Canon 7D Review - Focusing on Focus

As I said in a previous post, this blog is not a gear review blog. However, I just purchased a new Canon 7D DSLR and felt that I should add my two cents to various discussions that have been happening regarding this camera.

If you don't want to read about camera gear and the various arguments that photographers have about how many pixels can dance on the head of a pin, then you can stop reading right now and I won't be offended.

Some people, but certainly not all, have had issues with focusing and specifically image sharpness on the 7D. One of the more damning posts, from an image quality assessment, came from photographer Darwin Wiggett. He called the 7D "the camera we want to love". There are numerous other posts out there, particularly in the DP Review forums regarding softness issues with this camera.

Well, if this camera is so problematic, why did I purchase one? One reason I bought the 7D was to it use for wildlife photography. Though I don't do a lot of wildlife photography, it is something that clients occasionally request. My main body is a Canon 5D Mark II - a full frame body. Whereas a full frame camera gives you the advertised focal length of a lens (e.g., 400mm), a crop-framed camera such as the 7D gives you a 1.6X increase in effective focal length (e.g., 400mm = 640mm). That can be a great increase in 'reach' for relatively little money, particularly when a 600mm lens might cost $8000.

Of course, that wasn't the only reason I purchased a 7D; the improved autofocus and built-in wireless flash control were also important factors. Additionally, I knew that some people were getting outstanding results from their 7D's. I just hoped that mine would be sharp.

But I wasn't about to take that for granted, so since I received mine from B&H Photo last Thursday, I've been running tests on it.

This review focuses (no pun intended) solely on focusing and sharpness issues. Feature- and handling-wise, I think this is a great camera. But when it comes to 'pixel-peeping', it may be another story.

The first thing I decided to do was run the 7D up against my 5D Mark II.

Not a fair comparison, you say? Actually, all other things being equal, these two cameras should produce images with very similar resolutions.  The 5D II produces images of 5616x3744 pixels (21.1Mp); the 7D produces images of 5184x3456 pixels (18.0Mp). If you do the math, the 5D II theoretically produces 8.3% higher resolution than the 7D - which should not be discernible even at 100% enlargement.

The images produced by the 7D are only marginally smaller than the 5D II.

Note: all images in these tests were shot in RAW format and given default processing in Adobe Camera Raw. I did play around with sharpening, but it always introduced noise, so I considered that an unacceptable option.

Round 1: Photoshop Calls the FBI

I placed the camera on a tripod, set up a couple of flashes and took a few shots of a $20 bill in front of an old postal scale. I rarely use the scale, so it's covered by a layer of fine dust that comes in very handy in doing resolution tests.

I put the camera in One Shot AF mode, and took a bunch of photos at different f-stops. I did the same with the 5D Mark II. Even though the different sensor sizes produced a different crop, I adjusted the zoom on my lens (43mm for the 7D, 70mm for the 5D II) to give the same apparent subject size.

Images framed to give the same apparent subject size.

When I went open the RAW images in Photoshop, I made a very interesting discovery: Photoshop recognizes photographs of 'banknotes' and refuses to open them. It probably also reported me to the FBI. So I guess these photos might be illegal, but I swear I'm not trying to counterfeit anything!

Anyway, I ended up taking screenshots of the images at 100% in Adobe Camera Raw and then pasting those screenshots into another application that wasn't as 'sensitive' to potentially illegal activities.

One thing I wanted to test was the claim made by some that image quality for the 7D would begin to decline after about f/5.6 or so. I thought this concept was absolute bunk, so I tested the 5D II and 7D at various apertures.

All images were taken in manual mode, 1/80th sec, ISO 200, Canon 28-70mm f/2.8 L lens. The images below are 100% crops.

Canon 5D Mark II

Canon 7D

I notice two things from this test:

1. The people who say that you can't shoot over f/5.6 with the 7D are full of it. The f/11 image is clearly (if marginally) sharper than the f/5.6 image.
2. Holy moly, the 7D is as soft as Charmin compared to the crispness of the 5D II, particularly in the detail of Andrew Jackson's face.

Round 1 Verdict: I just wanted to toss the 7D in the trash.

Potential flaws in my technique: for this shoot I didn't move the camera, but rather zoomed the lens. Just as lens sharpness varies from aperture to aperture, it also varies from focal length to focal length, which may have affected the results.

Round 2: The Game is Afoot!

Actually, there wasn't much game afoot the day I took the 7D into the field. But I bought it mainly to use as a wildlife body, with it's fast and flexible AF and 1.6 crop factor.

Here is where I started to wonder what was up. Though it was a crappy overcast day, I got some decent shots with the 7D. The autofocus seemed sure and fast. But the results were as often horrible as they were decent.

Some shots came out nice, despite the crappy weather:

1/400th, f/5.6, ISO 800, AI Servo AF (view at large size for more detail)

Others were inexplicably soft all across the image. The image below has not one single area that appears perfectly in focus. I had a number of images like this.

1/400th, f/5.6, ISO 800, AI Servo AF (view at large size for more 'detail')

Those were about the only two birds I saw, but I did take a few photos of a sign in an attempt to gather more information about the camera.

1/200th, f/11, ISO 800, One Shot Focus AF, Canon 100-400 L IS, at 400mm

Even though I had to enlarge the 5D II image considerably to make up for the 1.6X magnification advantage of the 7D, the two images returned virtually identical levels of resolution, and the 7D has noticeably more noise.

100% crop of 7D image, enlarged crop of 5D II image (view full size for more detail)

Round Two Verdict: WTF?


Round Three: Tipping the Scales


So after that indecisive experience, it was back to the studio. This time, I tried repeating the first test, but this time without the $20 bill. I put the camera on a tripod, with a couple of flashes:

The results were all over the place. Some images seemed sharp, but most were soft. I had put the camera in One Shot AF and refocused between each shot when I changed the aperture. Each time the camera locked on the center part of the scale dial.

I couldn't figure out why some images seemed tack sharp while others were undeniably soft, when I suddenly noticed in one soft image that the dust on top of the scale was perfectly sharp. The problem snapped suddenly into sharp focus (OK, seriously, did you think I could resist that? Really?): the AF was not focusing at the same point on every shot!

Here are crops of two shots, taken one right after the other:

Both images are 1/80th, f/8, 100, Sigma 105mm macro lens

In both shots, the AF 'locked' on the center of the face of the scale, but only focused there on the first shot. The second shot is severely backfocused.

Round 3 Verdict: Aha!


Round 4: Back to Jackson

Now I had a theory to test, so I went back to my trusty Andrew Jackson $20 bill. But this time I wasn't going to photograph the whole bill - I was going to use my extremely sharp Sigma 105mm 1:1 macro lens and shoot only a tiny portion of the bill... and hope that Photoshop wasn't smart enough to tell that it was still a $20 bill.

My plan this time was to manually focus each shot, using live view at 10X to give me the sharpest focus possible. This time, I also moved the camera closer and farther from the subject in order to retain the exact same scale and framing for each camera (which is not as easy as you'd think).

I took two shots with each camera: one focused on the teensy tiny word "America" on the left side of the bill, and the other focused on the word "Jackson".

The two focus areas used for the images

The results confirmed my theory. When manually focused using live view at high magnification, images from the 7D are every bit as sharp and detailed as images from the 5D Mark II, if not more so.

7D on the left, 5D II on the right. Very difficult to discern any difference in resolution or sharpness. (100% crop for 5D II, 7D image slightly enlarged to match)

7D on left, 5D II on right. Here the 7D image appears clearly sharper than the 5D II image, possibly due to the fact that the 7D is using a better part of the lens. (100% crop for 5D II, 7D image slightly enlarged to match)

Round 4 Verdict: The sensor rocks, the AF sucks

Though it seems clear that the AF on my camera has 'issues', there are still a number of images I've taken with the 7D that are soft all the way around, with no part of the image appearing in focus (e.g., the egret photo). I don't have an answer for what's causing that.

But I do know that I have an RMA number from B&H and will be sending this copy of the 7D back. I'm hoping that the one that replaces it will have more reliable autofocus.

In truth, I almost hate to send it back since I was able to get such incredibly sharp images out of the 7D. But it took so much work to get those images, and I had so many more that were worthless, that there is no way I could rely on the 7D's AF for my daily business.