[kyteflyer]

Seems the K20 is the winner across the board, but I wonder if the differences in color aren't due to custom image settings. In any case, the K20's IQ appears better.

I was quite surprised at this result. Given that the K-x theoretically has a much better sensor etc, I expected that it would be at least as sharp and certainly not as noisy as some of those shots. That said, since buying mine, I have found that I have been leaving it at home in favour of using my K200D instead, because I just am not quite happy with the results from the K-x. Also, I miss the top LCD but thats a whole nother issue.

Thanks to rparmar for the test. I'm going to buy Noise Ninja. Magic.

I'd like to see a similar test between the K20D and the K-7, too. Anyone up for that?

[GeneV]

No it doesn't have focus adjust.
I will look into that, but have no knowledge of it.


I do not regularly shoot with flash, so am not as interested in tests that use flash. However, if I get the time I might try this.

I'll bet you don't regularly shoot at 1, 2 or 4 seconds, either. I suggested flash only because its brief duration acts as a fast shutter speed to minimize camera shake or mirror slap issues. It also gives you more aperture options which might eliminate any focus issues.

Debug for the K-x has some focus adjust ability, but it applies to all lenses. If you were to do this test with my DA40 on my K20d and my K-x both right out of the box, you would get the opposite result as far as sharpness. My K-x worked perfectly with the DA40 without adjustment, while the K20d did not. I think I dialed in something like +4 to get the K20d on target.

[GeneV]

I was quite surprised at this result. Given that the K-x theoretically has a much better sensor etc, I expected that it would be at least as sharp and certainly not as noisy as some of those shots. That said, since buying mine, I have found that I have been leaving it at home in favour of using my K200D instead, because I just am not quite happy with the results from the K-x. Also, I miss the top LCD but thats a whole nother issue.

Thanks to rparmar for the test. I'm going to buy Noise Ninja. Magic.

I'd like to see a similar test between the K20D and the K-7, too. Anyone up for that?

Before you accept Noise Ninja (NN) as the answer to all ISO issues, pay close attention to what Alexfoto posted. I own and regularly use NN, and it is an excellent product. But Alex is right about how the books, with large blocks of solid colors, are the best possible candidates for noise reduction. When you use it on more complex subjects--with hair, foliage, fabric texture--NN starts turning them to smudged or waxy replicas of the original. It is much better than the NR in CS3, but far from the spectacular performance seen on the books in Robin's test.

[rparmar]

I'll bet you don't regularly shoot at 1, 2 or 4 seconds, either. I suggested flash only because its brief duration acts as a fast shutter speed to minimize camera shake or mirror slap issues. It also gives you more aperture options which might eliminate any focus issues.

Actually if you look at my most recent waterfall series it might prove you wrong.

But I agree with your point. Not sure if I'll have time to try it out.

Before you accept Noise Ninja (NN) as the answer to all ISO issues, pay close attention to what Alexfoto posted. I own and regularly use NN, and it is an excellent product. But Alex is right about how the books, with large blocks of solid colors, are the best possible candidates for noise reduction.

I also commented that I do not use NN that often and that certain images work best with it. Well worth having in the toolbox. This thread was not supposed to be about noise reduction at all. This is by no means a "test" of Noise Ninja!

[Marc Sabatella]

Really odd what's going on with focus there, and that does throw off one's perception of the results - it's possible the K-x would have looked better in the noise department had there beem more detail to work with.

But FWIW, as for the idea of not worrying about higher ISO but instead working to get ISO 100 to 400 cleaner, I'd say that's just two different ways of looking at the same problem. The way you get cleaner high ISO is by getting cleaner base ISO. High ISO, after all, is implemented by simply amplifying the signal from base ISO. While perhaps the amplifier introduces noise of its own, I don't think that's where the bang for the buck is in terms of reducing the overall noise in the system - it's in reducing the noise at base ISO.

[er1kksen]

Hoo boy. Focus errors make it impossible to do a meaningful comparison of noise vs. detail here. A little perplexing from my standpoint, as my K-x has more consistent accuracy than my K20D did.

The biggest difference I've found between the K-x and K20D in IQ is the shadow noise. A well-exposed shot from any camera at high ISO is going to look alright. It's the shadows where things get tricky. I find that high-ISO shadows in K20D files tend to "disintegrate" if you try to manipulate them at all. The K-x files are much more pliable.

[rparmar]

Taking Gene's good advice I set up the K-x with trigger as shown in the next photo, determining focus manually as best I could. I used the FA43 again, but this time at f/11. I should note that the K-x viewfinder is noticeably less useful for MF than the K20D I am so used to. It's harder to see detail and it seems that the hexagon indicator has greater latitude. I could get it to light with an image that was way out of focus.

K-x with flash trigger


(Shot with the FA 77mm Limited.)


Next I set up my flash on a little tripod, using a convenient coffee table. I set it to Manual mode and made sure it was full power, then bounced the flash almost straight upwards for a room fill.

Metz-58 flash with trigger


(Shot with the FA 77mm Limited.)

I set the K20D on ISO100 and magically the exposure was perfect the first time. My procedure was then to increase sensitivity by one stop as I manually decreased the flash power by the same. ISO200 paired with 1/2 power, ISO400 with 1/4 power and so on. Exposures looked pretty darned consistent for this series.

Then I did the same for the K-x. This time, for some unknown reason, I needed to change the positioning of the flash to get ISO 100 exposed properly. I estimate it was 1 EV under. I then took a similar sequence of photos.

I'll post the results when I have them.

[rparmar]

Really odd what's going on with focus there, and that does throw off one's perception of the results - it's possible the K-x would have looked better in the noise department had there beem more detail to work with.

Hoo boy. Focus errors make it impossible to do a meaningful comparison of noise vs. detail here.

No doubt you are both correct. I apologise for my complete lack of chops when it comes to these sorts of tests. But maybe I am learning and the process will help others.

[rparmar]

OK, working at warp speed I have processed the images, managing to do it without actually comparing them myself. Only once they are posted here will I have a look. Oh, the suspense!

The crop area is different from before so as to emphasise shadow noise.

K20D ISO 100




K-x ISO 100




K20D ISO 200




K-x ISO 200




K20D ISO 400




K-x ISO 400




K20D ISO 800




K-x ISO 800




K20D ISO 1600




K-x ISO 1600




K20D ISO 3200




K-x ISO 3200




K-x ISO 6400




K-x ISO 12800

[rparmar]

This time the focus seems to be very close between the two cameras, for all intents and purposes eliminating that as a variable -- I hope! But the comparison is again difficult to make, for several reasons. Changing the angle of the flash for the K-x images has illuminated further into the shadow area, even though this was not a big positional change. The K20D images no longer have an annoying cast, but still the colour reproduction between the two cameras is very different. For what it's worth, with the books in front of me I can say that the darker richer tones of the K20D are far more realistic.

It appears to me that the K-x has less chroma noise, though just as much noise overall. Opinions?

[yeatzee]

to my eyes the K-x looks like it uses a lot of NR

[er1kksen]

The K20D images no longer have an annoying cast, but still the colour reproduction between the two cameras is very different. For what it's worth, with the books in front of me I can say that the darker richer tones of the K20D are far more realistic.

It appears to me that the K-x has less chroma noise, though just as much noise overall. Opinions?

You say richer, I say blocked up. You can always raise the black level of the K-x image, but I know from experience that you're not going to get anything pretty out of those K20D shadows if you want to go the other way.

As I've said, in well-exposed areas the two have similar noise, but in the shadows the noise "floor" of the K-x is at least a stop (I'd say more) lower.

[GeneV]

Much better comparison on the focus, though it is interesting that neither is as razor sharp as the K20d was on the first test at ISO 100. I totally understand what you are talking about with MF. The K10d/K20d are much preferred.

This test seems to show what many others have shown. The noise on the K20d forms what we would have called in film a larger grain. The K-x noise is finer and interferes less with sharpness, especially at ISO 3200. If these images were sharpened a bit, they might show this even more.

Still, the difference in this test does not seem as profound as in my actual use. I find ISO 3200 far more usable in the K-x than the K-20.

[rparmar]

You say richer, I say blocked up.

If it's "blocked up" then the books in front of me have their colour all "blocked up" as well.

Perhaps one can adjust the K-x image to get the perfect look. But once we start adjusting there is no way to baseline anything. It's clear that out of the box the K20D images are far more true to life.

I was not hoping for this result, but there you go.

[rparmar]

Out of interest I processed images at ISO 3200 through my normal workflow, except leaving aside the final sharpening step. I included Noise Ninja since I would no doubt use it in this case. I have upload full-size images for you to check out. Click through these reduced-size versions to get to them.

K20D ISO 3200 processed




K-x ISO 3200 processed



With my corrections in place the colours look a lot closer. If anything the K-x colours now look a little hyped. The K20D has a nasty green tinge in the background for some inexplicable reason.

But of course further adjustments could be made. There is little to choose between the two for real world applications. The fact that the entry-level K-x achieves the feat of matching the previous top-of-the-line body, plus gives two more stops if one needs them... well, it is quite remarkable.