[rparmar]

OK, the first thing I have to say is that I am not a tester; I am a photographer. My typical way of testing a camera or lens is to use it for a while and see how I feel about the pictures. That said, I do see the value in more formal testing, and sometimes even dip a toe into that area myself. But please do not expect this to be as comprehensive and exacting as a lab might do.

However, I am curious as heck about my new K-x. This was generously provided by Pentax and Pentax Forums as a prize in the competition held earlier this year. Thanks to Adam for this!

Besides being compact and well-specified, the most interesting thing about the K-x is its high ISO noise handling, reputed to be better than any previous Pentax camera, including the top-of-the-line K-7. After using it a week I wanted to explore for myself, a bit more formally, these characteristics. Since I am intimately familiar with the K20D, the comparison was made with that camera. (I have used the K-7 only slightly.)

I shoot the K20D at ISO 200 and 400 most of the time, going up to 800 in a pinch. I generally do not bother with noise reduction, since I find the noise characteristics in that domain to be pleasing. I do have a licensed copy of Noise Ninja for higher ISO situations.

To test I used the FA43 at f/4, a wide enough aperture, but one at which the lens excels. I used a tripod with 2s shutter release (hence SR off) and chose a section of my bookshelf as a suitable target. I did this at night with one overhead light bulb -- a typical dark room scenario. The distance was about 1 meter.

In Av mode I started at the lowest sensitivity, 100, and changed by one stop (eg doubled) for each shot. I found that both cameras underexposed in this situation, and so dialled in a constant +1EV compensation.

I let the camera auto-focus, which was maybe not the best decision in retrospect, though a typical one for most shooters. Strangely, the K-x refocused between each and every shot, while the K20D did not. I welcome an explanation of that phenomenon.

I used PEF files in sRGB colour space, though I have since changed my K-x to DEF for convenience. I retained auto white balance which produced some issues, as we will see. I am not one to fuss with grey cards since I have no problem correcting in post. (If I shot studio and strobes I would have a different practice.) The K-x Expanded Sensitivity custom setting (#3) was set On, so as to get a range of ISO 100 to 12800. However the Shadow Correction and Highlight Correction were left Off. The K20D sensitivity range was left unexpanded at ISO 100 to 3200.

This resulted in 8 shots from the K-x, at ISO 100, 200, 400, 800, 1600, 3200, 6400 and 12800. The recorded exposure times (in seconds) were 4, 2, 1, 1/2, 1/4, 1/8, 1/15 and 1/30. That's as close as can be to a perfect ratio of stops. The 6 shots from the K20D at ISO 100, 200, 400, 800, 1600, 3200 resulted in exposures of 4, 2, 1, 1/2, 1/4, 1/8. It's nice to see both cameras meter the same.

I processed through Adobe Camera Raw with all settings on default, no sharpening and the same brightness settings. Shots were layered and cropped in Photoshop, again with no processing. I could have obtained much "better" results following my normal procedures, but that is not the point here. Finally I cropped to 600 pixels square and produced JPGs (quality 10). Note that because of the difference in sensor size, these "100% crops" show a different amount of the subject.

The photos have been uploaded to Flickr and linked here. Since each has its own Flickr page, it makes it easier for you to compare one to another.

OK, here are the images.


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




Finally, just for kicks, I ran the highest ISO of each camera through Noise Ninja with my default settings (Strength, Smoothness and Contrast at 10, USM at 60/1.2, Turbo checked) just to see what would happen.


K20D ISO 3200 Noise Ninja




K-x ISO 12800 Noise Ninja



Conclusions in the next post.

---

Last edited by rparmar; 08-24-2010 at 10:26 AM.

[bymy141]

Nicely done.

The first thing that strikes me is that most of the K-x photo's are out of focus.
Also, to my taste you should have choosen a somewhat darker image to magnify the noise handling of the camera's.
Like for instance in this K-x / K-7 comparison: Google ½Ķ

- Bert

[yeatzee]

The Kx seems crazy soft....

and it would help if you put ISO 100 of the K20d next to ISO 100 of the Kx etc. so we can compare them without having to scroll so much up and down

[rparmar]

1. The K20D focused much better and more consistently. The K-x kept changing focus between shots and did a poor job overall. This is actually consistent with my daily usage so far. Next time I do a test like this, I must manually focus.

2. The K-x was significantly better at getting the white balance correct.

3. Every ISO setting is "usable" depending on your situation, tolerance for noise and state of desperation.

4. Even ISO 100 is not noise free. In future cameras I would rather see a cleaner ISO 100 through 400 than extending ISO further into the stratosphere. But I suspect that I am in the minority here. In fact, I am in the marketing category for the 645D in all ways except personal finance!

5. The noise characteristics of both cameras seem similar up to ISO 1600. I prefer the look of ISO 3200 on the K20D to that on the K-x. That result surprises me.

6. Above ISO 3200 I would use only in "emergencies", but at least the option is there.

It deserves emphasising that real photos viewed at realistic magnifications will be more appealing than these 100% crops.

[rparmar]

Wow, two posts before I was even finished! Thanks Bert and Yeatzee.

The first thing that strikes me is that most of the K-x photo's are out of focus.

Yep, that doesn't say good things about the camera. But next time I will manually focus. Thing is, there was so little light in the room, that would be error-prone as well.

Also, to my taste you should have choosen a somewhat darker image to magnify the noise handling of the camera's.

I deliberately chose a section with both shadow and light.

and it would help if you put ISO 100 of the K20d next to ISO 100 of the Kx etc. so we can compare them without having to scroll so much up and down

Yeah, I will re-order them. I was meaning to do that, but you guys got in comments first!

[yeatzee]

4. Even ISO 100 is not noise free. In future cameras I would rather see a cleaner ISO 100 through 400 than extending ISO further into the stratosphere. But I suspect that I am in the minority here.

+1 to you.... I have the exact same thought process. ISO 400 is way to noisy on both my K200d and K-7 for me..... I absolutely HATE having to shoot macro's with anything higher than ISO 200 becuase of the noise.

[rparmar]

OK, photos re-ordered. After I cook dinner I will think about re-shooting a comparison using manual focus. Before that I will check back here for any other suggestions.

[bdery]

Strangely, the K-x refocused between each and every shot, while the K20D did not. I welcome an explanation of that phenomenon.

That's wierd. When I (rarely) do bookshelf shots with my K20D, it does refocus. Maybe it just focused slightly and found that it was spot on? Otherwise, if you were in AF mode, something's wrong.

I'm flabbergasted (and I mean it) at the performance of the K20D with Noise Ninja at ISO 3200. Simply amazing. When I have to deal with noise I use ACDSee 10's noise removal tool, but I can see now that it's ridiculously underpowered. You just about convinced me to use Noise Ninja or some other similar soft (there was one with a free version, I forget it's name).

[rparmar]

At the time I bought it I thought Noise Ninja was the best product, but now I am not sure. Other software apps have come on in leaps and bounds. I originally got it when I was using a Fuji point and shoot, and it made the world of difference. How well it does is very dependent on what kind of noise exists. Ninja does a good job on some types, but once the noise gets into larger patterns (as seen on the K-x ISO 12800 shot) it seems to give up the ghost, producing an unattractive blotchy look. Still, the Home Bundle for $45 is a good deal. If anyone knows of a better open source product, I'm all ears.

[alexfoto]

Free product:
g'mic
gimp plugin that include even one function for denoise.

About these test i think:
- on the k-x you have some front focus issue;
- the subject used on this test haven't much detail so you can used a more stronger denoise but this false the result;
- the idea of change the time based on the iso used have no sense.
When you want shoot a static subject you simply use a tripod and shoot at the minium iso for obtain better performance.
While if you like simulate real shoot, for obtain real output, you increase the iso while the aperture and the shutter speed remain the same.
So i think make more sense a test using for example a lens at f/2.8 and 1/60 costant where you move the light until you obtain the correct exposition instead force the shutter speed.

I don't have a K-x myself, but I agree that something must be wrong. Probably focusing. Does the K-x have focus adjusting and have you checked the focusing on some grid, tape measure or something that would show the problem?

[GeneV]

Free product:
g'mic
gimp plugin that include even one function for denoise.

About these test i think:
- on the k-x you have some front focus issue;
- the subject used on this test haven't much detail so you can used a more stronger denoise but this false the result;
- the idea of change the time based on the iso used have no sense.
When you want shoot a static subject you simply use a tripod and shoot at the minium iso for obtain better performance.
While if you like simulate real shoot, for obtain real output, you increase the iso while the aperture and the shutter speed remain the same.
So i think make more sense a test using for example a lens at f/2.8 and 1/60 costant where you move the light until you obtain the correct exposition instead force the shutter speed.

I share some of those thoughts as well. From what is posted here, there is definitely a focus issue with the K-x. I've gotten better results than those ISO 200 and 400 hand-held on a living subject. It may be time to get out the debug utility.

Another thought that comes to mind is the possibility of camera shake and mirror slap. There was a good deal of noise about that when the K-x came out. I've found my tests of lenses on the K10d yielded too many results like this at low shutter speeds until I started doing all my tests with flash. The tripod does not cure all of these issues.

If I were doing this test, I would put the 540FGZ in wireless mode and place it at a distance from the subject that would properly expose at full power at ISO 100. I can then go 6 stops more sensitive with just the flash compensating for the increased sensitivity. By the time you get to the last couple of very high ISO settings on the K-x, you will be using available light, but the shutter speeds will be high enough that shake will not be a factor.

[rparmar]

Some good points here and a useful conversation developing -- thanks all.

- the subject used on this test haven't much detail so you can used a more stronger denoise but this false the result;

But I am not using strong denoise and don't wish to test that. The noise removal part was only an extra, really. If anything, areas of solid colour show up noise a lot more than areas of detail, so I think this is a more stringent subject.

When you want shoot a static subject you simply use a tripod and shoot at the minium iso for obtain better performance.

While if you like simulate real shoot, for obtain real output, you increase the iso while the aperture and the shutter speed remain the same.

But how else would you do an ISO test than hold exposure constant? Otherwise you are not really testing ISO noise. (Possible answer below, using flash.)

Does the K-x have focus adjusting and have you checked the focusing on some grid, tape measure or something that would show the problem?

No it doesn't have focus adjust.

It may be time to get out the debug utility.

I will look into that, but have no knowledge of it.

If I were doing this test, I would put the 540FGZ in wireless mode and place it at a distance from the subject and an aperture that would properly expose at full power at ISO 100. I can then go down 6 more stops with just the flash compensating for the increased sensitivity. By the time you get to the last couple of ISO settings on the K-x, you will be using available light, but the shutter speeds will be high enough that shake will not be such a factor.

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.

[Adam]

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.

[rparmar]

Seems the K20 is the winner across the board, but I wonder if the differences in color aren't due to custom image settings.

I just confirmed that both cameras are set to "Portrait" which I believe is the most neutral. I have always assumed that this does not affect RAW images -- am I wrong there?