[glanglois]

Or an alternative formulation I heard of recently, courtesy of an impressive Australian K-x photographer I came across via flickr and whirlpool: 'It's not the wand, it's the wizard'.

Somehow I think I've heard that in a context other than photography .....

[ogl]

Klaus has another opinion about AA filter and sharpness
Photozone - Pentax K5 ... ordered -

[ManuH]

Klaus has another opinion about AA filter and sharpness
Photozone - Pentax K5 ... ordered -

Remember, we're talking about 100% view, that means very big prints or... screens. USM 0.5 may produce halos but I think there are better algorithms anyway. The difference between weak and strong AA filters will never really be visible on anything but gigantic prints. OTOH moire can be visible even at small size...

I made a picture this morning with the DA* 50-135mm @f/5.6 and K-5 at ISO 80 on a tripod. I applied LR sharpening. I'm very satisfied with the level of detail produced. You can read the street name far away on a plate that is barely visible on a 24" screen. How much more details do you need to make a picture?

[JohnBee]

Remember, we're talking about 100% view, that means very big prints or... screens.

But don't forget crops...
I think this is quite possibly one of the most overlooked factors in assessments such as these. One good example of this is with wildlife photography.
Which almost always ends-up being subject to cropping(closing in) as a result of lens limitations.

USM 0.5 may produce halos but I think there are better algorithms anyway.

I agree!
USM belongs in the dark ages.
There are very good sharpening tools now that can sharpen images without haloe's(white and black peaks) and artifacts. So I think that from the angle, the argument simply doesn't work.

The difference between weak and strong AA filters will never really be visible on anything but gigantic prints. OTOH moire can be visible even at small size...

I also think this is true.
However... in the case of the D7K, I've found ACR/LR to play a diminutive role in moire artifacts. As can be seen here:

ACR 6.3 CROP:


NX 2 CROP:


Also, I have replicated the NX2 results with RPP as well(so it's no more special or anything).
The point being, that ACR over-inflates moire by inducing color registration errors where none should exist. Which leads me to say that we may be falsely lead into conclusions where none should exist.

Another good example of this can be seen here:

Click image for full size

Which is a good example of what the D7K(with weaker AA filter) looks like when processed with a better RAW processor(RPP). And before jumping-up and pointing out the moire in this image, I'd caution that the Pentax sample(which is JPG) looks somewhat worst than this.

So for me... the argument is not a very strong one in fact.
When this topic initially started(with the first K-5 softness thread), Falk pointed out some pretty convincing moire samples(with the dials). And I was on-board with that. But then... I started experimenting with difference RAW processors and found that the results were not consistent at all, and that it would require more investigating prior to a conclusion.

Anyways, that was probably too long an answer for the actualy comment, but I thought I would share my finding nonetheless.

I made a picture this morning with the DA* 50-135mm @f/5.6 and K-5 at ISO 80 on a tripod. I applied LR sharpening. I'm very satisfied with the level of detail produced. You can read the street name far away on a plate that is barely visible on a 24" screen. How much more details do you need to make a picture?

This is quite normal in fact.

I'm reminder when the transition between point and shoot camera's were quite blurred as DSLR became mainstream. At which point we had loads of people making statements like 'I made money with...' or 'I'm perfectly satisfied with...' etc etc. And there's nothing wrong with that. However, I don't think it is appropriate to the issue though.

I don't think anyone is saying the K-5 isn't good enough, or up to the task of doing anything. Mainly because I think the entire arguement exists solely based on the fact that another camera exhibits better than, performance(no more no less). Which is nothing new really. However, in this case, the outstanding issue is that the two camera's in question, are apparently using the same sensor. And I think this perpetuates the issue further than it normally should.

And because of this, I think those who care are either wondering why this discrepancy exists. I know in the case of the Imaging resource samples. There is detail in the D7K that can never be extracted with the K-5 no matter what. This can be visible at 100% and there is no moire issue in sight anywhere in the scene. And though I am not devistated by this(personally), a part of me does sort of feel as though this shouldn't be the case.

[ogl]

But don't forget crops...



Another good example of this can be seen here:

Click image for full size

To say honest, not better than K-5.



Here are some good samples of K-5
http://fotkidepo.ru/?id=album:32163

(full size inside)

[dankoBanana]

But don't forget crops...
I think this is quite possibly one of the most overlooked factors in assessments such as these. One good example of this is with wildlife photography.
Which almost always ends-up being subject to cropping(closing in) as a result of lens limitations.

or in macro .. not everyone has a Pentax 200mm macro lens

I agree with John Bee ... I won't say the K5 is bad, Its GREAT ... but it's not perfect and personally, I'll take this strong AA filter (if that is the case and it seems it is) as a con

[JohnBee]

To say honest, not better than K-5.

Now sure what your trying to say here tbh.
Perhaps you could elaborate a little.

Also, for the sake of others.
Do you think you could link the full size image under a smaller thumbnail or link?
It makes the entire thread very slow to load.

[dankoBanana]

maybe we should start a threat about sharpening?

[ManuH]

But don't forget crops...

I don't forget them, as it is true that more MP can help. But even in our AA filters case I would say that the difference is not visible enough to be of consequence, and that the sharpening method will have more incidence on the final result.

Also, I have replicated the NX2 results with RPP as well(so it's no more special or anything). The point being, that ACR over-inflates moire by inducing color registration errors where none should exist.

This is certainly interesting but personally for good or bad I have standardized my workflow on ACR/LR. Too bad if it doesn't support the D7K very well.

There is detail in the D7K that can never be extracted with the K-5 no matter what.

Are you sure it's actual detail? Something that would make a difference between a non-readable plate and a readable one? I'm a bit skeptical here. I was looking at the picture I made this morning at 200%. I just couldn't see how substantially more details I could extract from this 16MP file. And in the D7K/K-5 comparison are you sure it's not a matter of lens? Remember, anything can rob sharpness: lens not sharp enough, diffraction, correct focus, etc.

In the end, I don't think it's a big issue. I could live with the potentially "moirable" D7K as I could with the K-x which was prone to it too. I prefer the stronger AA filter because the overall picture is better but it's not a deal breaker. Actually I'm more concerned by the purple fringing more present in the K-5 than the K-7. Again not a deal breaker but I prefer less PF of course.

[JohnBee]

Are you sure it's actual detail? Something that would make a difference between a non-readable plate and a readable one? I'm a bit skeptical here.

I think so.
But your appeal to inquiry has motivated me to share my results.
However, I'm on the middle of re-installing my main workstation(arg) and so I won't be in a position to post anything for a few more hours. Though I will post something just as soon as I have my main tools back online.

I was looking at the picture I made this morning at 200%. I just couldn't see how substantially more details I could extract from this 16MP file.

Well it does look good!
But... (and please don't take this the wrong way, as I say it with the best of intentions), I see room for improvement in a number of areas(grain, edge detail, highlights etc). But that's more of a matter of PP than initial image integrity.

And in the D7K/K-5 comparison are you sure it's not a matter of lens? Remember, anything can rob sharpness: lens not sharp enough, diffraction, correct focus, etc.

That is the final issue in the entire affair imo.
Based on the going samples, something tells me that the examples provided by Imaging Resource and DPReview are both subject to optical discrepancies. And this makes it very hard to draw any sound conclusions. Though there does seem to be some consistency with regards to the D7K having the upper hand in overall resolution at this point. But again... I would be very hard to make a definitive conclusion based on these samples alone(its just not enough imo.)

Along with this, I was setting-up to test a D7K and K-5 this evening with an SMC Super Tak 55/1.8. But the person who was arranging the event had to work over the weekend(school teacher) and so we had to reschedule.
However I am really looking forward to getting some results and hopefully shed a little more light on the infamous AA filter issue.

I prefer the stronger AA filter because the overall picture is better but it's not a deal breaker. Actually I'm more concerned by the purple fringing more present in the K-5 than the K-7. Again not a deal breaker but I prefer less PF of course.

I would agree also.
There is much to like on the K-5 and the D7K just doesn't cut it in a number of other areas to be taken very seriously in many cases. Not that I have anything against the camera(it looks and feels nice). But it seems to have a number of quirks that just don't work for me very well.

However, on the issue PF, I am quite concerned as well. I just wish that more reviewers/testers out there would take the initiative and test the camera with better glass rather than the hit and miss mediocre type of conditions we've seen to date(very frustrating). However, I am very concerned about using the K-5 with some of my beloved older glass. Which I have grown to really love and use in part of my manual shooting fetish. Which makes it that much harder for me to part with the D700 given the incredible IQ I get from it with those older lenses.

Anyways, I could write for days about the seemingly endless grief surrounding my upgrade path and I guess this just isn't the place for that. However, I often wonder if am but one of several people in a similar circumstance?

Whatever the case, I'll post the D7K/K-5 RAW detail sample shortly.

[JohnBee]

maybe we should start a threat about sharpening?

We definitely should.
Sharpening has come a long way since the days of classic USM, and by the looks of things, I think alot of people could benefit from this also.

[eccentricphotography]

Personally I prefer using Raw Developer's sharpening output R-L deconvolution. Here's their explanation of how it works...The R-L Deconvolution option uses Richardson-Lucy deconvolution and is more an image restoration/reconstruction or "refocusing" technique compared to the other 3 options which are more traditional edge contrast enhancing filters. R-L Deconvolution is only available when using Mac OS X 10.4 or later. Richardson-Lucy gained some fame in use with correcting early images from the Hubble space telescope. Almost all digital cameras use an anti-aliasing filter which has the effect of slightly blurring the image. Richardson-Lucy in theory is well suited to undoing the effects of such a filter. Currently RAW Developer only provides basic controls for the number of Iterations to perform and the specification of a Radius for a Gaussian point spread function. R-L Deconvolution will become much slower as the number of iterations are increased and tends to give a more subtle effect than some of the other sharpening options. As with other sharpening options if the adjustment values are set too high R-L Deconvolution will begin to give an overprocessed look with "halo" type artifacts around edges.

[ManuH]

Well it does look good! But... (and please don't take this the wrong way, as I say it with the best of intentions), I see room for improvement in a number of areas(grain, edge detail, highlights etc). But that's more of a matter of PP than initial image integrity.

You're right, I just used LR on this one. TBH I don't think the image is strong enough to warrant spending time for the removal of minor imperfections. I also didn't use my favorite sharpening tool on it: Nik Sharpener. This kind of tool makes me not care about the strength of the AA filter.

Anyways, I could write for days about the seemingly endless grief surrounding my upgrade path and I guess this just isn't the place for that. However, I often wonder if am but one of several people in a similar circumstance?

I'm resisting that kind of upgrade path. Of course I'd prefer to get the best tools available for my budget but I'm also trying to remember that the big picture is more important than minute details between this or that sensor. That's why the minor differences between D7K and K-5 don't interest me too much. They're both in the same class as far as IQ is concerned. For myself the choice was pretty simple, the D7K cannot use K lenses.

[Jeff Charles]

Personally I prefer using Raw Developer's sharpening output R-L deconvolution...

In LR3, the Detail slider in the Detail panel controls the transition from USM to deconvolution. At 0, it is all USM. At 100, it is all deconvolution.

Jeff

[falconeye]

Klaus has another opinion about AA filter and sharpness
Photozone - Pentax K5 ... ordered -

Thanks for the quote. In that quote, Klaus mentions to be in private communication with me. This is correct and I think, Klaus and me have a much more similiar position than the forum dynamics makes people believe. E.g., consider that the term "aggressive post-sharpening" (for USM 0.5) could be meant in a general sense.

Below, I attach 100% crop, MTF and edge function (*).
__
(*) Gamma is set to 0.45 to linearize sRGB (imatest does the same (gamma=0.5), sfrmat uses gamma=1). At gamma=1, the MTF would show an insignificant sharpening effect in the blacks.

The data are from a K-5 with FA 31 Ltd. lens (center) at f/4.5 on a tripod. Focus determined by sharpest in a focus series.

The black and white levels are at 15% and 85% gray respectively. The rendering is from LR 2.x with standard settings except sharpening which is changed to r=0.5px and 100%.

As seen in the edge function, there is no overshooting in the edge (halo) and the MTF has no false maximum too. Both would be the typical signs of sharpening. So, this is a very good tuned amount of sharpening which shall be applied to raws from the K-5 (or K-7 and K20D for that matter). I don't consider the resulting image to be oversharpened (nor does Klaus AFAIK) and he gets the same images. If somebody sees sharpened artefacts in the image I would be glad to know (*).
__
(*) There are a few mild jaggies (only visible in direct comparison to the original) and USM with r=0.4 would be a better choice (if LR supported it). Or 80% rather than 100%. But in any real scenario, jaggies would be invisible.

The resulting K-5 resolution measurements are very good:
Sharpness (MTF50): 2709 LW/PH
Sharpness (edge): 3079 LW/PH
Pixels (the sensor): 3264 LW/PH
Resolution (MTF30): 3636 LW/PH
Resolution (MTF20): 4439 LW/PH

For comparison, original saved from photoshop (*):
Sharpness (MTF50): 2559 LW/PH
Sharpness (edge): 2743 LW/PH (edge function and original attached as last image)
Resolution (MTF33): 3264 LW/PH
Resolution (MTF30): 3427 LW/PH
__
(*) The original chart (5°) was reduced in size (25%) using bilinear to reduce softness caused by rotation. Using Bicubic (sharper) yielded worse results.

Discussion (what does it all mean):

• With optimum lens, setup and raw parameters, the K-5 delivers a sharpness which well reproduces the original. Sharper and softer both could be worse. There is no halo, almost no sharpening artefact or fuzzyness.
• However, there is mild color moiré and demosaicing mazes from a relatively weak AA filter. Weak if the goal would have been to prevent any color moiré. Just fine in a sense of a good compromise between acuity and fine texture color reproduction. Not weak in comparison to D7000 or Sony.
• Without the mazes (demosaicing), color moiré (AA), noise (sensor) and color aberration (lens), the image would look very close to the original.
• The contrast at the Nyquist frequency is 38%, close to and beating the original (a sign that less sharpening would do the job as well).
• The K-5 AA filter seems to be about the same strength as previous Penatx cameras (K20D and K-7). I didn't verify this though. It's just my feeling.

The K-5 has no strong AA filter or any softness "issue". This ends this discussion for me.

If Klaus can produce more reliable lens test figures using a weaker (or no) AA filter is another (interesting) question. But of no real concern to photographers.