When Canon announced the Powershot G11, they were the first camera maker to lower the resolution of the sensor when updating to a new model.  The G10 sported a 15mp sensor 1/1.7” sensor, whereas the G11 was outfitted with a 10.1mp 1/1.7” sensor. 

(I’m not sure what the 1/1.7” means, because it isn’t 1.7” in size.  This is the typical notation for camcorder sensors ... I’ll have to research to see exactly what it means.  The actual sensor size is 0.3”(7.6mm) x 0.25”(5.7mm).  This isn’t anywhere close to a 1.7” diagonal either, so I’m clueless as to what the 1.7” represents.)

Ever since the Canon 50D and G10 were introduced there seems to be some kind of internet ground swell complaining about camera makers and the megapixel race.  I’ve mentioned this in previous articles, and it seems these comments on many forums caught someone’s ear at Canon.  The reason you reduce the pixel count would be to trade possible resolution for gains in noise performance and dynamic range.  I’ve always had a problem chasing dynamic range, because I think cameras perform just fine.  Improving dynamic range is OK ... as long as you don’t have to give up some other aspect such as resolving detail.

I had already decided I wasn’t going to “upgrade” to the G11.  The G10, while pocketable, just wasn’t quite good enough for what I needed.  Rather than another camera with a tiny sensor, I decided I needed something better, and opted for the micro four thirds Lumix GF1.

Of course, all of this begs the question did Canon accomplish the objective of improving the cameras noise and dynamic range at the expense of reducing the pixel count?

As with any test I do, it isn’t super scientific.  Basically set up a scene, shoot at every ISO, and compare the results.  The only way to really do this is to vary the exposure time, since there isn’t enough latitude with the aperture on the small cameras to accomplish this.  Unfortunately in my mind this in some ways invalidates the test, because longer exposure times increase noise.  This would mean that my 1 second at ISO 100 would be at a disadvantage to my 1/15th of a second at ISO 3200.  Of course, this would exaggerate the differences if anything, and as long as each actual camera used the same shutter speed at the same f/stop the comparisons to each other are valid.

So here is the “scene” taken with each camera.  This is a mix of incandescent light as well as available light.  The day was overcast, and the amount of light stayed quite constant while I was shooting the test.  The small 3 step gray card was placed in the scene to be used for the noise evaluation.

I decided to try and “even” the field, and shoot everything in full manual mode including focus.  Using the Lumix I determined the base exposure was 1 second at f/8. The 7D and the GF1 were quite easy to focus, the other three were much more challenging (no surprise in this).

I used a tripod and 10 second timer, starting at the lowest exposure of each camera and moving through the entire range.  Everything was brought into Lightroom 3.0 beta, where all noise reduction and sharpening was turned off.  The only step performed in Lightroom was to set the white point based on the middle grey of the 3 step card.

Each cameras set of exposures was then opened in Photoshop as layers, and a 100x100 pixel section was cropped from the 3 step patch.  These were then all lined up and arranged in the resulting grid, saved as a jpeg at maximum quality to lower the chance of jpeg artifacts distorting the noise.

So here are a few observations.

First, note that the density of the images is different ... despite base exposures being identical, the resulting files vary.  If you look at the difference between the G10 and G11, the G11 exposure is actually brighter by more than 1/2 of a stop.  This means if you adjust the initial exposure when taking the pictures so the resulting densities match, you could use a lower ISO.  Shooting at ISO 100 on the G11 would give you almost the same density as shooting at ISO 200 on the G10.  If this is the case, it would imply the noise difference is about a stop more than the chart above, which to me appears to put the G11 at about a stop better. What I’m not sure of is whether this is due to the camera or if it results from the camera profiles used by Lightroom when imported.  I need to do some more research.

It is obvious that each camera at it’s maximum ISO is pretty noisy.  I’m sure each persons tolerance for noise is different, and subject matter is also a variable.  To me personally, I would say my max on the 7d is 1600, the GF1 is 800, the G11 and the s90 is 400, and the G10 is 200.  Using simple Lightroom noise reduction would all but eliminate the visible noise in these images. Obviously if I had to go beyond that for some reason I would, knowing I might have to work a little harder on the image to handle the resulting noise.

The s90 and the G11 actually both use the same sensor, and as expected performed very similarly.  I’m still unsure if the smaller lens of the s90 is optically as capable as the G11 ... or at least close to it. I hope to do some resolution tests with the two cameras soon.

One obvious question is whether upgrading to the G11 makes any sense?  I’m really not sure, because even though the G10 might have more noise, downsampling to an equivalent file size will help with noise.  The higher resolution sensor offsets some of the noise gain of the smaller sensor to a small degree.  The articulating display of the G11 is very nice, but other than that I’m not sure it offers anything at all.

In fact, since the s90 performs very similarly to the g11 and it will definitely fit in your shirt pocket it may be the logical choice for many looking at a higher quality point and shoot that can save RAW files.  At this point, the G11 looks like the odd man out ... a camera without a home.  If looking for a carry with camera, either go with something a little larger which offers significantly better quality, such as the Lumix GF1, or go ahead and get one that really will fit in your shirt pocket ... the Canon s90.

As I worked with these images it was obvious the 7D and the GF1 files were in a different category than the other 3 ... substantially superior. This really isn’t surprising, but seems to support my choice in the GF1 because it may even be an acceptable backup system.  The GF1 also appears to have less chroma noise than any of the Canon’s.  I can’t explain why, but luminance noise appears like natural grain and is more visually more acceptable in an image ... another reason to opt for a GF1 over a G10 or G11.

To me it appears Canon should face the music and realize the logical step now would be to increase the sensor size .. yes a completely new camera design.  Increasing the sensor size accomplishes all of the goals ... better noise characteristics, better dynamic range, and better resolution.  The micro four thirds cameras have definitely shown this. If you want a “shirt pocket” camera, might as well get a true shirt pocket camera .. the G11 won’t outperform the s90 by a significant amount.  With the success of the micro four thirds format, as well as Leicas new additions to the market, it appears this category may be more significant than previously thought, and currently neither Canon nor Nikon have a contender in that category ... in fact they don’t even offer an option.

For me, I realized that in my entire collection of my better images not a single one came from my G10.  I’ve already printed images from my GF1 and am confident it will yield high enough quality.  So anything not good enough for the GF1 will be taken with my other “pocket” camera ... yep, the iPhone .  I don’t need an s90 for just snapshots, the iPhone will be just fine for those.