Canon EOS 20D - is this review fair?

Peter Guest wrote:

Excerpts from a review of the EOS20D from a leading UK magazine:

What's unfair?

1- It has a known underexposure bias and is equipped with exp-comp, so
adjust as required.

2- Auto focus dependance is a bad habit, esp. if you use multiple
points. I doubt the camera scores high on mind reading either.

3- Post capture processing. Makes sense. It does depend on what
you're subject matter is and the end use, of course.

4- Spot metering: yes this is definite negative, esp. as highlights are
so easilly blown out. OTOH, for most shooting that most people do, a
review of the result in the monitor provides quick feedback.

Should you buy it? Sure. Or no.

Cheers,
Alan.

--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI] gallery & rulz: http://www.pbase.com/shootin
-- e-meil: Remove FreeLunch.

On Sat, 30 Jul 2005 13:18:36 GMT, wrote:

In message ,
Jan Böhme wrote:

No, this can't be implied. Whether the camera underexposes or not is
not logically connected to the ISO sensitivit as such.

Yes it is.

I have the deepest respect for your knowledge on sensors and noise,
but here I cannot back down, because what you say strikes me as
incorrect for purely logical reasons, and I have no reason to believe
that I am inferior to you in purely intellectual terms.

If the camera is set to ISO 1600, and it's metering for ISO
800, there is nothing 1600 about it, but a lie, and so is the "noise at
ISO 1600".

This is a non sequitur. There is no way to tell whether a camera which
is "set on ISO 1600, and [...] metering for ISO 800" underexposes,
overexposes or exposes correctly, based on only this information. If
the true ISO really is 1600, it overexposes. If the true ISO is 800,
it exposes correctly. Only if the true ISO is lower than 800, the
camera would underexpose. Thus, it is neither justified to claim that
an underexposed image is caused by the camera metering for a lower
ISO than set ISO.

AFAIAC, the *only* way to compare is to set the same f-stop and shutter
speed, on all cameras compared.

Yes, indeed. And in the comparisons of that kind that I have seen at
dpreview.com, Canon consistently omes out as slightly _more_ sensitive
than other cameras with the same stated ISO sensitivity. Hence, the
fact that Canon cameras in some instances underexpose cannot be
related to the their ISO sensitivity, as this would lead them to
overexpose, if there were a direct causal correlation.

The bottom line is that noise starts as a noise-to-signal ratio of the
sensor, and from there, is increased by digitization errors and
quantization.

This may very well be a very valid bottom line, but it is not a bottom
line that is pertinent to this particular discussion.

Jan Böhme
Korrekta personuppgifter är att betrakta som journalistik.
Felaktigheter utgör naturligtvis skönlitteratur.

wrote in message
...
In message ,
RichA wrote:

If you were taking a shot of something dark on a light background,
would you just guess and tack up the exposure to compensate?
This is the one thing that might prevent me from buying one.

Why would the 20D be any worse than any other camera for this? It's
probably one of the better DSLRs for dynamic range, because of the
relatively low noise.

It's good that the camera seems to underexpose to preserve
hightlights, but in certain sun-shadow situations, the exposure
difference can be up to 9 f-stops so blowing highlights to obtain
detail in something dark may be the only choice, but you have to be
able to expose for the darker object.

Would you really use the camera's default metering literally for this?
You can see for yourself what the camera really does, and work around
it. Set the contrast to -2 and look at the JPEG in the review. Or
bracket.
--


John P Sheehy


Almost 9 months with two 20Ds and previously 1 year with a 10D shows me the
EOS d series metering is highly questionable in it's overall accuracy. It is
also very likely when set to "matrix metering" to produce results similar to
over exposure. This could be mistaken for incorrect ISO rating.

If you use matrix metering in concert with a FX series speedlights in ETTL
mode, the exposures will look the reverse of using it without a flash in
matrix metering. This really weird behaviour is not limited to just one
camera. Both my 20Ds cannot be relied on to meter correctly all the time.
Canon are unable to 'fix' this situation either so it must be part and
parcel of the cameras. Despite this, I doubt that the ISO settings up to ISO
800 are inaccurate. Over that I think a lot of the claimed ISO is software
enhancement. I have never found the dynamic range of these cameras to be
greater than other brands. I don't think it equals some.
--
Douglas...
"You finally make it on the Internet
when you get your own personal Troll".
Mine's called Chrlz. Don't feed him, he bites!

In article ,
says...
I have never found the dynamic range of these cameras to be
greater than other brands.

Then again, you've never tested it objectively either.
--
http://www.pbase.com/bcbaird

Skip M wrote:
"RichA" wrote in message
...

On Sat, 30 Jul 2005 12:16:46 +1000, "pixby"
wrote:


Interesting stuff...
When I posted opinions like these about a month after the 20D was
released,
I got howled down by the disciples of EOS. Whenever I even suggested the
hallowed 20D might not be all it's cracked up to be, I get accused of
posting "non scientific" tests. OK...

I guarantee you that no matter what kind of test or how scientific you
make it, if you find fault with them you will be attacked.
"B-but, what was the white balance?"
"Did you use a tripod?!"
"You didn't use the right colour space!"
"If you had used raw, it would be different!!"

Even if you simply let two different manufacturer of cameras take
shots in "auto" mode. All of a sudden, there are excuses for why the
program mode just does not produce the results desired. There are
ALWAYS excuses.

The most desperate one of all, and it comes at the end is, "you must
have gotten a bad one." This is concerning products that are
manfucturered completely by computer and uniformly produced to the
0.001% mark. A "bad" one!
-Rich



Then, if he didn't get a bad one, then we got two very good ones, since we
haven't experienced any of the problems he has had with our 20Ds. So which
is it, since they are manufactured to the same tolerances?

From your previous post Skip, where you claim it's OK to have an $800
Speedlight sometimes work 1 - 1.3 stops out and sometimes work
correctly, depending on the lens you use and the fact your own cameras
were faulty from the factory suggests to me your expectation of getting
"two very good ones" is a little tainted by the God of EOS, wouldn't
you say?

If I spent as much on a car as I have on Canon gear and the car only
traveled at the speed limit if I had green seat covers on it, I'd be
equally as ****ed off as I am that I spent over $10k on cameras with
clear and frequently identified faults. You might open your eyes to some
of the complaints on DPreview about these cameras. I am not alone.

What do you do with your out of focus shots which had the little red
thingy light up on the correct point but the lens which cost 20% more
than the camera didn't focus on it?

--
Douglas,
Zero care factor for negative responses
from anonymous posters.

Alan Browne wrote:
Peter Guest wrote:

Excerpts from a review of the EOS20D from a leading UK magazine:


What's unfair?

1- It has a known underexposure bias and is equipped with exp-comp, so
adjust as required.

2- Auto focus dependance is a bad habit, esp. if you use multiple
points. I doubt the camera scores high on mind reading either.

3- Post capture processing. Makes sense. It does depend on what
you're subject matter is and the end use, of course.

4- Spot metering: yes this is definite negative, esp. as highlights are
so easilly blown out. OTOH, for most shooting that most people do, a
review of the result in the monitor provides quick feedback.

Should you buy it? Sure. Or no.

Cheers,
Alan.

Your answers are over simplistic Alan. The cameras are not predictable
in their wandering exposure values. Nor can they be relied on to focus
properly when the toggle is active for selection of focus points. The
"grip" now in the process of a recall is really a very poor substitute
for a decent body design.

It's not until you use a 'real' Professional camera like a 1Ds or Nikon
D2X that you discover how rough the consumer DSLR cameras actually are.
Unless Canon come up with some production controls and post assembly
testing procedures, any replacement for the 20D or 1D II will be no
better. With Nikon now a really viable alternative to Canon in the
Professional range, it won't be long before they get some stiff
competition in the consumer DSLR range too.

--
Douglas,
Zero care factor for negative responses
from anonymous posters.

"Pixby" wrote in message
...


The most desperate one of all, and it comes at the end is, "you must
have gotten a bad one." This is concerning products that are
manfucturered completely by computer and uniformly produced to the
0.001% mark. A "bad" one!
-Rich



Then, if he didn't get a bad one, then we got two very good ones, since
we haven't experienced any of the problems he has had with our 20Ds. So
which is it, since they are manufactured to the same tolerances?

From your previous post Skip, where you claim it's OK to have an $800
Speedlight sometimes work 1 - 1.3 stops out and sometimes work correctly,
depending on the lens you use and the fact your own cameras were faulty
from the factory suggests to me your expectation of getting "two very good
ones" is a little tainted by the God of EOS, wouldn't you say?

Good point, but not an usolvable problem. And one to which I did, indeed,
refer. But what I said was in reference to your litany of problems, none of
which we have had. So, in comparison to to yours, I'd still say we either
got two very good ones, or you got some very bad ones.

If I spent as much on a car as I have on Canon gear and the car only
traveled at the speed limit if I had green seat covers on it, I'd be
equally as ****ed off as I am that I spent over $10k on cameras with clear
and frequently identified faults. You might open your eyes to some of the
complaints on DPreview about these cameras. I am not alone.

What do you do with your out of focus shots which had the little red
thingy light up on the correct point but the lens which cost 20% more than
the camera didn't focus on it?

I haven't had to do anything with those "out of focus shots," because I
haven't had any that weren't my own fault. Certainly not the fault of a
lens that cost $200 less than the camera body, not 20% more. To what lens
are you referring?

--
Douglas,
Zero care factor for negative responses
from anonymous posters.

--
Skip Middleton
http://www.shadowcatcherimagery.com

Pixby wrote:

Alan Browne wrote:

Peter Guest wrote:

Excerpts from a review of the EOS20D from a leading UK magazine:



What's unfair?

1- It has a known underexposure bias and is equipped with exp-comp, so
adjust as required.

2- Auto focus dependance is a bad habit, esp. if you use multiple
points. I doubt the camera scores high on mind reading either.

3- Post capture processing. Makes sense. It does depend on what
you're subject matter is and the end use, of course.

4- Spot metering: yes this is definite negative, esp. as highlights
are so easilly blown out. OTOH, for most shooting that most people
do, a review of the result in the monitor provides quick feedback.

Should you buy it? Sure. Or no.


It's not until you use a 'real' Professional camera like a 1Ds or Nikon
D2X that you discover how rough the consumer DSLR cameras actually are.
Unless Canon come up with some production controls and post assembly
testing procedures, any replacement for the 20D or 1D II will be no
better.

You're comparing apples and oranges Dougie. The OP was asking about the
20D v. other 'consumer' Canon's such as the 350D and 10D.

--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI] gallery & rulz: http://www.pbase.com/shootin
-- e-meil: Remove FreeLunch.

In message ,
"pixby" wrote:

Almost 9 months with two 20Ds and previously 1 year with a 10D shows me the
EOS d series metering is highly questionable in it's overall accuracy. It is
also very likely when set to "matrix metering" to produce results similar to
over exposure. This could be mistaken for incorrect ISO rating.

Well, for the 10D, it is. It meters at about 2/3 the stated ISO. The
20D seems to be right on the mark, in my experience. The absolutes of
the two cameras are pretty close; I think full RAW number saturation in
the green channel at the ISO 100 setting is pretty much the same in
both, from memory (except that the 10D doesn't quite reach 4095; it
clips data at values something like 4006, 4005, 4003, 4002, and 3997,
depending on what vertical line a pixel is in - strange. These lines
are not scaled differently at all; just clipped differently. At the
higher ISOs, the data actually reaches 4095.

If you use matrix metering in concert with a FX series speedlights in ETTL
mode,

A seasoned photographer would use manual flash power if always at the
same distance from the subject. No influence from subject brightness.

the exposures will look the reverse of using it without a flash in
matrix metering. This really weird behaviour is not limited to just one
camera. Both my 20Ds cannot be relied on to meter correctly all the time.
Canon are unable to 'fix' this situation either so it must be part and
parcel of the cameras.

Maybe it a paradigm shift that you can't deal with, from what you were
accustomed to. I don't have any problems with flash, except that I have
to bump the compensation uniformly for all flash shots. The main
drawback is that I can't shoot something white with ETTL and "expose to
the right" by setting the +2 FC that I really want to.

It's annoying to lose a stop of + FC, but nothing that I would change
SLR systems over.

Despite this, I doubt that the ISO settings up to ISO
800 are inaccurate. Over that I think a lot of the claimed ISO is software
enhancement.

It is. On the 10D, "ISO 1600" has the same amplification as "ISO 800",
just metered a stop darker, with RAW numbers doubled. "3200" is
1600-level amplification, metered for 3200. The 10D does a gain-based
ISO 1600, but pushes its "3200".

This is a partially moot point, though, as it isn't really cheating, in
the sense that the same sensor voltage range is used as would be if
there were full amplification instead of multiplication. The capture
gets shortchanged by one bit of bit-depth. It would be slightly better
if full amplification were used, most likely, but this is probably at
the beginning of the rollof of the diminishing returns curve, and Canon
didn't want to bother with a better amplifier to do this, or just
thought that there would be no improvement (I think there'd be a small
improvement, useful mainly for binning or downsampling more accurately)

We could test all digital cameras by metering externally for a manual
exposure at an EI of 25,600 (which none, AFAIK, expicitly have), setting
all to their highest ISO, to compare their extreme low-light performance
(extreme only in a traditional sense; I could read a book in the type of
lighting that requires this EI). There is nothing unfair about this; we
would be comparing what the cameras can do in very low light.

I have never found the dynamic range of these cameras to be
greater than other brands. I don't think it equals some.

Dynamic range is just that; a range. It is not defined by one end from
an arbitrary middle point, but the distance between the 2 extremes that
meet some kind of image quality criteria. It is not "the highlight
headroom" of a particular metering mode. Given a common bit depth, the
image with the most dynamic range is going to come from the camera with
the least noise. Dynamic range also breaks down into full color and
monochrome ranges, based on white balancing. A 20D shooting under
magenta light will have the highest full-color dynamic range (and the
least chromatic noise). Shooting under green light would increase the
greyscale dynamic range, as the red and green channels digitize
completely different ranges of sensor voltages; the shadows coming
mainly from the green channel, and the highlights coming mainly from the
red channel, with blue falling in the middle.
--


John P Sheehy

On Sun, 31 Jul 2005 19:48:59 +1000, Pixby
wrote:

Alan Browne wrote:
Peter Guest wrote:

Excerpts from a review of the EOS20D from a leading UK magazine:


What's unfair?

1- It has a known underexposure bias and is equipped with exp-comp, so
adjust as required.

2- Auto focus dependance is a bad habit, esp. if you use multiple
points. I doubt the camera scores high on mind reading either.

3- Post capture processing. Makes sense. It does depend on what
you're subject matter is and the end use, of course.

4- Spot metering: yes this is definite negative, esp. as highlights are
so easilly blown out. OTOH, for most shooting that most people do, a
review of the result in the monitor provides quick feedback.

Should you buy it? Sure. Or no.

Cheers,
Alan.

Your answers are over simplistic Alan. The cameras are not predictable
in their wandering exposure values. Nor can they be relied on to focus
properly when the toggle is active for selection of focus points. The
"grip" now in the process of a recall is really a very poor substitute
for a decent body design.

It's not until you use a 'real' Professional camera like a 1Ds or Nikon
D2X that you discover how rough the consumer DSLR cameras actually are.
Unless Canon come up with some production controls and post assembly
testing procedures, any replacement for the 20D or 1D II will be no
better. With Nikon now a really viable alternative to Canon in the
Professional range, it won't be long before they get some stiff
competition in the consumer DSLR range too.

Speaking of quality control, what is "E99?" It popped up on
a Canon I was using (Rebel XT) and I keep hearing about it
from users. What does the code mean? You have to remove the lens,
turn the camera off and on again.
-Rich