Lumolabs: K-x vs. K-7 and the dynamic range champion of 2009

[falconeye]

Pentax K-x to be the new dynamic range champion!



Lumolabs has just published its lab test for the Pentax K-x (and Pentax K-7 to enable a side by side comparison).

Falk Lumo: Lumolabs: Pentax K-x vs. K-7, sensor and video

Of course, lumolabs, that's me. After I start to be disappointed by most standard lab tests, I am starting my own lab testing now.

I basically redo the DxO style full SNR, 18% gray and dynamic range tests which do make some sense. I will develop some more meaningful tests over time.


To make a long story short (this is a 6 part blog article series):

The K-7 is a decent camera with good enough noise performance.

But currently nothing can challenge the dynamic range and high ISO noise performance of the Pentax K-x. No Aps-C camera. And very few full frame cameras like the Nikon D3X only. Maybe. With more than 13 EV dynamic range, it has almost no peers.

Click on the link above for the full story.



Full SNR:

[Daemos]

You have basically come to the same conclusion as I have.

Lower ISO the K7 out performa the kx at high iso the kx outperforms the k7. Overall detail of the k7 is better.

in my limited testing the kx and k7's DR seems to be similar enough in most situations that it would be difficult without testing to notice a difference. It is unfortunate we have not developed any sensors that can replicate the range the human eye can see.

[falconeye]

Originally Posted by Daemos

in my limited testing the kx and k7's DR seems to be similar enough in most situations that it would be difficult without testing to notice a difference. It is unfortunate we have not developed any sensors that can replicate the range the human eye can see.

As you can see, when talking about average vs. outstanding DR, I talk about 11 vs. 13 stops. Everything beyond 10 stops is almost invisible if you don't push shadows in post processing. But high dynamic range scenes exist so this makes an important difference. I.e., you sometimes HAVE to push shadows.

BTW, the human eye is mediocre. It has a static DR of just 6.5 stops. The K-x sensor has twice as many stops DR as has the human eye!!

The eye tricks around its limit though: it uses an adaptice local exposure changing its iris when doing saccades. This could be emulated using a very fast HDR function.

The eye can only vary between f/8 and f/2 (4 stops), so the dynamic range using saccades is about 11 stops. Still less than the K-x.

Finally, the eye can adapt another 10 stops by chemical dark adaption which takes up to 30 min. Compare this to a change in ISO setting. 10 stops in ISO is like spanning the range from ISO 50 to 51200... I would say, the K-x is really close!

The fact you don't see the high DR from, e.g., a K-x is due to the fact that both printing paper and monitors cannot output it. A full 13 EV image when fully output could probably already hurt a viewer's eye

[Daemos]

I see. I hope I'll learn more about the physiology of the eye in the next xoming year in school.

But trying to learn now, when I look at a scene my eye can pick out details. but when you take a picture, the camera will have blown highlights. Could you explain to me how this works?

I am guessing its because the camera will take better details in the shadows than the eye can see (based on your ecplination) That and the eye will adapt to the intensities of light it sees.

[falconeye]

Originally Posted by Daemos

But trying to learn now, when I look at a scene my eye can pick out details. but when you take a picture, the camera will have blown highlights. Could you explain to me how this works?

I am guessing its because the camera will take better details in the shadows than the eye can see (based on your ecplination)

You gave half of the answer already (the shadow detail).

The other half is in the wording "pick out details": The eye really is moving around an image and constantly adjusting exposure to the region of interest (it has to do this anyway because the resolution outside its center of sharpest vision is lousy). So, you don't see the blown highlights because in the moment when you want to look at it, the eye recenters and adjusts exposure and the highlight isn't blown anymore. Now, when you want to see anything in the shadow, the eye moves again and adjusts exposure again.

To really see the limits of the eye, try to see ANYTHING at the edge of an incoming car beamlight at night ... You KNOW that you have to look elsewhere to not crashing


If you look at the eye from an engineers point of view, it is a rather lousy optical design with a brilliant image processing engine combining many images per second into one super image which is an illusion created by the visual cortex as a courtesy to the rest of the brain (makes them survive, you know ).

[zxaar]

well you have confirmed what other had been speculating about k-x, that at base isos it has more DR than k7 's sensor and at higher iso has lower noise. In other words it has superior sensor than k7.

thanks for efforts.

[Daemos]

Originally Posted by falconeye

You gave half of the answer already (the shadow detail).

The other half is in the wording "pick out details": The eye really is moving around an image and constantly adjusting exposure to the region of interest (it has to do this anyway because the resolution outside its center of sharpest vision is lousy). So, you don't see the blown highlights because in the moment when you want to look at it, the eye recenters and adjusts exposure and the highlight isn't blown anymore. Now, when you want to see anything in the shadow, the eye moves again and adjusts exposure again.

To really see the limits of the eye, try to see ANYTHING at the edge of an incoming car beamlight at night ... You KNOW that you have to look elsewhere to not crashing

Thanks! makes perfect sense. Im new to photography. Do you have any links to learn more about DR and how it relates to cameras and the eye?

[Daemos]

Originally Posted by zxaar

well you have confirmed what other had been speculating about k-x, that at base isos it has more DR than k7 's sensor and at higher iso has lower noise. In other words it has superior sensor than k7.

thanks for efforts.

I am unsure how it has a "superior" sensor espically when it loses out to the k7 in terms of resolution and definition at low to medium isos.

So the k7 is "superior" in terms of definition and resolution.
But the Kx is "superior" in terms of dr (which isnt really noticable by people) and at high iso 3200+ has less noise...so I am not sure how the kx is hands down superior.

[zxaar]

Originally Posted by Daemos

I am unsure how it has a "superior" sensor espically when it loses out to the k7 in terms of resolution and definition at low to medium isos.

So the k7 is "superior" in terms of definition and resolution.
But the Kx is "superior" in terms of dr (which isnt really noticable by people) and at high iso 3200+ has less noise...so I am not sure how the kx is hands down superior.

resolution advantage of 14mp over 12mp is very slight, between them DR has more impact.


edited to add: about losing out, i think i need to see some tests done by some of the reputable sites in controlled environment.
So far i have seen on this site is an badly focused image of kx compared with k7 and that too someone complaining that it sees uneven sharpness, which suggests that either the sensor is faulty or there is problem in testing procedure.

[LeDave]

I have to agree, the K-x does have the better sensor, but that doesn't mean the K-7 is bad at all, it does have the same sensor used in the older K20D which has been praised by many in the Pentax community. Sure the K-7 may have ISO 100 and better low ISO noise levels with more resolution, but it still doesn't change the fact that the K-x does have better high ISO noise levels and more DR. It's ridiculously easier to fix up a low ISO noise picture than to fix up a noisy picture at 3200 and 6400 ISO. But to make things easier for both sides, I will just add the pros for each one and what I think which camera is more suitable for. The K-7 may be the better choice for fashion and more business-like things because of having 100 ISO and better noise at lower ISO. But for street photography, and wild-life photography, the K-x may be the better one due to it's better high ISO noise levels.

[falconeye]

Originally Posted by zxaar

edited to add: about losing out, i think i need to see some tests done by some of the reputable sites in controlled environment.
So far i have seen on this site is an badly focused image of kx compared with k7 and that too someone complaining that it sees uneven sharpness, which suggests that either the sensor is faulty or there is problem in testing procedure.

Are you making a comment wrt my blog article? If so, can you detail?

If not, you may want to have a look at the resolution part of the article which is may be of interest to you as well.

The nominal resolution advantage of the K-7 is small, but both cameras have different characteristics close to their respective Nyquist frequencies. Personally, I prefer the cleaner image from the K-7, giving the K-7 a larger advantage than the pure number of pixels would suggest. Others may see a smaller advantage because they like very weak AA filters. YMMV. The details are in the blog article linked in the OP.

[zxaar]

Originally Posted by falconeye

Are you making a comment wrt my blog article? If so, can you detail?

If not, you may want to have a look at the resolution part of the article which is may be of interest to you as well.

na not at all, your findings are pretty much inline with what others have speculated about kx sensor.
i was talking about other thread comparing k7 and kx images, half of which is devoted to discussion whether kx images are even focused properly and then people playing with sharpness. And someone making conclusion that kx has uneven sharpness (probably meant resolution). Which i think that not possible for healthy sensor because for sensor all pixels should be equal. sensor only shows what reaches it through lense. so if there is unevenness either jpeg engine screwed it or testing procedure is faulty.

not about your op though.

[rawr]

Originally Posted by Daemos

...Im new to photography. Do you have any links to learn more about DR and how it relates to cameras and the eye?

Visit the DXOMark web site. They have an EXTENSIVE library of technical discussions on dynamic range, noise, ISO sensitivity, resolution and every other camera sensor variable you could imagine, as well as an extensive library of camera sensor measurements for hundreds of cameras.

Welcome to dxomark.com (beta), a free resource dedicated to RAW-based camera image quality

Perhaps it would be good for you to study these issues more before making such confident statements as you have made here about the performance of the K7 vs the K-x in terms of resolution, noise, DR or anything else.

[Daemos]

Originally Posted by rawr

Visit the DXOMark web site. They have an EXTENSIVE library of technical discussions on dynamic range, noise, ISO sensitivity, resolution and every other camera sensor variable you could imagine, as well as an extensive library of camera sensor measurements for hundreds of cameras.

Welcome to dxomark.com (beta), a free resource dedicated to RAW-based camera image quality

Perhaps it would be good for you to study these issues more before making such confident statements as you have made here about the performance of the K7 vs the K-x in terms of resolution, noise, DR or anything else.

Well I feel I can make the statements about the Kx vs K7 as subjectively the K7 has looked better in my eyes both times I've tested a Kx (2 different ones vs a K7). Everyone has opinions on what makes a picture 'look' better. Image quality is very subjective. It's not about pure numbers, it's about how someone feels when they look at a picture.

I make statements, this is the internet, it is quite difficult to convey emotion through the internet. I have been corrected about the DR on the eye, and based on what you are saying you are stating that I'm coming off as arrogant and cocky. Which I am not, I am an eager learner, if I wasn't I would of never posted twice now pictures comparing the Kx and K7. I fully posted them to see what other people thought and think, first time there were some issues with the testing, I went out learning what I did and tried re-testing.

Thanks for the link though. I'm sure there is a lot I can learn from the site. Again I hope I am not coming off as arrogant or anything like that.

Ultimately I would like to try and retest the K7 and KX again. However, other than the in-body sharpness thing, I'm not actually quite sure if I had done anything wrong in the way I've tested. I want to buy the KX for my gf, so I am trying to really like it.

Originally Posted by zxaar

na not at all, your findings are pretty much inline with what others have speculated about kx sensor.
i was talking about other thread comparing k7 and kx images, half of which is devoted to discussion whether kx images are even focused properly and then people playing with sharpness. And someone making conclusion that kx has uneven sharpness (probably meant resolution). Which i think that not possible for healthy sensor because for sensor all pixels should be equal. sensor only shows what reaches it through lense. so if there is unevenness either jpeg engine screwed it or testing procedure is faulty.

not about your op though.

Well unfortunately I've tested the K7 vs KX twice now, both using different KXs.

I am AWAITING someone else who would actually do the same thing. I invite you to do it.

[rawr]

Originally Posted by falconeye

But currently nothing can challenge the dynamic range and high ISO noise performance of the Pentax K-x. No Aps-C camera. And very few full frame cameras like the Nikon D3X only. Maybe. With more than 13 EV dynamic range, it has almost no peers.

Thanks for this detailed technical measurement of the K7 and K-x. Very useful. Especially as DXOLabs have yet to issue a K7 assessment of their own, let alone a K-x one.

When they do, it will be interesting to compare assessment(s).

Have you tested other cameras - eg the D90 - against the K-x or K7?

Or is your methodology similar enoght to DXOLabs for one to be able to directly compare your test results (eg for SNR) against DXO's?

Once again well done. LumoLabs will be very useful for a lot of people.