Canon Digital Rebel XSi Digital Camera Review

by Richard Baguley
Published on June 01, 2008

Navigation
Intro	Physical Tour
Testing/Performance	Components
Design / Layout	Modes
Control Options	Image Parameters
Connectivity / Extras	Overall Impressions
Conclusion	Sample Photos
Specs / Ratings

Overall, the Canon Digital Rebel XSi has excellent  image quality, with accurate color, low noise, and plenty of detail. The images also hold up well in low light.

Color (11.11)
Color is one of the most important aspects of the performance of a digital camera - how good is the camera at capturing and reproducing the colors we see in the real world? We test this by taking photos of a Gretag Macbeth color chart in precisely controlled lighting conditions, then running an image analysis program called Imatest that takes the captured photo and compares it with the original. Because the original is a known quantity (the charts are manufactured to have precisely the same colors on them), Imatest can work out exactly how accurately the camera captures these colors. It then produces a chart that shows the captured image (in the outer rectangle), the luminance corrected color (in the inner square), and the original (in the small rectangle on the right).

 

As you can see, the colors for the Canon XSi are fairly accurate; there are only minor differences between the original color and the captured one. This is better shown by the following chart, which Imatest also produces. In this chart, the captured colors are indicated by the circles, and the originals by the squares. The longer the line between the two, the more the captured color differs from the original.





Again, this shows how good the color on the XSi is; the lines are short. The only exception is a couple of the subtle green colors and some of the blues, which are very slightly off. But generally speaking, the color captured by the XSi is extremely good. It's definitely superior to the Nikon D60, which acheived a score of 5.09 and has much less accurate color.

One thing to note; we got the best color accuracy using the XSi's Faithful Picture mode, which (as the name suggests) is designed for color accuracy. The XSi offers several different color modes, including Standard, Faithful, and a Neutral color mode.

Resolution (12.1)
Resolution is another important aspect of the performance of a camera - great color won't help in your family photos if the camera captures such poor levels of detail that you can't tell Uncle Bob from Auntie Doris. We test the resolution of the camera by taking photos of a standard resolution test chart under tightly controlled lighting conditions, then running the resulting images through Imatest to analyze how much detail the camera captures.

Imatest produces a result from this in line widths per picture height (lw/ph), which indicates how many horizontal and vertical alternating black and white lines the camera can capture before the image turns into gray mush. This is a great indicator of how much detail the camera can capture; the higher this measure, the more fine detail will be represented in the final image.




The highest result we got for the Digital Rebel XSi was 1957 lw/ph horizontally and 1779 vertical, with 11.7 percent undersharpening. This is an excellent score; it is greatly improved from the XTi and is nearly as good as the much more expensive Nikon D300 (which managed 2031 horizontally and 1762 vertically). It's also about equal with the similarly-priced Nikon D60, which managed 1937 lw/ph horizontally. Using the XSi's kit lens, we found the best resolution is achieved with the camera in the middle of the zoom range (about 32mm) and in the middle of the aperture range (about f9). Using the lens at either end of the zoom or aperture range did make a significant difference; the resolution dropped off significantly at the wide and telephoto end of the range, and a similar drop-off happened with the aperture either wide open or stopped right down. This is not unusual with the cheap lenses we see in kits, and is one good reason to upgrade to a better lens if possible; the camera can capture a lot more resolution than the lens it comes with is capable of producing.

The XSi captures a great level of detail and compares well with other SLRs in this test. It is somewhat let down by the lens, though; the cheap kit lens limits the level of detail that can be captured.


Noise (8.43)

Manual ISO (10.16)
Noise is the bane of the digital photographer's life; it's the stuff that sneaks into images and adds patches of blotchy color or a speckled appearance, especially when you bump up the ISO to increase the camera's sensitivity. It's an inevitable fact of the photographer's life, but camera manufacturers try and minimize it in various ways. The Canon XSi does a decent job of keeping the noise down, but there is definitely noticeable noise in the images it captures. It also does a worse job at keeping the noise down than the similarly-priced Nikon D60.

We test the noise in the images captured by the camera by photographing a color chart and analyzing the captured image in our Imatest analysis program. We run the same test at all ISO settings the camera supports; our score is based on both how much noise there is and how much the noise increases as the ISO is bumped up. Our tests produce the graph below; the noise as a percentage of the image is on the X axis, and the ISO is on the Y axis.


This graph shows that the noise in the images increases as the ISO increases. This is no surprise, but what is a bit of surprise is that the noise climbs quite quickly, and that at the ISO 1600 setting, the noise is just less than 2 percent, which is a lot. The high ISO noise reduction feature does make a big difference here, though: that's shown by the pink line on the graph, and it reduces the amount of detected noise across the board.

However, the noises are significantly more noisy than we've seen in other cameras. The Nikon D60, for example, had much less noise in images; we found around 1.35 percent noise in images taken at ISO1600, and 1.14 percent in images with Nikon's noise reduction enabled. That's a significant difference that underlines that the XSi is noisier than the D60. However, it is worth remembering that the XSi's noise levels aren't a huge problem, as all cameras have made huge advances in reducing noise over the past few years. It's also worth remembering that the noise is about the same as we found in the previous Rebel (the XTi) despite the increasing resolution (12 megapixels on the XSi, up from 10 on the XTi). Increasing the resolution makes the image sensor more prone to picking up noise, but Canon seems to have done a decent job minimizing this.


Auto Noise (6.71)
We also test the amount of noise found in the images when using the auto ISO setting. In our studio lighting setup, the XSi picked an ISO level of 200, and the captured images had only moderate amounts of noise in them.


White Balance (7.82)
You might assume that the color of objects is the same no matter what the lighting is, but that's not the case; it depends on the color of the light that is illuminating the object. Your eye compensates for this, but the camera has to try and judge the color of the light, then work out how this affects the color of the objects in the photo. This is called white balance, and we do in-depth testing of how well cameras judge the light and adapt to it. Generally, we found the XSi does a very decent job judging the light; in most lighting situations, the automatic white balance setting is accurate. However, we did find the automatic setting is a bit off with some light sources, particularly incandescent/tungsten light sources, such as studio lights. With this light source, the preset option produces better results. This means if you are using the Canon XSi with tungsten lighting, you may be better off using the preset rather than the automatic white balance setting. But the converse is true with fluorescent light; in this case the automatic setting produces better results in our tests.

Auto (8.5)
For our first test, we set the camera to auto white balance and illuminate our color test chart with different light sources, then run the results through Imatest to see how close the camera gets to making the white parts of the chart appear as pure white. The results are shown in the images below, but remember that these are exaggerated results to show the difference; you would not expect real images to look this bad.

Exaggerated White Balance Errors

Auto WB - Flash Illumination

Auto WB - Florescent Illumination

Auto WB - Daylight illumination

Auto WB - Tungsten illumination

As you can see, the auto setting does a decent job with most of the light sources; the flash and shaded daylight results are nearly spot-on. But the tungsten image is very inaccurate; there is a serious color cast to the white parts of the chart, and the grays are almost purple.

Preset (7.13)
In our second white balance test, we illuminate the color chart with the same light sources, but then use the appropriate white balance presets. Again, the results are exaggerated to show the error; you would not expect to see this much of a white balance error and color cast in a real photo.

Exaggerated White Balance Errors (Presets)

Flash Preset WB - Flash Illumination

Florescent Preset WB - Florescent Illumination

Cloudy Daylight preset WB - Daylight illumination

Tungsten preset WB - Tungsten illumination

Still Life

To further test the Canon XSi, we took a variety of photos of a happily married couple and a still life scene at every ISO setting the camera supports. To view the original, full resolution version of any file, click on the image. However, remember that these files are very large (some are more than 4MB), so they may take a long time to download.

Low Light (8.97)
Most photos aren't taken in great lighting; if you take photos indoors or at night, you have to make the most of the available light. That's why we test the performance of the camera in low light, looking at how well the camera adapts both as the amount of light falls off, and as exposure time of the camera is increased.

We found that the XSi performs well in both of these tests; the amount of noise in the images remains low and the exposure is accurate, both with the lower light levels and the longer exposures. This means you can shoot in both low light situations and with long exposures without the images you capture turning into a grainy, grungy mess. The amount of noise does increase as the light level falls off and the exposure time increases, but less than it does with many other cameras we've tested.

Our first low light test involves taking pictures with the camera set at ISO 1600, but with different amounts of light, ranging from 60 lux (about the same as a well lit indoor room) right down to 5 lux (about the same amount of light you'd get from a single candle in a dark room).

Low Light Tests
60 Lux	30 Lux
15 Lux	5 Lux

As you can see from the images above, the XSi does a decent job capturing the images in these tough lighting situations; even in the 5 lux test, the colors are accurate, there is not too much noise in the image, and it is correctly exposed. Although there is a noticeable difference between the different light levels, it is significantly less than we see on many other cameras, especially point and shoot models like the Canon Powershot SD1100 IS. This is one reason to choose an SLR over a point and shoot; the extra processing these cameras can do on images means they get better results in tough shooting situations. The XSi has about the same performance in this test as the Nikon D60; both are able to make good use of the available light and take decent pictures in low light situations.

Our other low light test involves setting the camera ISO setting to 400, then testing a range of shutter speeds from 1 second up to 30 seconds. Again, the Digital Rebel XSi does well here; even with the longest tested exposure of 30 seconds, the images have good color, exposure, and relatively low levels of noise. You can see how much noise we found in the images in the graph below; the amount of noise is pretty much uniformly low across the exposure range.


We also tested how the long exposure noise reduction setting affects the noise level of the captured images. This is shown by the pink line in the graph above, and you can see this helped reduce the amount of noise in the images by a significant amount. However, there is a price to pay for this; the camera takes another exposure after the main one is complete with the shutter closed, then removes the noise present in this second exposure from the main image. This means the exposure time is effectively doubled; if you are using a 30-second exposure, you have to wait another 30 seconds after the shutter closes before you see the final image.


Dynamic Range (10.83)
In this test, we examine the dynamic range of the camera. What this means is how much shadow detail a camera can capture; the bigger the range, the wider the range of shades from pure white to darkest black a camera can capture, and the better an image will look. We test the dynamic range at a selection of ISO settings, because the dynamic range often degrades as the ISO increases.

The XSi did well on this test; not only does it have a wide dynamic range at low ISO settings, but the dynamic range doesn't fall off as quickly as the ISO increased as on other cameras. In fact, the dynamic range seems to be somewhat improved over the older XTi, and is on a par with the significantly more expensive Canon 40D.


As the graph shows, the dynamic range of the XSi does fall off as the ISO increases, but at a slower rate than many other cameras. In particular, it keeps a wider range at higher ISO settings than the Nikon D60, so images taken at these higher settings will have more shadow detail.



Speed/Timing – All speed tests are conducted using a 4GB Sandisk SDHC memory card, with the camera shooting large, superfine JPEGs. This combination means the memory card should not be a limiting factor; the speed at which the camera can capture and process images should be the only thing holding it back.

Startup to First Shot (7.2)
We measured the average time to turn the camera on and take a photo at 0.98 seconds, which is pretty quick. Some other SLRs are quicker (such as the Nikon D300, which starts up in 0.3 seconds), but it is significantly quicker than most point and shoots, which usually take two or three seconds at best. Being able to turn the camera on and take a photo in less than a second means you are less likely to miss a shot. However, remember that this time does not include the time to remove the lens cap, so don't forget to take that off before you start shooting.

Shot-to-Shot (3.3)
In this test of how quickly the XSi can shoot image after image, we found that the XSi lives up to Canon's promise; shooting large JPEG files at the maximum image quality, we measured the average gap between frames at 0.3 seconds. That works out to around 3.3 frames per second (fps); just a whisker below Canon's specified speed of 3.5 fps. We were also able to keep shooting at this speed for 53 images, which should be enough for pretty much any sort of shooting situation.

One thing to note here: if the high ISO speed noise reduction is enabled, the shooting speed is greatly reduced if you are at an ISO setting of 400 or above. So, remember to keep that turned off if you are looking to get the maximum shooting speed in a dimly-lit room.

Shutter-Shot (10.0)
Canon claims to have improved the delay between you pressing the shutter and the image being captured, and again we found in our tests that this is true; we measured the average delay between the shutter press and the image being captured at less than 0.2 seconds. This means there is no real perceptible delay, so you should be able to capture all but the fastest moving objects by quickly jabbing at the shutter.

Processing (3.83)
The XSi is a little slow at displaying images, though; we measured the time between pressing the shutter and the image appearing on the screen at 2.2 seconds. This is a bit slower than many other SLRs, and could be a problem if you are the sort of person who likes to shoot, check the image, and then shoot again. Fortunately, the live view means this sort of shooter can get an on-screen preview before the image is captured, so it is not a huge problem.