[JinDesu]

You should warn folks that that is a 22MB file....
When I processed the 6400 ISO DNG and exported it at 82% reduced JPEG (plenty good enough) it was still big, but only 7.85MB.

It would also be handy to show what settings you used for NR and sharpening on the LR sliders. I got good results with NR Luminance 40, NR detail 60, Sharpening 55, Masking 40 on the 6400 ISO DNG, as well as a few tweaks to clarity and exposure.

Incidentally, I also found that one of the reasons the Pentax JPG engine muddies up the reds on the swatch may be that it seems to read the dark pattern lines in the red as colour noise, and over-aggressively try to reduce that. By bumping up the colour noise reduction slider in LR you can just about replicate the effect.

Ah apologies - I just threw the RAW files into LR5 and just exported them full size and max quality.
I just got back from a trip so I was a bit tired and didn't want to play with the files. The jpegs are not edited in any way with regards to detail/NR.

[wahid_satu]

Edit - I found the RAWs for the other cameras - linked here: D7100 LR5 Edited: http://farm4.staticflickr.com/3787/1...74765b08_o.jpg k-5IIs LR5 Edited: http://farm6.staticflickr.com/5526/1...40b17387_o.jpg D600 LR5 Edited: http://farm8.staticflickr.com/7366/1...6a5516c7_o.jpg Looking at the RAW processed jpegs, my final take at ISO 6400 is: 1) The k-3 has less noise than the D7100. The k-3 retains similar detail to the D7100, with different color saturations. 2) The k-3 shows more noise than the k-5IIs. The k-3 shows greater detail than the k-5IIs. Color is about the same. 3) The D600 shows the least noise. The D600 also appears clearer/more detailed, but it is also a bit brighter than the k-3 image. Hope this is helpful!

I think you nailed it.

BTW, where is the link to the raws of the other cameras? I'd like to try them out with my weapon of choice: Rawtherapee.

Thanks

[rawr]

where is the link to the raws of the other cameras?

Just go to the Imaging Resource web site, then go to the review of each camera. There should be a 'Samples' tab, and from there a link to a 'Thumbnails' page, like this one for the D7100:

Nikon D7100 Review - Thumbnails

Buried in there are RAW files for many of their test images.

Same story for D600, K-5, D800 etc.

[JinDesu]

It would also be handy to show what settings you used for NR and sharpening on the LR sliders. I got good results with NR Luminance 40, NR detail 60, Sharpening 55, Masking 30 on the 6400 ISO DNG, as well as a few tweaks to clarity and exposure.

To update:
ISO 6400 with noise reduction performed on the RAWS:

D600: http://farm8.staticflickr.com/7388/10583745535_94e215ee3b_o.jpg
Constrast: 5
Clarity: 35
Vibrance: 10
Saturation: 10
Red Saturation: -10
Sharpening amount: 35
Radius: 1.0
Detail: 35
Masking: 10
Luminance NR: 55
Detail: 80
Color: 35
Detail: 80

k-3: http://farm8.staticflickr.com/7420/10583764306_c6333f199b_o.jpg
Constrast: -3
Clarity: 25
Vibrance: 10
Saturation: 10
Red Saturation: -15
Sharpening amount: 20
Radius: 1.0
Detail: 20
Masking: 15
Luminance NR: 55
Detail: 65
Color: 55
Detail: 80

D7100: http://farm3.staticflickr.com/2871/10584008073_74ea48c12c_o.jpg
Constrast: 10
Clarity: 45
Vibrance: 10
Saturation: 10
Red Saturation: -15
Sharpening amount: 25
Radius: 1.0
Detail: 25
Masking: 10
Luminance NR: 65
Detail: 70
Color: 55
Detail: 70

k-5IIs: http://farm4.staticflickr.com/3751/10583794264_dd23b3318d_o.jpg
Constrast: -3
Clarity: 35
Vibrance: 10
Saturation: 10
Red Saturation: -15
Sharpening amount: 35
Radius: 1.0
Detail: 35
Masking: 15
Luminance NR: 45
Detail: 65
Color: 55
Detail: 80

The reason why the settings are different are two-folds: I tried to edit the images to provide as much detail and reduce as much noise before artifacts pop up. WIth the k-3 RAW files, this was actually pretty difficult. I'm not sure if it's because my LR5 doesn't have the appropriate camera profile, but the noise started showing artifacts pretty early.

Thoughts after post-processing:
-The D600's noise cleans up very well and easily at ISO6400.
-The k-3's details in the fabrics is not retained as much as the D600's after noise reduction.
-The k-3's details are about equal with the D7100 at ISO6400. The D7100 seems a bit clearer in the scale on the right side, but that could just be focus differences. The k-3 resolves the words on the tiny thing on top that says "Faces Cameras - Pat Pending". The D7100 doesn't seem to resolve some of the words.
-The k-3 seems to have more noise per pixel than the k-5IIs, as expected. Sharpness in the scale seems equal visually - but since the k-3 has more pixels that's more detail there. Fabrics follow the same thought.
-The color table in the bottom: The D600 looks most consistent and un-oversaturated. The D7100 follows suit. The k-3 has more saturation in red and pink. The k-5IIs matches the k-3.
-The k-3 shows the salt better than the D7100.
-The k-3 image is marginally darker than the D7100, and there appears to be more shadows.

Also, these files lag my LR5 pretty hard!

[rawr]

Also, these files lag my LR5 pretty hard!

Thanks. Very interesting work.

In my own brief play with some K-3 DNG's yesterday, I must say I was quite happy with how well the image detail held together even as the luminance NR slider moved up, even with the 6400 ISO DNG.

However, after working with K-x and K-5 RAWs for so long now, the K-3 RAWs seem to behave very differently from what I am used to. Maybe that's because the tools I've used before (LR for example) haven't yet been profiled for the K-3 yet. Maybe it's just the new Prime III imaging engine at work. Who knows. But it seems working with the K-3 RAWs may require a new approach compared to working with K-5 RAWs.

Yet another learning curve ahead ...

[JinDesu]

Thanks. Very interesting work.

In my own brief play with some K-3 DNG's yesterday, I must say I was quite happy with how well the image detail held together even as the luminance NR slider moved up, even with the 6400 ISO DNG.

However, after working with K-x and K-5 RAWs for so long now, the K-3 RAWs seem to behave very differently from what I am used to. Maybe that's because the tools I've used before (LR for example) haven't yet been profiled for the K-3 yet. Maybe it's just the new Prime III imaging engine at work. Who knows. But it seems working with the K-3 RAWs may require a new approach compared to working with K-5 RAWs.

Yet another learning curve ahead ...

I'm mostly hoping it is the lack of profile information in LR - because otherwise it'll be a little harder for me to work with when I get the k-3. On the other hand... I am planning to get the k-3 sometime next year so maybe it'll be ironed out by then!

Also, I used LR5 which is new-ish to me. I just ran the k-3 file through LR3 and the noise reduction tool seems to handle the file a lot better than LR5 - but the bright red fabric lost all detail in LR3 so... it's kinda a toss up.

[wahid_satu]

Thanks RAWR.

This is a comparison of D600 and K-3 jpegs developed with Rawtherapee, using Amaze demosaic algorithm. No correction whatsoever (neutral profile.) I'd like to think this is as raw as it can get, since from what I know, Rawtherapee does not apply camera specific profile unless we choose to, CMIIW. We can see that the D600 is underexposed in accordance to its faster shutter speed (its histogram chart is skewed to the left.)

The K-3's noisier of course but not as much as one might think. However seeing as the noise pattern is more blotchy, it seems, as usual, there is raw applied NR on the Pentax file (not saying that it is bad practice.)

Attached Images

 

[wahid_satu]

Moving on to D7100 vs K-3, again we see the Nikon's image is underexposed. Due to the finer grain, I think I prefer the D7100's than the K-3's which is a little sad.

This is purely academic, of course, as we will never see images as raw as these in real world.

[rawr]

This is a comparison of D600 and K-3 jpegs developed with Rawtherapee,

Moving on to D7100 vs K-3,

Nice work!

It's interesting to see the Nikons are both underexposing. It's also good to see, as expected, that the red swatch has it's details and colours reproduced as accurately in the K-3 as in the Nikon's in the RAW.

It's also useful of course to keep reminding ourselves that these images are all 100% and 6400 ISO! Each camera is doing a very good job holding onto colour and detail despite this extremity of ISO.

The underexposure displayed in the two Nikons may relate to the tendency noted in DxoMark.com for those two Nikons to overstate their reported ISO compared to their actual, measured ISO - ie Dxomark says the measured ISO of the D7100 is actually about ISO 4500 when the manufacturer says it is 6400.

[weathermon]

I've put together a montage of 100% crops from ISO100-25600 for the coffee cup section of the image to show the noise in the shadows - quite impressive! Processed in LR 5.2 with 50% NR on all images. The file is 25mb

https://www.dropbox.com/s/b6e0b1ik4eklf3o/k3isotest.png

Click the download link at the top right of dropbox to get the original

[rawr]

I've put together a montage of 100% crops from ISO100-25600

Looks good.

Were these done from the IR JPGs, or from DNG's?

Either way, LR 5.2 did well. I suspect the coffee cup is probably an easy target for luminance NR though, given it's consistent colour and regular surface.

[ElrondElensar]

Nice work!
The underexposure displayed in the two Nikons may relate to the tendency noted in DxoMark.com for those two Nikons to overstate their reported ISO compared to their actual, measured ISO - ie Dxomark says the measured ISO of the D7100 is actually about ISO 4500 when the manufacturer says it is 6400.

That will not explain anything here as Nikon cameras are using faster shutter speed. If they are underexposed, that has nothing to do with dxomark ISOs but with faster shutter speed.

[weathermon]

Rawr These were from the DNG's

[Vylen]

I'm liking what I'm seeing.. ISO 6400 is probably the upper limit to what I'd use if I can help it... and that means a lot - my personal benchmark is a K20D which was horribly destructive at ISO 800 :P

[rawr]

That will not explain anything here as Nikon cameras are using faster shutter speed.

In these examples, aperture is fixed, and nominal or 'manufacturer' ISO is too. So the accuracy of the Nikon ISO metering may indeed have an influence on the cameras calculations about which shutter speed to employ under the lighting conditions in the IR test scene ...