K-3 studio test shots now on Imaging Resource

I was comparing the K-3 at ISO 51,200 with the K-5IIS at ISO 51,200. The K-5IIs worked OK with the Lightroom default "Color" noise reduction value of 25, but the K-3 didn't.

When my K-3 arrives in a week or so I will have to experiment to get a feel for how hard I can push it within the range that Lightroom can handle. My K-5IIs will become my back-up camera, always reasonably accessible, so if the K-3 turns out not to be as good in this case as the K-5IIs I won't be too upset, because I will have the latter to fall back on. But I would obviously prefer to be able to stick with just my K-3.

Colour noise reduction is far less destructive than luminance noise reduction, so having to crank it up shouldn't affect the images.

However, it's overall high ISO noise more than anything else that interests me, so it looks like I'll be skipping yet another model in the Pentax line-up. Hopefully the rumours of a FF digital LX are true.

I suppose the really important thing is not so much the exact numerical value of any processing correction, but rather how well the image responds to the processing - how easily the noise disappears and what after-effects remain.

Each different model may not have the same starting points (numerically) for any given processing parameter, or the reduction effect may not kick in until further up. I woudn't be concerned if the K3 needed 35 or 40 of ACR colour noise reduction, so long as there was no increased artifacts as a result of that input.

I'd question any assertion that if you need more NR input numerically for a particular sensor / camera model then the noise performance of that camera must be inferior. As I discovered with the K5II/s comparisons, actual pixel data responds differently and at different amounts of adjustment when the starting point of that data is different.

There are always increased artefacts as a result of increase NR settings. personally I prefer to have a bit of noise than the "hideous 10th generation VHS copy" smearing effect you get from too much NR

Could be that Lightroom needs an update to understand how to deal with K-3 files. The camera is not officially out in the wild just yet.

I'm still using Lightroom 3.6, and it handles the K-3 DNGs well. (In fact, it handles my K-5IIs DNGs well, even though it was released before the K-5IIs). My Lightroom is currently using the K-3's DNGs' embedded profiles.

I expect Adobe will eventually create their own camera profile for the K-3, but I never use Adobe profiles. I'll create my own for the K-3, just as I have for all my previous cameras.

Consider upgrading to LR4 or 5. Noise reduction is much better, as is the overall processing.

OK, downloaded the ISO 51200 DNG's of K3 and K-5 from Imaging resource....

Imported into Lightroom 4.2 and applied my standard NR settings for ISO52100 to each. I have to say the K-5 beats the K-3 hands down. And it's not just colour noise. the colour noise setting in LR4.2 on these 21/100. Even wacking it up to 100/100 on the k3 version, still left loads more luminance noise than the K-5 version.

The next step was to increase the luminance setting ( which is on 47/100 for these examples ) so that the K-3's luminance noise levels roughly equalled that of the K-5. Somewhere around 70/100 did the trick but left that hideous 69th generation VHS copy smear that high NR gives.

So far I conclude that the K-3's ISO performance has suffered because of the marketing need to keep up with the MegaPixel race. I hate this damn MP race, and would rather have better IQ than more MP. However before making a solid conclusion I aim to repeat his test at mid range ISO values, i.e ISO6400. I wager it will be the same proportional difference given Dynamic range/ISO specs are pretty linear. i.e you roughly lose 1 stop of dynamic range for 1 stop of ISO



Here's the full image and 100% pixel peep versions for each ( click on pics to avoid annoying dinosaur 800px restriction )


K3 ISO51200 - full pic





K5 ISO51200 - full pic





K3 ISO51200 - 100% pixel peep





K5 ISO51200 - 100% pixel peep

Does anybody actually shoot at these iso levels anyway? I don't think I've ever gone above 6400 and even then incredibly rarely.

Totally agree

This, as an argument for not worrying about the very high ISO settings, is putting the cart before the horse.

The reason we don't go above ISO 6400 or whatever is because the result is bad. If it wasn't, surely we would?

By the way, Smeggy, if you have time, please would you compare the still lifes from the K-3 and D7100 at a suitable range of ISO values?

Process 2010 was introduced with Lightroom 3, not 4. My understanding is that noise reduction, which was dramatically improved in Lightroom 3, is pretty much the same since then. (But the Basic panel was very much improved in Lightroom 4).

Yes, I will be upgrading to the latest versions of Lightroom and Photoshop when I have upgraded my Victorian-era Windows XP PC to something more recent! Thanks.

The 100% images result in different magnifications with the K-3 and K-5, thus introducing another variable.

What would we conclude the significance of that would be?

Unfortunately, that agrees with my own assessment. I was hoping I was wrong. (But I think up to about ISO 6400 there is much more similarity between the K-5IIs and the K-3).

I will soon know. The latest Track & Trace for my K-3 reads:
31-10-2013 - 06:32 - Manchester Piccadilly Depot - Loaded to vehicle for delivery