It'sCanon
 

Ideas to park in your photo-brain

3. Bigger is Always Better

What do you believe?

Who among us would actually kid themselves? Not me, bro. I would never fool myself, given the opportunity to encounter the real deal. And neither would you.

Except, of course, when we read specifications, numbers, reviews and threads in Internet forums.

Funny thing about beliefs. We acquire them left and right without going to the trouble of prodding them with a fork. Don't look a gift horse in the mouth! When some outstanding authority says it's this way -- well, dad gum if it isn't the gods' honest truth.

Such is the case of the curiously expanding megabyte. Now that Canon has introduced the successor to the 10D dSLR, many people are considering it as their next camera, having decided that they're at the end of the road with their basic Digital Rebel. Time to move on up!

Or is it?

Speed freqs

My Canon 20D is all metal. My 300D is all plastic. At least on the outside where it counts. Or does it count at all? Do sharper, better color balanced, better exposed images drop out of metal bodied cameras?

My 20D shoots an 8.2 MP image, but my 300D only collects a paltry 6.3 MP image. Obviously two extra megabytes will make all the difference in the world, won't it?

My 20D shoots long strings of shots at five frames per second and my DR only collects four images at 2.5 fps. Is that slow or what?

My 20D rips through its menu and review system like it was on steroids. What's my Digital Rebel's hang up?

My 20D doesn't have wireless remote triggering capability. My DR does.

My 20D captures pictures that take up more CF space than those inhaled by my 300D (same lens, same tripod, same light, same subject, same focus, same ISO, same WB, same compression), so obviously they contain more detail, right?

Which of these are beliefs and which of these are important deal-breaking bits of vital intelligence?

Self-evidence

Some qualities are beyond argument.

#1. The ability to shoot a dozen images in 2.5 seconds is one of those things. If you like the near-movie like rapid image gathering that the 20D brings you, then this becomes an important, and perhaps deal-breaking, priority for you. But every silver lining has its cloud, and the 20D's motor drive brings several notions worth considering:

A. You can't shoot at LESS than 5 fps with it in continuous mode (unless a long shutter speed slows you down) and

B. Keeping the camera in continuous frame mode--to let your finger do the choice on the fly--makes for a very large number of two-shot bursts. It's quite a bit harder to train your finger to let go of the shutter button after a single shot. Of course, you will bless the camera when shot #2 looks better than shot #1 and curse it when your CF card seems to be filling up prematurely.

OK, B is minor, and I'd much rather have too many images to choose from than too few.

#2. Fast running camera computers are a blessing. You'll get no argument from me on that. I like to pull my eye away from the viewfinder eyepiece and see my shot playing back to me instantly and the 20D does this about two full seconds faster than the Digital Rebel.

And a One, and a Two...

I'm an impatient dude, so this feels Important to me. It may not be a deal-breaker sort of quality but it has its advantages when I'm moving fast.

#3. Remote Possibilities. If you need to wire your digital camera to a simple contact closure remote or trigger it with a wireless button, those are things the DR does well, and the 20D can only dream about. Of course, the 20D can use Canon's cool wired TC-80N3 remote with built-in intervalometer so this could be another dead heat in the race for features. I take that back. After living with the TC-80N3 for a while I must say this is a Killer Accessory, miles better than a simple wired remote and it costs around $130, so if you have a 20D, go out and get one, today.

#4. Eight point two megapixels capture better images than 6.3 MP shots. Sure, that's the ticket. Or -- whoops, what's this: (Roll your mouse over it to see a change.)

Two images shot through the exact same lens within moments of each other. In fact, Canon's 50mm f/2.8 macro AF. Both are crops at 100%, meaning that the display on your computer is one pixel of camera to one pixel of computer image. Any blow-up from here would degrade the image and any reduction would be casting out information.

You see shot A and shot B. Which is "better?" It's only fair to tell you that one of the images was made with the 20D's 8.2 MP CMOS imager and the other one was made with the Digital Rebel's 6.3 MP CMOS imager. Moreover, the larger 8.2 MP image was shrunk from its native 3504 pixel width to 3072 pixels wide using Photoshop CS' Bicubic Sharper scaling engine. Presumably the down-sampled image plus the Bicubic Sharper interpolation would hold the larger original to a micro-hair of perfection.

Three things showed up:

1. One image is slightly smaller than the other, even though Canon's published specification for the two CMOS image chips is identical in size to the fraction of a millimeter. But from the full frames, it's clear that one of these imagers is collecting a tad more image than the other. About 1.5% more, thus making that frame seem slightly smaller. That chip's width captures a slightly wider view, so it had to be slightly physically larger to intercept a wider portion of the image circle projected by the lens. No big fizz, just an unexpected result.

2. One image looks microscopically less well defined than the other. The sun was not its steady best between shots, so small exposure-driven differences can be chalked up to test conditions, but is that the difference between MP count, Photoshop interpolation or chip performance?

3. Color and tonal handling differences between the two shots is visible, but minor. Either could be tweaked in Photoshop Elements, CS or virtually any editing program and be altered to more closely resemble the other's color and tone. Yet right out of the camera, the slightly sharper picture seems to have the color and tonality edge.

But which is which? Write your answer down on a piece of paper right now to preserve your first impression. Your whole world depends on it. Don't try to out-guess the process; just do it and read on...

You've probably seen this image before. The inner blue rectangle is a 6.3 MP pixel field and the outer pink rectangle is an 8.2 MP array of pixels graphically comparing the 300D to the 20D field of photosites. Obviously the pink one can capture more details than the blue one. Right? Are you positive? Final answer?

Conclusions

When you apply a string of logical comparisons to any idea, you can turn the logic crank and new ideas pop out the other side. Aristotle taught this thousands of years ago, and in spite of his best efforts, digital photography took all this time to evolve, idea by idea, point by point, up to images of today's quality. Canon has diligently been trying to catch up. It's encouraging.

Presumably, a two-year newer design of CMOS image chip would do a better job of defining detail, preserving color and tonality and operating faster. Certainly the 20D is much faster in its capture -> write and review processes. (Phil Askey's excellent comparative noise level charts at DPReview.com show that the 20D makes a significantly less noisy image than anything out there at the highest ISOs, and that's a big plus.)

Being in hot competition with other manufacturers doesn't hurt, either. We who buy these things are always ready to jump toward the next greatest thing at the drop of a new product announcement. We--The Market--want it all, and we're very demanding. "More, better, faster, cheaper!" we shout, repeating as needed.

In the evolution of features, hardware, software, computers and more advanced model cameras, we expect More at every step, and when the specs say More is available, we get all warm and fuzzy inside. What-has-been-promised now equals what-we-want-to-believe, so all's right with the world.

So you and I would never intentionally fool ourselves, eh? And we would never put our faith in something just because someone with an agenda all their own said something to us, right? What is it people say about the percentage of things you can trust in newspapers? How does that translate into advertising?

No matter what your level of cynicism, you have to trust something, somewhere, sometime, right? Or are we at the mercy of our beliefs? All this chat about the concepts of belief, trust, truth and photography seems scattered, but when you read the paragraph after next, you may end up smacking yourself in the head again. It's a theme that runs through these Advanced Concepts articles.

Besides, I had to write several paragraphs here so you couldn't just scan down the Internet page and read the exciting conclusion to our test.

A:

The microscopically lower quality image in the rollover example above (B) is from the 8.2 MP Canon 20D. The slightly higher quality image (A) is from a Digital Rebel that is selling for just about exactly half of the price of the 20D body. Its tiny edge of detail is only part of the story. Tonality and color seem better, too, eh? So much for believing that 8 MP trumps 6 MP. If you ask me, and I realize you didn't, it's a virtual dead heat in the image quality department.

By the way, both cameras were firmly affixed to a tripod for this test. No fair thinking that some sort of extraneous camera-handling factor has degraded the 8.2 MP image. If you form a thought like that, it could be the irresistible pull of self-delusion trying to re-establish its priorities among your little gray cells.

Read the more thorough reviews from Imaging Resource and DP Review to gain a more in-depth insight under a wider variety of conditions.

From the test charts seen in these reviews, one notes that the 20D does about 10/9ths as well as the 300D Digital Rebel. Meaning that under laboratory conditions, the 14.1% larger imager delivers about 8.8% more actual fine detail. How do we know that? It seems that finely scribed black and white lines on a sheet of paper show it that way. The 20D portrays 2000+ image lines in this sort of test where the 300D tends to peak out around 1850.

Real world experience suggests that the actual performance between the two imagers and processing is substantially moot. It's so close between them that I have no hesitation at all to use the camera that is handiest for any given situation. If the lens I want to shoot with next happens to be on the DR, it's not worth the dust-risk to swap lenses in order to convince myself that the big --woo hoo-- 8 MP image is just what is vital right now.

Of course, as of March, 2005, this whole page may be moot. The Rebel XT uses the exact same image chip as the 20D but masks it down to 48 x 32 fewer image pixels. We think that was done for Bragging Rights marketing purposes, but they may have had some underlying technical reason (yeah, right) to deliver the microscopically smaller image. At any rate, the Rebel XT is a killer camera by any measure and pixel for pixel matches the image quality of the original Rebel, while delivering more of them. That makes handy crops easier to deal with.

Bottom Line:

The 300D is now a $700 camera with the Kit Lens. The 350D is now an $850 camera with the Kit Lens. The 20D is now a $1250 camera with a Kit Lens and the handling qualities, ergonomics, speed of operation, options and features of these cameras is closely reflected in their current prices.

Who Needs What?

  • For 20D owners--relax. You have a faster, more ergonomic camera with higher ISO images that show very low noise. If you are a pro who needs a second camera body, don't hesitate to buy a 350D in its black body option.
  • For 300D owners--relax. Your pictures are within a smidgen of the image quality a 20D will give you. Unless you need the extra fast operation and 5 fps motor drive, just keep that 300D banging away, collecting excellent images. You owe it to your sense of Right and Truth to check out the 350D.
  • If you are a 10D owner--relax. The upgrade cost to a 20D will set you back around $800 (including recycling your 10D on eBay if you are lucky) and the main reward is a faster operating camera with fewer delays in startup, review and write speed. Some 10D photographers I know are thinking of getting the 20D as their #1 cam shuffling their older 10D as their #2 body, and that's a strategy that makes complete sense. Perhaps you should check into the 350D. Not as fast in some ways, but way faster in others.
  • If you are a Nikon D70 owner--relax. Your camera has a slight edge on the 20D in ergonomic and operational aspects that you don't need to get bent out of shape over. Your D70+Kit Lens used is worth maybe $700 today (used or trade-in) and the 20D with 17-85 IS lens will cost you +$1,300 over that. Some future camera may fit your needs better. No sense in recycling all your optics.

About the Test

As noted earlier, there has been a small adjustment from the 8.2 MP image: It was reduced to 3072 pixels wide, a loss of 432 pixels, about 12.3% smaller, in order to superimpose over the DR image.

Transformations of this sort can be responsible for small degrees of image degradation, even though we have used Photoshop's Bicubic Sharpen option to preserve as much detail as is possible.

When the two images are cropped side by side, the smaller DR image is easy to spot, but are the pixels of the 20D doing as well with every little dot of color? You decide.

I plan to do more testing with the two shooting through the same lenses. I'll report back here when I find out new things.

My theory is that many people buy the 20D not for its speed and continuous frame capabilities--they buy it for that big, honking 8.2 MP image that will take them to places no mere 6.3 MP image could ever handle. The big winner here is the marketing department. In my macro lens at an optimum f-stop of f/9, they are virtually equivalent at best. Advantage: DR. But only due to its price.

Of course, by not doing this test with your own hands, aren't you taking some third party's word for it again?

-iNova (January, 2005)


Addendum:

Once again (April, 2005) we have tested the image differences between the 20D and 300D cameras. This time we were comparing the per-pixel resolution of the newer EOS 350D Digital Rebel XT to the 20D's chip coverage with the 300D thrown in for comparison.

Here's the test: A, B and C images are presented. This time, all were scaled to the same pixel count. Since the 20D image is largest in pixels, but only 98.9% of the physical dimensions of the 300D image, the 300D shot was scaled up to 101.1122346% (the reciprocal of 98.9% or 3543 pixels wide) in order for the actual subject detail to drop on top of the larger frame. The 350D image was simply dropped intact on top of the 20D image and since pixel sizes are identical, no subject size difference is seen at all.

Results are interesting. In the first test, we down-sampled the larger 20D frame to mate with the unchanged 300D image. The theory being that down-sampling operations at this scale would not harm the 20D image much at all, and from the later 1:1 comparison (About the Test, above) we saw a consistent slightly lower detail level in the 20D image.

A=300D (straight), B=350D (rollover) and C=20D (click/hold). All images are crops showing the final 1:1 pixels on your computer screen.

The new test was made at f/10 and several freshly auto-focused frames were made of the subject to both increase DOF and avoid any missed focus variables. Indeed, all three cameras delivered very sharp images. You could blow any of the frames up to 12 x 18 on a Canon i9900 printer and nobody could tell them apart other than from slight differences of color. Even the manual exposures (ISO 100 - 1/100sec - Daylight WB - f/10 - Parameter 1) produced results that look like nearly sequential exposures from a single camera.

But this time, the ever so slight edge goes to the 8MP cameras, and if push came to shove, the 20D has the slimmest of advantages over the 350D. It shows up not as greater resolution or finer detail but as a tiny bit tighter control of the red channel. It's really the sort of thing you can only see under a 200% blow up when flipping back and forth between images of subjects that have bright red, orange or yellow colors in high contrast contours. The 300D shows the same micro plus in its red channel as the 20D.

Any small level scaling of an image can compromise high-frequency detail (fine textures, sharp contours, small line detail) and that holds true when scaling up or down. That's what made the 20D image look microscopically lower in resolution than the 300D image. When a reciprocal enlargement is applied to a 300D image, the reverse shows up and the 300D image now looks slightly less detailed than the 20D/350D frame.

When small image scaling occurs, it's prudent to add sub-pixel unsharp masking to the image to help make up from the low-level pixel compromises. In the blown-up 300D image, a small (30% of one pixel at 100% influence) Unsharp Mask tweak was added, not to mess with your mind, but to refine the image the way it would be in any practical use. I typically add this degree of small refinement to images that are manipulated in the ±10% scale range. Here, the technique brings the 300D image up to 20D/350D level in natural images like this. If you performed the same adjustment to an image of a test chart, it would not produce the higher definition of the larger chips.

Where our first test involved slinging pixels around, the second comparison sees them at actual size. Here are all three side by side (300D, 350D, 20D) without any scaling or Unsharp Masking involved:

So what do we know now that we didn't know before? Probably these:

1. Canon makes great pixels and you can't go wrong with any of these three cameras.

2. They had nailed down virtually all the top attributes of image quality before the 300D was introduced and haven't skimped on per-pixel detail even though you get more of them for your camera dollar in the current models.

3. Your own verification of ideas like this will still mean more to you than taking my--or anybody else's--word for it.

-iNova (April 2005)


Get the eBooks

For your Canon Digital Rebel- DSLR: Canon Digital Rebel EOS 300D eBook is already on the shelf. DSLR: Canon Digital Rebel XT is next. Then DSLR: Canon 20D follows. These are great cameras with a great deal inside them, so we put a great deal of information into their eBooks. You'll find things here that nobody else ever told you about and not even Canon knew or could discuss.

The existing iNova eBooks, DSLR: Nikon D70 has become the most popular premium eBook of all time. Mastering Nikon Compact Digital Cameras and The SONY Advanced Cyber-shot eBook are also available. Click on the image for the appropriate order page.

 
--Peter iNova (peter.inova@mac.com)

itscanon is © 2005, Peter iNova. All rights reserved. Images by the author. They're © 2006, too.
Note: This is a private series of page not supported by or associated with Canon Inc.