tow.com

I have been performing a number of tests of the Canon 1D Digital SLR. This document is focuses on the following topics:

March 21, 2002: This document was originally written analyzing an early-production EOS-1D. Since then, Canon has released a number of firmware updates and a hardware fix that upgrades the performance of these early 1D’s performance to the level of currently shipping units. This document will remain online (for the time being) as document for historical purposes. For the most up-to-date information on my 1D’s performance, please visit the photography section of this web site.

• Comparisons between the 1D and the D30 image quality
• 1D image banding
• Moire
• Long Exposure
• Printing Observations
• Conclusion

Note: For my initial impressions of the camera, read this article.

1D vs. D30

This section pits the Canon 1D against the Canon D30. I have been shooting with the D30 for well over a year. As a result, I am very familiar with its image quality for web and printing purposes. The following set of images compares the 1D with the D30 at various ISO levels. I used the following settings and equipment for both cameras:

• Lens: Canon EF 50mm f/1.4
• Aperture: 8.0
• Metering: Matrix-Metering
• AF: Auto
• Quality: JPEG High
• White Balance: Auto

The 1D was set to use Color Matrix 3, which is the closet to the D30’s color reproduction. The shutter speeds were kept constant at the same ISO setting, using the 1D as the baseline. This proves interesting, because the D30 registered a 1/2 stop underexposure for the same scene that the 1D registers perfect exposure. I moved the tripod forward and back to compensate for the focal length magnification factor differences between the 1D and the D30:

• 1D Effective Focal Length (EFL): 65mm
• D30 EFL: 80mm

I shot the following pictures at ISO200, 400, 800, and 1600. Click on the links below to see a 750×500 pixel sized image or click on the original link to see the full high-resolution image. Other than resizing, no additional image processing functions, such as levels, curves, or sharpening, were performed on the files.

As Phil has noted in his 1D Review, there is less noise in the 1D images than in the D30 images. At higher ISO levels, however, the noise pattern becomes banded on the 1D, whereas on the D30, the noise remains random or gaussian to the eye.

The images below were taken with the D30 at the optimum shutter speed calculated by the camera at a given ISO speed. Remember in the examples above, the D30 registered a 1/2 stop underexposure compared to the 1D.

I’m torn between which camera’s images I like best, but from these well-exposed images, I’ll give the nod to the 1D’s image over the D30’s. Your opinions may vary, however.

1D Banding

The topic of 1D banding in high ISO and pushed images has been a heated one among the Canon digital camera community. Here’s a preliminary analysis using in a controlled environment. The real test will come when using the 1D in real-world photography situations. Until I can perform those tests, check out the following images. All of the photos were taken with the following settings and equipment:

• Lens: Canon EF 50mm f/1.4
• Aperture: 8.0
• Metering: Matrix-Metering
• AF: Auto
• Quality: JPEG High
• EV Compensation: -2.0

It’s clear that the noise pattern is clumped into horizontal bands in nearly all the underexposed images above. The images are all dark, but if you boost the levels or curves, the horizontal banding is readily apparent. Take a look at a crop of the ISO200 shot below:

Banding at pushed ISO200

It’s interesting to note that the horizontal banding appears more prominent in the underexposed ISO200 shot than in the ISO400 shot. I’m of the opinion that photographers don’t always get their exposures spot on, even with digital photographers. I often underexpose an image so as to not burn out the highlights. Underexposing a 1/2-1 stop at ISO800 on my D30 gave me an effective shot at ISO1600 without the increased noise I would have gotten has I shot the image at ISO1600. Maybe this is a bad habit of mine that I should stop with the 1D, but it’s a habit I’ve grown accustomed to with the D30. I’m curious to find out if there are other D30 owners who do this on a regular basis.

Pushing It Revisited

Added: December 18, 2001, 10:03 am

Brian on DPReview suggested I compared a 3 stop pushed ISO200 image with a normal ISO1600 shot to see if there are any differences in the image. I actually took some -3.0 EV shots yesterday evening, but didn’t post them because the pushed images were too dark. Let’s take a look at them again:

I tried to make the pushed ISO200 image look as close to the ISO1600 image as possible by adjusting the levels in the image. I selected the white eye dropper and click on the scene until I got a reasonable approximation of the ISO1600 level settings. Clearly, there’s not much comparison between the two. You’re better off shooting at ISO1600 than trying to push ISO200. A better test would have been to shoot at ISO800 -1.0 EV and compare it to the ISO1600 image. Alas, the lighting conditions of the scene have changed at the time of this writing, which would throw off any critical analysis.

Dark Horizontal Stripe

If you look at the ISO1600 shot, you’ll see something even more interesting, a dark horizontal stripe across at the top of the image! Here is a 100% crop from that image:

Horizontal Strip at underexposed ISO1600

Egads, what’s that?!? I’ve seen this in a number of 1D sample shots around the Internet. Something that looks like it is even present in one of Canon’s own sample 1D images on their web site. Load up sample image two in Photoshop and examine the blue channel. You’ll see a dark vertical stripe to the left of the woman. Here’s a crop from the sample image’s blue channel:

Canon Sample Image #2

In a post from Rob Galbraith’s Canon 1D Forum, Chuck Westfall said that this was a pre-production issue that shouldn’t affect production 1D’s. In Canon’s sample image, you have to go to the blue channel to see this vertical line. On my ISO1600 underexposed shot, however, the horizontal line is plainly visible without having to look in the blue channel. As a result, I’m unsure whether the same issue is causing these vertical/horizontal line or if it’s really two issues.

You don’t have to push the camera to see this horizontal line on my 1D. Check out the following ISO1600 shot below:

Horizontal stripe at correctly exposed ISO1600

The image above also exhibits the banding issues discussed earlier in this document.

Some Like It Raw

Added: December 18, 2001, 8:00 pm

Macintoshes are great for graphic design and digital photographers, but Canon’s support for Macs is pitiful at best. One of the reasons why I haven’t shot RAW in the past with the D30 or the 1D has been the lack of good software for handling RAW files. On the PC side, there are plenty of third-party alternatives, but on the Mac, we’re stuck with Canon’s bundled TWAIN plug-ins and the Remote Capture application. I use MacOS X exclusively, with Photoshop running in Classic mode, but Canon’s software forces me to reboot into MacOS 9.2.1 Boo hoo, Canon!

One cool feature of 1D RAW files is that each image has 4 stops (-2 to +2) of exposure compensation. This means you can shoot a scene and adjust the exposure when transferring the file to your computer. When you shoot JPEG, the exposure is set when you pressed the shutter button. Having this ability is great for saving highlights and shadow detail in images. This feature alone should get me to use RAW more often, but man, the software on the Mac really blows (which rhymes with slow).

The forums at DP Review and Rob Galbraith’s are great sources of information about digital photography. If you are new to digital photography and have a question to ask, you may want to log in and do a search there. Chances are, someone’s already asked the question and has gotten some good responses. Earlier today, Andrew Grant suggested that I shoot the scene using RAW with +1EV, returning the image back to 0EV when transferring the RAW file to my computer.

Here are my results:

It appears to me that the +1.0EV image, with the stop of exposure dialed back in the RAW conversion, has less noise than either the 0.0EV or the -1.0EV images. Take a look at a crop from the three images:

Looking at these crops, it’s clear to me that the compensation +1.0EV images are much cleaner. Banding is visible in all three images, but it’s less noticeable in the +1.0EV compensated shot. I don’t know if this means everyone will be shooting +1.0EV RAW files, but it’s something worth considering, given the results I’ve seen here.

Update: December 18, 2001, 8:00 pm

It’s been noted on the forums that shooting at ISO1600+1.0EV is effectively shooting at ISO800.

Horizontal Line Revisited

What about that horrid horizontal line I saw in my previous images. I was at Stanford’s Memorial Church this afternoon, and I shot a number of images at ISO1600 and ISO800 with -1.0EV (JPEG, not RAW). Even in the small image below, you can see this dark, horizontal band along the top of the image. Note: I adjust the curves in the image below for web display purposes.

ISO1600, f/2.8 at 1/45 second @ 17mm

I didn’t have a chance to shoot this image in RAW with exposure compensation. I am perhaps more troubled by this horizontal line than I am the banding. If anyone has an idea as to what’s happening there, I’d appreciate your comments!

Moire

Added: December 19, 2001, 10:54 am

I’m noticed more moireing in my 1D images than I’m used to from my D30. Look carefully at the four windows in the picture of the Packard Engineering Building below:

Long Exposure

The Canon EOS-1D has exceptional long exposure performance. Here is a link to some night-images I shot at Stanford University in February, 2002:

• Stanford At Night

Printing Observations

Update: December 19, 2001, 21:55 pm

I printed out two ISO800 shots from the 1D and the D30 today. There was really no contest. The 1D’s image was remarkably cleaner than the D30’s. The 1D’s image did exhibit banding and the horizontal dark line. These artifacts, however, are not as annoying on the printed page as they are on screen.

The D30’s image exhibits much more color noise than the 1D’s. It’s more random than the 1D’s noise pattern, but it doesn’t make the image look any better. I’ll perform more printing tests in the future, but right now, it appears that the 1D’s prints are better than the D30, which is no slouch itself!

Conclusion

Just as different films react differently to light, perhaps the 1D’s noise pattern is just different from the D30’s. Many of us have grown accustomed to the look and feel of D30 images, so it’s a bit of a shock to see images from the 1D. The 1D’s noise pattern is definitely not as random as the D30, which has been praised countless times in the past for its film look. When creating photographs with the 1D, one must pay particular attention to the exposure settings of the camera. Perhaps there’s a reason why the 1D doesn’t allow the user to zoom in on the images — they would see the horizontal banding! Jokes aside, this type of noise is difficult to get rid of using Photoshop. It remains to be seen whether or not this noise pattern can be “fixed” via a user-installable firmware update or a send-to-manufacturer hardware update. The worst-case scenario, in my opinion, is that the noise pattern is a characteristic of 1D images. Just as people can tell if an image is a D30 image through a variety of means — saturated colors, buttery smooth tones, etc. — perhaps we’ll be able to tell 1D images apart by looking for horizontal noise banding and/or dark horizontal or vertical stripes.

All that being said, I think that if you get the exposure dead on accurate, you’ll be creating images that are superior to the D30 in terms of noise and color fidelity.

Next up on my plate is field testing the 1D. I’ll be using it just as I use my D30 today, as my main take everywhere digital camera. Stay tuned for more!