Blackmagic Forum

Previously had noise issues with my new 6K FF but put it down to the lens (Panny 24-105) but this shadow banding is with a new lens (sigma 28-70 2.8) Not a very challenging scene IMO. Any thoughts/suggestions would be appreciated. 8:1 800iso f10.2 Exposed for the sky, but you can see the banding on the wall.

Link to braw file:https://go.wetransfer.com/t-C7h7kdhdz3

Have you tried recalibrating the sensor? I heard that everytime you replaced the lens the sensor has to be recalibrated. I don't know if this is a bug?

I recalibrated the sensor before I used the new lens

Is the lens information added before sensor recalibration?

Torture test of the FF vs G2:

5:30 mark. He pushes the sensor, but the G2 doesn't exhibit banding.

Did you try a few seconds of BRAW Q0 as part of the test rather than use 8:1? I don’t know if Q0 would make a difference, but it might.

rick.lang wrote:Did you try a few seconds of BRAW Q0 as part of the test rather than use 8:1? I don’t know if Q0 would make a difference, but it might.

Thanks, Rick. I'll try.

rick.lang wrote:Did you try a few seconds of BRAW Q0 as part of the test rather than use 8:1? I don’t know if Q0 would make a difference, but it might.

Tried Q0 vs 8:1. The banding is still there in the shadows, but not as pronounced. If you lift the shadows you can see it.

8:1 https://go.wetransfer.com/t-RP3fVYhW2q

Q0 https://go.wetransfer.com/t-TvlX8uCXUq

Thanks for the update. Occasionally I lift shadows modestly but usually go the other way to get a cleaner look.

I’m not able to download BRAW files right now.

Are you talking about banding like what you get from an electrical issue?

Are you talking about banding that’s from a discharge light source?

Are you talking about banding like from CMOS pixel overflow (blooming) in a high contrast situation? Clip ore for a RED camera but the same idea…


JB

John Brawley wrote:I’m not able to download BRAW files right now.

Are you talking about banding like what you get from an electrical issue?

Are you talking about banding that’s from a discharge light source?

Are you talking about banding like from CMOS pixel overflow (blooming) in a high contrast situation? Clip ore for a RED camera but the same idea…


JB

None of these, John. Sensor noise in the shadows. It's there even in lower ISOs. There's been some chatter about it online. Here is YT post that demonstrates it, although he's shooting in very dark conditions and I wasn't.

Actually looks like sensor blooming to me. It’s only present when there’s clipped photosites. (Like in the video clip I linked third)

http://www.optique-ingenieur.org/en/cou ... nu_14.html

JB

Doesn't look like any pixels are saturated when he films outside though. And looks to also happen when he films inside when nothing's clipped but that's hard to tell due to YT compression.

John Brawley wrote:Actually looks like sensor blooming to me. It’s only present when there’s clipped photosites. (Like in the video clip I linked third)

http://www.optique-ingenieur.org/en/cou ... nu_14.html

JB

Hmm. Looks like classic sensor noise to me...

scorsesefan wrote:

John Brawley wrote:Actually looks like sensor blooming to me. It’s only present when there’s clipped photosites. (Like in the video clip I linked third)

http://www.optique-ingenieur.org/en/cou ... nu_14.html

JB

Hmm. Looks like classic sensor noise to me...

What do you mean by classic sensor noise?

In the linked video that I saw, it’s a smear that horizontally tracks off clipped highlights and moves with camera movement. Basically an echo of a highlight tipping over into dark areas of the frame that have no exposure. A phenomena called sensor blooming that most CMOS sensors suffer from to some degree.

As opposed to classic sensor noise.

JB

But what clipped highlights, I don't see any, outside or inside?

Michel Rabe wrote:But what clipped highlights, I don't see any, outside or inside?

Me neither.

John Brawley wrote:

scorsesefan wrote:

John Brawley wrote:Actually looks like sensor blooming to me. It’s only present when there’s clipped photosites. (Like in the video clip I linked third)

http://www.optique-ingenieur.org/en/cou ... nu_14.html

JB

Hmm. Looks like classic sensor noise to me...

What do you mean by classic sensor noise?

In the linked video that I saw, it’s a smear that horizontally tracks off clipped highlights and moves with camera movement. Basically an echo of a highlight tipping over into dark areas of the frame that have no exposure. A phenomena called sensor blooming that most CMOS sensors suffer from to some degree.

As opposed to classic sensor noise.

JB

I don't see any clipped highlights in that video, John. But it's not my video. If you take a look at the three braw clips I uploaded, I exposed correctly for the highlights and the noise bands are in the shadows.

I’m not in a position right now to look at BRAW files.

I’m going off the clipped highlights. The light sources in the exterior. Those lights are clipped.

I’ll try to look at the BRAWs when I can but life is in the way right now.

JB

scorsesefan wrote: ... 800iso ...

I don't know what's causing your issue, but FYI: With many Blackmagic camera models, ISO 800 can be far noisier than ISO 1250. Refer to page 43 in the BMCC 6K user manual.

Cheers,

- Peter

scorsesefan wrote:Previously had noise issues with my new 6K FF but put it down to the lens (Panny 24-105) but this shadow banding is with a new lens (sigma 28-70 2.8) Not a very challenging scene IMO. Any thoughts/suggestions would be appreciated. 8:1 800iso f10.2 Exposed for the sky, but you can see the banding on the wall.

Do you mean that band of darkness in the bottom left of the frame? I took a look at the OCF you linked to and yes, you're seeing CMOS smearing. My 2¢:

I. Don't take this the wrong way, but are you actually processing your BRAW clips correctly? Because while it's definitely there if you look for it in BMDFilm, once you actually bring it in to a display corrected gamma (e.g., Rec709) it's not visible. Unless you're taking a very unorthodox approach to colour correction, I don't think you're going to be watching your work in BMDFilm so I wouldn't stress too much about it. I've included a still frame below of your clip with the Extended Video look applied.


II. I'm not here to litigate any creative decisions you're making or to call this a skill issue but I will say that, with the settings you chose, this sensor is not getting very much light. Hell, the portion of the frame with this issue is hovering around, like, 4 IRE with the Extended Video LUT. If you push the sensor like this then you're going to see these kinds of artifacts, plain and simple. Considering that there's still plenty of highlight room in this frame, I would encourage you to open your aperture up and lower your ISO. That should deal with it without losing any detail.

Now, I don't have first hand experience with the BMCC6K but my experience with other dual-gain BMD bodies is that 800-1000 ISO pushes that first gain stage very hard. It would appear that the BMCC6K is very similar in that regard, so I would only recommend using 800-1000ISO when you need that extra dynamic range above middle grey and the rest of the frame has some level of base exposure. Also, I'm not sure if you're monitoring with a LUT on but I would sincerely recommend at least putting the Extended Video look on, otherwise the BMDFilm gamma is going to give you a misleading interpretation of where your exposure actually is.

Good luck!

Thanks for the insight/critique Alex. Totally welcome your feedback and take no offense. Just grabbed a quick shot of the sunrise and definitely had more latitude to work, which would have protected my shadows. And I haven't been in "raw land" for awhile (been shooting slog for years) so my skills def need polishing. But the banding is concerning though and I think BM needs to address it, no?

scorsesefan wrote:Thanks for the insight/critique Alex. Totally welcome your feedback and take no offense. Just grabbed a quick shot of the sunrise and definitely had more latitude to work, which would have protected my shadows. And I haven't been in "raw land" for awhile (been shooting slog for years) so my skills def need polishing. But the banding is concerning though and I think BM needs to address it, no?

Concerning? I dunno. I don't own an BMCC6K and haven't put it through its paces myself, but based on the footage you uploaded I'm not prepared to dust off my pitchfork yet.

I'm never going to say no to a cleaner image but I will say that I wouldn't hold your breath for a software fix. This level of smear isn't an issue that is unique to the BMCC6K or even BMD generally; every camera out there suffers from some kind of smear and—based on what I've seen from exposure latitude tests—the BMCC6K isn't that far off from other brands. You've gotta remember that this is a $2600 USD camera and, while you're getting a lot of value for that money, there are very real limits to how resilient it is going to be when you push it. There's a reason we work with Arri Alexas whenever we get the chance.

Don't have this problem with the Pocket 6k S35. I guess this problem never showed up with testers cameras.

scorsesefan wrote:I recalibrated the sensor before I used the new lens

Have you tried switching on for 30 minutes then recalibrate the sensor?

WahWay wrote:Don't have this problem with the Pocket 6k S35. I guess this problem never showed up with testers cameras.

Yeah, AFAIK on the pocket 6k/pro it's not an issue...

WahWay wrote:

scorsesefan wrote:I recalibrated the sensor before I used the new lens

Have you tried switching on for 30 minutes then recalibrate the sensor?

No. I thought that was a Red thing

Did you try to replicate the sensor blooming John suggested?

scorsesefan wrote:

rick.lang wrote:Did you try a few seconds of BRAW Q0 as part of the test rather than use 8:1? I don’t know if Q0 would make a difference, but it might.

Tried Q0 vs 8:1. The banding is still there in the shadows, but not as pronounced. If you lift the shadows you can see it.

8:1 https://go.wetransfer.com/t-RP3fVYhW2q

Q0 https://go.wetransfer.com/t-TvlX8uCXUq

I checked your footage. Definitely sensor blooming. But it only visible after I changed ISO from 400 to 1000 boost the Exposure from 0 to 4. That's a lots of lifting! Your footage is extremely dark and you nearly shoot a silhouette, the highlights just around the midtones. If you try to recover any details from the shadows it will not go unpunished. You have so much room for the highlights you don't need to shoot at f10 if you want details in the shadows.



I had similar effect once with my BMPCC 6k when I quickly tested the focus on a new lens under similar situation. I would say everything works as expected with your camera.

On these dual ISO sensors I still leave the ISO at the ‘native’ 400 in the lower range. In the upper range, in the past I’ve used ISO 3200 but mostly ISO 1250. My last few shoots I’ve settled on ISO 1600 successfully. Definitely try to avoid increasing ISO in the lower range beyond 400.

In the sensors that are not dual ISO, I leave the setting at the native ISO which is 800. Never increase it beyond 800 in camera.

In DaVinci Resolve, I use a combination of Camera raw Exposure adjustments if needed but use the HDR set of six primary controls to balance the Lift, Gamma, Gain primary settings of traditional SDR controls.

bandibandi wrote:

scorsesefan wrote:

rick.lang wrote:Did you try a few seconds of BRAW Q0 as part of the test rather than use 8:1? I don’t know if Q0 would make a difference, but it might.

Tried Q0 vs 8:1. The banding is still there in the shadows, but not as pronounced. If you lift the shadows you can see it.

8:1 https://go.wetransfer.com/t-RP3fVYhW2q

Q0 https://go.wetransfer.com/t-TvlX8uCXUq

I checked your footage. Definitely sensor blooming. But it only visible after I changed ISO from 400 to 1000 boost the Exposure from 0 to 4. That's a lots of lifting! Your footage is extremely dark and you nearly shoot a silhouette, the highlights just around the midtones. If you try to recover any details from the shadows it will not go unpunished. You have so much room for the highlights you don't need to shoot at f10 if you want details in the shadows.

black-image-example_.jpg

black-image-example_2.jpg

I had similar effect once with my BMPCC 6k when I quickly tested the focus on a new lens under similar situation. I would say everything works as expected with your camera.

Ok. So what exactly is the difference between sensor blooming, sensor (cmos) smearing and banding?

At 400 (native) the bulk of the sky is exposed about a half stop over mid grey. I personally wouldn’t define that as “exposed correctly” but tastes vary.

The cmos blooming is apparent when the car passes through the shot, look at the right side of the frame.

The thick band is a little concerning but overall I’d say it’s underexposed at least 2 stops.

Good Luck

I’ll try to answer the easiest one in a simplification because I’m exhausted at the moment.

I believe banding is generally the correct term to describe two unrelated issues.

The first issue is a consequence of the lack of bit depth used to record the media or used in the grading of the media when selecting the colour space.

The BMD sensors are all capable of reasonably high bit depths from the sensor. But the image processing that takes its input from the sensor defined a bit depth depending upon the codec the camera will create. BMD raw write 12 bits of information (BMDFilm) whereas as BMD ProRes uses 12 bits per colour for 444 XQ and 444 but 10 bits for all the 422 variations. Usually bands of colour won’t appear as long as you have 12bit media and are unlikely with 10 bit media. Your colour space supports HDR or SDR grades.

However the post processing in which you select an output codec, may not use all the bits. If your output was BMD WideGamut, you’re familiar with the ‘milky’ look which makes full use of the camera’s media file. But if you grade in SDR Rec.709 h.264, with 420 rather than 422, areas with significant variations in brightness may not support a smooth transition from the lightest and darkest areas. So a sky shot midday that your eye perceives as a smooth blue, may display bands of blue of varying intensity. Our vision makes these band transitions more pronounced at the borders where the two different brightness levels meet.

The second banding typically appears on the walls of a room illuminated with a light source that has a slow cycle. The interaction of when the light pulses with the frame rate or scanning frequency of the camera creates alternate brighter and darker bands of light on a smooth surface such as a wall. This effect can be avoided using lights that have a higher frequency and may be called flicker-free lights. Such lights may have cycles of several thousand hertz so the shutter of normal frame rates will always display the illusion of even illumination.

Howard Roll wrote:At 400 (native) the bulk of the sky is exposed about a half stop over mid grey. I personally wouldn’t define that as “exposed correctly” but tastes vary.

The cmos blooming is apparent when the car passes through the shot, look at the right side of the frame.

The thick band is a little concerning but overall I’d say it’s underexposed at least 2 stops.

Good Luck

No, I definitely underexposed the sky. The band on the wall is what concerns me. I’m OK with noise, hopefully organic noise, but this is different

Checked it and was more surprised about the crazy moiré on the sliding gate across the street

Michel Rabe wrote:Checked it and was more surprised about the crazy moiré on the sliding gate across the street

Yeah the olpf should have mitigated that

rick.lang wrote:I’ll try to answer the easiest one in a simplification because I’m exhausted at the moment.

I believe banding is generally the correct term to describe two unrelated issues.

The first issue is a consequence of the lack of bit depth used to record the media or used in the grading of the media when selecting the colour space.

The BMD sensors are all capable of reasonably high bit depths from the sensor. But the image processing that takes its input from the sensor defined a bit depth depending upon the codec the camera will create. BMD raw write 12 bits of information (BMDFilm) whereas as BMD ProRes uses 12 bits per colour for 444 XQ and 444 but 10 bits for all the 422 variations. Usually bands of colour won’t appear as long as you have 12bit media and are unlikely with 10 bit media. Your colour space supports HDR or SDR grades.

However the post processing in which you select an output codec, may not use all the bits. If your output was BMD WideGamut, you’re familiar with the ‘milky’ look which makes full use of the camera’s media file. But if you grade in SDR Rec.709 h.264, with 420 rather than 422, areas with significant variations in brightness may not support a smooth transition from the lightest and darkest areas. So a sky shot midday that your eye perceives as a smooth blue, may display bands of blue of varying intensity. Our vision makes these band transitions more pronounced at the borders where the two different brightness levels meet.

The second banding typically appears on the walls of a room illuminated with a light source that has a slow cycle. The interaction of when the light pulses with the frame rate or scanning frequency of the camera creates alternate brighter and darker bands of light on a smooth surface such as a wall. This effect can be avoided using lights that have a higher frequency and may be called flicker-free lights. Such lights may have cycles of several thousand hertz so the shutter of normal frame rates will always display the illusion of even illumination.

Thanks, Rick. I was referring more to the narrow horizontal noise bands than lack of bit depth type banding

That looks like FPN that my Production 4k exhibits. I would contact customer service.


You can use NeatVideo to remove it.

ShaheedMalik wrote:That looks like FPN that my Production 4k exhibits. I would contact customer service.


You can use NeatVideo to remove it.

Thanks, Shaheed. Looks like it. I still remember the hubbub the bmpc FPN caused. I’m in discussions with BM support

I’m not looking at footage so grain of salt.

Rolling banding independent of what’s happening in the image is sensor noise / processing noise.

Bands that change with clipped highlights and move with the camera movement is sensor blooming.

JB

Posting this on behalf of a friend who joined 3 days ago but is awaiting approval. He is on his 2nd camera and is anxious to get the noise issue resolved:

These are my findings considering the “censor streaking/blooming” issue of the Bmcc6K FF.

The first camera I received had very noticable streaking/ “banding” (its not actual banding) in low light situations at ISO 200-1000, and less noticable from 1250 and up, but still there, even with a lut I could see it flickering in the deep shadows. Yes I was pushing the sensor, a lot of purple in false color, but with the pocket 4K (and other cams) this never resulted in this type of streaking. The images from the Bmcc6K were basically unusable.

A link to some quick shots I did to test:
https://vimeo.com/881413582/44ea5776ae?share=copy

First LOG, then with lut but lifted exposure a little bit to make it clear after compression. Note: Even without lifting I could see the flickering bands with the lut applied.

I returned this first camera since I was convinced it was a faulty sensor (since pocket 4k was fine). The new camera arrived today, but unfortunately it suffers from the same issue, although the second ISO stage (1250-up) is a little bit cleaner, and made it barely usable. So there is that.

To summarise:

- heavy streaks or flickering bands in image when underexposed.
- It was most affected by light hitting the lens from the side or off center.
- Blocking this light affected the streaks (disappear).
- Lens cap on or complete dark results in no streaks. Slight FPN but not an issue.
- MOD focus also softened the streaks a little bit.


I do think it’s strange that some people seem to imply that this is normal en users just shouldn’t underexpose the sensor (even when they’re not underexposing heavily). I think this is a bad excuse and ignores the fact that there is in fact an issue with the sensor. I’ve shot many things with many cameras, and never have I seen something as “extreme” as this while underexposing an image. And you should be able to underexpose parts, or even most of the frame without it ruining the image.

Besides this issue, I think the image is really nice. But what I want to know, is if this issue can be fixed or mitigated by a future firmware update. This will be a deciding factor for me to keeping the cam or not.

If anyone can offer an answer to this, or speculate the options, it would be greatly appreciated.

Regards,

Stefan van Diest
stefanvandiest.com

Hi, I am waiting for my BMCC6K which I will often use in low light conditions.
Do you have any updates? You also wrote you were talking to the BM Support?

Tech support asked me to send the Braw files which I did just now.
Hopefully they can provide some answers regarding this problem. I did another test in lower daylight settings and its still in the image, although kinda negligible, it leaves absolutely no room for grading/corrections. So it's still a big problem.
I really hope it can be mitigated via a firmware update because I took it out today for daylight tests and the image/colors are absolutely beautiful.

Ill share any extra info I get here asap.

I just sent BM support a low light test I did with my 6K last night: https://vimeo.com/881940044?share=copy

I also tested my fx6 under the exact same conditions (yes, yes, I know the 6K isn't a low light camera): https://vimeo.com/882072038?share=copy

I couldn't see any CMOS smearing in the fx6 footage

It's the old issue: if you want dark images, don't starve a camera of light.
Set up good lighting and pull down in post.

Uli Plank wrote:It's the old issue: if you want dark images, don't starve a camera of light.
Set up good lighting and pull down in post.

If it were just noise I'd agree, this looks pretty bad however.

I'd set up a low key shot and take another look at the shadows. If it's going to smear with that tiny bit of light I'd be curious to see what the shadows look like with a properly lit subject.

Good Luck

A bad sample is always possible too.

Uli Plank wrote:It's the old issue: if you want dark images, don't starve a camera of light.
Set up good lighting and pull down in post.

I don't think it can be blamed on that here.

Uli Plank wrote:A bad sample is always possible too.

Multiple cameras from multiple users on different continents.

What lens were you using? Wondering if this is relevant too.

I’ve never heard of this before but it seems repeatable.

JB

viewtopic.php?f=2&t=191446&p=997187#p997169