[jgredline]

Well at least I am not getting the green line.

I exposed my camera for 73mins and no lines.

Only I am encountering comes up on videos under certain lighting conditions, but so far nothing under regular stills, even after LONG exposures.

So it is not completely green line less?

[Christopher M.W.T]

So it is not completely green line less?

It's actually a pinkish line and ONLY shows up in video mode, when you had a wide open apperture, you lock the exposure on something really bright, then you put the camera in the dark, it's clearly visible.

Thats the only time I get unwanted artifacts, in my still images there is nothing and i've even been trying.

[Christopher M.W.T]

Another shot, taken at f2.8, 40mm @ 1000ISO, minor NR applied in PS.

[pentaxian_tmb]

I'll play


Exposure: 0.05 sec (1/20)
Aperture: f/5
Focal Length: 16 mm
ISO Speed: 1600
Exposure Bias: 0/10 EV
Flash: Flash did not fire

[Marc Sabatella]

That seems to be the consensus about the K7 vs the K20.

I hope it is something Pentax can improve via firmware. If it is a problem tied to the hardware directly that the firmware can't improve, that would be a big pity.

Actually, the consensus from people who have extensively analyzed the contents of the RAW files is that the reason the K20D appears to do a little better at high ISO is that it actually performs noise reduction on its RAW files whereas the K-7 doesn't. Doing the NR yourself in PP - or even using the NR options in the JPEG processing - pretty much eliminate the differences. Given that many people were upset that the K20D performed NR on its RAW files, I think most of them are happier witht he approach taken by the K-7.

[Christopher M.W.T]

I'll play


Exposure: 0.05 sec (1/20)
Aperture: f/5
Focal Length: 16 mm
ISO Speed: 1600
Exposure Bias: 0/10 EV
Flash: Flash did not fire

Thats freaky, at my old high school there is a chapel that look 100% identical to that.

[rawr]

Actually, the consensus from people who have extensively analyzed the contents of the RAW files is that the reason the K20D appears to do a little better at high ISO is that it actually performs noise reduction on its RAW files whereas the K-7 doesn't.

Hmmm. That's not how I am reading the 'consensus'. (We sound like politicians!) Either way, at the end of the day the K7 out of camera RAW is visibly different from the K20D out of camera RAW.

Going further in the discussion is challenging due to the variables involved in assessing RAW - how it is recorded, how it is rendered etc - which is probably out of scope in this thread.

[Marc Sabatella]

Hmmm. That's not how I am reading the 'consensus'.

Depends on who you listen to - the folks who just look at a few images and see the K-7 doing a little worse but don't understand why and therefore start complaining about it, or the folks who have looked into the RAW data itself and seen for themselves that the difference is the NR being done to the K20D RAW files. That's not something everyone knows. Most who complain about K-7 noise relative to the K20D don't realize that it is an apples-to-oranges comparison, because one has NR applied and the other doesn't.

[jgredline]

Regardless of how the high ISO images are processed out of the K20D and K-7, All I know is the K20D is better.

[xixco]

Based on the look of it, I probably did some PP, but nothing in terms of exposure compensation. I remember being really pleased that I finally seemed to get the hang of Manual exposure in crappy lighting like that. Manual focusing on the other hand...

135mm 2.8, probly wide open
1/25 sec

[rawr]

That's not something everyone knows. Most complains about K-7 noise relative to the K20D don't realize that it is an apples-to-oranges comparison, because one has NR applied and the other doesn't.

I think you are too confident on this point. Merely repeating a statement a lot doesn't make it true

Technically, no one can tell what is going on unless they have access to the K20D and K7 firmware source code, and/or the sensor microcode, and/or any other information about the way the K7 sensor and associated circuitry works compared to the K20D's. No one here or on dpr has any of that information. And looking at the hex code or metadata of the RAW file itself won't tell you. So claiming the K20D's RAW output has NR in it whilst the K7's doesn't can only be done, I think, through some process of deduction based on looking at the RAW output of the cameras.

Making these deductions on the basis of visual output is difficult though because you are always looking at RAW output that has been rendered by a RAW processor (ACR, dcraw, Capture One, Silkypix, whatever) that usually applies its own 'personality' to the RAW output of any camera, depending on what it might decide is appropriate for what it knows of a particular camera model or not, or other decisions of the RAW processor programmers about interpreting stuff like the DNG specifications (in the case of RAW output in that format) etc etc.

[Gooshin]

FA43 Limited
F2.8
1/15

ISO 3200

[Marc Sabatella]

I think you are too confident on this point. Merely repeating a statement a lot doesn't make it true

Technically, no one can tell what is going on unless they have access to the K20D and K7 firmware source code, and/or the sensor microcode, and/or any other information about the way the K7 sensor and associated circuitry works compared to the K20D's. No one here or on dpr has any of that information. And looking at the hex code or metadata of the RAW file itself won't tell you.

I think you'd be surprised at what you can tell just by looking at th RAW data itself. NR algorithms will leave a tell-tale trace on the data that you can see by atching what happens to the actual standard deviation of the pixel values as you increase ISO (I *think* that's the basic idea of process, anyhow - probably oversimplified). People who make it their business to analyze this kind of data are pretty confident on this point. We're talking about some amazingly astute individuals putting a *lot* of effort into this task here, not some random people just sort of guessing.

Making these deductions on the basis of visual output is difficult though because you are always looking at RAW output that has been rendered by a RAW processor (ACR, dcraw, Capture One, Silkypix, whatever) that usually applies its own 'personality' to the RAW output of any camera, depending on what it might decide is appropriate for what it knows of a particular camera model or not, or other decisions of the RAW processor programmers about interpreting stuff like the DNG specifications (in the case of RAW output in that format) etc etc.

The people I am talking about have written their own software to analyze the data, and at least one of them ("GordonBGood") has created his own utility to demosaic the data himself using *no* additional "personality". Like I said, we're talking about going *FAR* beyond simply looking a some pictures from a geneic RAW converter and guessing.

I think unless you've looked at the data yourself with the same degree of scrutiny and have some specific numeric arguments to counter the very convincing analysis by Gordon and the others who have partaken in this analysis, there is really no reason to doubt their findings.

Bottom line: if K-7 images using default settings look noisier than K20D images using default settings, it is primarily because there is no NR applied to the RAW data on the K-7. Meaning that when you add the NR yourself in PP - or even perhaps with the in-camera JPEG's if you don't care as much about IQ at this level - you should be able to get results that are more or less exactly as good from the K-7 as from the K20D.

[rawr]

I've followed (as best anyone can given their length) the discussions GBG has participated in about the K20D (and the D90/D5000/D300 too) vs the K7 and all I will say is that the man is a saint.

He keeps getting badgered by people who are fervently convinced of various issues to confirm their views, and asked to provide black/white responses based on their latest fuzzy snapshots, or the latest image file someone found on a web site in Poland, but he patiently, politely and cautiously replies to all of their relentless hectoring. Amazing.

But unless I've missed a pronouncement of his somwhere within the morass of the DPR forum, nowhere does he emphatically declare that the K20D, or the Nikons for that matter, add NR into their RAW (whatever that may in practice mean).

He can't be conclusive because he doesn't have the engineering data and he doesn't work at Pentax. Aside from perhaps Adobe or DXO Labs, few people have the resources to seriously analyse these issues and reverse engineer what is going on.

[Marc Sabatella]

But unless I've missed a pronouncement of his somwhere within the morass of the DPR forum, nowhere does he emphatically declare that the K20D, or the Nikons for that matter, add NR into their RAW (whatever that may in practice mean).

I dont know that he's said anything about Nikon, but he *has* said this about the K2D, on several occasions. See, for example:

Re: K7 vs K20D direct comparison: Pentax SLR Talk Forum: Digital Photography Review

Relevant quote:

Oleg_V and I proved long ago that the K20D applies Noise Reduction (NR) to the raw data for ISO 1600 and above, where it reduces the noise magnitude at ISO 1600 to about 70%. The K-7 does not do this noise reduction at ISO 1600 although it does at higher ISO's. When I was looking at the black read noise, it was already so high at ISO 1600 that I lost interest in higher ISO's such as 3200. With an equivalent NR algorithm, the K-7 images at ISO 1600 can be made to look like the K20D images at the same ISO sensitivity.

Note he says "proved", not "guessed", and puts specific numbers to it. He is looking at the actual data, measuring standard deviations and other statistical stuff that goes way over my head (although back when I was a math major in college, I'm sure I'd have followed it better!).