Canon PowerShot G9

Color (11.05)
Every camera reproduces colors differently, which is why we test color accuracy. For example, every camera renders the same blue sky a slightly different shade of blue. Many will shift the blues so the sky looks more vibrant, though some will dull the blue or oversaturate it so much it looks unnatural. Some cameras even turn blue skies purple. The principle goes for all colors, especially skin tones. We test color accuracy by photographing an industry standard color test chart, the GretagMacbeth ColorChecker. The ColorChecker is made up of 24 different color tiles, each a common color from around the color spectrum. The image below represents the G9’s color accuracy. The outer squares show the colors the camera reproduces, the inner squares show the actual color of the ColorChecker corrected for exposure, and the small inner rectangle shows the actual color of the chart under an even exposure. The reason the inner rectangles are brighter than the squares is because the G9 was most accurate when underexposed by -1/3 stop.

As you can see in the image above, many of the inner squares blend right into the outer squares. This indicates the colors are very accurate. The biggest problems appear to be in the blues and yellows; though as mentioned above, manufacturers often shift blues to make skies look more vibrant. In the G9’s case, the blues are shifted toward purple, making skies look richer without making them too purple. The graph below shows the color accuracy more quantitatively. The background of the graph is the color spectrum. The values of the ideal ColorChecker chart colors are shown as squares, and the colors the camera reproduced are shown as circles. The longer the lines connecting the circles and squares are, the worse the color error.

The graph shows how accurate the G9’s color is; only one blue tile is significantly shifted from its ideal hue. Colors are also slightly oversaturated, at 103.2 percent. The G9 will reproduce the colors in your photos remarkably well, and will make them slightly more saturated and vibrant. The camera scored extremely well here, similar to other high-end Canon point-and-shoots, and significantly better than the Nikon Coolpix P5000.

Resolution (9.06)
With 12.1 megapixels packed onto its sensor, the Canon G9 has the highest resolution in a point-and-shoot camera to date. We put the camera to the test by photographing an industry standard resolution test chart and varying the aperture, focal length, and exposure to determine the G9’s sharpest image. All resolution test images are run through Imatest, which evaluates resolution in terms of line widths per picture height (lw/ph) and sharpening level. The unit lw/ph represents the number of equally spaced alternating black and white lines that can fit across the entire image frame before becoming blurred.

Click to view the high-resolution image

The G9 has the best resolution at ISO 80, f/3.5, at a focal length of 14.8mm. Imatest measured 2009 lw/ph horizontally with 1.8 percent oversharpening, and 1926 lw/ph vertically with 11.7 percent undersharpening. This is more undersharpening than we would like to see, but the resolution numbers are still very impressive. However, the corners and sides of the images appear softer than the centers, and slightly blurred and washed out. Unfortunately, you will only get the impressive resolution of the G9 at the centers of your photos. Overall, the G9 had the second highest resolution score of any camera tested in 2007, only behind the Fuji FinePix F40fd.

Noise – Manual ISO (5.65)
One of the major problems with the megapixel race is that while cameras may have better resolution, their noise levels also significantly increase. More and more pixels are being crammed onto sensors that aren’t being increased in size, so the pixels are becoming smaller. The signal to noise ratio decreases with pixel size, meaning smaller pixels generally yield more noise. We test noise by photographing our test chart under even, bright studio light at every ISO sensitivity a camera offers. We run the photos through Imatest, which measures noise in terms of the percentage of image detail it drowns out.

The 12-megapixel G9 keeps noise levels moderately low up to ISO 400, but noise levels are very high at ISO 800 and 1600. The noise itself looks fairly uniform and has fine splotches of yellow and blue. It’s not the ugliest noise ever seen, but it certainly isn’t desirable. To its credit, the G9 avoids smoothing the noise significantly, which would also smooth over detail. The 12-megapixel camera manages to maintain good detail even at a high ISO sensitivity. The G9 edges out its lower-megapixel predecessor, the G7, in the manual noise score.

Noise – Auto ISO (1.56)
We also test the noise levels when the camera is set to Auto ISO. The G9 fired at ISO 200, yielding a moderate amount of noise, visible when viewed at full resolution (see the still life images below). This is more noise than we would like to see under such bright light, and thus the G9 earns a poor auto noise score.

White Balance (9.33)
Because all light sources have different color tints, it is important that cameras can recognize this and adjust color reproduction accordingly. This process, called white balance, is necessary for accurate color reproduction. Manually white balancing is almost always the most accurate way to go about it, but if you don’t happen to have a white card on you then you will be using the Auto setting or one of the presets. We test white balance by photographing the ColorChecker test chart under four types of light: flash, fluorescent, outdoor shade, and tungsten, using both the Auto setting and the appropriate presets.

Auto (9.33)
Set to Auto white balance, the G9 is very accurate using the flash and under fluorescent light. In fact, the Auto setting is even more accurate than the flash preset. However, under outdoor shade and tungsten light, Auto is very inaccurate, giving photos a blue cast in outdoor shade and a yellow cast in tungsten light.



Preset (9.34)
The Tungsten preset is much more accurate than the Auto setting, though the Cloudy preset does not help accuracy in outdoor shade. The Fluorescent preset is very accurate in fluorescent light, though surprisingly not as accurate as Auto. With the G9 it is a good idea to use the presets in indoor shooting, but stick with Auto when outdoors.

Flash WB - Flash Illumination

Fluorescent WB - Fluorescent Illumination

Shade WB - Shade Illumination

Tungsten WB - Tungsten Illumination

Still Life Sequences
Click to view the high-resolution image.

Still Life Scene
ISO 80	ISO 80
ISO 100	ISO 100
ISO 200	ISO 200
ISO 400	ISO 400
ISO 800	ISO 800
ISO 1600	ISO 1600

Low Light (7.97)
Not all photographs are taken in ideal lighting, so we evaluate the color accuracy of cameras in low light, as well. To test color accuracy in low light, we photograph the ColorChecker at light levels of 60, 30, 15, and 5 lux; 60 lux corresponds roughly to a room lit softly by two lamps, and 5 lux corresponds approximately to the amount of light that reflects on your face from your video iPod. All shots are taken at the camera’s highest ISO setting, which in the case of the G9 is ISO 1600.

Low Light Tests
60 Lux	30 Lux
15 Lux	5 Lux

Color accuracy was very good in low light, hanging around a mean color error of 7 or 8, even at 5 lux. Noise levels, however, were a different story. ISO 1600 is very, very noisy in the G9, and this certainly does not go away in low light.

We also test long exposure performance in low light at ISO 400. The G9 can be set to shutters speeds as long as 15 seconds. One thing to note about shooting long exposures is that it is frequently nearly impossible to manually white balance. You need to increase the shutter speed in order for the G9 to properly manually white balance. Other than this annoyance, the G9 has very good color accuracy in long exposures. Noise levels are moderately high, but more manageable than at higher ISO levels. The G9 scores identically in low light to the Canon PowerShot G7, its predecessor.

Dynamic Range (5.12)
Dynamic range is a measurement of how well a camera can discern detail at many tonal levels. A camera with good dynamic range will pick up information in both bright and dark areas of an image. This is especially useful in wedding photography (white dress and black tux) and landscape photography in bright sunlight (bright highlights and dark shadows). Like noise, dynamic range is also affected by the size of the camera’s pixels – the G9’s pixels are quite small because there are so many of them on the sensor.

We test dynamic range by photographing a backlit Stouffer test chart, which consists of a row of rectangles that vary in tone from the brightest white to the darkest black. The more rectangles the camera can distinguish, the better the dynamic range.

The G9 has decent dynamic range at ISO 80 and 100, but at higher sensitivities it falls off quickly (see the graph above). At ISO 800 and 1600 dynamic range is severely limited. Note that the graph indicates the best possible dynamic range with the camera, and these values my not be achievable in normal shooting conditions. The G9’s dynamic range performance is worse than average for 2007 cameras, though better than the Canon PowerShot G7 and Canon PowerShot S5 IS, and identical to the Nikon Coolpix P5000.

Speed/Timing – All speed tests are conducted using a Kingston Ultimate 120X 2GB SD Card, with the camera set to the highest resolution and best quality JPEG.

Startup to First Shot (7.6)
The G9 takes 2.4 seconds to start up and fire its first shot. It is slowed by its rather long autofocusing time.

Shot-Shot (9.3)
The G9 can be set to two different Burst modes, Continuous and Continuous AF. In Continuous mode, the camera takes full resolution photos every 0.7 seconds until the card is full. This is nice, and should help capture some action shots. In Continuous AF mode, the camera takes photos every 1.3 seconds, autofocusing between each.

Shutter-to-Shot (9.0)
The G9 has a lag time of less than 0.1 seconds when the shutter is held halfway down and prefocused, and a lag of 0.8 seconds when not prefocused.

Processing (6.8)
It takes 1.6 seconds for one 5.4MB photo taken at ISO 125 to be processed.

Video Performance (5.60)
Bright Indoor Light – 3000 lux
We test the video color accuracy by recording footage of the ColorChecker under bright studio lights at 3000 lux. The G9 has a mean color accuracy of 23.4, which is terrible, but actually quite normal for a camera shooting under tungsten light set to auto white balance. Noise levels are quite low, however.

Low Light – 30 lux
We also evaluate video performance in dim light. At 30 lux, the G9 has fantastic color accuracy, much better than many cameras have when shooting stills and manually white balanced. Noise levels are also kept quite low at 30 lux, drowning out only 1.5 percent of image detail.


Resolution
We record footage of our ISO chart to see how well resolution holds up in Movie mode. Sure, the camera uses the same optics, but the processing is quite different than with still images, and video compression changes video resolution tremendously. The G9 records 316 lw/ph horizontally with 6.8 percent undersharpening, and 456 lw/ph vertically with 22.3 percent oversharpening. This is more sharpening than necessary and introduces some image artifacting. However, resolution still stays quite sharp, as you can see in the crops below.


Motion
We take cameras out of the lab to shoot some video footage of moving cars and pedestrians to see how the motion looks. Motion captured by the G9 looks great, with excellent color, very good exposure, and sharp detail. However, motion stutters a little, and you’ll see some moiré on fine grid patterns, like brick walls. The Movie mode looks almost identical to the Canon PowerShot S5 IS, which has an excellent Movie mode. The G9 fell a little short of the S5 because its color isn’t quite as accurate.