-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
Banding issue with K10D
- Thread starter CapitaineAbitibi
- Start date
... the band, not banding... Come on, photographers, what is gone wrong- years ago we sat in darkroom and were happy and now without slightest willing to sit as long in lightroom (bibble, ACR etc.) we prefer pixel-peeping... shame on these peep-shows and let us all get stuck to the real world and capture the emotion, light, colors etc.- IMO it is a PHOTOGRAPHIC community not TECH whining- OK issues MUST be mentioned and discussed... but to some reasonable extent IMO :-(
Best and happy SHOOTING, JR
Best and happy SHOOTING, JR
*isteve
Veteran Member
I though EXIF said ISO800?
--
Steve
Pixel peepers miss the big picture.
http://www.pbase.com/steve_jacob
Yeah, we had ASA 400 film noisy as a punk rock concerts. But it was uniform noise. Now we have pattern noise. Why should we be happy?
Pattern noise is not easy to remove without removing details. Or teach us the opposite. Just claiming that it can be removed is not helpful at all.
Especially when it was proven that the pattern noise is at least partially to a design error on Pentax's side.
No. Noone does. It is just a matter of fact that scrutiny is now easier for the average Joe&Jane than it was with film. Asking questions is the human nature. Ignoring problems is, too.
Yeah, capture it, forget the noise bands. Who cares if they ruin an otherwise mediocre shot?
dpreview is a tech-community. Love it or hate it. Some of the less-cat-picture ladden dSLR forums here on dpreview don't want picture posts that serve no purpose than show pictures. I don't say that's better, but as a matter of fact, these forums were intended as a technical resource.
And you decide what is reasonable? Do you think this is reasonable.
Best wishes and an open-minded discussion:
For example, we still do not know how these images looked straight from the camera. A very essential point. Without it, it is all speculation.
Jens
--
'Well, 'Zooming with your feet' is usually a stupid thing as zoom rings are designed for hands.' (Me, 2006)
http://www.jensroesner.de/
--=! Condemning proprietary batteries since 1976 !=--
Jonas B
Forum Pro
Right, correct, obvious. I'm a victim of constant confusion. I don't know how I got it to ISO100. That also makes half of my post above pretty odd. Exchange my ramblings about ISO 200 and 400 with ISO 1600 and 3200 instead.
The pictures being ISO 800 pictures doesn't change this a whole lot though. If we had that noise level at ISO100 I really would have been worried.
Thank you for correcting me before I continued on the same track!
--
Jonas
Jonas B
Forum Pro
Some misunderstanding here... This is a forum for discussing technicalities and related things on the topic Pentax dSLR cameras.
"TECH whining" Lol. There are a lot of different pictures here, luckily. But don't try to change the basic idea about the forum. There are several other places where pictures are the primary target for the discussions.
--
Jonas
*isteve
Veteran Member
OP said ISO 100 - I think it was a misprint. I would have been surprised too.
Dont know if you read this...
http://www.bythom.com/d80review.htm
Scroll down to the "noise" section. ISO1600 at 1/15 pushed using autolevels - the noise structure is the same so this is obviously a sensor issue. Nikons chroma NR probably reduces it somewhat but its still there.
Interestingly, silkypix chroma NR seems to work better than ACR. Dont know if its worth trying different RAW developers to see which works best for banding supression?
I would try but I have no decent examples. Im sure I'll find one at some point.
--
Steve
Pixel peepers miss the big picture.
http://www.pbase.com/steve_jacob
GordonBGood
Veteran Member
Well, Steve, not exactly a sensor issue but rather a combination of the features the sensor offers and the necessary processing to offset dark levels.
In short, the problem is that the 10 MP CCD sensor doesn't have a very wide masked band across the bottom of the sensor (only eight photosites high) and "vertical pattern noise" results because the camera algorithms average too few photosites to completely eliminate random variations in their "on-the-fly as acquired" black level offset algorithms.
In the case of the K10D, this is made even worse because they use a facility of the NuCORE NDX 2240 AFE chip to treat all colour channels separately and this results in them averaging even less dark level photosites - only 32 per colour to be exact. This is insufficient to eliminate small variations in offset that are systematically applied to columns as dark offset values, causing the vertical pattern noise we see when push processing images, even as low as ISO 100. This noise can be as little as one raw output level in size, but being systematic still causes a recognizable vertical pattern when push processed.
It appears that Nikon D80 has some problem with this as well from Thom's posted review (via your link), but it may well be less if they average more photosites to obtain a better reduced-noise average for black offset per column.
The best way of handling this would be to preserve enough data on the subtraction applied to be able to correct the pattern in post processing the raw data after it has been acquired. The Sony Alpha 100 may do this, as I have not been able to see any vertical banding with that camera (although it has other noise problems).
Looks like I need to prepare a thread on "How the K10D works (and doesn't)" to fully explain this.
Regards, GordonBGood
Jonas B
Forum Pro
Joshua Hakin
Well-known member
Hey Captain!
I have the same problem of banding at 1600 when shooting musicians/bands.
The shadows show it up really bad... which is the majority of the frame when shooting in low lit pubs and such.
You are not imagining it no matter who says there's no problem.
Like you said, your Ds can handle extreme post processing, and same with my D. I've used it extensively at 3200 and could process the hell out of it! I cant even shoot a properly exposed image with K10D at 1600 without noise pattern/banding.
And I've tried all sorts of software: NeatImage, Noise Ninja, special Photoshop tricks... nothing works. You need a priest to get rid of this thing!!!
I've resigned myself to the fact that I can't use the K10D to shoot in low light for musicians, I'll have to use my istD until Pentax produces some sort of magical fix.
--
Joshua Hakin Photography
http://www.hakinphoto.com
I have the same problem of banding at 1600 when shooting musicians/bands.
The shadows show it up really bad... which is the majority of the frame when shooting in low lit pubs and such.
You are not imagining it no matter who says there's no problem.
Like you said, your Ds can handle extreme post processing, and same with my D. I've used it extensively at 3200 and could process the hell out of it! I cant even shoot a properly exposed image with K10D at 1600 without noise pattern/banding.
And I've tried all sorts of software: NeatImage, Noise Ninja, special Photoshop tricks... nothing works. You need a priest to get rid of this thing!!!
I've resigned myself to the fact that I can't use the K10D to shoot in low light for musicians, I'll have to use my istD until Pentax produces some sort of magical fix.
--
Joshua Hakin Photography
http://www.hakinphoto.com
GordonBGood
Veteran Member
Joshua, have you tried the new firmware 1.1 upgrade. Although there is still some vertical banding, some have reported that this update reduces its high frequency random component so that the vertical patterns may be less objectionable (and more correctable). But if you are shooting at both ISO 1660 and with fairly slow shutter speeds in the order of a tenth of a second to a second, you are likely still seeing vertical banding/pattern noise that Pentax have not yet fixed. I am continuing to look into the problem and how it might be solved.
Regards, GordonBGood
wll287622
Leading Member
I did not see banding, only noise ?
wll
wll
Joshua Hakin
Well-known member
Hey Gordon,
I've been reading a lot of your posts and want to thank you for the effort you are putting into this, very much appreciated!
I have upgraded to ver1.1 but haven't done much testing except for some shots around the house, gonna do some more tonight.
I recently saw some samples from someone using the K10D at 1600 and the results were far better than anything I've seen so far. He was shooting in JPEG mode Bright with the new ver1.1
I've only been doing RAW DNG's... so I'm wondering (and gonna test a few) if perhaps they included in the firmware some special "Al Gore Rhythm" (lol) with in-camera processing of JPEGs that doesn't take affect with the RAW files... does that make sense? is it possible? I don't how these things work, I'm just guessing.
I've also wondered if it has something to do with the fact that the RAW converter I'm using isn't "tweaked" for K10D DNGs. I'm using AdobeCR ver 2.4 and there's a few weird things that happen with the DNGs from this camera, but nothing big - e.g. when I shoot at WB daylight, the ACR says the images are around 7000K, but the colours looks fine anyway. I do know that Aperture 1.5.2 won't even accept the DNGs yet as it hasn't been updated to support the K10D yet (weird I know).
Perhaps the Raw converter needs to be tailored for this camera. Again, I have no idea about this stuff, I'm only grasping at straws.
But I'm going to compare the JPEG capture against the RAW capture and see what happens.
--
Joshua Hakin Photography
http://www.hakinphoto.com
I've been reading a lot of your posts and want to thank you for the effort you are putting into this, very much appreciated!
I have upgraded to ver1.1 but haven't done much testing except for some shots around the house, gonna do some more tonight.
I recently saw some samples from someone using the K10D at 1600 and the results were far better than anything I've seen so far. He was shooting in JPEG mode Bright with the new ver1.1
I've only been doing RAW DNG's... so I'm wondering (and gonna test a few) if perhaps they included in the firmware some special "Al Gore Rhythm" (lol) with in-camera processing of JPEGs that doesn't take affect with the RAW files... does that make sense? is it possible? I don't how these things work, I'm just guessing.
I've also wondered if it has something to do with the fact that the RAW converter I'm using isn't "tweaked" for K10D DNGs. I'm using AdobeCR ver 2.4 and there's a few weird things that happen with the DNGs from this camera, but nothing big - e.g. when I shoot at WB daylight, the ACR says the images are around 7000K, but the colours looks fine anyway. I do know that Aperture 1.5.2 won't even accept the DNGs yet as it hasn't been updated to support the K10D yet (weird I know).
Perhaps the Raw converter needs to be tailored for this camera. Again, I have no idea about this stuff, I'm only grasping at straws.
But I'm going to compare the JPEG capture against the RAW capture and see what happens.
--
Joshua Hakin Photography
http://www.hakinphoto.com
guyoliver
Well-known member
--
guy
For crying out loud why not just go back to shooting film and leave poor old Pentax off the hook.
guy
For crying out loud why not just go back to shooting film and leave poor old Pentax off the hook.
GordonBGood
Veteran Member
Even with firmware version 1.0, the JPEG's have hardly any visible vertical pattern noise at any ISO sensitivity (only vertical banding at high ISO's/longer exposures), as you can see by the absence of vertical pattern noise in the JPEG sample images posted as samples in the K10D review on this site. The reason banding doesn't show up in JPEG's much is that the Dynamic Range is reduced so that very small patterns such as this wouldn't be noticed and that the limited 8-bit tonality of JPEG's does not have enough precision to be affected by these really minute effects.
Raw conversion as you are using are much more prone to showing the vertical pattern noise effect because it's so easy to boost exposure and especially shadow exposure so that minute effects become clearly visible.
No, the vertical pattern noise has nothing to do with the raw convertor as I've explained above. I use my own raw convertor that I can tailor any way I want, and it shows the same effect if I "push process" the shadows, as the source problem is in the raw data itself.
As to weird problems with ACR support of DNG's and in particular with the reporting of colour temperature and even worse, the WB conversions when you use other than the As Shot (camera) WB, that is also a problem with version 3.6. The problem has been described before if you search for ACR WB K10D. The problem is as follows:
1) DNG contains all the information to be able to calculate the As Shot true Colour Temperature.
2) From knowing the true As Shot Colour Temperature, one can change WB accurately to any other Colour Temperature without problems.
3) Up to at least ACR 3.6, Adobe do not properly do this. Instead, they base colour temperature on the assumed ratios between the RGB channels, which varies between camera models.
4) When ACR encounters a camera model that is not in their data base, they arbitrarily assign 1:1:1 ratios between the RGB channel gains and base their calculations on that. In most cases this is wrong, as it is for the K10D.
Hopefully in some future version, Adobe in their infinite wisdom might see fit to properly calculate DNG colour temperatures, which will then mean they can do WB manipulations showing the correct colour temperature and be able to manipulate WB freely without errors, and without having to have the camera model in their data base.
Since you have upgraded to firmware 1.1 and assuming that it does properly fix the vertical banding at longer time exposure problem properly, you will see almost no vertical pattern noise from in-camera produced JPEG's, even boosting exposure or manipulating tone curves to boost shadows a lot. It seems to take a raw convertor to be able to boost this minute vertical column errors enough to become really objectionable. However, raw conversion is also the only way to extract maximum Dynamic Range from your images, so this is still a real problem.
Regards, GordonBGood
CapitaineAbitibi
Well-known member
Ok here's a new sample, this one at ISO 1600 no retouching in PS except for noise ninja TWICE... For concert pics I think the max is ISO 800 with the K10D, ISO 1600 is just awful and the banding... if you can't see it I need a new pair of glases lol... Look in the upper portion of the picture, there are some horizontal bands (vertical if you prefer)
CapitaineAbitibi
Well-known member
Note : this one was taken as .DNG... I shot jpg at the same time and jpg looks fine, no grain or banding... WTF?!?
The JPEG may have a higher contrast setting. This draws the shadows towards black and may hide the problem?
Jens
--
'Well, 'Zooming with your feet' is usually a stupid thing as zoom rings are designed for hands.' (Me, 2006)
http://www.jensroesner.de/
--=! Condemning proprietary batteries since 1976 !=--
Gordon already explained this JPEG vs RAW discrepancy in this same thread:
http://forums.dpreview.com/forums/read.asp?forum=1036&message=21813047
--
http://www.pbase.com/klopus
I did a few tests with my *ist D and the K10D and it would seem as thought the D sensor is about a stop more sensitive than the K10D at the same ISO200, same aperture setting, the same lens, for the same camera histogram spike when taken of a grey card. In other words, the D would give a shutter speed of say 1/200sec and the K10D would give a shutter speed of 1/100sec.
--
Lance B
http://www.pbase.com/lance_b
GMT +10hours
Keyboard shortcuts
- f
- Forum