[awaldram]

[quote name='Class A' timestamp='1290767877' post='4463']

I think you are right with your "going well into the false positive realm" comment, but don't quite understand why you regard it as "subjective" rather than "inappropriate"?





Do we agree that there is no point in looking for false detail and that this means that the K10D figure is inflated? I don't think we should be looking into pushing the K-5 figure but rather bringing down the K10D figure. Agreed?



I still think there is some clarification needed as to whether we are talking about B&W LW/PH in the case of the K10D or not.

[/quote]

Well I think if your going to use visual assessment of the resoltuion then it will always be subjective you will tend to see what you wish to see to prove/disprove your point.



Given these test charts are designed to allow computational assessment I'm surprised that more 'review' sites don't use this objective method.



Though I think Klaus's figure is to high I do accept his right to set the bar where he sees fit, But that bar should really remain constant for all cameras and not slide around to fit the story at the time.

[Klaus]

[quote name='awaldram' timestamp='1290766085' post='4462']

It would appear to me applying the same criteria as applied to your k10d figure would see the k5 hitting 2800 with a vanishing point around 3000.



As I can easily see false detail well past 2800.

[/quote]



The results from the different websites are not cross-comparable.



The locally provided numbers here were based on converted DNGed with the same (old) RAW converter. It is extremely unlikely that the others sites have done exactly the same. Some do their comparisons based on JPEGs, some on native RAW converters, some on the same converter but different versions, etc. pp.



I'm sometimes wondering how other sites measure the "extincting" resolution in order to make them comparable. This is all very dependent on how you convert the RAWs.

[Klaus]

[quote name='awaldram' timestamp='1290768394' post='4464']

Though I think Klaus's figure is to high I do accept his right to set the bar where he sees fit, But that bar should really remain constant for all cameras and not slide around to fit the story at the time.

[/quote]



Camera-based tests are never cross comparable. They never were and they never will be. Neither here at PZ nor elsewhere.

The tests provide a base on which you can rank a lens vs other lenses tested on the same camera-lens combination. This rank will remain the same or very similar on a different camera.



If I didn't remove the AA filter on the K5 I could as well stick to the K10D. The only point of restarting the tests is to do it on a higher (effective) resolution. This is a completely different question than whether the K5 is good or bad. We will not test the K5 here at PZ - the whole point of our sample is to do lens testing with it.

[awaldram]

[quote name='Klaus' timestamp='1290770932' post='4465']

The results from the different websites are not cross-comparable.



The locally provided numbers here were based on converted DNGed with the same (old) RAW converter.

[/quote]



Is it not possible then that your old Raw converter is the limiting factor in your results.?



may be an expensive exercise to remove the AA filter only to find the converter was the problem.

[mst]

[quote name='Class A' timestamp='1290765043' post='4461']

Fact is, as you say, that moiré represents damage to the image that can never be fully removed (not without losing detail anyhow).

[/quote]



In order to remove it, it needs to be in the shot. And as a matter of fact: there is none in most of my images (with the M9).



I think you put way too much weight on the issue of moire.



Apart from that, I definitely see it as Klaus: it's a matter of taste.



-- Markus

[awaldram]

[quote name='Klaus' timestamp='1290771568' post='4466']

Camera-based tests are never cross comparable. They never were and they never will be. Neither here at PZ nor elsewhere.



The tests provide a base on which you can rank a lens vs other lenses tested on the same camera-lens combination. This rank will remain the same or very similar on a different camera.



If I didn't remove the AA filter on the K5 I could as well stick to the K10D. The only point of restarting the tests is to do it on a higher (effective) resolution. This is a completely different question than whether the K5 is good or bad. We will not test the K5 here at PZ - the whole point of our sample is to do lens testing with it.

[/quote]

But you compared the k5 to the k10 and even the a33!



K5 @ 16mp:

max. 2500 LW/PH (RAW)

max. 2050 LW/PH (JPEG ****)



For comparison:



A33 @ 14mp:

max. 2850 LW/PH (RAW)

max. 2500 LW/PH (JPEG)



Maybe be I'm missing something .?

[Klaus]

Regarding Nyquist - you can look a bit beyond the barrier if you're test target is known. You cannot if you've no idea of what you're measuring of course. The K10D does naturally not deliver a resolution beyond that - especially considering Bayer. It's just about measuring a lens potential (FWIW, the big one provides resolution figures beyond Nyquist in their lens tests as well).

I will certainly not go into the details of the local testing procedure. Leave us our little secrets, please. :-) However, the basic procedure is described at imatest.com. Their documentation is very extensive and their method is used across the whole industry (including the some of the camera manufacturers). They also describe a little how they can measure beyond Nyquist if you're interested.

I reckon that the resolution numbers are not overly accurate anymore beyond Nyquist but that's my personal guess here.



As far as "false detail" is concerned:

http://www.maxmax.com/nikon_d200hr.htm

http://www.maxmax.com/nikon_d300HR.htm

http://www.maxmax.com/nikon_d700hr.htm



There's certainly some amount of false details in the "hot rod" versions. However, there's also a fair amount of extra detail as far as I can tell. Sharpening during post-processing will not recover all this and it would just come at cost of more noise anyway.



Just to clarify this - I DO NOT RECOMMEND THIS TO NORMAL USERS and most will not invest the bucks anyway.



And another note - I'm just guessing that the K5 has a strong AA filter. The "softness" may also originate in RAW NR. Only Pentax does not for sure (so far).



And one more - I know Falk personally. He lives just around the corner actually. A highly knowledgeable guy and I agree with him about the AA. The question is about the thickness, not about its existence. I would, of course, prefer to have some sort of AA filter like e.g. in the Pana G series. This is preferable to none at all. However, I doubt that a very strong AA filter is the right solution. The service company will "exchange" the AA filter so they'll not just remove it - they claim a gain of effective resolution of 30% max. so I suspect that there's still some filtering in there.

[Klaus]

Here's a real world scene without AA filter (Oly E-520):

http://photozone.smugmug.com/photos/551754081_WwNJs-O.jpg



Yes, there're moirees in there. Yes, there're some "false details". Yet there are vastly more details in there than in an AAed image.



Again - I do not recommend this to normal users.

[awaldram]

[quote name='Klaus' timestamp='1290773926' post='4470']

Regarding Nyquist - you can look a bit beyond the barrier if you're test target is known. You cannot if you've no idea of what you're measuring of course. The K10D does naturally not deliver a resolution beyond that - especially considering Bayer. It's just about measuring a lens potential (FWIW, the big one provides resolution figures beyond Nyquist in their lens tests as well).

[/quote]

Now I'm really confused

Everything I've ever done in electronics shows that a signal cannot be recovered below the Shannon sampling limit without multiple signal copies and pre-stacking.



Your statement "You cannot if you've no idea of what you're measuring of course" implies to me that you are using your mind to interpolate the missing data from the random dots above Nyquist.



This isn't usable data as without the reference you cannot reconstruct so is of zero practicable use in signal recovery.



It would also mean your data is not reproducible by a third party and cannot therefore be ratified or quantified as to accuracy.



The fact others do it, the same applies if your reporting data above Nyquist then you are interpolating that data (ie it isnt there) and as such are no longer producing a photograph but a piece of art or for a better word a lie.



ie the final image is a poor approximation of the original not a copy.

[Class A]

[quote name='Klaus' timestamp='1290773926' post='4470']

I will certainly not go into the details of the local testing procedure. Leave us our little secrets, please. :-)

[/quote]

Of course. I'm not after any trade secrets (although I wonder what competitive advantages you want to protect. Everyone can run the Imatest software?).



[quote name='Klaus' timestamp='1290773926' post='4470']

However, the basic procedure is described at imatest.com. Their documentation is very extensive and their method is used across the whole industry (including the some of the camera manufacturers). They also describe a little how they can measure beyond Nyquist if you're interested.

[/quote]

Do you have a pointer to that bit in their documentation handy?

I found this "Sensor response above the Nyquist frequency is garbage.". They also write "Since MTF is the product of the lens and sensor response, demosaicing algorithm, and sharpening, and since sharpening typically boosts MTF at the Nyquist frequency, the MTF at and above the Nyquist frequency is not an unambiguous indicator of aliasing problems. It may, however, be interpreted as a warning that there could be problems.".

I understand that you want to keep trade secrets but on the other hand it would be great to learn how are making sure that you are avoiding aliasing problems with the K10D. If you don't want to disclose it you are asking people to simply trust you.



Also, are you positively sure that your K10D vs K-5 comparison is appropriate?

I'd really like to see more backing up of your claim -- which I understand is applicable only to your specific needs.



[quote name='Klaus' timestamp='1290773926' post='4470']

And one more - I know Falk personally.

[/quote]

I know. <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/wink.gif' class='bbc_emoticon' alt='' /> He wrote he may will give you a ring. Hopefully a chat between the two of you will help to clear up the K10D vs K-5 mystery. I can speculate as to why your claim may not be quite as unreasonable as it may seem at first sight (and hinted at this in previous posts), but a) I'd really love to hear a justification from you rather than speculate myself and <img src='http://forum.photozone.de/public/style_emoticons/<#EMO_DIR#>/cool.gif' class='bbc_emoticon' alt='' /> I think even after some explanations, your claimed difference will not be justifiable to the extent it is currently stated.



But I'm open to learn, so it will be great to hearing from you again.