Canon EOS Digital Rebel XT
(350D) Review
by Bob Atkins; created 2005
Initial Impressions of the Canon Rebel XT
The Canon Rebel XT is small. I know the specifications give you the size, but
there's nothing like picking up the camera and holding it. In fact for me the
Canon Rebel XT is on the verge of being too small. Now I'm a 6ft male,
so my hands are, I guess, fairly large. If I was a 5ft female, maybe the size
would be perfect. The image below shows the Canon Digital Rebel XT compared to
several other cameras:
As you can see, the Rebel XT is significantly smaller than the
EOS 20D. It's almost hard to hold in my right hand. It's comparable in size to
the HP945 digicam and the old Canon IX lite APS (film) camera. The Digital Rebel
is 4.98" x 3.71" x 2.63", the EOS 20D is 5.7" x 4.2" x 2.8".
The Canon Digital Rebel XT feels less "sturdy" than the Canon EOS
20D. Part of this I'm sure is due to it being lighter. It's also finished in
slightly rough hard plastic. The 20D has a softer rubberized finish in the grip
areas, which is easier to hold. Looking through the viewfinder it's evident that
the image is pretty small. The illustration below gives you some idea of the
relative sizes of the viewfinder image in the EOS-3, EOS 20D and Canon Rebel
XT:
Some people don't like small viewfinders, but personally I could
live with it. Sure it's smaller, and larger would be nice, but it's by no means a
deal breaker for me.
Operationally the major differences between the Rebel XT and the
Canon 20D are in terms of convenience and ergonomics rather than functionality.
For example setting flash exposure compensation is done on the EOS 20D using a
button and control dial. On the Canon Rebel XT it is done by going into the menu,
selecting the appropriate page, scrolling down to the appropriate entry, hitting
"set", selecting the amount of compensation using the scrolling buttons and
hitting the "set " button again. Both work, but making changes is faster and
easier on the Canon 20D. Actually in one case the use of the menu turns out to be
an advantage though. Setting white balance on the Canon Rebel XT is done via the
menu screen, and the options are given as symbols with descriptive text. With the
20D you set WB via a button and control dial, but the display is on the data LCD
and all you see is a tiny symbol, which is not only difficult to make out, but
remembering which symbol corresponds to which WB setting!
The cursor buttons also act as dedicated buttons for ISO, AF, WB
and metering mode. You press one of the buttons, it brings up the menu screen,
you select the option with either the control dial or the cursor buttons, then
you have to remember to push "SET". If you don't push "SET", the new setting will
not take effect - and you won't know it hasn't unless you check on the data LCD
or in the appropriate menu.
The menu screen of the Canon Digital Rebel XT is a little harder
to see in daylight than that of the 20D due to the structure. The Rebel XT uses a
"Powershot" style menu with 5 selectable pages. However the page text is somewhat
"grayed out" in a low contrast mode until the cursor is moved onto the page. The
Canon 20D uses a longer, scrolling menu screen, but all the entries are at full
contrast all the time.
Another Rebel XT/20D difference is in the battery charger. The
Rebel XT charger glows red while charging and green when charged. The Canon EOS
20D charger shows the state of the charge in 4 levels by means of flashing LEDs.
No big deal, they both charge batteries, but the 20D charger does does it with
more user feedback. Since the battery used in the Canon Rebel XT is the same as
that used in many of the lower priced "Powershot" digicams, I assume the same
(less expensive) battery charger is used. It works fine of course.
Shutter Noise
The shutter noise of the Rebel XT is lower and less "harsh" than that of the
20D. Since I can think of no reason anyone would want a louder shutter, score one
for the Canon Rebel XT! However the 20D shutter is faster, has a higher sync
speed, has less viewfinder blackout time and a shorter lag time, so you are
trading off noise for performance. Below are plots of the recorded audio of each
shutter.
If you click on the links you should hear a sequence of 4 shutter firings for
each camera. They were recorded with the microphone about 2" behind the camera.
[Note: these are 250KB .wav files]
Image Noise
One of the critical parameters of any digital sensor is its noise level.
Consumer digicams suffer badly from noise due to the small size of their pixels
and most aren't really usable above ISO 400. DLSRs are much better, with usuable
ISO ratings of ISO 1600 or even ISO 3200. The Canon Rebel XT has a maximum ISO
setting of 1600 (the 20D goes to ISO 3200).
An obvious comparison to make is the noise level of the Canon Rebel XT vs.
that of the EOS 20D. The images below are 100% crops from images of a grey card
shot at each ISO setting
As you can see, noise levels are very similar. If there are differences,
they're too small to matter. For all practical purposes the noise level of the
Canon Rebel XT is the same as that of the Canon EOS 20D.
Sensor Resolution
In theory the resolution of the Canon Rebel XT sensor should be the same as
that of the EOS 20D sensor, since they are approximately the same size and have
approximately the same number of pixels. Below are two shots which are 200% crops
from the center of the image of a resolution test chart, shot with a Canon EF
300/4L at f8 using a Canon Rebel XT and an EOS 20D
It's pretty clear that, as expected, there's no significant difference in
resolution between the 8.2MP EOS 20D and the 8MP Canon Digital Rebel XT.
Continuous shooting Frame Rate
Canon specifies that the Canon Rebel XT should be capable of shooting a 3
frames/second (vs. the 5 frames/second of the EOS 20D). To check this I shot a
series of images of a running stopwatch. Below are the first and last shots of a
9 frame burst (large/fine JPEGS). That's as much as the XT buffer holds.
As you can see the 9 frames took just about 3 seconds, confirming Canon's
specification of 3 fps.
White Balance
The image below shows the white balance under tungsten and fluorescent
lighting for both the EOS 20D and Canon Rebel XT. These are crops of a shot of a
Kodak 18% gray card.
The bar below the images shows what a theoretical 18% grey patch should look
like. The small numbers in white are the Red, Green and Blue (RGB) intensity
values. As you can see neither camera gets it quite right under any of the white
balance settings! All Canon cameras seem to bias towards red/yellow for tungsten
in both AWB and Tungsten modes. Some have suggested that this is because people
expect indoor shots under Tungsten lighting to be a little red/yellow, so Canon
built in that bias. The best neutral performance is shown by the 20D using AWB
under fluorescent lighting.
The images below show the same gray card shot under daylight
illumination from a cloudy sky.
Again there are differences, but they are not so noticeable. In
both cases, Auto WB does a slightly better job than selecting Cloudy white
balance. The aim is for all three numbers to be the same (which is what makes a
neutral gray). It's noticeable that the Canon Rebel XT images in all these tests
are slightly lighter than those from the 20D. In theory, a shot of an 18% gray
card in sRGB color space (gamma=1.8) should be RGB 117:117:117.
Note that in all cases and under any given lighting, if you use the custom
white balance function you will get a neutral color balance, by definition. You
shoot a gray (or white) card under the desired lighting and use that as a
reference. This results in similar shots of the gray (or white) card being
exactly neutral.
The next two shots are the exact same scene shot with the Rebel XT and the
Canon EOS 20D so you can see what a somewhat neglected lawn looks like in the
rain! White balance was set to auto.
As you can see, on a real world image (rather than a gray card), color balance
looks similar (I'd say slightly magenta). The 20D shot is still a little darker.
If you convert both images to grayscale, then average the whole image, the
average RGB value of the XT image is 108, that of the 20D image is 98 so in this
case both meter slightly darker than 18% gray (117) on this scene. Both look
better with slight level and color tweaks though, as shown below:
In these and other "real world" exposure tests I saw the Canon 20D give, on
average, about 1/3 stop less exposure than the Canon EOS Digital Rebel XT. This
is with both cameras using exactly the same lens at the same aperture and the
same parameter sets and other settings, and it was independent of metering mode
(evaluative, partial etc.). This appears to be due to metering differences, not
ISO differences, since at a given aperture the Canon 20D seems to select a
shutter speed about 1/3 stop slower. Whether you regard this as a defect or a
feature could, I guess, depend on your point of view. The 20D images are a little
darker - on the other hand they give you 1/3 stop more "headroom" for highlights
under default conditions. Whether this difference is intrinsic to the camera
metering algorithms and was designed that way by Canon or simply represents
sample to sample variation in these two particular cameras, I don't know since I
only have one sample of each to test.
I don't currently have the ability to check or compare color accuracy in any
scientific way. I can say that the XT and 20D seem pretty similar in normal use
on normal subjects in normal lighting. I've been using the 20D for a while now
and have had no real color problems, so I'd expect the same to be true for the
Canon Rebel XT.
Infrared Capabilities
While no DSLR is specified for IR use, many people are using them for IR work.
Sensitivity isn't high and exposures are long, but I have seen some pretty good
examples shot with cameras like the 20D and Digital Rebel.
Below is a comparison of shots taken with the EOS 20D and Canon
Digital Rebel XT using a Hoya R72 filter. This blocks most visible light and
allows most IR over 720nm to pass. There wasn't a lot of IR around for these
shots. Conditions were overcast
. 
There are a couple of things worthy of comment. First, to get
roughly comparable image brightness the Canon Rebel XT needed an exposure of 2.5s
when the EOS 20D needed 15s, so the Rebel XT seems more sensitive to IR. However
this may not be as good as it seems. As you can see from the histograms, the
Canon Rebel XT's blue and red pixels seem to respond to whatever gets through the
R72 filter, while in the EOS 20D most response seems to come from the red pixles.
Whether this results in lower contrast is hard to say, but the Rebel XT's image
is lower in contrast than that of the 20D. Note these histograms are for images
taken using Auto White Balance. However the same general conclusions hold true
for the various fixed white balance options. Using custom white balance is a
better option, but since it uses an image shot with the camera as a reference,
channel response differences cancel out, so it's not the best option if you
wantto know what's going on!
Of course IR images are usually converted into black and white (as shown
above), and with enough manipulation both image can be made to look similar,
though the XT's image needs stronger corrections Note the central "hotspot" in
the XT's image. This is a phenomenon which has been observed with many DLSRs and
which depends on the lens in use. Some "hotspot" much more than others. It
appears to be due to some sort of back reflection from the surface of the sensor
or sensor filters interacting with the lens. Whatever the cause, it seems to be
significantly worse for the Rebel XT than for the 20D. So from these tests it
seems that though the sensitivity of the Rebel XT to IR is higher (and exposure
times therefore shorter), the intrinsic image quality (contrast) is lower and the
hotspot sensitivity is higher.
Real World Images
What I'm sure many people want to know most is how the image
quality of the Canon Digital Rebel XT compares to that of the EOS 20D. I shoot a
couple of comparison shots at ISO 100 and a couple at ISO 1600. Both cameras were
set to the same parameter set (set 2, all setting "normal") and the same lens was
used. Lighting was normal outdoor daylight.
My conclusion from looking at these few images is that image
quality is very very close. In fact I really couldn't see any significant
difference in terms of sharpness or noise levels, even at 200% in PhotoShop. I'm
not saying there's no difference. What I'm saying is that if there is a
difference, it was to small for me to see based on "normal" shots of a "normal"
subject.
Above are sample crops at 100% from the images. The top row were
shot at ISO 1600, the bottom row at ISO 100. Can you tell the difference between
camera "A" and camera "B"? One is the EOS 20D and the other is the Canon Rebel
XT. If there's enough of a difference for you to worry about, the you're too
picky!
Conclusions
It looks very much like the image quality of the Canon Rebel XT
is up to that of the EOS 20D. It's also evident from using the Rebel XT, that
it's a Rebel, i.e. a camera aimed at the consumer entry level, while the Canon
20D is clearly aimed at the more experienced and serious photographer. Ultimately
in many respects the cameras will be capable of yielding almost identical
results, it's just that doing it with the 20D will be a little easier. Of course
there are some things the Rebel XT doesn't do as well. It has a slower frame
rate, a smaller JPEG image buffer, a lower capacity battery, the AF system is
different and, as mentioned, the viewfinder is smaller. However for many people
saving $400-$500 will make up for all those things. Personally, I'd buy the 20D,
but that's just me. Many people will buy the Canon Digital Rebel XT and be very
happy with it. Given the image quality and sensor resolution and long with the
low price, I'd say that it's currently the best performing camera in the "entry
level" digital SLR group and I can certainly fully recommend it.
Where to buy the Canon Rebel XT
Photo.net would like to thank
ADORAMA for making a camera available to us for testing. There are 4 models
available:
Links
All original text ?2005
Bob Atkins (www.bobatkins.com)
Add a comment
