Blackmagic Forum

I'd hate to stay with the old firmware but, I'll try with 3.1 as well and keep you updated.
Maybe I was on 3.3 on the old footage I checked, with the grid still there.

EDIT: just rolled back to 3.1, issue is still there, and clearly visible ! 4.6k RAW, 400% zoom in Camera Raw.

ISO 800


ISO 1600

I still want to know if this is straight up a design flaw in their sensors and every camera has this problem (though some people don't seem to notice in their cameras, or haven't yet shot in an environment that makes the crosshatching easily visible).

I emailed BM and straight up asked them if there are any cameras that don't have this problem (because they wan't me to RMA again), but I'm still waiting on an answer.

rick.lang wrote:Olan, olé, good find.
Phillip, I understand but was hoping it might lead to better results. Would you consider trying 3.1 firmware again?


Sent from my iPad using Tapatalk

Hey Rick,

Yea rolled through all versions of the firmware and it's present in all them. Though it is the least visible in 3.0. Then gradually gets worse up to 4.0 B2.

Good to know; thanks, Phillip. For professional use, that 3.0 firmware may be an alternative. But I just can't let 4.0 go. You know the old saying, "How are you going to keep them down on the farm, after they've seen Paris?"


Sent from my iPad using Tapatalk

Actually I'm quite confident, that BMD is working on this, regardless of them committing it or not... In my opinion, getting this sorted out, is far more important than trying to post-implement global shutter or other features...

Of course it would be REALLY nice to hear a simple "Oh, yeah, we're aware of the issue and working on it..." from BMD... I hope they respond something to Phillips question...

I can't imagine they`re planning to RMA those cams forever...

rick.lang wrote:Good to know; thanks, Phillip. For professional use, that 3.0 firmware may be an alternative. But I just can't let 4.0 go. You know the old saying, "How are you going to keep them down on the farm, after they've seen Paris?"


Sent from my iPad using Tapatalk

Yea exactly. And in all honesty, it's pretty ridiculous to have an awesome new UI that you can only use if you're willing to accept odd patterns in all of your footage. Like, what's the point?

And I guess in this case, I've only actually seen Paris by looking through a screen door....I want to see it clearly! lol

My hope is that either the final release of 4.0 or 4.1 firmware will put this issue to bed.


Sent from my iPad using Tapatalk

rick.lang wrote:My hope is that either the final release of 4.0 or 4.1 firmware will put this issue to bed.


Sent from my iPad using Tapatalk

Considering they keep RMAing cameras for this issue, I'm guessing they don't think firmware will fix the problem

Yeah, you have to wonder... maybe it really is a hardware/sensor issue? And if it is, I'm sure BMD is doing their damndest to avoid having a product recall. However, if that's the case,how long can they go on applying firmware bandaids for every issue that crops up (and possibly creating new ones in the process)... is it really worth it for them to keep RMA'ing cameras over and over? It certainly doesn't leave a very good taste in the end-users mouth... or help their reputation at all.
---
c

chris.white wrote:Yeah, you have to wonder... maybe it really is a hardware/sensor issue? And if it is, I'm sure BMD is doing their damndest to avoid having a product recall. However, if that's the case,how long can they go on applying firmware bandaids for every issue that crops up (and possibly creating new ones in the process)... is it really worth it for them to keep RMA'ing cameras over and over? It certainly doesn't leave a very good taste in the end-users mouth... or help their reputation at all.
---
c

My guess is until they come out with a new camera and can just sweep this one under the rug. It is very curious though. I mean, the new sensors for the main URSA aren't even ready yet, and those were suppose to come out before the Ursa Mini. I'm really wondering if they realized that there's a horrible flaw in their sensor design and that's why the URSA turret upgrade is in the wind and there's no real response to people who inquire about the sensor problems. I'm kicking myself for not having invested in a RED instead.

My Camera is still with BMD. When I last inquired about when will I see it back I was told that the camera is being "re calibrated". whether or not it is a Firmware or a hardware issue... go figure.

Hi folks. I've been doing some investigating in a similar thread at BMCuser (http://www.bmcuser.com/showthread.php?18452-Cross-hatch-screen-door/page8) and everything seems to indicate the crosshatching issue originates from a debayering artifact. The artifact then becomes visible in the form of scaling moire when the images are resized.

It's possible to fix this with CornerFix using a RAW workflow and overriding the "BayerGreenSplit" parameter of the DNG (explained at https://sites.google.com/site/cornerfix/using-cornerfix/maze-patterns-1). However, this won't help anyone shooting ProRes, as the artifact gets recorded to that format. It would require BMD to acknowledge the issue and fix it in the firmware or the camera calibration.

Using ExifTool instead of CornerFix is much easier I think. But regardless which application you use to set the BayerGreenSplit parameter you won't see any difference in Resolve unfortunately. Only in Adobe Camera RAW.

Good that BMD is recalibrating RMA:d cameras!

For anyone who needs it, I've created an infographic explaining how the "Bayer Green Split" issue causes the scaling moire. It also demonstrates why people using the 4K cameras are less likely to see the phenomena than people using the 4.6K camera, and how people with the 4.6K camera will get different amounts depending on which shooting resolution they're using.

Phillip Bergman wrote:I still want to know if this is straight up a design flaw in their sensors and every camera has this problem (though some people don't seem to notice in their cameras, or haven't yet shot in an environment that makes the crosshatching easily visible).

I emailed BM and straight up asked them if there are any cameras that don't have this problem (because they wan't me to RMA again), but I'm still waiting on an answer.

The cross-hatching and mazing are a demosaic artifact caused by green imbalances in the CFA. It's not specific to the BM cameras. If I recall, the LMMSE demosaic didn't exhibit this problem, though it has its own issues. You can try opening a frame in RawTherapee. The demosaic in Resolve is closest to AHD modified by Paul Lee.

You'll also get scaling artifacts as described by Earl since 4.6K -> 1080p doesn't work out in integer numbers.

Out of curiosity, I did another test with one of the sample DNG's posted in this thread to see how much of a difference there was between the odd and even rows/columns. My measurements put it at about 4% on average.

I have this with my BMMCC. It's visible in Resolve, but not Premier Pro CS6. But when I export from either, it is still visible. When I pause the video and adjust the viewer size (VLC, Windows) the cross hatching will change, get larger, blur, disappear, ect... Once again this happens with the video paused. So definitely some sort of scaling issue inherent in Black Magic camera images. Thing is, mine doesn't show up in all clips. I think it has something to do with how light is hitting the sensor. Possibly hitting the receptors in a way that triggers the outline of them to show up?

Larry Sullivan wrote:When I pause the video and adjust the viewer size (VLC, Windows) the cross hatching will change, get larger, blur, disappear, ect...So definitely some sort of scaling issue

Just to be clear, it's not directly a scaling issue. The scaling issue is a secondary side effect of the actual problem, which is that the image isn't being debayered properly. Scaling just makes the underlying artifact more or less visible.

True, true. Scaling only enhances or mitigates the underlying problem.

By the way, any updates on the recent RMAs?

Simon Schubert wrote:By the way, any updates on the recent RMAs?

My Camera is on its way back. I am suppose to receive it today according to UPS. No "heads up" from support on shipping the item...
I'll report back when I get it to let you guys know.


Eli.

Simon Schubert wrote:By the way, any updates on the recent RMAs?

Shipping mine back today, I hope it won't take too long...

I'm still going back and forth with the BMD support folks on whether my camera needs to be RMA'd. They seem to be unwilling to believe that there's any issue with my camera. Last time around, I sent DNG's shot at 4 different resolutions... all of which exhibit the crosshatch. The attached jpeg with enlarged examples is from 1920x1080 RAW DNG. Yet, here was the last response I recieved...

"I took a look at your images from the camera and I dont see any patterns in these images that would justify RMA. You can still send it in for us to test but I think that would be a waste of your time."

Am I just hallucinating? That does show the crosshatch issue, right?
---
c

Yep, it's there and it's ugly :p Kinda surprised that the guy from the support doesn't see it, it's so obvious.

chris.white wrote:Am I just hallucinating? That does show the crosshatch issue, right?

Yes its there and its definitly ugly and unusable. My Camera also has the exact same problem like shown here. support says they have to look at the camera to see where the problem is.... i asked if they dont think that the next firmware will correct it, but they cant say. So ill probably also send it in, what else should we do?

I wonder if all the new cameras that shipped after the first batch with the Magenta problem have this, this would explain why there are people that dont expirience this problem. There was a stop in delivery and now that they started shipping again suddenly everyone has this crosshatch problem.

So I got my camera back and did a really quick test : throw the battery plate back on, lens and card and shot all flavors of pro res HQ: 4.6k 4k dci, 3k anamorphic, UHD all the way down to 1080.

Now before sending the camera in, when I looked at the footage in QT right out of the camera I would see the crosshatch pattern in all of the above mentioned resolution (except 1080).

After getting it back:
I am happy to report that I am not seeing it anymore out of the camera.

I attached an image from a raw lossless. when I enlarged it to 400% I saw the pattern but ever so slightly I am not sure it is not something else....

let me know what you think.

Eli hershko wrote:So I got my camera back and did a really quick test...let me know what you think.

That looks fine, Eli. Seems like they fixed it for you.

chris.white wrote:Am I just hallucinating? That does show the crosshatch issue, right?

It definitely does, and the response you got from BMD is what I've been fearing -- that they're not seeing this problem for what it is.

If you can attach that DNG file, I can probably put together some explanatory images highlighting the problem so you can send those to them.

Yeah, this is frustrating for sure. My camera is one from the original batch that went out. The magenta actually didn't seem too horrible compared to some, but it was there.

I'm really not sure I have the time or patience for this anymore (and I'm a really patient person, typically... I pre-ordered and waited 10 months without complaint for this friggin camera). I'm not sure that it's worth sending it back right now if they don't even really have a real fix for it. However, I also don't want the warranty to run out so that I'll have to pay to get something known fixed later on. And having to send things multiple times is even less appealing.

I really love the look of this cameras footage and it's great to work with, but I'm regretting this purchase more each day. It's feeling like one disappointment after another (between originally announced features removed and now continual performance issues). Plus, BMD's lack of transparency/communication/accountability certainly doesn't foster any good feelings either.
---
c

thomas bruegger wrote:

chris.white wrote:Am I just hallucinating? That does show the crosshatch issue, right?

Yes its there and its definitly ugly and unusable. My Camera also has the exact same problem like shown here. support says they have to look at the camera to see where the problem is.... i asked if they dont think that the next firmware will correct it, but they cant say. So ill probably also send it in, what else should we do?

I wonder if all the new cameras that shipped after the first batch with the Magenta problem have this, this would explain why there are people that dont expirience this problem. There was a stop in delivery and now that they started shipping again suddenly everyone has this crosshatch problem.

Earl R. Thurston wrote:

chris.white wrote:Am I just hallucinating? That does show the crosshatch issue, right?

It definitely does, and the response you got from BMD is what I've been fearing -- that they're not seeing this problem for what it is.

If you can attach that DNG file, I can probably put together some explanatory images highlighting the problem so you can send those to them.

Thanks, Earl. But I'm just going to point the support agent at this thread. I already explained to them that you have to zoom in to see more clearly see the issue.
---
Chris

Eli, blowing something up to 400% may not be the single best way to judge the image. Can you also attach a few images at normal view? Not seeing anything in that shot that would look like it was shot through a screen door.


Sent from my iPad using Tapatalk

chris.white wrote:Thanks, Earl. But I'm just going to point the support agent at this thread.

Are you sure? I was thinking of making something like this for you, which will demonstrate the point more directly:


The key point is that it IS easy to fix, but BMD needs to acknowledge it's there and fix it in the camera, otherwise it makes any UHD ProRes files you shoot useless. (This can only be fixed at the debayering stage.)

Seems this issue is not just isolated to the 4K and 4.6K cameras. Another user has posted a DNG with this issue from the Micro Cinema Camera on BMCuser. In their case it's very minimal, but it's still there regardless.

http://www.bmcuser.com/showthread.php?18709-Is-this-noise-normal-Micro-Cinema-Camera

I can confirm that i see same issue on BMMCC camera too, but it is way less visible than in 4.6k examples. It is slightly visible only in the Purple-Red-Orange-Yellow colors area. It is also highly depends of added unsharp mask level.
So let's force BM to fix it in future version of Resolve. ACR is not the way to go because poor overall debayer quality and slow workflow.

rick.lang wrote:Eli, blowing something up to 400% may not be the single best way to judge the image. Can you also attach a few images at normal view? Not seeing anything in that shot that would look like it was shot through a screen door.


Sent from my iPad using Tapatalk

You are right Rick... The normal view doesn't show any "screen door" pattern. I really do think the BMD fixed the problem with my RMA.
I just saw the image a bit pixelated at 400% and was curious if the pattern is "hidden"

Thanks anyway for the input.

Eli.

Eli, all art is illusion. "The willing suspension of disbelief." Best judged in the intended way: a moving image with a story that moves us. All else will be forgiven if you achieve that (as you have done in your film).


Sent from my iPad using Tapatalk

Dmitry Shijan wrote:So let's force BM to fix it in future version of Resolve.

We really need them to fix it in the camera. When this gets recorded to ProRes, it can't be fixed properly in any program. (Plus, some of us can't use Resolve.)

EDIT: I used 80A filter (the most cool), not a 80B.
Today I was able to access a full version of Resolve in local VFX studio, so i made some very clear tests of BMMCC. Shoot with Mosaic OLPF and Hoya UV-IR filters on, with 50mm at f5.6 and f4 to compensate B80 80A filter light loss. Expose was slightly compensated in Resolve to match shots. 300% crops.

RAW WB corerction:
RAW Color Temp 3310, Tint 28.70 (shoot with 3200K halogen light)
RAW Color Temp 4823, Tint 58.66 (shoot with 3200K halogen light with B80 80A filter)

Settings used:
RAW Sharpness 10
Spartial NR Node - Faster, Small, Treshold Chroma 10
Temporal NR Node - Frames 5, Better, Medium, Treshold 30, Motion 10.7
Color Space Transform Node - Timeline to Rec.2020/Cineon Film Log
Contrast boost Node with Curves tool
LUT Node generated in LUTCalc - Cineon REC.2020 to Alexa-X-2 in-Data out-Legal
Sharpen Node - Radius 0.47

As you can see the Cross Hatching became way more visible when you shoot with warm light and WB RAW file to netural grey. If you use B80 80A filter or 5000K light you get way better results.

Same frames with R G B channel separation with Splitter Node.
I also tested RAW 3:1 vs RAW with this method and find NO any difference.
Open images in new Tab to see in original sze

Dmitry Shijan wrote:Today I was able to access a full version of Resolve in local VFX studio, so i made some very clear tests of BMMCC.

Dmitry, are you able to post one DNG file from each test? I'd like to try some experiments in Adobe.

Cinema5D just posted a comparison review of the URSA Mini 4K vs. 4.6K. They also discovered the debayering problem and posted highlighted images titled "pattern noise," but it's the same problem we've been discussing here.

https://www.cinema5d.com/blackmagic-ursa-mini-4k-vs-4-6k/

I just debayered test files in Raw Photo Processor (VCDMF interpolation) and find that RPP is almost free of Cross Hatching. Instead of fixed pattern it makes something more like a random pixel noise. RPP has its own different problems with BMD RAW files so it is not perfect.
Just one more proof that there is a software problem.

Earl R. Thurston wrote:Dmitry, are you able to post one DNG file from each test? I'd like to try some experiments in Adobe.

Thanks for sending me those DNG's, Dmitry.

I have also been able to confirm that the "BayerGreenSplit" fix in CornerFix (or ExifTool) improves the output on these DNG files. I first chose one notable color square from each of the darker sample images:



Where I really saw a difference, though, was in the shadow on the wall behind the case:



This is the kind of natural part of an image that could really suffer from this problem, as it should otherwise be a smooth, soft gradation.

Dmitry's test also was good in that it contained a full color chart. I wanted to see what happened if the image was converted to L*a*b color (which is similar to YUV). The pattern appears entirely in the luminance channel. The two color difference channels are incredibly clean:

Eli hershko wrote:So I got my camera back and did a really quick test : throw the battery plate back on, lens and card and shot all flavors of pro res HQ: 4.6k 4k dci, 3k anamorphic, UHD all the way down to 1080.

Now before sending the camera in, when I looked at the footage in QT right out of the camera I would see the crosshatch pattern in all of the above mentioned resolution (except 1080).

After getting it back:
I am happy to report that I am not seeing it anymore out of the camera.

I attached an image from a raw lossless. when I enlarged it to 400% I saw the pattern but ever so slightly I am not sure it is not something else....

let me know what you think.

Eli, this is exciting news! I need to send my camera in for a second time for them to fix the crosshatch issue. The first time they just sent me a new camera with the same issue, but like it looks like with you're camera, there's some sort of re-calibrating they need to do to the camera to fix this issue. Could you post a link to download some of the clips you shot at high ISO at 4.6k prores? I would love to compare to my camera. And also, can you tell me which Blackmagic RMA center you sent your camera to? I get the feeling that the different RMA locations don't communicate with each other, and so I'm going to specifically site your case in my email to them so they can actually relay what they did to the RMA center here in california! IF you have your RMA number that would be extremely helpful as well! Hoping I can get them to fix it this time rather than just send me a new camera with the same problems!

Do we know if the new firmware fixes this problem?


Sent from my iPhone using Tapatalk

This issue reminds me of the Pocket camera "bloom" image issue, which was fixed with a camera recalibration.
Cheers

James Parker wrote:Do we know if the new firmware fixes this problem?


Sent from my iPhone using Tapatalk

It did not fix the issue on my camera

Phillip Bergman wrote:

Eli hershko wrote:So I got my camera back and did a really quick test : throw the battery plate back on, lens and card and shot all flavors of pro res HQ: 4.6k 4k dci, 3k anamorphic, UHD all the way down to 1080.

Now before sending the camera in, when I looked at the footage in QT right out of the camera I would see the crosshatch pattern in all of the above mentioned resolution (except 1080).

After getting it back:
I am happy to report that I am not seeing it anymore out of the camera.

I attached an image from a raw lossless. when I enlarged it to 400% I saw the pattern but ever so slightly I am not sure it is not something else....

let me know what you think.

Eli, this is exciting news! I need to send my camera in for a second time for them to fix the crosshatch issue. The first time they just sent me a new camera with the same issue, but like it looks like with you're camera, there's some sort of re-calibrating they need to do to the camera to fix this issue. Could you post a link to download some of the clips you shot at high ISO at 4.6k prores? I would love to compare to my camera. And also, can you tell me which Blackmagic RMA center you sent your camera to? I get the feeling that the different RMA locations don't communicate with each other, and so I'm going to specifically site your case in my email to them so they can actually relay what they did to the RMA center here in california! IF you have your RMA number that would be extremely helpful as well! Hoping I can get them to fix it this time rather than just send me a new camera with the same problems!

HI there... My RMA # was 123328. I sent the camera to Fremont CA. I am shooting for the next 3 days so Can't post anything but will try to next week.

Eli

Eli hershko wrote:

Phillip Bergman wrote:

Eli hershko wrote:So I got my camera back and did a really quick test : throw the battery plate back on, lens and card and shot all flavors of pro res HQ: 4.6k 4k dci, 3k anamorphic, UHD all the way down to 1080.

Now before sending the camera in, when I looked at the footage in QT right out of the camera I would see the crosshatch pattern in all of the above mentioned resolution (except 1080).

After getting it back:
I am happy to report that I am not seeing it anymore out of the camera.

I attached an image from a raw lossless. when I enlarged it to 400% I saw the pattern but ever so slightly I am not sure it is not something else....

let me know what you think.

Eli, this is exciting news! I need to send my camera in for a second time for them to fix the crosshatch issue. The first time they just sent me a new camera with the same issue, but like it looks like with you're camera, there's some sort of re-calibrating they need to do to the camera to fix this issue. Could you post a link to download some of the clips you shot at high ISO at 4.6k prores? I would love to compare to my camera. And also, can you tell me which Blackmagic RMA center you sent your camera to? I get the feeling that the different RMA locations don't communicate with each other, and so I'm going to specifically site your case in my email to them so they can actually relay what they did to the RMA center here in california! IF you have your RMA number that would be extremely helpful as well! Hoping I can get them to fix it this time rather than just send me a new camera with the same problems!

HI there... My RMA # was 123328. I sent the camera to Fremont CA. I am shooting for the next 3 days so Can't post anything but will try to next week.

Eli

Thanks Eli! Yea that's the same location I send my camera, so I'll be telling them they need to reference your RMA to figure out what they did so they can fix mine as well!