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

pxpaulx · 10-29-2009, 12:33 PM

Originally posted by Sew-Classic

Interesting.

In those samples, I do think the K-7 has more noise even with the NR from LR (It looks good, especially when view like a normal photo and not pixel peeped, just more noise in comparison to the K-x), BUT try and read the numbers on the lens for both photos. Why is the K-x photo so hard to read? Is it more of the same noise vs. detail loss, blah, blah, blah - pick your poison?? Dunno....

I resized the K-7 photo to the same size as the K-x, and then I did a 600x600 crop at the lens to both photos.

K-x 100% crop from the iso 1600 photo:

K-7, resized to match K-x, then 100% cropped from the 1600 iso photo:

Honestly the K-x photo looks out of focus in these examples. As far as actual noise is concerned, they appear pretty close to my eye, but with more chroma noise in the K7 shot - I would definitely take the K-x photo out of these two examples (OOF not being considered)

Sew-Classic · 10-29-2009, 12:47 PM

Originally posted by pxpaulx

Honestly the K-x photo looks out of focus in these examples. As far as actual noise is concerned, they appear pretty close to my eye, but with more chroma noise in the K7 shot - I would definitely take the K-x photo out of these two examples (OOF not being considered)

You can see the original K-x photo here and here. It looked like it was in focus to me- all except for the lens. weird

Here's the K-7 full shot at the same iso from that site:
http://fotkidepo.ru/photo/132601/24445pwW2GP2SlT/GZa8syV6SN/412281.jpg

Eruditass · 10-29-2009, 01:02 PM

Leads me to believe the K-x was front focused possibly due to lighting without the K-7's new AF sensor or FF problem. I suggested f/8.

*isteve · 10-29-2009, 02:01 PM

Falk, I have read through your review, great job.

Having already got a K7, I plan to trade my old K20/K10 pairing for a Kx as my backup camera, and the one I use for gigs and street shots with fast primes.

Having both cameras is really the best of both worlds. To me this is a superior option to a single FF camera and still cheaper than an A850 or D700.

And although the push processing of K7 shadows may reveal some noise, its not much effort to apply noiseware to the shadow areas first as there is little detail in the last 2 stops anyway. This way you can fake another couple of stops quite easily.

falconeye · 10-29-2009, 06:49 PM

Originally posted by *isteve

its not much effort to apply noiseware to the shadow areas first as there is little detail in the last 2 stops anyway. This way you can fake another couple of stops quite easily.

This is true and many testers including DxO forget that dynamic range should be measured in combination with a much lower resolution.

I propose to use the same resolution used for the DoF scale which is about 1.4 Mpixel. Or if people prefer, a more modern HD norm like 1920x1280.

E.g., the dynamic range of a black&white text page is whatever be the difference between black and white, e.g., one stop. But still, when printed out and hung on a wall, one may reveal that all the tiny letters combine into a large artwork and you see another image, not letters anymore. And this artwork actually may have great dynamic range. Just at a coarser level of resolution.

So, dynamic range is a concept not well understood. It reminds me at the depth of field confusion

Code:

MZZZZZZZZZZZZZ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ZMZZ++
MMMZZZZZZZZZZZZ+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ZMZZ++
MMZZZZZZZZZZZZZ+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ZMZZ++
MMMMZZZZZZZZZZ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ZMMZ+ 
MMMZZZZZZZZZZZ+++++++++Z+++++++++++++++++++++++++++++++++++++++++++++++++++MMZ+ 
MMMMZZZZZZZZZ++++++++++++ZZ++++++++++++++++++++++++++++++++++++++++++++++++MMZZ 
MMMMZZZZZZZZZZ++++++ZZZ++++++++++++++++++++++++++++++++++++++++++++++++++++MMZZ 
MMMMMZZZZZZZZ+ZZZZZZZZZZZZZZZZZ+Z++++++++++++ZZZZZZZZZZZZZZZZZ+ZZZZZ+++++++ZZZ+ 
MMMMMMMMMMZZZZZZZZZZZMMMMMZZZZZZZZZ++++++++++++ZZ+ZMMZZZZ++Z++++++ZZZ++++++ZZZZ 
MMMMMMMMMZZZZZZZZMMMMMZZMMMZZZZZZZZ++++++++++++++ZMMMMMMZZZZZZZ++++ZZ++++++MMZZ 
MMMMMMZZZZZMMMMMMM+MZ +ZMZ+MMZZMMZZ++++++++++++ZZMZZZMM++M+ ZZZZMZ+++++++++MMMZ 
MMMMMZZZZMMMZZMMMMMMZ  +ZZ++MMMMMZZZ++++++++++ZZMZ ZZMMMZMZ  ZMZMZ+++++++++MZZZ 
ZZZZZZZMMMMZZZ+ZZZZ   ++ZZZMZZMMMMZZZ++++++++++ZZZ+++ZZZZ+  +ZZMZZ++++++++ZMZ++ 
ZZZZZZZMMMMZZZZZZZ+++ZZ++ZZZZZMMMMMZZ++++++++++ZZZZZ+++++  +ZZZZZZ++++++++ZZZ++ 
Z++ZZMZZZMMMMZZZZZZZ+ZZZZ+++ZZZMZMMMZZZ+++++++++++ZZZZZZZZZZZZZ+++++++++++ZZZ++ 
ZZZZZZZZZZZZZZZZ++++++++++++ZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z+++Z 
ZZZZZZZZZZZZZZ+++++++++++++ZZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z+++Z 
ZZZZ++ZZ++++++++++++++++++++ZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z++MZ 
ZZZZZZZ+++++++++++++++++++++ZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z++ZZ 
ZZZZZ++++++++++++++++++++++ZZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z++++ 
ZZZZZ++++++++++++++++++++++ZZZZZMMMZZZZ+++++++++++++++++++++++++++++++++++++Z++ 
ZZZZZZ+++++++++++++++++++++ZZZZZMMMMZZZ+++++++++++++++++++++++++++++++++++Z+Z++ 
MZZZZZZ++++++++++++++++++++ZZZZMMMMMZZZ++++++++++++++++++++++++++++++++++ZZZZ++ 
MZZZZZZZ+++++++++++++++++++ZZMMMMMMZZZZ+Z+++++++++++++++++++++++++++++++++ZZZ++ 
MZZZZZZZZZ+Z++++++++++++++ZZMMMMMMMMZZZ++++++++++++++++++++++++++++++++++++++++ 
MMZZZZZZZZZ+++++++++++++++ZMMMZMMMMMZZZ++++++++++++++++++++++++++++++++++++Z++  
MZZMZZZZZZ+ZZ++++++++++++ZMMMZZZMMMMZZ++++++++++++++++++++++++++++++++++++Z+++  
MMMMZZZZZZZZ++++++++++++ZZMMMZZZMMMMZZ++++++++++++++++++++++++++++++++++++Z++   
MMMMZZZZZZZ++++++++++++ZZZMMMMZMMMMZZZZ+++++++++++++++++++++++++++++++++++Z++   
MMMMZZZZZZZZ++++++++++++ZZZMMMMMMMMMZZZZ++++++++++++++++++++++++++++++++Z+Z++   
MZMMMZZZZZZZZ++++++++++ZZZZMMMMMMMMMMZZZ++++++++++++++++++++++++++++++++++Z++   
MZZMMZZZZZZZZ+++++++++++ZZZZZMMMMMMMMMMMZZZZZZZ+++++++++++++++++++++++++++Z++   
MMZZMZZZZZZZZ+++ZZZZ++ZZZZZZZZZMMMMMMMMZZZZZ+++++++++++++++++++++++++++Z++++Z   
MZZZZZZZZZZZ+++ZZZZZZZZZZZZZZZZZZZZZZZ+ZZZ++++++++++++++++++++++++++++++++++Z   
MMZZZZZZZZZ+++ZZZZZZZ++ZZZZZZZZZZZZZZ+++ZZ++++++++++++++++++++++++++++++++Z+Z   
MMZZZZZZZZZZ++ZZZZZZ++ZZZZZZZ++++++++++++++++++++++++++++++++++++++++++++ZZ+    
MMMZZZMZZZZ+++ZZZZZZZZZZZZZZZZZZZZZZZ++++++ZZZZZZZZZZZZ++++++++++++++++++ZZZ    
MMMZZZMMZZZ+ZZZZZZZZZMMMMMMMMMMMMZZMMMZZZZZZZZZMMZMMMMZZ++++++++++++++++++ZZ    
MMMZZZMMZZZZZZ++++++++ZZMMMMMMMZZMMMMMMMMMMMMZZZZZZ++++++++++++++++++++++++Z    
MMMZMZMMZZZZZZ++++++++ZZZZMMMMMMZZZZZZZZZZZZZZZZZZ++++++++++++++++++Z+++++++    
MMMMMMMMMZZZZZ+++++++++ZZZZZZZZZZZZZZZ++++ZZZZZZ+++++++++++++++++++Z+++++++Z    
MMMMMMMMMZZZZZ++ZZ++ZZZZZZZZZZZZZZZZZZZ+++Z+++++Z++++++++++++++++++++++++++     
MMMMMMMMMMZZZZZ++++ZZZZZZZZZZZZZZZZZZZZZZZZZZZZ++++++++++++++++++Z++++++++Z     
MMMMMMMMMMMZZZZZZZZZZZZZZ+ZZ+++++++ZZZZ+ZZ++++++++++++++++++++++Z+++++++++      
MMMMMMMMMMMZZZZZZZZZZZZZZZZ+++++++ZZZZ+ZZ++++++++++++++++++++++Z+++++++++       
MMMMMMMMMMMZZZZZZZZZZZZZZZ+++++++++++++++++++++++++++++++++++Z+++++++++++       
MMMMMMMMMMMMZZZZZZZZZZZZZZ+++++++++++++++++++++++++++++++++++Z+++++++++++       
MMMMMMMMMMMMMMZZZZZZZZZZZZ+Z++++++++++++++++++++++++++++++ZZ++++++++++++Z       
MMMMMMMMMMMMMMMZZZZZZZZZZZZZ+++++++++++++++++++++++++++++ZZ+++++++++++++        
MZZMMMMMMMMMMMMMMMMZZZZZZZZZZ++++++++++++++++++++++++++ZZ+++++++++++++++        
MZZMMMMMMMMMMMMMMMMMZZZZZZZZZZ++Z+++++++++++++++++++++ZZ++++++++++++++++

Daemos · 10-29-2009, 07:56 PM

Originally posted by falconeye

This is true and many testers including DxO forget that dynamic range should be measured in combination with a much lower resolution.

I propose to use the same resolution used for the DoF scale which is about 1.4 Mpixel. Or if people prefer, a more modern HD norm like 1920x1280.

E.g., the dynamic range of a black&white text page is whatever be the difference between black and white, e.g., one stop. But still, when printed out and hung on a wall, one may reveal that all the tiny letters combine into a large artwork and you see another image, not letters anymore. And this artwork actually may have great dynamic range. Just at a coarser level of resolution.

So, dynamic range is a concept not well understood. It reminds me at the depth of field confusion

Code:

MZZZZZZZZZZZZZ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ZMZZ++
MMMZZZZZZZZZZZZ+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ZMZZ++
MMZZZZZZZZZZZZZ+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ZMZZ++
MMMMZZZZZZZZZZ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ZMMZ+ 
MMMZZZZZZZZZZZ+++++++++Z+++++++++++++++++++++++++++++++++++++++++++++++++++MMZ+ 
MMMMZZZZZZZZZ++++++++++++ZZ++++++++++++++++++++++++++++++++++++++++++++++++MMZZ 
MMMMZZZZZZZZZZ++++++ZZZ++++++++++++++++++++++++++++++++++++++++++++++++++++MMZZ 
MMMMMZZZZZZZZ+ZZZZZZZZZZZZZZZZZ+Z++++++++++++ZZZZZZZZZZZZZZZZZ+ZZZZZ+++++++ZZZ+ 
MMMMMMMMMMZZZZZZZZZZZMMMMMZZZZZZZZZ++++++++++++ZZ+ZMMZZZZ++Z++++++ZZZ++++++ZZZZ 
MMMMMMMMMZZZZZZZZMMMMMZZMMMZZZZZZZZ++++++++++++++ZMMMMMMZZZZZZZ++++ZZ++++++MMZZ 
MMMMMMZZZZZMMMMMMM+MZ +ZMZ+MMZZMMZZ++++++++++++ZZMZZZMM++M+ ZZZZMZ+++++++++MMMZ 
MMMMMZZZZMMMZZMMMMMMZ  +ZZ++MMMMMZZZ++++++++++ZZMZ ZZMMMZMZ  ZMZMZ+++++++++MZZZ 
ZZZZZZZMMMMZZZ+ZZZZ   ++ZZZMZZMMMMZZZ++++++++++ZZZ+++ZZZZ+  +ZZMZZ++++++++ZMZ++ 
ZZZZZZZMMMMZZZZZZZ+++ZZ++ZZZZZMMMMMZZ++++++++++ZZZZZ+++++  +ZZZZZZ++++++++ZZZ++ 
Z++ZZMZZZMMMMZZZZZZZ+ZZZZ+++ZZZMZMMMZZZ+++++++++++ZZZZZZZZZZZZZ+++++++++++ZZZ++ 
ZZZZZZZZZZZZZZZZ++++++++++++ZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z+++Z 
ZZZZZZZZZZZZZZ+++++++++++++ZZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z+++Z 
ZZZZ++ZZ++++++++++++++++++++ZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z++MZ 
ZZZZZZZ+++++++++++++++++++++ZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z++ZZ 
ZZZZZ++++++++++++++++++++++ZZZZZZMMMZZ++++++++++++++++++++++++++++++++++++Z++++ 
ZZZZZ++++++++++++++++++++++ZZZZZMMMZZZZ+++++++++++++++++++++++++++++++++++++Z++ 
ZZZZZZ+++++++++++++++++++++ZZZZZMMMMZZZ+++++++++++++++++++++++++++++++++++Z+Z++ 
MZZZZZZ++++++++++++++++++++ZZZZMMMMMZZZ++++++++++++++++++++++++++++++++++ZZZZ++ 
MZZZZZZZ+++++++++++++++++++ZZMMMMMMZZZZ+Z+++++++++++++++++++++++++++++++++ZZZ++ 
MZZZZZZZZZ+Z++++++++++++++ZZMMMMMMMMZZZ++++++++++++++++++++++++++++++++++++++++ 
MMZZZZZZZZZ+++++++++++++++ZMMMZMMMMMZZZ++++++++++++++++++++++++++++++++++++Z++  
MZZMZZZZZZ+ZZ++++++++++++ZMMMZZZMMMMZZ++++++++++++++++++++++++++++++++++++Z+++  
MMMMZZZZZZZZ++++++++++++ZZMMMZZZMMMMZZ++++++++++++++++++++++++++++++++++++Z++   
MMMMZZZZZZZ++++++++++++ZZZMMMMZMMMMZZZZ+++++++++++++++++++++++++++++++++++Z++   
MMMMZZZZZZZZ++++++++++++ZZZMMMMMMMMMZZZZ++++++++++++++++++++++++++++++++Z+Z++   
MZMMMZZZZZZZZ++++++++++ZZZZMMMMMMMMMMZZZ++++++++++++++++++++++++++++++++++Z++   
MZZMMZZZZZZZZ+++++++++++ZZZZZMMMMMMMMMMMZZZZZZZ+++++++++++++++++++++++++++Z++   
MMZZMZZZZZZZZ+++ZZZZ++ZZZZZZZZZMMMMMMMMZZZZZ+++++++++++++++++++++++++++Z++++Z   
MZZZZZZZZZZZ+++ZZZZZZZZZZZZZZZZZZZZZZZ+ZZZ++++++++++++++++++++++++++++++++++Z   
MMZZZZZZZZZ+++ZZZZZZZ++ZZZZZZZZZZZZZZ+++ZZ++++++++++++++++++++++++++++++++Z+Z   
MMZZZZZZZZZZ++ZZZZZZ++ZZZZZZZ++++++++++++++++++++++++++++++++++++++++++++ZZ+    
MMMZZZMZZZZ+++ZZZZZZZZZZZZZZZZZZZZZZZ++++++ZZZZZZZZZZZZ++++++++++++++++++ZZZ    
MMMZZZMMZZZ+ZZZZZZZZZMMMMMMMMMMMMZZMMMZZZZZZZZZMMZMMMMZZ++++++++++++++++++ZZ    
MMMZZZMMZZZZZZ++++++++ZZMMMMMMMZZMMMMMMMMMMMMZZZZZZ++++++++++++++++++++++++Z    
MMMZMZMMZZZZZZ++++++++ZZZZMMMMMMZZZZZZZZZZZZZZZZZZ++++++++++++++++++Z+++++++    
MMMMMMMMMZZZZZ+++++++++ZZZZZZZZZZZZZZZ++++ZZZZZZ+++++++++++++++++++Z+++++++Z    
MMMMMMMMMZZZZZ++ZZ++ZZZZZZZZZZZZZZZZZZZ+++Z+++++Z++++++++++++++++++++++++++     
MMMMMMMMMMZZZZZ++++ZZZZZZZZZZZZZZZZZZZZZZZZZZZZ++++++++++++++++++Z++++++++Z     
MMMMMMMMMMMZZZZZZZZZZZZZZ+ZZ+++++++ZZZZ+ZZ++++++++++++++++++++++Z+++++++++      
MMMMMMMMMMMZZZZZZZZZZZZZZZZ+++++++ZZZZ+ZZ++++++++++++++++++++++Z+++++++++       
MMMMMMMMMMMZZZZZZZZZZZZZZZ+++++++++++++++++++++++++++++++++++Z+++++++++++       
MMMMMMMMMMMMZZZZZZZZZZZZZZ+++++++++++++++++++++++++++++++++++Z+++++++++++       
MMMMMMMMMMMMMMZZZZZZZZZZZZ+Z++++++++++++++++++++++++++++++ZZ++++++++++++Z       
MMMMMMMMMMMMMMMZZZZZZZZZZZZZ+++++++++++++++++++++++++++++ZZ+++++++++++++        
MZZMMMMMMMMMMMMMMMMZZZZZZZZZZ++++++++++++++++++++++++++ZZ+++++++++++++++        
MZZMMMMMMMMMMMMMMMMMZZZZZZZZZZ++Z+++++++++++++++++++++ZZ++++++++++++++++

THAT is neat!

*isteve · 10-29-2009, 08:20 PM

Originally posted by falconeye

This is true and many testers including DxO forget that dynamic range should be measured in combination with a much lower resolution.

I propose to use the same resolution used for the DoF scale which is about 1.4 Mpixel. Or if people prefer, a more modern HD norm like 1920x1280.

E.g., the dynamic range of a black&white text page is whatever be the difference between black and white, e.g., one stop. But still, when printed out and hung on a wall, one may reveal that all the tiny letters combine into a large artwork and you see another image, not letters anymore. And this artwork actually may have great dynamic range. Just at a coarser level of resolution.

So, dynamic range is a concept not well understood. It reminds me at the depth of field confusion

You are the first person to bring this up and of course you are right. Measured DR (the way DXO does it) is largely academic in relation to achievable subjective DR (the range of patches that a camera can actually record in RAW mode which is resolution independent and can therefore be normalised for noise as in your rather neat demonstration

). In this sense, Phil Askey's measurement technique is the most useful if not the most accurate. Sadly, he only seems to apply it to JPEG files.

The K7 does fine even with minimal NR (Silkypix Default) as in this sample...

http://ic2.pbase.com/o6/89/257389/1/115505969.bo4n95rr.RAWdr.jpg

tr13 · 10-29-2009, 09:35 PM

falconeye

You could try to repeat measurements with DISABLENR set to 1 for both DSLRs.
And look is RAW is now not affected bvy noise reduction.

falconeye · 10-30-2009, 02:38 AM

Originally posted by tr13

You could try to repeat measurements with DISABLENR set to 1 for both DSLRs.

That is a good suggestion. I'll try to follow it with my next test. Another reason to retest the K-x (the other reason being to test it at -10EV

).

jamesm007 · 10-30-2009, 12:27 PM

The K-x does use RAW NR starting at ISO3200 as the K-7 this NR is twice as strong as used on the K20D at ISO3200 and probably much stronger than the K-7. This of course does not answer why your ISO1600 pic from the K-7 came out the winner. Try the test at ISO3200 and see what happens. The K-x does measure high DR but its also using much stronger NR.

Remember the D300 was not as good per the numbers at DXO versus the D90 and D5000 it was becuase mainly those two were doing things to the RAW data such as RAW NR at ISO200!

rawr · 10-30-2009, 12:50 PM

Originally posted by jamesm007

The K-x does use RAW NR starting at ISO3200 as the K-7 this NR is twice as strong as used on the K20D at ISO3200 and probably much stronger than the K-7.

Huh?

I seriously doubt that information about the K-x's use of NR in RAW. And even the statement about K-x/K7 NR being twice as strong as K20D seems wildly off the mark. No one who has looked into the K-x and K7 or even the K20D RAW has drawn the same conclusions.

georgweb · 10-30-2009, 01:21 PM

Thanks for sharing Falk !

I went through some of your blog's pages, very readable and eye-relaxing. I can only imagine how much work this was - you are saying it is in fact windfall from your work projects but nevertheless.

I am foremost interested in real world (well my world) low light testing, and I understand you are doing this with tungsten light and shutter speeds down to 1/15 or 1/10. This would comply with my needs (stage and social stuff without flash) and I just want to ask you whether your high ISO testing includes those settings and is therefore represented adequately in your charts.

Vielen Dank !
Georg (the other German)

jamesm007 · 10-30-2009, 03:34 PM

Originally posted by rawr

Huh?

I seriously doubt that information about the K-x's use of NR in RAW. And even the statement about K-x/K7 NR being twice as strong as K20D seems wildly off the mark. No one who has looked into the K-x and K7 or even the K20D RAW has drawn the same conclusions.

Its a well known fact that Pentax uses NR in RAW files at high ISO, DXO found this and penalized Pentax for it.

DxOMark review for Pentax cameras

A quote from DXO

"Noise filtering processing is present on the Pentax K10D (ISO 800 and ISO 1600), the Pentax K20D (ISO 3200 and ISO 6400), and the Pentax K200D (ISO 800 and ISO 3200).
As the Low-Light ISO scores for the Pentax K10D and Pentax K200D changed slightly after taking noise filtering into account, their global dxomark.com rankings changed as well. However, Pentax K20D Low-Light ISO scores were not impacted by noise filtering."

They made an error for the K20D its NR starts at ISO1600 not ISO3200.
I found high levels of chroma NR in the K-x samples I have very few now but just judging them I could see very high levels of NR at ISO3200 to the point of robbing detail; this can change if I see other samples. Gordon B Good a sensor scientist at drpreview confirmed my findings and has it all measured; basically the K-x uses twice as much NR at ISO3200 versus the K20D, if you where to open to identical normal scenes the K20D would have and does have a finer noise pattern versus the K-x you can begin to see it affects in the pics above but according to Gordon B Good the K-x does not apply NR at ISO1600.

Just for your info in RAW the K10 applies NR starting at ISO800, the K20D at ISO1600, the K-7 at ISO3200, the K-x at ISO3200.

Gordon B Good also discovered some interesting things in Nikon's RAW files, Gordon B Good has found 90% for sure NR in RAW at base ISO (200) with the Nikon D90 and D5000. Ever wonder why they would beat the D300 in DXO? Because it was better sensors, lol nope because Nikon tricked DXO the D5000 and D90 do not have more DR than the Nikon D300 at ISO200 they have less, and less than the K20D.

There is no one reliable source of info for sensors from PRO Web sites as falconeye says (maybe his blog can be the source), not DXO, not dpreview, not..., you did not even know the Pentax K20D applies NR in RAW starting at ISO1600!!! its a very small amount however. Now people still keep knocking the Samsung sensor the one at least in the K20D may be one of the best high ISO APS sensor out at most second by a small amount to the D300 I lean toward the K20D Samsung being the overall best APS sensor (have not seen enough of the Canon 7D yet); it all depends on what light level you test, if you test at 18% grey or pics like the queen at dpreview the K20D takes the win as dpreview shows us versus the D300, and it would also beat the K-x, Now in real dark tests the K-x with its better dark read noise, or the D300 would be better. I think a standard measure of high ISO noise, really 3 standards should be made. Low-light, mid-light, high-light.

And now a quote for sensor scientist Gordon B Good who has known for years of non-defeatable NR in Pentax RAW files and some othe folks who are really into sensors. It is a subject not talked about much and easy for me to understand how and why you or anyone would not know

"James, as I already posted further up the thread, the K-x does apply raw Noise Reduction (NR) to the K-x raw data for ISO's of 3200 and up, and I measure from the French contributor's posted raw files that this raw NR is a little bit heavy handed to reduce the noise by a factor or two in standard deviation. I don't believe that any NR is applied to the raw data for ISO's of 1600 and lower. This NR applied by default (can't be turned off) by the K-x to high ISO raw data looks to be about twice as strong as that applied to the raw data of the K20D for ISO's of 1600 and above, as in the K20D only reducing the standard deviation of the noise by a about a factor of about 0.7 instead of by a factor of 0.5 for the K-x. Of course, it is this fairly heavy handed raw NR that makes the ISO 6400 and 12800 reasonably usable without further application of NR during JPEG processing. The strength of the NR applied to the K-x raw data for ISO 3200 and above is about the same as the NR applied to the K10D, my K200D, and the K-m/K2000D for ISO 800 and above, so it isn't all that bad! I suspect that one could make very good prints from even ISO 3200 to 6400 up to about 8 X 12 inches from the K-x which is amazing in that I wouldn't want to print this size from anything much over about ISO 800 for my K200D (equivalent image quality to the K-m/K2000D).
Remember, if this raw NR weren't applied by the K-x, then there likely wouldn't likely be a ISO 12800 option as otherwise the noise would make it look fairly bad. Pentax has to try to tread the fine line between keeping the fine-detail-at-full resolution junkies such as yourself happy as well as the ever increasing drive to make higher and higher ISO's somewhat usable, and for this entry level camera they seem to have felt that a little loss of detail is a worth while trade-off in order to have reasonably clean looking high ISO images"

rawr · 10-30-2009, 04:16 PM

Hmm. Interesting. Any K-x NR above 3200 is very subtle then, since lots of usable detail remains even at 12800 ISO, whilst still retaining good colour sensitivity and tonal range.

I await DXOLabs eventual examination of the K7 (and K-x) with even more interest now.

PS: I don't think Nikon would agree with GBG's views about their 'sensor tricks'.

jamesm007 · 10-30-2009, 05:48 PM

Originally posted by rawr

Hmm. Interesting. Any K-x NR above 3200 is very subtle then, since lots of usable detail remains even at 12800 ISO, whilst still retaining good colour sensitivity and tonal range.

I await DXOLabs eventual examination of the K7 (and K-x) with even more interest now.

PS: I don't think Nikon would agree with GBG's views about their 'sensor tricks'.

The neat part is you can see in pics anything I say or GBG says. I can see very course noise and lack of detail in the deep shadows in pics from the D5000 and D90 compared to the D300 at ISO200. The D300 had a hands off approach to RAW PP in the EXPEED engine remember the respective image engines from different manufactures as the PRIME for Pentax can just as easily work with RAW as JPEG in fact they do. So I see with my own eyes more noise and smoothed over details in the Nikons D5000 and D90 compared to the very fine D300; its there 90% for sure according to me and lots of others. I will try to post for you some pics with examples as I see your interested and kinda don't know what to believe which is good. Do some searches on dpreview for Gordon B Good, and you should along with his posts find some of my work occasional, that I submit to him and I my thoughts as to whats going on, for his analysis I just look and try to figure out with help on how to analyzie pics he has freely given to others and use the tools he says to use.

The trick is the RAW converter its needs to be a neutral one like Raw Therapee, ACR and many others will add NR even with the sliders at 0, RT does not. So I can import DNG (files ran through Adobes DNG converter) open them in RT and compare areas of my interest. This is how I found high course NR in the K-x ISO3200 pics.

BTW DXO is not 100% accurate they have yet to catch Nikons NR in RAW and many of the results are skewed. There is no one single source of accurate data for questions of performance we want answered. Remember the D90 according to DXO was much better than the D300 in many ways yet when I looked at real pics with a neutral converter I saw the D300 cleaning the D90s house.

If interested keep reading, Gordon B Good has written programs such as a program to remove tint from the corners of the K20D at high ISO. GBG also writes his own raw converters he has some serious sensor knowledge. Its healthy for you to be skeptical however.

You know after all this heavy pixel peeping of many different cameras and lots of tests I have come to one conclusion about APS sensors, I would use any of todays APS sensors and not feel bad. If given a choice I prefer higher-mega pixels because I am a cropper, but as far as IQ at normal viewing sizes there is not much to talk about.

10-30-2009, 01:21 PM	#42
georgweb Veteran Member Join Date: Apr 2008 Location: Berlin, Germany Photos: Gallery Posts: 1,386	Thanks for sharing Falk ! I went through some of your blog's pages, very readable and eye-relaxing. I can only imagine how much work this was - you are saying it is in fact windfall from your work projects but nevertheless. I am foremost interested in real world (well my world) low light testing, and I understand you are doing this with tungsten light and shutter speeds down to 1/15 or 1/10. This would comply with my needs (stage and social stuff without flash) and I just want to ask you whether your high ISO testing includes those settings and is therefore represented adequately in your charts. Vielen Dank ! Georg (the other German) Last edited by georgweb; 10-30-2009 at 04:41 PM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Dynamic Range on the K-x	Shahmatt	Pentax DSLR Discussion	33	04-22-2010 03:51 PM
Lumolabs: Nikon D700 vs. D5000 vs. Pentax K-x, Dynamic range and noise	falconeye	Pentax News and Rumors	12	12-18-2009 05:34 AM
dynamic range K20D	dafiryde	Pentax DSLR Discussion	4	01-08-2009 10:51 AM
Dynamic Range	Rene`	Pentax DSLR Discussion	6	09-25-2008 07:46 AM
Dynamic Range	Vlad	Photographic Technique	5	06-01-2008 10:40 PM

10-29-2009, 12:33 PM	#31
pxpaulx Veteran Member Join Date: May 2009 Location: Twin Cities, MN Posts: 2,867	Originally posted by Sew-Classic Interesting. In those samples, I do think the K-7 has more noise even with the NR from LR (It looks good, especially when view like a normal photo and not pixel peeped, just more noise in comparison to the K-x), BUT try and read the numbers on the lens for both photos. Why is the K-x photo so hard to read? Is it more of the same noise vs. detail loss, blah, blah, blah - pick your poison?? Dunno.... I resized the K-7 photo to the same size as the K-x, and then I did a 600x600 crop at the lens to both photos. K-x 100% crop from the iso 1600 photo: K-7, resized to match K-x, then 100% cropped from the 1600 iso photo: Honestly the K-x photo looks out of focus in these examples. As far as actual noise is concerned, they appear pretty close to my eye, but with more chroma noise in the K7 shot - I would definitely take the K-x photo out of these two examples (OOF not being considered)

10-29-2009, 12:47 PM	#32
Sew-Classic Senior Member Join Date: Jun 2009 Location: Ohio, USA Photos: Gallery Posts: 238	Originally posted by pxpaulx Honestly the K-x photo looks out of focus in these examples. As far as actual noise is concerned, they appear pretty close to my eye, but with more chroma noise in the K7 shot - I would definitely take the K-x photo out of these two examples (OOF not being considered) You can see the original K-x photo here and here. It looked like it was in focus to me- all except for the lens. weird Here's the K-7 full shot at the same iso from that site: http://fotkidepo.ru/photo/132601/24445pwW2GP2SlT/GZa8syV6SN/412281.jpg

10-29-2009, 01:02 PM	#33
Eruditass Veteran Member Join Date: Jul 2009 Photos: Gallery Posts: 2,207	Leads me to believe the K-x was front focused possibly due to lighting without the K-7's new AF sensor or FF problem. I suggested f/8.

10-29-2009, 02:01 PM	#34
*isteve Veteran Member Join Date: Nov 2006 Location: London, England Posts: 1,187	Falk, I have read through your review, great job. Having already got a K7, I plan to trade my old K20/K10 pairing for a Kx as my backup camera, and the one I use for gigs and street shots with fast primes. Having both cameras is really the best of both worlds. To me this is a superior option to a single FF camera and still cheaper than an A850 or D700. And although the push processing of K7 shadows may reveal some noise, its not much effort to apply noiseware to the shadow areas first as there is little detail in the last 2 stops anyway. This way you can fake another couple of stops quite easily.

10-29-2009, 08:20 PM	#37
*isteve Veteran Member Join Date: Nov 2006 Location: London, England Posts: 1,187	Originally posted by falconeye This is true and many testers including DxO forget that dynamic range should be measured in combination with a much lower resolution. I propose to use the same resolution used for the DoF scale which is about 1.4 Mpixel. Or if people prefer, a more modern HD norm like 1920x1280. E.g., the dynamic range of a black&white text page is whatever be the difference between black and white, e.g., one stop. But still, when printed out and hung on a wall, one may reveal that all the tiny letters combine into a large artwork and you see another image, not letters anymore. And this artwork actually may have great dynamic range. Just at a coarser level of resolution. So, dynamic range is a concept not well understood. It reminds me at the depth of field confusion You are the first person to bring this up and of course you are right. Measured DR (the way DXO does it) is largely academic in relation to achievable subjective DR (the range of patches that a camera can actually record in RAW mode which is resolution independent and can therefore be normalised for noise as in your rather neat demonstration ). In this sense, Phil Askey's measurement technique is the most useful if not the most accurate. Sadly, he only seems to apply it to JPEG files. The K7 does fine even with minimal NR (Silkypix Default) as in this sample... http://ic2.pbase.com/o6/89/257389/1/115505969.bo4n95rr.RAWdr.jpg

10-29-2009, 09:35 PM	#38
tr13 Veteran Member Join Date: Jul 2009 Location: Russia Posts: 343	falconeye You could try to repeat measurements with DISABLENR set to 1 for both DSLRs. And look is RAW is now not affected bvy noise reduction.

10-30-2009, 02:38 AM	#39
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871 Original Poster	Originally posted by tr13 You could try to repeat measurements with DISABLENR set to 1 for both DSLRs. That is a good suggestion. I'll try to follow it with my next test. Another reason to retest the K-x (the other reason being to test it at -10EV ).

10-30-2009, 12:27 PM	#40
jamesm007 Veteran Member Join Date: Jul 2008 Location: Detroit MI, USA Posts: 508	The K-x does use RAW NR starting at ISO3200 as the K-7 this NR is twice as strong as used on the K20D at ISO3200 and probably much stronger than the K-7. This of course does not answer why your ISO1600 pic from the K-7 came out the winner. Try the test at ISO3200 and see what happens. The K-x does measure high DR but its also using much stronger NR. Remember the D300 was not as good per the numbers at DXO versus the D90 and D5000 it was becuase mainly those two were doing things to the RAW data such as RAW NR at ISO200!

10-30-2009, 12:50 PM	#41
rawr Veteran Member Join Date: Jun 2009 Posts: 11,913	Originally posted by jamesm007 The K-x does use RAW NR starting at ISO3200 as the K-7 this NR is twice as strong as used on the K20D at ISO3200 and probably much stronger than the K-7. Huh? I seriously doubt that information about the K-x's use of NR in RAW. And even the statement about K-x/K7 NR being twice as strong as K20D seems wildly off the mark. No one who has looked into the K-x and K7 or even the K20D RAW has drawn the same conclusions.

10-30-2009, 03:34 PM	#43
jamesm007 Veteran Member Join Date: Jul 2008 Location: Detroit MI, USA Posts: 508	Originally posted by rawr Huh? I seriously doubt that information about the K-x's use of NR in RAW. And even the statement about K-x/K7 NR being twice as strong as K20D seems wildly off the mark. No one who has looked into the K-x and K7 or even the K20D RAW has drawn the same conclusions. Its a well known fact that Pentax uses NR in RAW files at high ISO, DXO found this and penalized Pentax for it. DxOMark review for Pentax cameras A quote from DXO "Noise filtering processing is present on the Pentax K10D (ISO 800 and ISO 1600), the Pentax K20D (ISO 3200 and ISO 6400), and the Pentax K200D (ISO 800 and ISO 3200). As the Low-Light ISO scores for the Pentax K10D and Pentax K200D changed slightly after taking noise filtering into account, their global dxomark.com rankings changed as well. However, Pentax K20D Low-Light ISO scores were not impacted by noise filtering." They made an error for the K20D its NR starts at ISO1600 not ISO3200. I found high levels of chroma NR in the K-x samples I have very few now but just judging them I could see very high levels of NR at ISO3200 to the point of robbing detail; this can change if I see other samples. Gordon B Good a sensor scientist at drpreview confirmed my findings and has it all measured; basically the K-x uses twice as much NR at ISO3200 versus the K20D, if you where to open to identical normal scenes the K20D would have and does have a finer noise pattern versus the K-x you can begin to see it affects in the pics above but according to Gordon B Good the K-x does not apply NR at ISO1600. Just for your info in RAW the K10 applies NR starting at ISO800, the K20D at ISO1600, the K-7 at ISO3200, the K-x at ISO3200. Gordon B Good also discovered some interesting things in Nikon's RAW files, Gordon B Good has found 90% for sure NR in RAW at base ISO (200) with the Nikon D90 and D5000. Ever wonder why they would beat the D300 in DXO? Because it was better sensors, lol nope because Nikon tricked DXO the D5000 and D90 do not have more DR than the Nikon D300 at ISO200 they have less, and less than the K20D. There is no one reliable source of info for sensors from PRO Web sites as falconeye says (maybe his blog can be the source), not DXO, not dpreview, not..., you did not even know the Pentax K20D applies NR in RAW starting at ISO1600!!! its a very small amount however. Now people still keep knocking the Samsung sensor the one at least in the K20D may be one of the best high ISO APS sensor out at most second by a small amount to the D300 I lean toward the K20D Samsung being the overall best APS sensor (have not seen enough of the Canon 7D yet); it all depends on what light level you test, if you test at 18% grey or pics like the queen at dpreview the K20D takes the win as dpreview shows us versus the D300, and it would also beat the K-x, Now in real dark tests the K-x with its better dark read noise, or the D300 would be better. I think a standard measure of high ISO noise, really 3 standards should be made. Low-light, mid-light, high-light. And now a quote for sensor scientist Gordon B Good who has known for years of non-defeatable NR in Pentax RAW files and some othe folks who are really into sensors. It is a subject not talked about much and easy for me to understand how and why you or anyone would not know "James, as I already posted further up the thread, the K-x does apply raw Noise Reduction (NR) to the K-x raw data for ISO's of 3200 and up, and I measure from the French contributor's posted raw files that this raw NR is a little bit heavy handed to reduce the noise by a factor or two in standard deviation. I don't believe that any NR is applied to the raw data for ISO's of 1600 and lower. This NR applied by default (can't be turned off) by the K-x to high ISO raw data looks to be about twice as strong as that applied to the raw data of the K20D for ISO's of 1600 and above, as in the K20D only reducing the standard deviation of the noise by a about a factor of about 0.7 instead of by a factor of 0.5 for the K-x. Of course, it is this fairly heavy handed raw NR that makes the ISO 6400 and 12800 reasonably usable without further application of NR during JPEG processing. The strength of the NR applied to the K-x raw data for ISO 3200 and above is about the same as the NR applied to the K10D, my K200D, and the K-m/K2000D for ISO 800 and above, so it isn't all that bad! I suspect that one could make very good prints from even ISO 3200 to 6400 up to about 8 X 12 inches from the K-x which is amazing in that I wouldn't want to print this size from anything much over about ISO 800 for my K200D (equivalent image quality to the K-m/K2000D). Remember, if this raw NR weren't applied by the K-x, then there likely wouldn't likely be a ISO 12800 option as otherwise the noise would make it look fairly bad. Pentax has to try to tread the fine line between keeping the fine-detail-at-full resolution junkies such as yourself happy as well as the ever increasing drive to make higher and higher ISO's somewhat usable, and for this entry level camera they seem to have felt that a little loss of detail is a worth while trade-off in order to have reasonably clean looking high ISO images"

10-30-2009, 04:16 PM	#44
rawr Veteran Member Join Date: Jun 2009 Posts: 11,913	Hmm. Interesting. Any K-x NR above 3200 is very subtle then, since lots of usable detail remains even at 12800 ISO, whilst still retaining good colour sensitivity and tonal range. I await DXOLabs eventual examination of the K7 (and K-x) with even more interest now. PS: I don't think Nikon would agree with GBG's views about their 'sensor tricks'.

10-30-2009, 05:48 PM	#45
jamesm007 Veteran Member Join Date: Jul 2008 Location: Detroit MI, USA Posts: 508	Originally posted by rawr Hmm. Interesting. Any K-x NR above 3200 is very subtle then, since lots of usable detail remains even at 12800 ISO, whilst still retaining good colour sensitivity and tonal range. I await DXOLabs eventual examination of the K7 (and K-x) with even more interest now. PS: I don't think Nikon would agree with GBG's views about their 'sensor tricks'. The neat part is you can see in pics anything I say or GBG says. I can see very course noise and lack of detail in the deep shadows in pics from the D5000 and D90 compared to the D300 at ISO200. The D300 had a hands off approach to RAW PP in the EXPEED engine remember the respective image engines from different manufactures as the PRIME for Pentax can just as easily work with RAW as JPEG in fact they do. So I see with my own eyes more noise and smoothed over details in the Nikons D5000 and D90 compared to the very fine D300; its there 90% for sure according to me and lots of others. I will try to post for you some pics with examples as I see your interested and kinda don't know what to believe which is good. Do some searches on dpreview for Gordon B Good, and you should along with his posts find some of my work occasional, that I submit to him and I my thoughts as to whats going on, for his analysis I just look and try to figure out with help on how to analyzie pics he has freely given to others and use the tools he says to use. The trick is the RAW converter its needs to be a neutral one like Raw Therapee, ACR and many others will add NR even with the sliders at 0, RT does not. So I can import DNG (files ran through Adobes DNG converter) open them in RT and compare areas of my interest. This is how I found high course NR in the K-x ISO3200 pics. BTW DXO is not 100% accurate they have yet to catch Nikons NR in RAW and many of the results are skewed. There is no one single source of accurate data for questions of performance we want answered. Remember the D90 according to DXO was much better than the D300 in many ways yet when I looked at real pics with a neutral converter I saw the D300 cleaning the D90s house. If interested keep reading, Gordon B Good has written programs such as a program to remove tint from the corners of the K20D at high ISO. GBG also writes his own raw converters he has some serious sensor knowledge. Its healthy for you to be skeptical however. You know after all this heavy pixel peeping of many different cameras and lots of tests I have come to one conclusion about APS sensors, I would use any of todays APS sensors and not feel bad. If given a choice I prefer higher-mega pixels because I am a cropper, but as far as IQ at normal viewing sizes there is not much to talk about.