[Pentaxor]

That's because the dumbs at DPR set D90 to NR = High (how could it be a fair comparaison with K20d btw ?) which effectively remove details. But with NR= off it's better (page 21 show differences between NR reduction)

High iso is not my main complain about K-7 though, in fact it's more the low iso noise and overall sharpness/micro constrast that annoy me, I've the same feeling about the canon 50D for ex. I believe there is too much pixels, if only they could have made a 10 mp sensor instead...

I think it's not a good comparison either for Pentax. it wasn't set-up for NR in ISO testing. if they are going to make a side to side comparison, atleast they should had taken sample shots with both cameras with NR's on/off.

right now, there is no real comparison as we are left with different results and different function used. it's all an assumption. but it is clear that Pentax have some noise on higher res. for detail.

[claude21]

claude21 could be right and D90 does have a better sensor. I am really curious to know.

But from samples I can find in public, it looks we are comparing noise reduction algorithms, not sensors. And if Nikon puts them into their raw files, dxo will run into the same trap. Is there any serious source studying the underlying qualities of both sensors, like full well capacities etc.?

Here is another comparison, from imaging-resource. Left is the K20D, right the D90, both 100%, both ISO 1600.

Besides that the D90 has somewhat fewer pixels (K20D would gain in downsizing), it is very visible that the D90 sacrificed detail for smoothness. I can see many hairs which are plain missing in the Nikon shot ... Altogether, the K20D shot looks snappier and I assume when printed, it would look better (somebody should actually do the printing test )

On the same site, you can download raw files from D90 and K20D. Theses samples at iso 1600 show the opposite : better sharpness from the D90. What I mean is that so many factors are influencing what we see on review sites. Lens (and sample variation between same lens), focusing, camera settings, lighting etc

So it's a bit misleading. The best way to compare 2 bodies should be raw format, no NR, the exact same lens (so an M42 would fit on both bodies), enough DOF to avoid misfocusing, controled lighting etc

[falconeye]

Interesting discussion.

Just because it actually is about my various blog entries, I am a bit afraid to bring it OT too much. Ok, just this bit...

That's because the dumbs at DPR set D90 to NR = High (how could it be a fair comparaison with K20d btw ?) which effectively remove details. But with NR= off it's better (page 21 show differences between NR reduction)

High iso is not my main complain about K-7 though, in fact it's more the low iso noise and overall sharpness/micro constrast that annoy me, I've the same feeling about the canon 50D for ex. I believe there is too much pixels, if only they could have made a 10 mp sensor instead...

I looked at NR=off samples now. Then, the noise looks about the same (at 800, at 1600 a bit better for D90) but still with a bit more details for K20D. Both sensors now just look more similiar.

Just right now, I've finished my micro-contrast analysis for K-7. Stay tuned

However, I don't understand why you care about #pixels. You can always downsample and sharpen to normally get better results than from the fewer pixel sensor.

better sharpness from the D90. What I mean is that so many factors are influencing what we see on review sites.

Yes. But also, the scale on the background is out of focus for the Pentax and surprisingly, the noise looks not very different (a bit better for D90, most visible in the red channel).

I agree, this all quite subjective and claims are hard to verify.

This is what our entire discussion is all about: Is the claim that D90 has a sensor superior to K20D true? How do we know? And if so, how much in x0.1EV, really? At the end of the day, I need to know three values: read-out noise, leakage and full well capacity. Just, these figures are hard to get. Now, that I looked into it, I feel that K-7 will be able to compete better than anticipated with D90.

[Steelski]

Hello Falk man.
You shared with me some early Raw files a while back. whilst they were at low, I was able to discern that the K-7 did not contain as many artifacts at 3200 and 6400 ISO..
What I mean was that the K-7 did not have the magenta cast around the corners when the images were pushed, and there were not any horizontal patterns.
I know you are doing a very thorough analysis. But please could you confirm that my biggest gripes with the K20D are fixed.

[GLXLR]

Honestly, the results are very disappointing. The time Nikon had from the D300 to the D90 sensor, there were already large improvements as the D300 was barely useable at 3200 while the D90 was an absolutely great performer. Pentax however had even MORE time from the K20D to the K-7 and from your results, there was barely any improvemnt from the K20D's sensor to the K-7 sensor. Does the K-7 even have a different sensor from the K20D? I guess from the results, that is questionable now.

It isn't a deal breaker for me (as I will still get the K-7), it's just that I am disappointed that even with so much time, little was improved in ISO performance (I meant that instead of IQ)

Thanks Falk for your great testing. The Pentax community here is very proud of what you are doing!

[yakiniku]

Fortunately, dpr has a page showing noise from D90 and K20D side by side (because it is JPG, the K-7 could score better here if it has an improved NR; something I didn't look at at all!). Anyway, it is here:
Nikon D90 Review: 20. Photographic tests (Noise): Digital Photography Review

What jumps out is the coin at ISO 100: It is known to have background texture but only K20D shows it (and D80 a bit). Esp. the D90 destroys this texture for ALL values of iso. Not something that I would call a progress...

As I wrote in my blog: details rules over noise. And I am glad that K-7 accepts this heritage.

Don't get me wrong ... the D90 may have a better sensor than K-7. I won't defend Pentax because it is Pentax. It is just that I need more than the simple claim that this be fact.

I totally agree with this!! And I think that's very important and is miss out by so many users and reviewers. Most people just enlarge to 100% and say "look there is so much noise!" and totally ignore the detail.

[Class A]

Fortunately, dpr has a page showing noise from D90 and K20D side by side

This shows the same trend as the Nikon D50 vs Pentax K100D. The Pentax shows noise and detail and the Nikon just smudgy smear. One of the reasons why I chose the K100D over the D40.

That could be down to different settings, but I have a feeling that this is not down to chance. If you look at the grey blocks, the Canikons always look like clear winners. To DPR's credit one has to acknowledge that they respect the Pentaxes for retaining greater detail.

High iso is not my main complain about K-7 though, in fact it's more the low iso noise and overall sharpness/micro constrast that annoy me

I'm still not a 100% sure whether I should be concerned about this or not. I've seen great shots from the K20D so I'm optimistic. Just a bit concerned about the many soft shots I keep seeing from the K-7.

On the same site, you can download raw files from D90 and K20D.

The shots differ slightly in lighting and field of view but the main issue is that the right shot (presumably K20D) looks to be slightly out of focus.

[claude21]

Yes. But also, the scale on the background is out of focus for the Pentax and surprisingly, the noise looks not very different (a bit better for D90, most visible in the red channel).

Not only the scale but also the entire frame is way less sharp, foreground included. Considering the shoot is f/8, I doubt it is really out of focus. My guess is the lens is at fault or another problem with severe back/front focusing.

[Yohan Pamudji]

Why don't any of the review sites do resolution tests at every ISO? Wouldn't this be more representative of noise vs. detail than just noise numbers? Noise numbers by themselves aren't even good enough to describe how noise looks.

[raz]

Why don't any of the review sites do resolution tests at every ISO? Wouldn't this be more representative of noise vs. detail than just noise numbers? Noise numbers by themselves aren't even good enough to describe how noise looks.

correct, the problem with high iso is the loss of detail/resolution, not exactly the noise itself.

[Rondec]

The problem really is that people have gotten used to the smoothness of digital, to the point that they want that, even if it means wiping away detail. Maybe Pentax can change that, but I doubt it.

When I shot film, I would see graininess at ISO 400 in the right situation and ISO 800 was as high as I ever went. To me, even the 3200 looks equivalent to some of the grain I got on film. Not saying I wouldn't like to have it better, but it just shows how times change.

[cbaytan]

Yohan-

What do you shoot at 3200? I never go above 400. I'm not being sarcastic, I just never understand why people need these high ISO numbers. I'm curious.
Thanks!

What kinda question is that? It shows you have never been into Coal Mine Photography.

Bests.
Can.

[Pentaxor]

Yohan-

Two questions.

What do you shoot at 3200? I never go above 400. I'm not being sarcastic, I just never understand why people need these high ISO numbers. I'm curious.

Where are you in Mississippi? I was born there; Cleveland. My grandmother still lives in Jackson. I was just over there last week visiting.

Thanks!

High ISO is needed when you shoot in very low light conditions where max lens openings is not enough and slow shutter speed is not an alternative in shooting moving objects in the dark nor flash should be used in that particular scene. it is needed when you need to shoot something in a hurry (in most occasions). in other words, it is one method of exposure control or exposure assistance.

you will learn to use of this more often once you get acquainted with night photography.

other uses of high ISO is for artistic value.

[melander]

Cloud21: Remember that D90 got the same flaw like every other brand. Iso 1600 is no way near iso 1600. D90 Iso 1600 is in fact iso 1125 while pentax Iso 1600 is 1383. So crank up the iso on the D90 to iso 2000 and compare there instead.
Same goes at iso 3200. D90 = 2300 K20D = 2700 D300 = 2100!!

Start comparing real iso instead of in camera Iso numbers. D300 needs somewhere around iso 4500 to compete with K20d iso preformance, who wins then?

[Pentaxor]

Cloud21: Remember that D90 got the same flaw like every other brand. Iso 1600 is no way near iso 1600. D90 Iso 1600 is in fact iso 1125 while pentax Iso 1600 is 1383. So crank up the iso on the D90 to iso 2000 and compare there instead.
Same goes at iso 3200. D90 = 2300 K20D = 2700 D300 = 2100!!

Start comparing real iso instead of in camera Iso numbers. D300 needs somewhere around iso 4500 to compete with K20d iso preformance, who wins then?

hmmmm...... does that imply that Nikon is somewhat cheating with regards with real ISO values? that may also explain why there is less noise in the results coupled. the question would be, how effective is their NR feature with regards to the actual ISO value. that would be interesting.