[pefoto01]

Портрета на Асахифлекс е много хубав,но няма достатъчно рязкост/острота/.

[paperbag846]

Is this sharp enough for you???

https://www.pentaxforums.com/forums/1276033-post46.html

[Class A]

The resulting JPEG was resized to match the K200D's resolution at the longest side: 3872 pixels.

What resizing method did you use?

Typically, it is necessary to sharpen after resizing, the amount will depend on the resizing method (bilinear, bicubic, sinc, etc.) used.

[Class A]

I tried really hard not to say that, or make that comparison, but since you chose to..............................
carry on.

What is the intended contribution of such posts?

Seems like you are having fun at the expense of another person. If a moderator, oh, wait...

---

Last edited by Class A; 11-25-2010 at 05:54 PM. Reason: typo

[rawr]

While I do like my K200D, and know it can produce sharp shots, I do know it has limitations.

Just for fun, I put the following chart together, based on the resolution charts from the camera reviews in photoreview.com.au, which have been done in a more or less consistent fashion for several years.

Although there are lots of methodological issues etc that complicate making comparisons like this, I think the chart shows several things, including that although resolution does (generally) increase as sensor megapixels increase, there isn't really as big a spread as you would think between the resolution one is likely to see in cameras at 10MP and 24 MP at base ISO, even factoring in full-frame vs crop.

However once ISO climbs you see a big divergence in resolution, with the full frame (and the Kx!) sensors holding onto resolution quite well as ISO increases while most APS-C sensors lose resolution faster. And the poor K200D certainly starts well for a 10MP camera but suffers a lot of resolution loss as ISO increases - something I can attest to.



I expect the K-5 results, when photoreview.com.au looks at it, will sit between the K-x and 5DII lines.

[falconeye]

Sorry, the thread is spinning too fast for me and I had to skip its 2nd half now.

I'm sorry you feel that way, especially in light of you saying in an earlier post that his measurements are "meaningless".
A couple of things, one being that many times I've found his results to be completely flawed in relation to what I know about the equipment in use, and that if he is putting his results out there as an expert witness, his test methods need to be transparent, flawless and above reproach, all of which you yourself say they are not.

Wheatfield, that's a more appropriate tone, thanks.

I commented about Klaus short post in his forum. I agree things could be more transparent on his site. For me, they actually are more transparent because he told me a bit about. I'd say as long as you stick to the photozone dogma to not compare cross platform everything is fine.

A forum post may not make transparent anything though and that's ok.

Falk, Klaus writes "The RAW quality is Okayish - 25% blur - which is significant.", while you write "the measured value is a very good balance between sharpening artefacts and detail." and "26% 'blur' is the mathematically perfect value.".

It, therefore, strikes me that Klaus' comment is not only misleading but simply inappropriate.

We talk about a forum post by Klaus. Not a photozone article or lens test. Come on. He found that the AA filter may be stronger than on his K10D where it actually is very weak. He just likes it thin for high SFR edge figures.

It was ogl who decided to copy his post out of context here and create a thread with it, giving it false exposure. ogl is to be blamed, not Klaus.

My conspiracy theory is that the AA filter strength were increased intentionally to hide shutter induced blur

Oleg_V

Another poster ogl choose to copypaste ...

Nobody hides shutter induced blur with me. And Pentax knows me, they wouldn't even try

Also, I don't think the K-5 has a stronger AA filter than K-7. Stronger than K10D or D7000, yes. But not stronger than K-7. I don't know it yet though because measuring resolution with 0.1 pixel accuracy is not for the faint hearted ... (if you have good Bavarian beer, Klaus could tell you many stories about )

[Class A]

I commented about Klaus short post in his forum.

Where? I checked here and elsewhere but couldn't find it.

We talk about a forum post by Klaus. Not a photozone article or lens test. Come on.

I realise that it wasn't a formal test result he officially published but the fact that this thread and similar threads over at DPR exist are testimony to the fact that it is not possible to make "private"/"informal" comments of this calibre and expect them to not do the rounds. Klaus should know how well-known he is and that everything he says -- no matter where and no matter how informally -- will be discussed.

I don't believe in the notion that JCPentax can appear as a private person at PF, that Ned Bunnell's blog is his own private affair where he has no responsibility to keep it up to date, or that Klaus can make pretty harsh statements about the K-5 and expect this to be interpreted as a loose comment of a random forum member. All these public, famous figures -- as sadly as it may appear -- AFAIC, cannot shake off their roles and publicly make comments that are not meant to be associated to their known roles.

He found that the AA filter may be stronger than on his K10D where it actually is very weak.

I don't think anyone disputes "stronger". I think people are taking issue with his comment about "RAW IQ" in general. Also, (in subsequent "clarification" posts at DPR) he is saying that the K-5 only has 6% resolution advantage over the K10D (EDIT: To be precise, he didn't mention the term "resolution" but only provided his LW/PH figures for the K10D & K-5 respectively and used the expression "pixel-to-pixel sharpness"). I'm not buying this.

It was ogl who decided to copy his post out of context here and create a thread with it, giving it false exposure. ogl is to be blamed, not Klaus.

I don't think this is an "out of context" problem at all because Klaus continues (over at DPR) to defend the notion that without removal of the AA filter, the K-5 would not represent a useful improvement over the K10D. Are you in agreement with him on this matter?

---

Last edited by Class A; 11-25-2010 at 10:34 PM.

[falconeye]

This is an MTF measurement from a shot with the FA 31 at f/4.5. Focussed with phase AF (~2 m distance) and on a tripod. Unselected shot (the very first one with the lens actually).

Testing parameters defined by my White Paper, using the "Sharp Setting" which is RAW from LR 2.x with all settings default except sharpening which is 100% r=0.5px.

The result is MTF30 exceeding the Nyquist limit and MTF50 of 2580 LW/PH. Note that the test chart original exported from Photoshop in K-5's resolution w/o applying extra sharpening to the output would only score at 2460 LW/PH!.

The MTF has almost no oversharpening maximum (>1) and scores like a perfect image. A weaker as well as a stronger AA filter wouldn't be good. There already is mild color moiré near the Nyquist frequency. The gray moiré patterns in the zone plate is from the printer, not the camera. The print resolves up to the outer circle but not without moiré due to a varying line width. You see the wrong maze pattern at the Nyquist limit which unlike color moiré, can be avoided by a better demosaicer if it exists.

The bottom two images show the effect of output sharpening as is commonly applied prior to SFR edge measurements when people expect to see figures close to their camera resolution.

Now, all of a sudden, the K-5 has an MTF50 resolution of 3525 LW/PH which kindly exceeds the number of pixels in the sensor

Attachments:
- 100% crop
- MTF
- 100% crop (resharpened with r=1)
- MTF (resharpened with r=1)

[nanok]

rawr: your comment might be a bit rushed. keep in mind resolution/sharpness is typically measured linearly, while megapixels is effectively a two-dimensional measure (a measure of virtual surface, if you will). to make a long story short, taking that into account (and not taking the "ff factor" (?) into account), if you take k-7 as reference, d3x would work out to

(sqrt(24.5)/sqrt(14.6))*2500
3238.52029461452302107500

instead of the 2900 it starts at on the graph. keep in mind this is a very dumb/blind way to calculate, but my point is the gap is not as huge as you might think (24mp is not that much more than 14 in linear terms..), even if you look at megapixels as a pure source of resolution (physics of light discarded )

[falconeye]

I don't think this is an "out of context" problem at all because Klaus continues (over at DPR) to defend the notion that without removal of the AA filter, the K-5 would not represent a useful improvement over the K10D. Are you in agreement with him on this matter?

Ok, I didn't see Klaus carry the argument forward. His bad. I don't have time to involve myself into this but may give him a ring

No, I don't agree. MTF50 isn't measuring resolution. It's a good lens measure but it measures contrast as much as resolution. Some lens tests use MTF20 to measure resolution. I recommend MTF30 together with 0.5px USM sharpening for Bayer sensors in raw development (no sharpening for Foveon w/o AA filter) and allow for >100% Nyquist results in case of an excellent lens. As for K-5/FA31 in my post above.

[jpzk]

This is an MTF measurement from a shot with the FA 31 at f/4.5. Focussed with phase AF (~2 m distance) and on a tripod. Unselected shot (the very first one with the lens actually).

Testing parameters defined by my White Paper, using the "Sharp Setting" which is RAW from LR 2.x with all settings default except sharpening which is 100% r=0.5px.

The result is MTF30 exceeding the Nyquist limit and MTF50 of 2580 LW/PH. Note that the test chart original exported from Photoshop in K-5's resolution w/o applying extra sharpening to the output would only score at 2460 LW/PH!.

The MTF has almost no oversharpening maximum (>1) and scores like a perfect image. A weaker as well as a stronger AA filter wouldn't be good. There already is mild color moiré near the Nyquist frequency. The gray moiré patterns in the zone plate is from the printer, not the camera. The print resolves up to the outer circle but not without moiré due to a varying line width. You see the wrong maze pattern at the Nyquist limit which unlike color moiré, can be avoided by a better demosaicer if it exists.

The bottom two images show the effect of output sharpening as is commonly applied prior to SFR edge measurements when people expect to see figures close to their camera resolution.

Now, all of a sudden, the K-5 has an MTF50 resolution of 3525 LW/PH which kindly exceeds the number of pixels in the sensor

Attachments:
- 100% crop
- MTF
- 100% crop (resharpened with r=1)
- MTF (resharpened with r=1)

I'm dumb ... I'm not.
I'm smart ... I'm not!

Falk ... please!

Can you summarize in one plain English language sentence what this all mean?

JP

[jpzk]

Is this sharp enough for you???

https://www.pentaxforums.com/forums/1276033-post46.html

Yep!

Where did you get it sharpened?

JP

[falconeye]

Can you summarize in one plain English language sentence what this all mean?

JP, my entire post was meant to illustrate that "MTF" shouldn't be used in any single plain English language sentence. Not in mine and not in the OP. Nor in any "review" hiding "MTF" behind "resolution".

And that the K-5 renders tack-sharp images.

[ogl]

OK then, here's one of my shots with the K-5 + FA77 combo. It was shot at f/2.5. The RAW was developed in ACR with the sharpening & noise reduction sliders set to 0. The resulting JPEG was resized to match the K200D's resolution at the longest side: 3872 pixels.

Have fun pixel peeping!

http://www.camerac.nl/foto/kaakslag-big.jpg

Thank you. Such photo are no good for me at all.

[ogl]

Is this sharp enough for you???

https://www.pentaxforums.com/forums/1276033-post46.html

No. it's not sharp.