an 80mm lens----the reciprocal of 1.6.) The amount by which it is shorter is thus the complement of .625, or .375.

In other words, it's a wash. Check the math.

No, YOU check your math! LOL

and on the issue of 1.6X shorter, I know I'm right.

Wanna bet $20? ;)

Your crusade only serves to confuse people and worse. If one person decides not to purchase a 10D because of your 'theory,' then you've done great disservice to Canon and the person who otherwise would have owned what is arguably the finest camera on the market for the price.

My "crusade" is to call things by their names. Flaw is a flaw. It will benefit all of us if Canon takes action and redesigns AF. You are doing great disservice by taking "silence of the lambs" position and effectively stagnating AF development.

Suggestion: run for governor of California. Right or wrong, at least you try. That's more than can be said for the bulk of our politicians.

Can't compete with Arnold. He's Terminator! LOL

[bnpndxtr]

mishkin wrote:
1.6x crop factor requires 1.6x larger magnification of the images to final print size. Thus, any imperfection is magnified 1.6x more. Including misfocus blur.


That's the point- it is just magnification, but no different from just sticking a zoom lens on the camera. So....your point is minor.

See-

http://home.kc.rr.com/bnpndxtr/download/10D_AF_flaw.pdf

[Mark Kemp]

Who cares?

[mishkin]

bnpndxtr wrote:
That's the point- it is just magnification, but no different from just sticking a zoom lens on the camera. So....your point is minor.

See-

http://home.kc.rr.com/bnpndxtr/download/10D_AF_flaw.pdf


First example is correct. Second example is incorrect. If it were, a 600mm lens will give 30x larger blur than 20mm lens. Which is not true. The misfocus blur (CoC) is determined only by AF error dx (a characteristic of a given AF sensor and a lens) and F-stop: CoC = dx/F. The blur should be the same in the second example with or without teleconverter.

And... for me personally, a 2.56x loss of resolution is not a minor point.

[ilya]

I think that this thread is the ideal illustration that the circle of confusion is inversely proportional to the X-Factor.:cool:

[SeanH]

I'm at a loss as far as all the math is concerned. But I am buying it only due to my image quality comparing my old 10D (which I always thought was a bit soft) to my new 20D. To me there is a night & day difference between the image quality of the 2 cameras.......and yes I am using the exact same lenses. At least that's my 2 cents.