D7100 'streaking'

Started Apr 7, 2013 | Discussions
Horshack
Veteran MemberPosts: 4,068
Like?
D7100 angle shot
In reply to BlueBomberTurbo, Apr 7, 2013

BlueBomberTurbo wrote:

Something else to think about: this may be a form of lens flare, considering the intensity of the light source.  Try shooting the same shot again, but at an angle.

You might be right, although it's curious that the D7100 exhibits it so much more than the D5100 with the same lens. Maybe there's flare/reflection inside the D7100's chamber. Here's an angle shot with the D7100 - all I see is the typical D5200/D7100 banding.

Reply   Reply with quote   Complain
mosswings
Veteran MemberPosts: 5,827Gear list
Like?
Re: D7100 angle shot
In reply to Horshack, Apr 7, 2013

One thing that has changed in the angle shot is the relationship of the bright light bands with the geometry of the sensel rows.  You're now illuminating all of the rows rather than a subset of them.

Again, changing the aperture should affect any internal reflections occurring in the optical path.

Try the angle shot with the D5100/D800. If it uniformly disappears this could deprecate a mirror box reflection in the D5100.

 mosswings's gear list:mosswings's gear list
Olympus XZ-1 Nikon D90 Nikon D7100 Nikon AF-S DX Nikkor 18-105mm f/3.5-5.6G ED VR Nikon AF-S Nikkor 70-300mm f/4.5-5.6G VR +1 more
Reply   Reply with quote   Complain
mosswings
Veteran MemberPosts: 5,827Gear list
Like?
Re: D7100 angle shot
In reply to mosswings, Apr 7, 2013

mosswings wrote:

Try the angle shot with the D5100/D800. If it uniformly disappears this could deprecate a mirror box reflection in the D5100.

Sorry, I recant that.  Even if it uniformly disappears, we can't discount either blooming or angle-of-view. Both could be happening simultaneously.

 mosswings's gear list:mosswings's gear list
Olympus XZ-1 Nikon D90 Nikon D7100 Nikon AF-S DX Nikkor 18-105mm f/3.5-5.6G ED VR Nikon AF-S Nikkor 70-300mm f/4.5-5.6G VR +1 more
Reply   Reply with quote   Complain
James JC
Regular MemberPosts: 135Gear list
Like?
Re: D7100 'streaking'
In reply to Jussi Mattila, Apr 7, 2013

This precise streaking phenomenon is widely reported in plasma televisions but not LCDs. I think it has to do with the technology of the Toshiba sensor. Thanks for posting… very interesting.

 James JC's gear list:James JC's gear list
Nikon D610
Reply   Reply with quote   Complain
Kerry Pierce
Forum ProPosts: 17,244Gear list
Like?
reminds me of the d200 problem
In reply to Jussi Mattila, Apr 7, 2013

The first d200's would do this and you didn't have to have to boost the shadows to see it.  It could be very pronounced in almost any scene where there was very bright lighting with dark areas and the overall scene exceeded the sensor's DR.

Nikon eventually fixed the problem.  IIRC, it had something to do with the way they were moving the data off of the sensor.

I'd suggest that you report this, with samples, to your regional service center.  Nikon won't fix what they don't know about.

good luck

Kerry

-- hide signature --

my gallery of so-so photos
http://www.pbase.com/kerrypierce/root

 Kerry Pierce's gear list:Kerry Pierce's gear list
Nikon D300 Nikon D700 Nikon D3S Nikon D7000 Nikon D5100 +11 more
Reply   Reply with quote   Complain
mosswings
Veteran MemberPosts: 5,827Gear list
Like?
Re: reminds me of the d200 problem
In reply to Kerry Pierce, Apr 7, 2013

Kerry Pierce wrote:

The first d200's would do this and you didn't have to have to boost the shadows to see it.  It could be very pronounced in almost any scene where there was very bright lighting with dark areas and the overall scene exceeded the sensor's DR.

Nikon eventually fixed the problem.  IIRC, it had something to do with the way they were moving the data off of the sensor.

I'd suggest that you report this, with samples, to your regional service center.  Nikon won't fix what they don't know about.

good luck

Kerry

-- hide signature --

my gallery of so-so photos
http://www.pbase.com/kerrypierce/root

The D200 was also CCD-based.

 mosswings's gear list:mosswings's gear list
Olympus XZ-1 Nikon D90 Nikon D7100 Nikon AF-S DX Nikkor 18-105mm f/3.5-5.6G ED VR Nikon AF-S Nikkor 70-300mm f/4.5-5.6G VR +1 more
Reply   Reply with quote   Complain
_sem_
Veteran MemberPosts: 4,139
Like?
Re: D7100 'streaking'
In reply to mosswings, Apr 7, 2013

mosswings wrote:

BlueBomberTurbo wrote:

Yup, the D5200/D7100 will band if you push shadows too far, though it's not consistent fom one scene to another.  If you use mid/high ISO, the effect can go away, sometimes completely.

BlueBomberTurbo, this is very likely NOT deep shadow pattern noise ("banding").  That has completely different characteristics that what are presented in this image.  The width and spacing of deep shadow pattern noise bands is much narrower and is not related to any subject geometry.  The streaking in this image is clearly subject geometry related.

In the CCD age you could expect row or column overflow from extremely bright light sources, but the effect was gross and obvious. And it's not possible in CMOS based sensors.

I think the images here show exactly what I've spotted in the banding debate. CreaDVty noticed that banding is more obvious in some samples than others. And in one sample I've noticed it gets much more obvious in the horizontal level of intense highlights. I think "streaking" is a good term.

http://www.dpreview.com/forums/post/51215789

Has anybody checked how well the usual banding removers respond to this (Nik Dfine, Topaz Denoise)?

(Do we remember the nasty CMOS vertical sensor bleeding, hehe?)

Reply   Reply with quote   Complain
mosswings
Veteran MemberPosts: 5,827Gear list
Like?
Re: D7100 'streaking'
In reply to _sem_, Apr 7, 2013

_sem_ wrote:

I think the images here show exactly what I've spotted in the banding debate. CreaDVty noticed that banding is more obvious in some samples than others. And in one sample I've noticed it gets much more obvious in the horizontal level of intense highlights. I think "streaking" is a good term.

http://www.dpreview.com/forums/post/51215789

I downloaded Mattila's NEFs and gave them a 4 stop push in LR 4.4 to check out any change in banding behavior.  The shadows in the original scene are extremely dark, less than 1% of FS in all three channels in the darkest parts, so banding should be expected. In this screenshot of a 100% blowup of the window around the blind pull, the top is the original scene and the bottom is the pushed image.

You can see a weak banding throughout the shadows on the right, but a tonal shift from reddish in the areas adjacent to the brightest light bands, and a greenish shift adjacent to the darker bands and dark structures seen through the windows. What is interesting here, though, is that the most obvious banding is NOT adjacent to the brightest light bands in the windows, but the darkest. In this regard, then, it appears that if there is any bleeding/blooming going on, it is increasing the fill level of the pixels in those rows enough to mask true dark level banding, not enhance it.

From DSC_0332

Here's a larger region screenshot to show the global pattern:

 mosswings's gear list:mosswings's gear list
Olympus XZ-1 Nikon D90 Nikon D7100 Nikon AF-S DX Nikkor 18-105mm f/3.5-5.6G ED VR Nikon AF-S Nikkor 70-300mm f/4.5-5.6G VR +1 more
Reply   Reply with quote   Complain
Jack Hogan
Senior MemberPosts: 3,698Gear list
Like?
D7100 'streaking' - doesn't look like blooming or flare
In reply to Jussi Mattila, Apr 8, 2013

Jussi Mattila wrote:

I put the images available at: D7100 streaking NEFs

Hi Jussi,

A few observations:

1) All images were taken at the very same exposure, 12 bits, compressed Raw
2) The ISO 100 image has perceivable streaking even without pushing, and its maximum Raw values are at less than 1/4 of full scale - this would exclude blooming
3) The ISO 1600 image has no streaking even when pushed several stops - this would exclude most input related sensor issues including flare (same lens, same exposure)
4) The difference in the mean raw values from the darker and lighter streaks to the top right of the window in the ISO100 image is less than 1 ADU (on average about 0.6), Raw values 12.6 stdev 1.3 next to the bright light, 11.9 stdev 1.2 next to the blind at an SNR of about 10 - this corresponds to a difference of about 5 electrons
5) The mean raw values from the same spots at the top right of the window in the ISO1600 image are around 195 with stdev around 20, also with an SNR of about 10.  If there is a slight difference it is too difficult to measure/see.
6) One can also see streaks at ISO100 in the slightly brighter vertical riser just to the right of the measurements in 4), with mean values around 18/19 and differences of about 0.7-1 ADU between the brighter and darker streaks.

It could be something in the processing (row specific dynamic offset, gain, ADC parameter or ?).  Suggestion: take the same 5 shots at 14 bit lossless compressed to see if it makes any difference (more noise might dither things out) just to eliminate one more variable.

Jack

Reply   Reply with quote   Complain
mosswings
Veteran MemberPosts: 5,827Gear list
Like?
Re: D7100 'streaking' - doesn't look like blooming or flare
In reply to Jack Hogan, Apr 8, 2013

Jack Hogan wrote:

Jussi Mattila wrote:

I put the images available at: D7100 streaking NEFs

Hi Jussi,

A few observations:

1) All images were taken at the very same exposure, 12 bits, compressed Raw
2) The ISO 100 image has perceivable streaking even without pushing, and its maximum Raw values are at less than 1/4 of full scale - this would exclude blooming
3) The ISO 1600 image has no streaking even when pushed several stops - this would exclude most input related sensor issues including flare (same lens, same exposure)
4) The difference in the mean raw values from the darker and lighter streaks to the top right of the window in the ISO100 image is less than 1 ADU (on average about 0.6), Raw values 12.6 stdev 1.3 next to the bright light, 11.9 stdev 1.2 next to the blind at an SNR of about 10 - this corresponds to a difference of about 5 electrons
5) The mean raw values from the same spots at the top right of the window in the ISO1600 image are around 195 with stdev around 20, also with an SNR of about 10.  If there is a slight difference it is too difficult to measure/see.
6) One can also see streaks at ISO100 in the slightly brighter vertical riser just to the right of the measurements in 4), with mean values around 18/19 and differences of about 0.7-1 ADU between the brighter and darker streaks.

It could be something in the processing (row specific dynamic offset, gain, ADC parameter or ?).  Suggestion: take the same 5 shots at 14 bit lossless compressed to see if it makes any difference (more noise might dither things out) just to eliminate one more variable.

Jack

To put Jack's suggestion into context a little bit, the D5200 writes files ONLY in 14 bit visually lossless compressed format.  Just as the D5100 before it.  This is apparently to avoid posterization in the lowest levels.  The D7100's default is 14 bit "uncompressed". High DR scenes should always use 14 bit formats. The question is now whether Horshack's test shots were shot in 14 or 12 bit format. He had to resort to a 5-stop pull to reveal the streaks, so the answer may be yes.

I'd also note that Lightroom doesn't just do anything simply even though it uses a simple name to describe it.  I raised the "exposure" knob to reveal the streaks, but it may be playing games with the tonal balance in the shadows as it does so. YRMV.

 mosswings's gear list:mosswings's gear list
Olympus XZ-1 Nikon D90 Nikon D7100 Nikon AF-S DX Nikkor 18-105mm f/3.5-5.6G ED VR Nikon AF-S Nikkor 70-300mm f/4.5-5.6G VR +1 more
Reply   Reply with quote   Complain
AngryCorgi
Contributing MemberPosts: 913Gear list
Like?
Re: D7100 'streaking' - doesn't look like blooming or flare
In reply to mosswings, Apr 8, 2013

Well, I'm feeling good about not being an early adopter on the D7100/D5200!!

-- hide signature --

-AC-

Reply   Reply with quote   Complain
_sem_
Veteran MemberPosts: 4,139
Like?
Re: D7100 'streaking' - doesn't look like blooming or flare
In reply to mosswings, Apr 8, 2013

mosswings wrote:

To put Jack's suggestion into context a little bit, the D5200 writes files ONLY in 14 bit visually lossless compressed format.  Just as the D5100 before it.  This is apparently to avoid posterization in the lowest levels.  The D7100's default is 14 bit "uncompressed". High DR scenes should always use 14 bit formats. The question is now whether Horshack's test shots were shot in 14 or 12 bit format. He had to resort to a 5-stop pull to reveal the streaks, so the answer may be yes.

Yep one should really use 14bit uncompressed when studying this.

The "usual" banding requires +5EV exposure and over +50 shadows to become prominent. The "streaking" not as much.

I'd also note that Lightroom doesn't just do anything simply even though it uses a simple name to describe it.  I raised the "exposure" knob to reveal the streaks, but it may be playing games with the tonal balance in the shadows as it does so. YRMV.

Certainly, LR's controls are heavily coupled under the hood.

Question: does streaking show up if the highlights are not blown (according to RawDigger or some other reliable raw histogram, not LR's which manages to recover some blown highlights)

Reply   Reply with quote   Complain
krikman
Regular MemberPosts: 292
Like?
Re: D7100 'streaking' - doesn't look like blooming or flare
In reply to _sem_, Apr 8, 2013

I spend some stupid time and found yet another possible reason: 12-bit raw.

There may be bug in quantification algorhytm or bug in ADR.

We badly need to retest with 14-bit lossless compressed and without dynamic range extension.

P.S. Sorry I'm last with such a message $). You guys already found it. Great!

Reply   Reply with quote   Complain
sshoihet
Senior MemberPosts: 2,250Gear list
Like?
Re: D7100 'streaking'
In reply to Jussi Mattila, Apr 8, 2013

what happens to the pattern when you tilt the camera so the window is at an angle within the frame?

 sshoihet's gear list:sshoihet's gear list
Canon PowerShot S95 Nikon D7000 Canon EOS M Nikon D600 Nikon D7100 +14 more
Reply   Reply with quote   Complain
krikman
Regular MemberPosts: 292
Like?
Re: D7100 'streaking' - doesn't look like blooming or flare
In reply to krikman, Apr 8, 2013

Some reasoning:

ADR algorhytm works with native data (in 14-bit space?) by somehow analysing of large blocks of data, then RAW compression defines which data will cut in conversion to 12-bit, but coders forgot to cut unusable bits from precise calculations so we got rounding error in green channel.

(i.e. pixel has value 10100011, we cut last 2 binary digits '11' but they appear in each and every cell in row and add regular 0,5% for whole row)

But VERY IMPORTANT DISCOVERY:

Until ISO400 we presumable have clear raw data, but from ISO800 Nikon 'cooks' RAW data by different datapath with intensive onchip (?) NR. It is common sense but first time we had clear instrumental proving.

RAW cooking alghorhytms works at low ISO for high DR too.

That is the reason why maximum DR so low on D4 - Nikon don't touch data too much on their only pro body.

And another suggestion to test:

JUST ROTATE CAMERA 180 degrees!

Reply   Reply with quote   Complain
Jussi Mattila
New MemberPosts: 13
Like?
Re: D7100 'streaking'
In reply to sshoihet, Apr 9, 2013

I'm swamped with work right now, but I will take more test photos next weekend. I will try at least:

14 bit lossless compressed

14 bit compressed

Different lens

Tilting the lens by a few degrees / more

Reply   Reply with quote   Complain
krikman
Regular MemberPosts: 292
Like?
Re: D7100 'streaking'
In reply to Jussi Mattila, Apr 9, 2013

Jussi Mattila wrote:

I'm swamped with work right now, but I will take more test photos next weekend. I will try at least:

14 bit lossless compressed

14 bit compressed

Different lens

Tilting the lens by a few degrees / more

And don't forget 12-bit with camera rotated 180 dg

Reply   Reply with quote   Complain
dougster1979
Regular MemberPosts: 106
Like?
"I love how it forces me to become a better photographer"
In reply to Jussi Mattila, Apr 9, 2013

Just curious why you think having a better camera will force you to become a better photographer. I`m not saying Nikon havn`t advanced the tech since the D90, but i don`t see how this relates to Photography skills.

Reply   Reply with quote   Complain
Jack Hogan
Senior MemberPosts: 3,698Gear list
Like?
Re: D7100 'streaking' - doesn't look like blooming or flare
In reply to _sem_, Apr 9, 2013

_sem_ wrote:

Question: does streaking show up if the highlights are not blown (according to RawDigger or some other reliable raw histogram, not LR's which manages to recover some blown highlights)

Highlights at the given Exposure are far from being blown on the sensor - they don't even reach 1/4 of FWC, see my previous message for details.

Reply   Reply with quote   Complain
Jack Hogan
Senior MemberPosts: 3,698Gear list
Like?
Re: D7100 'streaking' - doesn't look like blooming or flare
In reply to krikman, Apr 9, 2013

krikman wrote:

Some reasoning:

ADR algorhytm works with native data (in 14-bit space?) by somehow analysing of large blocks of data,

If you mean ADL, we can exclude it as the culprit because I see streaking in RawTherapee and RawDigger which do not do ADL.

then RAW compression defines which data will cut

Raw compression is just a lookup table, no intelligence/definition involved

in conversion to 12-bit, but coders forgot to cut unusable bits from precise calculations so we got rounding error in green channel.

(i.e. pixel has value 10100011, we cut last 2 binary digits '11' but they appear in each and every cell in row and add regular 0,5% for whole row)

I doubt that Nikon would do this differently in the D7100 than in the D800 or that it would depend on whether part of the line is much brighter than another part.

But VERY IMPORTANT DISCOVERY:

Until ISO400 we presumable have clear raw data,

Yet streaking is visible at ISO100 without pushing

but from ISO800 Nikon 'cooks' RAW data by different datapath with intensive onchip (?) NR. It is common sense but first time we had clear instrumental proving.

Can you provide details on your analysis and the data to support your claim?

Jack

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads