7d FPN and iso 6400

Started by N/A, January 31, 2015, 11:45:09 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

N/A

I've found that fpn (fixed pattern noise, or vertical stripes in shadows) is blatantly obvious at iso 3200, somewhat at 1600, but at iso 6400 it becomes hardly noticeable, and there is a significant difference in sharpness and color accuracy at 6400, which is to be expected. But is iso 6400 a digital push from 3200, from a lower iso or an actual analog iso? The difference can clearly be seen going from iso 4000 to 5000 in 5x zoom. It almost looks as though 6400 is a push from 1600?
7D. 600D. Rokinon 35 cine. Sigma 30 1.4
Audio and video recording/production, Random Photography
Want to help with the latest development but don't know how to compile?

Audionut

The data from Roger Clark suggests that ISO 3200 is the last analog ISO, although ISO 1600 is probably the last useful ISO.

http://www.clarkvision.com/reviews/evaluation-canon-7d/index.html


FPN generally reduces with increasing ISO, so not sure what you are experiencing.

N/A

Good read, thanks. I've found that dark frame subtraction right in AE takes care of most, sometimes all of these stripes. I had some pretty nasty clips from a video shoot in a dark club last month, so that combined with shooting in 50p (1728x516) on the 7d and having to scale the footage even more had me worried that the footage would be unusable. I used MLVFS to convert the footage, with only bad pixel fix and vertical stripe fix selected. Even those options didn't touch this noise. 
I recorded another 50p clip, same exact settings but with the fps override set to 2 with the lens cap on. I used the VisionLog dcp on both clips, did some slight NR on the original and matched the dark frame setting to it, except with no NR. Then added it to a layer above the original footage, blend mode on subtract and opacity ~40-60%. That took care of most of the stripes, and Neat Video fixed the rest, with hardly any loss in quality and sharpness. Sharpening the low luma frequencies in Neat Video helps to make underexposed footage look nicer without that digital harshness as well, I've found.
Kind of cumbersome and time-consuming but it works in a pinch, and a little bit of grain and contrast smooths everything over nicely.
7D. 600D. Rokinon 35 cine. Sigma 30 1.4
Audio and video recording/production, Random Photography
Want to help with the latest development but don't know how to compile?

arrinkiiii

Hi N/A

Thanks for the walk through =)  I shoot a lot in clubs and the MlRawViewer with the strips correction on do the trick, at least as been doing.  Baldand also have a 7D  so the strips/noise correction is more good for the 7D owners.

...but, i think each camera got is own noise pattern, it would be nice that would that MlRawViewer  or MLVFS have the option to upload this sample video for subtracting the noise wend you convert the MLV files.

It's possible ?

N/A

No problem, it seemed like way more stripes than usual, I may have had auto lighting optimizer or high iso nr on by accident. I'll have to test and see if I can duplicate it and figure out.

I had the same idea, you only need one frame for each video but it must match the settings of it (same resolution, iso, tv, av, etc.) so the option to select a specific dark frame for each would be nice. I believe dark frame subtraction works best before debayering so this would probably be even more helpful, maybe one of the gurus can chime in and add their knowledge.
7D. 600D. Rokinon 35 cine. Sigma 30 1.4
Audio and video recording/production, Random Photography
Want to help with the latest development but don't know how to compile?

Audionut

Try the dark frame subtraction in mlv_dump.  Put the lens cap on and record a few frames with the same settings.  Use the average feature in mlv_dump on a black frame clip, then use that averaged MLV for the dark frame subtraction.

If noise in the shadows is still an issue, you might want to think about sacrificing (more) highlights (use higher ISO).

N/A

Found an interesting explanation for the vertical banding on the 7d-

Another source of banding is the ADC. There are potentially two culprits here. In the case of a camera like the 7D, which uses split parallel readout (where four readout channels are directed to one DIGIC 4 chip and another four are directed to another DIGIC 4 chip in an interleaved fashion), a fairly pronounced but even vertical banding can occur, even in the midtones, thanks to different response of the DIGIC DSP image processors that house four ADC units each. As even bands are sent to one DIGIC's ADC units, and odd bands are sent to the other DICIC's ADC units, 100% identical processing is unlikely, and slight differences manifest as vertical bands.

The final potential source is high frequency components. High frequency logic has a tendency to be noisy. Using the 7D again as an example, it is an 18 megapixel sensor, which a grand total of eight ADC units must process, at a speed fast enough to support an 8fps shutter rate. (Technically speaking, the 7D has even more than 18 million pixels...it is actually a 19.1 megapixel sensor, as Canon always masks off a border of pixels for bias offset and black point calibration.) At 8fps, total pixels processed per second must be at least 152,800,000, and since there are eight ADC units, each unit must process 19.1 million pixels each and every second. That requires a higher frequency, which can (via a variety of mechanisms I won't go into here) introduce additional noise.
7D. 600D. Rokinon 35 cine. Sigma 30 1.4
Audio and video recording/production, Random Photography
Want to help with the latest development but don't know how to compile?

Audionut

Quote from: N/A on March 19, 2015, 06:30:02 PM
As even bands are sent to one DIGIC's ADC units, and odd bands are sent to the other DICIC's ADC units................................................and since there are eight ADC units, each unit must process 19.1 million pixels each and every second.

The explanation contradicts itself.  Can you share a link?

Yes, the 7D does have 8 ADC's, so does the 5D3, but being at an substantially larger price point (and a slower readout, and newer), you would expect the better noise control of the 5D3.
In terms of noise, if you were thinking about overclocking a camera, you would actually want to reduce the FPS. 

raw_diag.mo allows some analysis of banding.  You may have some luck playing with adtg_gui (send me an PM if you need a build) to reduce banding.  I think a1ex was looking at banding PP as part of his ISO research.

You should try the dark frame subtraction in mlv_dump.  The banding noise will be in the dark frames.  If you record 120 frames or so of dark frames and use the average feature in mlv_dump first, you will have a very good master frame to use with the actual subtraction step.

N/A

I found that on another forum, I thought it sounded a bit too interesting. I'll pm you, I broke my compiler somehow, the files seem to build ok but the camera won't boot with the autoexec from my machine.

I ran a dark frame mlv through mlv_dump with

$ mlv_dump -a -o dark.mlv <input mlv>

but it was only 16 frames, and opening it with mlvfs produced one dng with no noise or stripes, so I'm assuming I need more frames and to save it to dng instead of another mlv.

:edit:
Just found the actual step-by-step here.

:edit.1:
Worked like a charm, thanks for the hint! Here are the mlv_dump commands-

# sum up all frames of the darkframe.MLV and write single frame into the averaged output
mlv_dump -o avg.mlv -a <darkframe.MLV>

# process into a dng, optional
mlv_dump --dng avg.MLV

# process fpn mlv into a clean mlv by subtracting the frame in avg.mlv from every single fpn frame
mlv_dump -o clean.mlv -s avg.mlv <fpn.MLV>
7D. 600D. Rokinon 35 cine. Sigma 30 1.4
Audio and video recording/production, Random Photography
Want to help with the latest development but don't know how to compile?

Audionut

The more dark frames you average, the better.  But once you get to around 120 frames, the law of diminishing results has well and truely set in.