Big Mike's Digital Photography Blog

Autofocus: DSLR vs. compact digital camera

There’s a web page at the Nikon website which explains the difference between phase detection autofocus and contrast detection autofocus.

Contrast detection autofocus is pretty simple. The camera moves the lens around until the image hitting the sensor is in focus. In focus is determined by evaluating the contrast of the area within the focusing square. This is the type of focusing system used in all non-SLR digital cameras.

DSLRs use the much more complicated phase detection system. To quote the website:

Most single-lens reflex cameras use an autofocus method called the "phase detection system." Using a separator lens in autofocus module, this system produces two images from the image information of the subject captured through the lens. It then measures the distance between those two images using a line sensor, and detects defocus amount. If the focal plane is in front of the camera lens relative to the image surface (front focus), the distance between the two images will be narrower than the optimum value. If the focal point is behind the subject (rear focus), the distance between the two images will be wider than the optimum value (Fig. 1). In other words, if the distance between the two images is narrower than the corresponding distance in an in-focus condition, the autofocus system determines that the focal point is in front of the subject. If the distance is wider than the reference value, the system determines that the focal point is behind the subject. Then, the autofocus system computes instantaneously how much the lens position should be moved and in which direction, and moves the lens accordingly. This focusing operation is extremely fast.

I can’t say that I understand all that, but what I get out of it is that phase detection autofocus requires separate hardware for determining focus, and that the separate hardware and the lens all have to be calibrated perfectly, otherwise the system will misfocus.

This explains why DSLRs are often plagued with focusing problems, while you never hear about such problems on compact digital cameras. With complexity comes the potential for more things to go wrong.

Here are some examples of messages at a popular internet photography forum complaining about focusing problems on Canon DSLRs:

After two 40Ds and consistently soft focus, I sent my camera into Canon. I recently photographed an award ceremony on Capital Hill and every shot was soft. I never had this problem with my 20D. Occasionally, I do get sharp photographs out of the 40D, but rarely. I am using all L glass and am a true believer in Canon. I am anxious to see what Canon does to the camera. I see other posts with sample pics and they look so sharp. It can't be operator error all the time. [link ]

* * *

Too be honest my faith in Canon (Not Cannon) was completely lost when I bought my 40D... I seem to have non stop problems with the camera, here is a list of things that have gone wrong.

-When I let the camera choose the focus point, my fence is in focus and not the bird.

-When I set it to Auto Mode or Sports Mode, my pictures still come out blurry even if I try to push the button faster...

-When I tested the focus on a paper with lines on it, the lines weren't in focus, I even tried it manually by setting the white balance to shade, and shutter speed to 1/10s but it is still blurry, I tried fixing the focus with setting the camera sharpening to 7 but it still has blurry pictures. Clearly it is back focusing and a bad camera... [link ]

* * *

After many failures at Canon Service to calibrate my 350d focus I reached the conclusion that it is impossible to have it calibrated.

I continuously moved from strong back-focus to strong front-focus (as it is today).

Never inside the deep of field, with all the 3 lenses I have.

For this reason, I'm thinking to the possibility to buy another camera, rather than sending again my 350d for calibration (after 7 times).

Actually I don't need another camera. I just would like to solve my focus issue. [link]

So you see, DSLRs may have faster autofocus, but the autofocus system on compact digital cameras is more reliable.

Central Park South (looking east)

This photo was taken only a few feet away from where I took the test shots for the G9/E-500/P200 comparison. This demonstrates that if you walk a few feet, point the camera in a different direction, and maybe wait for better light, you then have a whole new picture.

This photo was taken with my Canon G9 at a 36.78mm focal length, which is equivalent to 174mm on a full frame camera.

Whether or not the G9’s 6X zoom range, from 35mm to 210mm (equivalent) is a good thing or a bad thing is much debated on internet photography forums. There exist the purists who believe that any focal length unobtainable by a Leica rangefinder is just a useless gimmick. They believe that a camera with a zoom range from 28mm to 90mm would be a much more desirable camera. Or even better, just a camera with a 28mm prime lens such as the overpriced Sigma DP1 (because zoom is evil and you are supposed to zoom with your feet). Well if I had the camera that the purists say I would be better off with, then I wouldn’t have been able to get this particular shot (or I only would have gotten it with considerable cropping). From the way I’m looking at things, the longer zoom makes the G9 a much more versatile camera than previous cameras in the Canon G series.

This shot also demonstrates an amazing feature of the G9: image stabilization. The traditional rule is that the minimum safe shutter speed when handholding the camera is the reciprocal of the focal length. Under that rule, I would have had to take this photo at a shutter speed of 1/174 second. But in fact, it was taken at 1/25 second, which is about three stops slower than the recommended safe shutter speed. Yet the photo shows no camera shake blur (at least none that I noticed). This is quite amazing! If I was using my supposedly more versatile Olympus E-500 DSLR, in order to get this shot I first would have had to change lenses (in order to obtain the telephoto focal length), and then I would have had to turn the ISO up to 800 in order to get the necessary shutter speed. At ISO 800, the E-500 has more noise than the G9 at ISO 80. Thus the G9, in this particular situation, is actually a better low light camera than the DSLR with its much bigger sensor. This is the power of image stabilization, and I am now a believer. Never again will I buy a camera without this feature.

* * *

The photo has annoying sharpening halos. I will write about why that happened in a future post.

Canon G9 vs. Olympus E-500 vs. Sony P200

I took my Olympus E-500 DSLR, Canon G9, and Sony P200 to Columbus Circle, and while facing southeast, attempted to take the same photo with each camera. All cameras were set to cloudy white balance (because the picture was mostly in the shade, although there was some hazy sun), contrast -2, and sharpening -2, except for the Sony P200 which only has -1 settings. All cameras were set to - 1/3 EV. The Olympus and Canon were both set to F4.5. The Sony is at F6.3.

On the Olympus E-500, I'm using the 14-54mm F2.8-3.5 lens. This is the upgrade lens (around $425) and not the cheap kit lens.

I successfully captured approximately the same field of view with each camera. The Sony focal length is 51mm (in "full frame" equivalence), the Olympus and the Canon G9 both 50mm, so I was pretty close. Furthermore, the camera with the fewest megapixels is at the narrower focal length, so the P200 isn't at an unfair disadvantage.

Canon G9

Olympus E-500	Sony P200

I observe two things from the above photos:

(1) The Canon G9 has a yellower interpretation of the color of the brick building compared to the other cameras. As I said above, I set the cameras to cloudy white balance, so this is contributing to the yellowness. I believe the photo needs to have a slight cooling filter applied to it.

The Olympus photo has a magenta tint to it, which I think is an incorrect interpretation of the scene.

(2) The Olympus photo is darker. The Olympus seems to be underexposing by about 1/3 EV.

(Or since I did set all the cameras to - 1/3 EV, perhaps the other cameras are overexposing by 1/3 EV and the Olympus has the correct exposure. But don't conclude that the Olympus has better light metering--the Olympus will often produce a photo in which the exposure is way off, and that never happens with the Sony P200. Consistently overexposing by 1/3 EV is a good thing because it's easy to correct.)

Now, let's examining the resolution by looking at a street sign:

Sony P200	Olympus E-500	Olympus E-500 with sharpening	Canon G9

The Sony P200 has the obviously worst performance. The Olympus has only one more megapixel than the Sony, yet the No Parking sign is much more legible. This is especially true after sharpening is applied to the Olympus photo (the E-500 seems to apply the least in-camera sharpening of the three cameras). If I were only comparing the Sony and the Olympus, I might wrongly conclude that this proves that the large sensor of a DSLR outperforms the small sensor of a compact digital camera. But the G9 photo on the right demonstrates that this is not the case. The photo from the G9 shows more detail than the Olympus photo.

This explains why I was never satisfied with the Sony and its "Carl Zeiss" lens. The Carl Zeiss Foundation is obviously whoring its brand name out to Sony to be put on cheap lenses.

The Sony is so bad compared to the other two cameras that I'm no longer going to include it in the comparison.

Now let's look at another comparison demonstrating the superior resolution of the Canon G9:

Olympus E-500	Canon G9

In the crops above, the bricks mush together on the Olympus camera, but on the Canon G9 you can better make out the individual bricks. Once again, we see the superior resolution of the G9.

Now for some crops from the extreme upper right corner:

Olympus E-500	Canon G9

We see that the G9 has more chromatic aberration than the Olympus E-500 with the 14-54mm lens.

However, after I apply some chromatic aberration correction and "defringing" in ACR (yes, you can open a JPEG file in ACR), the G9 shot looks a lot better:

Canon G9 after CA correction and "defringing"

This flaw in the Canon G9 lens can be mostly corrected with some post-processing.

And now finally I get to a problem that irks me quite a bit, blown out highlights. Because the sun is shining, there are various places in the photo where there's a blown out highlight. Let's look at the street lamp (and please note that I increased the midtones brightness in the E-500 photo because the blown highlights look worse when next to dim midtones, and I didn't want to unfairly make the E-500 photo look worse):

Olympus E-500	Canon G9

I was expecting the Olympus to outperform the G9 in this comparison, because the much bigger sensor of the DSLR is supposed to have greater dynamic range. Furthermore, the Oly was underexposed by -1/3 EV compared to the shot from the G9, and that once again favors the Olympus. Further handicapping this comparision in favor of the Olympus, the Olympus photo is developed from a RAW file, while the Canon G9 is simply a JPEG. But surprisingly, I like the look of the crop from the Canon G9 photo better. The DSLR should have performed better here, but it did not rise to the challenge.

CONCLUSION

The Canon G9 is producing better image quality than I get with the Olympus E-500 DSLR. This is despite the fact that the E-500 has a much larger imaging sensor than the G9. From now on, I'm never going to bother to use the E-500, except when I need some special attribute of the E-500 or one of its multiple lenses. What attributes might I be talking about? A wider angle than 35mm, or shallow depth of field which gives a photo the look of a more expensive camera. The E-500 also does ISO 800 pretty well, and the G9 really falls apart at ISO 800. The E-500 is definitely better for photographing people indoors. But for regular pictures of buildings, landscapes, still lifes, and street photography, the G9 produces the higher quality image. On top of that, the G9 is less conspicuous and there's no shutter noise. And the G9 has image stabilization. Yes, I will no longer be getting much use out of the E-500.

SOME OTHER OBSERVATIONS

In the course of doing this comparison, I also noticed the following two things:

(1) The G9 produces sharper photos at F4.5 than it does wide open, so I suggest stopping down to at least F4.5 if you want to maximize image quality.

(2) With this particular scene, there is no benefit at all to shooting in RAW.