[FHPhotographer]

So, the new K20 is here. Rather than just assuming the K20 will be superior, I want to compare the IQ to my K100 in the most objective way I can. Is the following the best way, or am I missing something:

same lens (43mm) and manual settings, full size RAW image, opened without editing (including turning off sharpening) in Photoshop; then crop at 4x6" (or so)out of middle and view them at full resolution on screen and printed... or do I need to compensate in sizing for the difference in resolution from 6 to 14.2MP? and if so, how?

What about some capture sharpening? I know what works best in USM on the K100, would the same setting work for the K20, or does it need something else?

Thanks for any/all input,
Brian

[kenyee]

I'd just print out the entire image into an 8x10 or 11x17...

[falconeye]

then crop at 4x6" (or so)out of middle and view them at full resolution on screen and printed...

This would be an unfair and unequal comparison.
As kenyee said, print'em out, both giving the best treatment you can. But you'll need 17" x 24" to really see the difference.

If you want a method to compare on screen:
Resample the K100D image to 14.6 MPixels (use the exact pixel dimensions of the K20D) using a good upsampler (bicubic at least -- most photoeditors can do this). Then compare at 100% magnification. No need to crop anything.

Notice: Some (good) upscalers produce wonderful images. So, watch out for lack of detail / texture in the upsampled images. This holds particularly true for images not in ISO 100 -- as larger pixels have less noise and upsampling doesn't create noise.

Notice 2: This method is even more important to evaluate tele converters. Because sometimes, pure upsampling outperforms them.

[FHPhotographer]

This would be an unfair and unequal comparison.
As kenyee said, print'em out, both giving the best treatment you can. But you'll need 17" x 24" to really see the difference.

If you want a method to compare on screen:
Resample the K100D image to 14.6 MPixels (use the exact pixel dimensions of the K20D) using a good upsampler (bicubic at least -- most photoeditors can do this). Then compare at 100% magnification. No need to crop anything.

That sounds counter intuitive... enlarge the K100 image by 240% and compare that to a non-interpolated K20 image? I'm going to guess the K20 will look better. Maybe I'm missing something,
Brian

[Alfisti]

Yeah i'd just downsample the K20D image to 6mp and have a look. At the same time do a few upsamples of the K100D image at larger print sizes and see what happens.

It's all about output size in the end.

[*isteve]

That sounds counter intuitive... enlarge the K100 image by 240% and compare that to a non-interpolated K20 image? I'm going to guess the K20 will look better. Maybe I'm missing something,
Brian

Precisely. This is why the K20D has better print quality because you dont HAVE to upsize so much to print at a given size. Thats what the extra 8.6MP is for.

[jsherman999]

Shooting RAW and converting with the same settings is the fairest thing for making a comparison, but for kicks you should also try JPEG/bright with both cameras and see how you like both. After all, one of the benefits of the K20D is the excellent JPEG engine, so you should take a gander at that benefit also while you're at it.

[Alfisti]

Precisely. This is why the K20D has better print quality because you dont HAVE to upsize so much to print at a given size. Thats what the extra 8.6MP is for.

Agree in principle but sometimes the extra pixels result in worse dynamic range and a noisier image .... i hav eno idea if this is the case with the 20D though.

By the time you get to 11*14 the 20D ought to cane the 100D you'd think.

[Michaelina2]

So, the new K20 is here. Rather than just assuming the K20 will be superior, I want to compare (it) to my K100 in the most objective way I can.

When I acquired a K20D, I retained my K100D for back-up purposes. Using IQ to determine which is best seems to miss the point IMHO. Both cameras can produce stunning images. Where the K20D shines is in ease of use, general convenience and it expands the range of possible capture conditions for marginal subjects.

Given that you are new to the K20D and experienced with the K100D, may I suggest that you use with good cheer the K20D for (say) a year, then ask your research question, design your testing procedure, and draw a conclusion. It's only human nature to bias toward what you know. As such, its possible to unknowingly create a test that tilts the table and thereby produces a misleading result.

Best of success...

[Mister Guy]

As such, its possible to unknowingly create a test that tilts the table and thereby produces a misleading result.

But see, that's important too. It's been rolling around in the back of my head for a bit a way to create a series of completely biased tests to compare my current k110d to the k20d to the hypothetical-hopefully-soon-to-be-release k300d. Maybe it's just how I look at things, but I'm personally trying to create a list of outlier situations, and then compare that to how I shoot to help me weigh my upgrades.

I'm less concerned with how two cameras handle identical circumstances that both cameras can handle reasonably well unless the more expensive one actually does worse. And example would be that I've found I end up taking a lot of shots pushing my ability to hand hold. An extra stop or two from stabilization is an edge case that I encounter often.

Another case I'm considering is burst refresh. I'm encountering more situations than I had anticipated where I'm waiting on the buffer in my k110d.

I'm trying to figure out the rest of those edge cases, so biased tests aren't always bad!

[Michaelina2]

...biased tests aren't always bad!

Agree... so long as the bias is clearly identified and the result is a fair representation of the subject's performance.

I look forward to learning more about how the OP plans to control the variables.

Best of success...