Cross Hatching Wont Go Away

The place for questions about shooting with Blackmagic Cameras.
  • Author
  • Message
Offline

Travis Inghram

  • Posts: 8
  • Joined: Wed Jul 20, 2016 3:03 am

Re: Cross Hatching Wont Go Away

PostSat Oct 01, 2016 7:14 pm

So I just tried the steps provided earlier in how to fix your image using CornerFix, it is somewhat confusing but I will outline them here.

The first thing I tried to do was convert by already filmed prores 4.6k footage into a .dng file somehow. I have come to the conclusion that your source material has to be raw for this fix to work (I did not try 3:1 or 4:1).


Imported the pr4.6k into premiere and exported a frame that was in .tiff format. I then converted that into a .dng using Adobe lightroom. I downloaded the CornerFix application and found the website to be down. Since the website was down I was on my own.

At first CornerFix is not easy to understand at all. Eventually I figured out the way that program works is that you take a photo or video frame from a reference grey card with a particular lens make it look bright and even slightly over expose.

I will copy and paste the instructions from the website here as a backup. I had to reference the website using internet archive.
https://web.archive.org/web/20150921195 ... ns-profile

To generate a devignetting profile:

1. Generate a flat field reference file. This is an image of a grey card taken with the lens, filter and combination of your camera's settings (e.g., on the Leica M8 or M9, lens detection on/off, IR filter on/off) that you want to correct for. If your camera has automatic lens detection, you should typically, have this off. You should overexpose by between 1 and 2 stops – enough to get the brightest possible field, without the highlights in any channel blowing out. It's important that the lighting be as even as possible, and the camera be exactly aligned with the grey card.

2.Double-click on CornerFix to start it

3.Load the flat field reference file by going to File->Open

4.Create a lens profile by going to Lens Profile->Create. Note that this can take several minutes.
When the profile is ready, a thumbnail of the corrected flat field reference file will show on the right hand side of the screen.

5.To save the profile, go to Lens Profile->Save As. CornerFix will suggest a filename which is the same as the reference image, but with the extension “.cpf”


After that you need to bring in your .DNG files that correspond to the lens profile you made. I just made a generic one to see if this worked and used a Canon 24mm on white paper, I was in a rush and didn't spend too much time making it exactly right, just close enough. I followed the instructions found here https://web.archive.org/web/20150920224929/https://sites.google.com/site/cornerfix/using-cornerfix/maze-patterns-1.

Maze patterns
An artifact that sometimes comes up when you convert images from cameras with Bayer sensor are maze patterns. An example (image courtesy of Brad Vineyard) is below. Typically these patterns show up most clearly in large areas with the same color, e.g., sky, but actually can exist anywhere in the image.


Crop from Sony NEX-5 image converted by Lightroom 3.2 at default settings (200%).
While these maze patterns have nothing to do CornerFix - if they exist, you'll see them regardless of whether you've corrected the file via Cornerfix or not - CornerFix has the ability to correct for the problem, assuming you're using an Adobe raw converter (Adobe Camera Raw or Lightroom).

Why do maze patterns happen on Bayer sensors?

In non-technical terms, the problem comes up because of the way that modern Bayer demosiacing algorithms go about extracting the maximum detail possible from an image. Bayer sensors typically have two green channels, and to extract the maximum detail, raw converters like to assume that the two green channels are very closely matched. In other words, they assume that any difference between adjacent pixels from different green channels is actually the result of a real object in the image. Problem is, due to the technology of sensor manufacturing and the way data is read by the camera electronics, that's not always so. There can be differences between the two channels that have nothing to do with anything in the image.
Fixing the maze pattern problem

In Adobe raw converters, there's a mechanism to tell the demosiacing algorithm how closely or not the two green channels track each other. This is the "BayerGreenSplit" parameter. The Adobe default for the BayerGreenSplit for the NEX image shown above is too low. However, here's what happens if we set the BayerGreenSplit to 500 in a DNG version of the image, and then convert:


BayerGreenSplit set to 500, then converted by Lightroom 3.2 (200%).
Magically the maze pattern vanishes, without any significant impact to image detail.

Unfortunately there's no easy way to change the BayerGreenSplit directly; for whatever reason, Adobe chose not to include BayerGreenSplit as a parameter in their DNG Camera Profiles (DCP files). However, CornerFix has a "hidden" ability to manipulate the BayerGreenSplit value in a DNG file.

Using CornerFix to correct for maze patterns

CornerFix can use special CornerFix profiles (.cpf files) to override the BayerGreenSplit value in any DNG file it processes. However, please note:

You must be using CornerFix V1.4.0.2 or later.
This only works for Lightroom or Adobe Camera Raw. Other raw converters, so far as I am aware, ignore the BayerGreenSplit parameter in DNG files.
When CornerFix changes the BayerGreenSplit parameter, it also changes the Camera Model name by appending "/CornerFix". This is to prevent Lightroom or Camera Raw from recognizing the model name, and using the Adobe default anyway. However, this means that you will not be able to use Adobe Camera profiles with the converted image unless you edit the profile to have the new model name.
The process you need to follow is:

Get a CornerFix BayerSplit profile - there are a variety of pre-built profiles covering a wide range of BayerGreenSplit settings in the ZIP file attached at the bottom of the page.
If your image file isn't already in DNG form, convert it, as described on this page.
Load both the DNG file, and the profile into CornerFix, just as you would if you were devignetting the image, as described on this page. Note that you won't see any change in the image; CornerFix's image display doesn't show enough detail to see the change.
Save the image.
Then load the new DNG file that you saved from CornerFix into Lightroom or Adobe Camera Raw.
Selecting a value for the BayerGreenSplit

Unfortunately, there's no way to know what value will work for the BayerGreenSplit - you just have to use a process of trial and error. Note that the existing BayerGreenSplit, which you can access by setting CornerFix to "Verbose messages" mode is not a good guide to start with; Lightroom and Camera Raw override the value for cameras they recognize, so it is usually set to a high value. Starting with zero is generally a better way. You should then just try increasing values until the maze pattern goes away. Too high a value might lead to some loss of detail.
Combining Devignetting and BayerGreenSplit correction into one step
If you're using CornerFix to devigentte your images as well as to deal with maze problems, the recommended method is to do so in one step, with a single combined profile:

Generate your devignetting profile as usual (you can also use any existing profile, of course). This process is described on this page.
Open your existing devignetting profile with a text editor, and add a line that looks like the following to it:
BayerGreenSplit 500
You add this just above the "end" line. Replace the "500" with whatever value you have found to work well. Adobe states that "useful values are from 0 to 5000". Leave everything else as is.
By way of example, here's an entire devignetting and BayerSplit profile, with the additional BayerSplit line shown in red:

profile_version 1
cornerfix_version 1.3.0.7
description ""
creator ""
copyright ""
licence "This work is licensed under the Creative "
licence "Commons Attribution 3.0 License. To view a copy of this license, "
licence "visit http://creativecommons.org/licenses/by/3.0/ or send a letter "
licence "to Creative Commons, 171 Second Street, Suite 300, San Francisco, "
licence "California, 94105, USA."
LeftHorizonal 0 equation 2 1.000000e+00 -1.701929e-07 2.114596e-14 -2.043448e-21
RightHorizonal 0 equation 2 1.000000e+00 -1.232488e-07 1.355467e-15 -8.729449e-24
TopVertical 0 equation 2 1.000000e+00 -2.903816e-07 1.169707e-13 -3.647586e-20
BottomVertical 0 equation 2 1.000000e+00 -8.926998e-08 -1.246919e-13 3.090077e-20
color 0 equation 2 1.000000e+00 -1.883246e-07 2.122495e-14 -1.341883e-21
BottomRight 0 equation 2 1.000000e+00 -1.388336e-07 6.044672e-15 -1.349633e-22
TopRight 0 equation 2 1.000000e+00 -1.592902e-07 1.097803e-14 -4.965810e-22
BottomLeft 0 equation 2 1.000000e+00 -1.618955e-07 1.256597e-14 -5.688725e-22
center 0 location -10 -122
LeftHorizonal 1 equation 2 1.000000e+00 -1.466988e-07 1.253272e-14 -1.329167e-21
RightHorizonal 1 equation 2 1.000000e+00 -1.063008e-07 -6.122975e-15 5.324453e-22
TopVertical 1 equation 2 1.000000e+00 -2.453555e-07 5.663471e-14 -1.710023e-20
BottomVertical 1 equation 2 1.000000e+00 -6.840060e-08 -1.590840e-13 3.897991e-20
color 1 equation 2 1.000000e+00 -1.610041e-07 1.274781e-14 -7.009662e-22
BottomRight 1 equation 2 1.000000e+00 -1.266264e-07 4.942876e-16 3.316971e-22
TopRight 1 equation 2 1.000000e+00 -1.376776e-07 3.474097e-15 9.193446e-23
BottomLeft 1 equation 2 1.000000e+00 -1.463081e-07 7.585889e-15 -2.149813e-22
center 1 location 1 -122
LeftHorizonal 2 equation 2 1.000000e+00 -1.511571e-07 1.448267e-14 -1.518410e-21
RightHorizonal 2 equation 2 1.000000e+00 -1.084289e-07 -4.049769e-15 2.831658e-22
TopVertical 2 equation 2 1.000000e+00 -1.481634e-07 -4.667805e-14 8.543223e-21
BottomVertical 2 equation 2 1.000000e+00 -1.528736e-07 -5.966654e-14 1.691715e-20
color 2 equation 2 1.000000e+00 -1.488420e-07 9.904034e-15 -4.600982e-22
BottomRight 2 equation 2 1.000000e+00 -1.327082e-07 3.845138e-16 4.335249e-22
TopRight 2 equation 2 1.000000e+00 -1.243614e-07 1.776910e-15 1.259678e-22
BottomLeft 2 equation 2 1.000000e+00 -1.572535e-07 9.234088e-15 -3.137841e-22
center 2 location -2 -9
LeftHorizonal 3 equation 2 1.000000e+00 -1.328712e-07 1.449994e-14 -1.904813e-21
RightHorizonal 3 equation 2 1.000000e+00 -8.700673e-08 -5.979029e-15 1.470460e-22
TopVertical 3 equation 2 1.000000e+00 -1.387800e-07 -3.942716e-14 3.294063e-21
BottomVertical 3 equation 2 1.000000e+00 -1.261547e-07 -7.936528e-14 2.198361e-20
color 3 equation 2 1.000000e+00 -1.368453e-07 1.006282e-14 -6.291270e-22
BottomRight 3 equation 2 1.000000e+00 -1.087999e-07 -4.487942e-15 7.153057e-22
TopRight 3 equation 2 1.000000e+00 -1.071818e-07 1.119964e-17 1.127300e-22
BottomLeft 3 equation 2 1.000000e+00 -1.384568e-07 5.786922e-15 -1.186023e-22
center 3 location -3 -1
BayerGreenSplit 500
end


So I Did all that and used the number 10 for BayerGreenSplit because that is what the earlier post mentioned. It worked on the DNG from the camera but did not work on any prores footage I had roundabout converted into a dng. So for now the best option is to film everything in raw and use CornerFix on all your footage until there is a better option.

Here is with CornerFix Applied zoomed into 200% in Photoshop.
Screen Shot 2016-10-01 at 2.25.44 PM.png
Corner Fix 200%
Screen Shot 2016-10-01 at 2.25.44 PM.png (188.22 KiB) Viewed 19592 times


Here is without CornerFix , 200%
Screen Shot 2016-10-01 at 2.25.48 PM.png
No CornerFix 200%
Screen Shot 2016-10-01 at 2.25.48 PM.png (232.68 KiB) Viewed 19592 times


It is better but for some reason it still doesn't feel right I feel like it has a destructive quality in whatever it is doing. Hopefully we hear something from BlackMagic, just them acknowledging the problem would be nice. I hope this issue get's resolved quickly, it is kind of a big deal.
Offline

Kyle Gordon

  • Posts: 405
  • Joined: Thu Dec 26, 2013 4:06 am

Re: Cross Hatching Wont Go Away

PostSat Oct 01, 2016 7:37 pm

Id say 10 is too low. Experiment with the highest value you can use before you see the cross hatching?
Kyle Gordon
Professional Singer/Composer/Producer and Director/Editor/Colorist
Offline

Travis Inghram

  • Posts: 8
  • Joined: Wed Jul 20, 2016 3:03 am

Re: Cross Hatching Wont Go Away

PostSat Oct 01, 2016 8:15 pm

I just tried going from 1-5000. There is almost no difference between 1-10-5000. Idk what that means, I have no idea what those numbers are changing tbh. Should I try more odd numbers?
Offline

Kyle Gordon

  • Posts: 405
  • Joined: Thu Dec 26, 2013 4:06 am

Re: Cross Hatching Wont Go Away

PostSat Oct 01, 2016 10:31 pm

I had thought it had to do with the depth of the effect, and I was interested to learn what the change in numbers would yield.
Kyle Gordon
Professional Singer/Composer/Producer and Director/Editor/Colorist
Offline
User avatar

roger.magnusson

  • Posts: 3354
  • Joined: Wed Sep 23, 2015 4:58 pm

Re: Cross Hatching Wont Go Away

PostSun Oct 02, 2016 6:34 am

Travis, when you're doing the grey card thing for CornerFix you're using it for the app's main purpose, to combat vignetting. You don't need do to that. Instead, download the prebuilt profiles from the site which won't affect vignetting at all. They're available at the bottom of this page: https://sites.google.com/site/cornerfix/using-cornerfix/maze-patterns-1

The reason it only works on RAW files is because BayerGreenSplit is an instruction to the debayering algorithm. Since ProRes is already debayered it won't work.

Anyway, instead of CornerFix, try using ExifTool, it's much easier. http://www.sno.phy.queensu.ca/~phil/exiftool/

Code: Select all
"exiftool(-k).exe" -ifd0:bayergreensplit=10 "d:\Footage\2016-08-15\*.dng"

In my mind, both of these apps are band aids for the problem, hopefully BMD will have a solution.
Last edited by roger.magnusson on Sun Oct 02, 2016 6:47 am, edited 1 time in total.
Offline

Travis Inghram

  • Posts: 8
  • Joined: Wed Jul 20, 2016 3:03 am

Re: Cross Hatching Wont Go Away

PostSun Oct 02, 2016 6:44 am

Is there anyway to modify the Prores codec to give it additional instructions? That would seem like a quick fix for Blackmagic at least.
Offline
User avatar

roger.magnusson

  • Posts: 3354
  • Joined: Wed Sep 23, 2015 4:58 pm

Re: Cross Hatching Wont Go Away

PostSun Oct 02, 2016 6:48 am

Not really, you would want this to be fixed at the debayering process or earlier for best results. Debayering is permanent.
Offline
User avatar

Jamie LeJeune

  • Posts: 2012
  • Joined: Tue Feb 26, 2013 4:33 am
  • Location: San Francisco

Re: Cross Hatching Wont Go Away

PostSun Oct 02, 2016 7:54 am

Did I miss something here? BMD support is issuing fixes for this problem. Contact them directly to:

1. RMA your camera to prevent any future cross hatching.

2. For files you've already shot that have cross hatching, BMD support can send .drx files that will remove the crosshatching specific to your workflow to add as a node to clips in Resolve. It's an annoying extra step, but it is easier than fussing with other software fixes.

Why BMD isn't posting this info themselves is odd, but not really surprising given their track record so far when it comes to communicating necessary information about the Ursa Mini.
www.cinedocs.com
http://www.imdb.com/name/nm4601572/
Offline
User avatar

Gary Yost

  • Posts: 87
  • Joined: Fri Jan 17, 2014 2:03 pm

Re: Cross Hatching Wont Go Away

PostSun Oct 02, 2016 9:41 pm

Kyle Gordon wrote:Gary! Fantastic news! I hope we start to see all the cameras are like this, or that BM will RMA them and then return them like this! Thanks so much for your detailed tests and sharing.

I have read some wild theories about the magenta vignetting and this issue. At the very least they are probably both related to the green photosites, so I wonder if you'd mind sharing a test of that so we can see what your good sensor looks like under those conditions.

The test is a non compressed DNG of a white wall, hopefully somewhat evenly lit, a 35mm or 50 mm lens, Sigma 18-35 at 35 is best if you have it, and f/8. Post it here or in the magenta thread, if you are able to do it.

Thanks so much for sharing all this, it really helps to understand what's going on!


Kyle, I apologize for the delay in getting this test to you... had a shoot last week and lots of other things going on, whoops.

Here ya go. In this file you'll find two uncompressed raw shots of a white card:

https://www.dropbox.com/s/iyzsisw9b45x8yn/UM%20white%20card%20tests.zip?dl=0

Once you unzip them the labeling is pretty clear. Both are shot with the Sigma 18-35 @35mm as you requested... first is at f1.8 of the card in deep shade. Second is at f8 in full sun with a 0.3ND on it.

Let me know what you think!

-g
Offline
User avatar

Gary Yost

  • Posts: 87
  • Joined: Fri Jan 17, 2014 2:03 pm

Re: Cross Hatching Wont Go Away

PostSun Oct 02, 2016 9:43 pm

Man that's odd that they'd just send another unit out without testing that. I've requested at least some information about what they did to mine but haven't heard anything back yet. Of course I'll let everyone know if/when I do.

Phillip Bergman wrote:I already emailed them about it and they responded pretty quickly offering another RMA so props to them for that, but I have a lot of shoots this month so i'll probably have to wait to send it back in until November. It's just frustrating having waited a year for this camera to come out in the first place, promising all these cool things, then getting rid of a lot what was promised (Global shutter, gyroscope, 160fps), and then releasing the camera 8 months late only to have horrible magenta issues and sensor problems that are still yet to be addressed. At this point I know I'm going RED for my next camera. I've tried to stick it out with blackmagic but I need an actual professional camera at this point.
Offline

Kyle Gordon

  • Posts: 405
  • Joined: Thu Dec 26, 2013 4:06 am

Re: Cross Hatching Wont Go Away

PostMon Oct 03, 2016 2:57 am

Gary, no apologies necessary, on the contrary, I really appreciate you taking the time to do this test and share it.

Why/How are these DNGs 1920 x 1080? Shouldn't they be 4.6k if they are DNGs?
Kyle Gordon
Professional Singer/Composer/Producer and Director/Editor/Colorist
Offline
User avatar

Gary Yost

  • Posts: 87
  • Joined: Fri Jan 17, 2014 2:03 pm

Re: Cross Hatching Wont Go Away

PostMon Oct 03, 2016 3:03 am

Kyle Gordon wrote:Gary, no apologies necessary, on the contrary, I really appreciate you taking the time to do this test and share it.

Why/How are these DNGs 1920 x 1080? Shouldn't they be 4.6k if they are DNGs?


Ah, good grief. I'm using firmware 4b2 and when I selected RAW I forgot to check the resolution, assuming 4.6k. But of course it was set to window 1920x1080. Grrrrr..... I'll try to get some time to run the tests again tomorrow at full res.
Offline

Kyle Gordon

  • Posts: 405
  • Joined: Thu Dec 26, 2013 4:06 am

Re: Cross Hatching Wont Go Away

PostMon Oct 03, 2016 3:31 am

hahahaha no WONDER the vignetting was PERFECT!

Thanks Gary, I'm looking forward to it!

if you need to set exposure, its cool to dial back the shutter angle, rather than add a ND filter, and one frame is good enough, no need to upload more than one frame.

Thanks SO much!
Kyle Gordon
Professional Singer/Composer/Producer and Director/Editor/Colorist
Offline
User avatar

Gary Yost

  • Posts: 87
  • Joined: Fri Jan 17, 2014 2:03 pm

Re: Cross Hatching Wont Go Away

PostMon Oct 03, 2016 4:37 pm

Kyle Gordon wrote:hahahaha no WONDER the vignetting was PERFECT!

Thanks Gary, I'm looking forward to it!

if you need to set exposure, its cool to dial back the shutter angle, rather than add a ND filter, and one frame is good enough, no need to upload more than one frame.

Thanks SO much!


Yup it sure does explain the absolute lack of vignetting. Ha. It's raining here this morning so I set up inside with a light. Here's a 4.6k dng, 35mm f8, 180 shutter.

https://www.dropbox.com/s/h6mcoc97bc935ku/A004_10030949_C008_000022.dng?dl=0

I just compared this to a white card I shot after I got the camera and the vignetting (which wasn't bad before) has been improved.
Offline

Phillip Bergman

  • Posts: 159
  • Joined: Mon Nov 02, 2015 6:36 pm
  • Location: Fresno, CA

Re: Cross Hatching Wont Go Away

PostMon Oct 03, 2016 10:27 pm

So, I'm starting to get pissed with BlackMagic and their refusal to discuss this issue. Like I posted earlier, after sending in my camera for RMA, I was sent a replacement unit which has the same crosshatching problem as the one I RMA'd. When I emailed them about it, they said they were sorry and would set up another RMA. So I emailed back discussing my concern about this issue and whether or not blackmagic had a fix for it, or if we just had to live with it. I told them I was weary of spending another $60 to send in my camera again, when the brand new unit they just sent me has the same issue. I also spoke of Gary Yost's experience of sending his camera in and apparently getting it fixed with no more crosshatching. I sent them 3 prores clips and 5 screen shots showing the issue. The email I got back read....

"Hi Phillip

If we set up another RMA, I would send you a shipping label to use to send this to us. We wouldn't want you to have to pay for shipping again.

Would you like to proceed?"

Soooo, completely ignoring all of the concerns (other than the cost of shipping it back) and all my questions regarding the issue. So I replied saying...

"I really can't do it until the end of October because we have a heavy shoot
schedule throughout October and I need the camera for that. Unless you want
to ship me a non crosshatched camera and I will send this one back upon
receipt of the new one. (I doubt that's an option, understandably). So I'd
really have to wait until the end of October.

But is there really no explanation of whether or not this is fixable
problem?"

And then the response I got back from that was.....

"Hi Phillip

When a camera is showing unusual behavior and we have the customer send it in, we will always do whatever we can to get it working properly.

We have not been given information on what that process is for a camera showing this type of unusual behavior. If we are not able to get the camera working in a timely manner, then we would normally set up a replacement as long as the camera is under warranty. That is what happened with the last RMA. Unfortunately, the replacement is showing similar behavior. If and when you decide to set up an RMA for the replacement unit, let us know and we will try to get that turned around for you as quickly as possible.

Let us know if you have any other questions.

All the Best"

Soooo apparently, even though they were able to fix said issue with Gary Yost's camera, they don't feel the need to share that information with the rest of their RMA department. So I guess this is an acknowledgment that this issue is "unusual behavior", and also an acknowlegment that they really don't test their new units before shipping them out as replacements to customers.

Really irritated. This is ridiculous.
Offline
User avatar

Jamie LeJeune

  • Posts: 2012
  • Joined: Tue Feb 26, 2013 4:33 am
  • Location: San Francisco

Re: Cross Hatching Wont Go Away

PostTue Oct 04, 2016 1:12 am

Phillip Bergman wrote:This is ridiculous.


I totally agree that it is crazy and completely frustrating. It is also, sadly, par for the course with BMD. We (their customers) appear to be the only real quality control process in place for their cameras. They seem to operate on the same model as many software companies who release a product before it's debugged or even fully featured and then they iterate updates over time. Since BMD has zero competition at the price point and is selling cameras as fast as they can make them, there is no reason for BMD to change their approach. In fact, it seems to be working out quite well for them even if it drives many of their customers nuts. I've owned three different models of BMD cameras so far and the situation has been the same with each. When the cameras work, they work well, but to get a working camera often takes multiple returns and/or repairs. In a sense, their products are permanently in beta and buyers just have to beware. If I had the capital, I'd buy an Alexa. But, instead I spend my time testing BMD cameras and sending them back when they (too often) fail.
www.cinedocs.com
http://www.imdb.com/name/nm4601572/
Offline

Travis Inghram

  • Posts: 8
  • Joined: Wed Jul 20, 2016 3:03 am

Re: Cross Hatching Wont Go Away

PostTue Oct 04, 2016 3:42 am

So who wants to build there custom firmware/OS for this camera and release it to the public? An open source camera would be pretty cool.
Offline

John Brawley

  • Posts: 4267
  • Joined: Tue Aug 21, 2012 7:57 am
  • Location: Los Angeles California

Re: Cross Hatching Wont Go Away

PostTue Oct 04, 2016 3:48 am

Travis Inghram wrote:So who wants to build there custom firmware/OS for this camera and release it to the public? An open source camera would be pretty cool.


It already exists.

https://www.apertus.org

For all those that think that this is what's holding these camera back from reaching their full potential.

JB.
John Brawley ACS
Cinematographer
Currently - Los Angeles
Offline

Kyle Gordon

  • Posts: 405
  • Joined: Thu Dec 26, 2013 4:06 am

Re: Cross Hatching Wont Go Away

PostTue Oct 04, 2016 4:53 pm

Gary, thank you. I still see a little, and I eagerly await the final version of FW 4 to see if they are going to completely remove it.

I'll stop talking about this issue now, on this thread, since this thread is about the cross hatching issue.

The only reason I posted here was that your camera got a cross hatching fix, and I wondered what the effect on the vignetting was.

Thanks again so much for your time sharing these.
Kyle Gordon
Professional Singer/Composer/Producer and Director/Editor/Colorist
Offline
User avatar

Valentin Remy

  • Posts: 282
  • Joined: Thu Mar 24, 2016 3:52 pm
  • Location: Belgium

Re: Cross Hatching Wont Go Away

PostWed Oct 05, 2016 12:22 am

Aaand I just noticed I'm experiencing the issue as well...
Guess I'll have to RMA as well, it fixed the issue for some people here.

But I don't understand why some people never experienced it with the default firmware, and then got it with a firmware update. That means it's a software issue, nothing to do with hardware, so no need for a RMA ?

I really hate that BMD is always going silent every time there's a problem, seriously wtf ?
http://www.instagram.com/valentinremy.be/
Offline

Francisco Rodriguez

  • Posts: 127
  • Joined: Tue Sep 24, 2013 8:48 pm

Re: Cross Hatching Wont Go Away

PostWed Oct 05, 2016 7:10 am

I was under the impression that it was just a display-scaling issue...
I never shoot over 2K so I haven't been following this... what's the official verdict?

I see the pattern at UHD and above... the thing is that I soon as I click on "show at full size" within resolve the pattern disappears...
Offline
User avatar

Jamie LeJeune

  • Posts: 2012
  • Joined: Tue Feb 26, 2013 4:33 am
  • Location: San Francisco

Re: Cross Hatching Wont Go Away

PostWed Oct 05, 2016 7:50 am

Francisco Rodriguez wrote:I was under the impression that it was just a display-scaling issue...
I never shoot over 2K so I haven't been following this... what's the official verdict?

I see the pattern at UHD and above... the thing is that I soon as I click on "show at full size" within resolve the pattern disappears...


Problem isn't in the display, it's in the camera. It crops up most often when the recorded sensor frame size matches the timeline frame size. For example, full sensor recorded in cam as HD or 2K hasn't been a problem. Windowed sensor recorded at 2K or HD is a problem. Plenty of info and some example clips to download in the posts earlier in the thread. It's worth a read from the start, especially the wonderfully detailed posts by Gary Yost who took a lot of care in documenting the issue.

If you see the issue in your camera files, contact BMD for RMA.

EDIT: I forgot that Gary's examples of the issue are in the original thread here:
https://forum.blackmagicdesign.com/viewtopic.php?f=2&t=50175
www.cinedocs.com
http://www.imdb.com/name/nm4601572/
Offline

Francisco Rodriguez

  • Posts: 127
  • Joined: Tue Sep 24, 2013 8:48 pm

Re: Cross Hatching Wont Go Away

PostWed Oct 05, 2016 2:25 pm

Just sent an email...

RMA All over again :?

Thanks for the advice Jamie..
Offline

chris.white

  • Posts: 185
  • Joined: Fri Aug 15, 2014 9:34 pm

Re: Cross Hatching Wont Go Away

PostWed Oct 05, 2016 3:07 pm

Yup, I have this issue too. I noticed it early on when testing, but was sitting on it for a while since I wasn't really shooting anything over HD... hoping that firmware would clear it up down the road.

Not too excited about the RMA process, since even that seems to be hit or miss, but my EVF needs to go back too... so might as well send it all :(
---
c
Offline

David Hessel

  • Posts: 274
  • Joined: Mon Mar 10, 2014 7:53 pm

Re: Cross Hatching Might've Gone Away

PostWed Oct 05, 2016 9:38 pm

Gary Yost wrote:I received my UM 4.6k back from being RMA'd this morning and have run it through a series of tests... sharing all the results here.

Short story is that, upon first look I'm loving what they did. I'm not seeing any of the regular cross-hatching patterns on any of the 20 different tests I've run (both on firmware 3.0 and firmware 4b2).

My basic quick test is to bring the file up in quicktime and then scale it, looking for the "harmonics" that show the crosshatching artifact. On all of my previous footage this was the fastest and most reliable test.

Without getting into a huge post, I'm just going to give you dropbox links to various sets of files and you can see for yourselves. Of course I'm curious as to what everyone else thinks.

Firmware 3.x. - 6 tests. (raw 4.6k full, raw 4.6k windowed, etc as labeled in the screenshot files) Absolutely no changes of any kind applied to any of the tests.

1) https://www.dropbox.com/s/p2i5zvgklk9j3k7/v3%20screenshots.zip?dl=0 - This is a set of 1000% zoomed-in screenshots pulled from each of the six firmware 3.x tests, all inside of Resolve with the resolution of the timeline natively matching the resolution of the shot so there's no rescaling.

2) https://www.dropbox.com/s/hchvy691oxrfedj/UM46%20tests%20post%20fix%20-%20firmware%203.zip?dl=0 - Here are the 6 actual files, both raw and prores. They are labelled as they came out of camera so for you to know what each specific test is you'll have to listen to the audio track.


Firmware 4b2 - 14 tests. (raw 4.6k full, raw 4kDCI windowed, raw UHD windowed, etc as labeled in the screenshot files) Absolutely no changes of any kind applied to any of the tests.

1) https://www.dropbox.com/s/29rz5bggvkgfhex/v4b2%20screenshots.zip?dl=0 - This is a set of 1000% zoomed-in screenshots pulled from each of the 14 firmware 4b2 tests, all inside of Resolve with the resolution of the timeline natively matching the resolution of the shot so there's no rescaling.

2) https://www.dropbox.com/s/n5xxhmtia9l5r8y/UM46%20tests%20post%20fix%20-%20firmware%2042b.zip?dl=0 - Here are the 14 actual files, both raw and prores. They are labelled as they came out of camera so for you to know what each specific test is you'll have to listen to the audio track.

Every single one of these files passed my "harmonic artifacting" test and it seems to me as if the issue has been put to bed. I'm looking forward to hearing what y'all think.


Just tested one of the DNG's and they look far more normal than the one I tested before that had the crosshatch.

With Crosshatch
UM-green-zoom.jpg
Crosshatch
UM-green-zoom.jpg (73.44 KiB) Viewed 19193 times


Without Crosshatch (DNG from this post)
UM-no-hatch.jpg
No Crosshatch
UM-no-hatch.jpg (90.39 KiB) Viewed 19191 times
David Hessel
Offline

olan_collardy

  • Posts: 175
  • Joined: Sat Apr 05, 2014 10:59 am
  • Location: London

Re: Cross Hatching Wont Go Away

PostFri Oct 07, 2016 10:32 pm

Don't think I've ever exported a 4K video so the cross hatch never really bothered me.

Was just looking at some 4K footage I shot and handed over to a client - no questions asked about pattern on footage even though I can see the pattern.

When graded (i.e. shadows adjusted), the pattern fades away - so is this some kind of noise pattern?

See attached - you can see the pattern when viewed in the Rendered file

Ungraded: https://www.dropbox.com/s/y8kabsqji3c9p ... d.png?dl=0
Graded: https://www.dropbox.com/s/ppra10wctzzb2 ... d.png?dl=0

However, on uploading to Vimeo - pattern seem to have disappeared or at least greatly reduced (looking more like noise)

https://vimeo.com/186029936/29606b293d
Offline
User avatar

James Alexander Barnett

  • Posts: 220
  • Joined: Sat Jan 09, 2016 5:48 pm
  • Location: UK

Re: Cross Hatching Wont Go Away

PostSun Oct 09, 2016 5:13 pm

So I received my camera on Friday and after reviewing one of the clips i noticed cross hatching, however it only appeared when scaling the image in odd multiples or divisions of the actual frame size, so 25% 50% 75% and 100% were fine.

After reading this thread I was a little concerned so I conducted a few tests, I set the codec to ProRes HQ, shot at every resolution, every ISO and also did some clips in cropped sensor mode and I could not get any sort of pattern to appear at all.

I don't really know how to use resolve yet so I was just viewing in QT & FCPX, is this problem more evident in raw?
James Alexander Barnett
www.jamesalexanderdp.co.uk
URSA Mini 4.6k
Pocket 6k Pro
Offline

Phillip Bergman

  • Posts: 159
  • Joined: Mon Nov 02, 2015 6:36 pm
  • Location: Fresno, CA

Re: Cross Hatching Wont Go Away

PostSun Oct 09, 2016 5:46 pm

James Barnett wrote:So I received my camera on Friday and after reviewing one of the clips i noticed cross hatching, however it only appeared when scaling the image in odd multiples or divisions of the actual frame size, so 25% 50% 75% and 100% were fine.

After reading this thread I was a little concerned so I conducted a few tests, I set the codec to ProRes HQ, shot at every resolution, every ISO and also did some clips in cropped sensor mode and I could not get any sort of pattern to appear at all.

I don't really know how to use resolve yet so I was just viewing in QT & FCPX, is this problem more evident in raw?



You initially said after reviewing your clips that you noticed crosshatching, then you said you did a test shooting at every ISO and resolution and couldn't see any crosshatching at all. I'm confused, does your camera show the patterning or not? Do you think you could shoot an ISO 1600 at 4.6k prores short clip (preferably inside your house with just the lighting from the lights in the room) and post it on here? I would like to compare your cameras images to mine and see if yours is crosshatch free.
Offline
User avatar

Dustin Albert

  • Posts: 406
  • Joined: Wed Mar 27, 2013 3:27 am

Re: Cross Hatching Wont Go Away

PostSun Oct 09, 2016 5:55 pm

Phillip Bergman wrote:
James Barnett wrote:So I received my camera on Friday and after reviewing one of the clips i noticed cross hatching, however it only appeared when scaling the image in odd multiples or divisions of the actual frame size, so 25% 50% 75% and 100% were fine.

After reading this thread I was a little concerned so I conducted a few tests, I set the codec to ProRes HQ, shot at every resolution, every ISO and also did some clips in cropped sensor mode and I could not get any sort of pattern to appear at all.

I don't really know how to use resolve yet so I was just viewing in QT & FCPX, is this problem more evident in raw?



You initially said after reviewing your clips that you noticed crosshatching, then you said you did a test shooting at every ISO and resolution and couldn't see any crosshatching at all. I'm confused, does your camera show the patterning or not? Do you think you could shoot an ISO 1600 at 4.6k prores short clip (preferably inside your house with just the lighting from the lights in the room) and post it on here? I would like to compare your cameras images to mine and see if yours is crosshatch free.


I think he was referring to ONLY seeing it IN POST when scaling to an odd percentage.


Sent from my iPhone using Tapatalk
Never stop learning and trying new things…
Offline
User avatar

James Alexander Barnett

  • Posts: 220
  • Joined: Sat Jan 09, 2016 5:48 pm
  • Location: UK

Re: Cross Hatching Wont Go Away

PostSun Oct 09, 2016 8:08 pm

Phillip Bergman wrote:
James Barnett wrote:So I received my camera on Friday and after reviewing one of the clips i noticed cross hatching, however it only appeared when scaling the image in odd multiples or divisions of the actual frame size, so 25% 50% 75% and 100% were fine.

After reading this thread I was a little concerned so I conducted a few tests, I set the codec to ProRes HQ, shot at every resolution, every ISO and also did some clips in cropped sensor mode and I could not get any sort of pattern to appear at all.

I don't really know how to use resolve yet so I was just viewing in QT & FCPX, is this problem more evident in raw?



You initially said after reviewing your clips that you noticed crosshatching, then you said you did a test shooting at every ISO and resolution and couldn't see any crosshatching at all. I'm confused, does your camera show the patterning or not? Do you think you could shoot an ISO 1600 at 4.6k prores short clip (preferably inside your house with just the lighting from the lights in the room) and post it on here? I would like to compare your cameras images to mine and see if yours is crosshatch free.



Yeah I was confused too, it was defiantly there the day I got the camera however when I did checks the next day to test for the pattern I couldn't get it to appear at all.

Very strange.
James Alexander Barnett
www.jamesalexanderdp.co.uk
URSA Mini 4.6k
Pocket 6k Pro
Offline
User avatar

Valentin Remy

  • Posts: 282
  • Joined: Thu Mar 24, 2016 3:52 pm
  • Location: Belgium

Re: Cross Hatching Wont Go Away

PostMon Oct 10, 2016 2:43 pm

I'm having a email exchange with the BMD Support.

The guy said that he played my ProRes file through a BMD DeckLink Playback Device, and the issue was not there, and that my behaviour was related to video scaling. I sent a RAW file for further testing, but pattern is still there for me in Adobe Camera Raw when zoomed in.

And the weird workaround found here that adds an additional step and screw up the quality => no thanks

EDIT: here's a comparison, between my DNG and BMD's downloadable DNG (from the parking scene).


Image
Image
http://www.instagram.com/valentinremy.be/
Offline

David Hessel

  • Posts: 274
  • Joined: Mon Mar 10, 2014 7:53 pm

Re: Cross Hatching Wont Go Away

PostMon Oct 10, 2016 4:02 pm

Yeah that looks like you have the crosshatching problem, the image should not look like that zoomed in. Depending on what resolution they were viewing it at compared to their display resolution the pattern may not have been noticeable if they were not zooming in on the image like you have.
David Hessel
Offline
User avatar

Valentin Remy

  • Posts: 282
  • Joined: Thu Mar 24, 2016 3:52 pm
  • Location: Belgium

Re: Cross Hatching Wont Go Away

PostMon Oct 10, 2016 4:21 pm

Well I mentioned a few times in the emails that they have to zoom in, but the guy I have contact with keeps telling me that he believes it's related to video scaling and not the camera... Which obviously isn't, you can see it in my screen above.

Still have to get his feedback with the RAW file I provided a few hours ago (from which I took the screenshot above), but I'm kinda afraid that he won't allow a RMA... I'd be really mad.
http://www.instagram.com/valentinremy.be/
Offline

David Hessel

  • Posts: 274
  • Joined: Mon Mar 10, 2014 7:53 pm

Re: Cross Hatching Wont Go Away

PostMon Oct 10, 2016 4:48 pm

If you want to post the DNG here I can see if it looks like the other ones that had crosshatching.
David Hessel
Offline
User avatar

Valentin Remy

  • Posts: 282
  • Joined: Thu Mar 24, 2016 3:52 pm
  • Location: Belgium

Re: Cross Hatching Wont Go Away

PostMon Oct 10, 2016 4:55 pm

Sure ! => https://we.tl/IR3E4BnwR3

But I'm 100% sure it's there, I'm just waiting for BMD to acknowledge it and send me a normally working camera...
http://www.instagram.com/valentinremy.be/
Offline

David Hessel

  • Posts: 274
  • Joined: Mon Mar 10, 2014 7:53 pm

Re: Cross Hatching Wont Go Away

PostMon Oct 10, 2016 5:06 pm

I am seeing the same varying brightness of rows of pixels in the green channel like the other DNG I tested in this thread that has crosshatching in ProRes. It doesn't look clean like the DNG from the fixed camera also posted in this thread. Definately keep pushing for an RMA.

remy-green-channel-zoom.jpg
Green Channel
remy-green-channel-zoom.jpg (291.22 KiB) Viewed 18932 times
David Hessel
Offline
User avatar

roger.magnusson

  • Posts: 3354
  • Joined: Wed Sep 23, 2015 4:58 pm

Re: Cross Hatching Wont Go Away

PostMon Oct 10, 2016 5:13 pm

I really hope the different support offices have the same information about this issue. It worries me that they would blame this on a scaling issue. Sure, an affected image will have scaling issues, but that's not the root cause. If there's already an embedded pattern in a zoomed in image like this, of course there will be scaling issues.

It would be one thing if this was only visible when shooting high frequency detail, but it can even be seen when the subject is completely out of focus.

I'm getting ready to contact support myself, just need to formulate a perfect description first. :)
Offline
User avatar

Valentin Remy

  • Posts: 282
  • Joined: Thu Mar 24, 2016 3:52 pm
  • Location: Belgium

Re: Cross Hatching Wont Go Away

PostMon Oct 10, 2016 5:57 pm

David Hessel wrote:I am seeing the same varying brightness of rows of pixels in the green channel like the other DNG I tested in this thread that has crosshatching in ProRes. It doesn't look clean like the DNG from the fixed camera also posted in this thread. Definately keep pushing for an RMA.


Thanks for checking on your side as well :) Looking forward to BMD's answer.
http://www.instagram.com/valentinremy.be/
Offline

Simon Schubert

  • Posts: 10
  • Joined: Tue Oct 11, 2016 8:58 am

Re: Cross Hatching Wont Go Away

PostTue Oct 11, 2016 9:39 am

So, let me join the happy complaining...

obviously, I'm not the only one with this issue... We got our UM46k in march. And man, I can't deny it's a real pleasure shooting with this cam! BUT lately we've been expieriencing the mentioned issue extremely. We don't produce much content above HD, but in our case, it gets really annoying when shooting HD >60p on a windowed sensor (no matter the codec or raw). The grid (or crosshatch) pattern - especially at ISO800 and above - is simply awful. I tested it on resolutions above HD, and it's definitely there for almost all resolutions in windowed mode. I'm actually not getting it, shooting on full sensor. (Maybe the position of the cropped area is badly calculated?)

(screenshots on request, but I think, we've seen the issue more than enough)

I Tried fiddling with all the settings in Resolve, Scaling, debayer quality, bla bla,... no real improvement.
Rendering in native timeline, leaves the issue in the final file...
The only thing that got rid of the grid (while sacrificing quality), was the "0.5 pan/tilt hack" (which implies a scaling problem - again, maybe wrong calculation of crop area? - but still leaves the question, why the heck it should be so difficult, to capture images, that don't provoke scaling issues.... c'mon ... every smartphone can do it!)

I reviewed footage, I recorded before updating the firmware to V4b2 and it's the same thing...
Then I went back to footage recorded before updating to V3.3 and voilà... nothing... Looks like 3.3 introduced the issue (maybe someone's able to confirm that). So I don't really believe in a hardware problem.
Checking back the release notes, V3.3 was intended to "Improve sensor response to lens shading". So maybe they screwed up something, or had to decide between improved sensor response and getting rid of the (let's call it) "Bayer issue"...

Sadly, BM doesn't seem to be very willig to solve this properly, so I'm stuck, unable to decide pro or contra RMA... But I guess the more related RMAs they have to deal with, the sooner they might provide official statements...

Greets
Offline

Phillip Bergman

  • Posts: 159
  • Joined: Mon Nov 02, 2015 6:36 pm
  • Location: Fresno, CA

Re: Cross Hatching Wont Go Away

PostTue Oct 11, 2016 7:09 pm

Simon Schubert wrote:So, let me join the happy complaining...

obviously, I'm not the only one with this issue... We got our UM46k in march. And man, I can't deny it's a real pleasure shooting with this cam! BUT lately we've been expieriencing the mentioned issue extremely. We don't produce much content above HD, but in our case, it gets really annoying when shooting HD >60p on a windowed sensor (no matter the codec or raw). The grid (or crosshatch) pattern - especially at ISO800 and above - is simply awful. I tested it on resolutions above HD, and it's definitely there for almost all resolutions in windowed mode. I'm actually not getting it, shooting on full sensor. (Maybe the position of the cropped area is badly calculated?)

(screenshots on request, but I think, we've seen the issue more than enough)

I Tried fiddling with all the settings in Resolve, Scaling, debayer quality, bla bla,... no real improvement.
Rendering in native timeline, leaves the issue in the final file...
The only thing that got rid of the grid (while sacrificing quality), was the "0.5 pan/tilt hack" (which implies a scaling problem - again, maybe wrong calculation of crop area? - but still leaves the question, why the heck it should be so difficult, to capture images, that don't provoke scaling issues.... c'mon ... every smartphone can do it!)

I reviewed footage, I recorded before updating the firmware to V4b2 and it's the same thing...
Then I went back to footage recorded before updating to V3.3 and voilà... nothing... Looks like 3.3 introduced the issue (maybe someone's able to confirm that). So I don't really believe in a hardware problem.
Checking back the release notes, V3.3 was intended to "Improve sensor response to lens shading". So maybe they screwed up something, or had to decide between improved sensor response and getting rid of the (let's call it) "Bayer issue"...

Sadly, BM doesn't seem to be very willig to solve this properly, so I'm stuck, unable to decide pro or contra RMA... But I guess the more related RMAs they have to deal with, the sooner they might provide official statements...

Greets


I think the crosshatching is there regardless of the firmware (at least in my cameras case), but I do notice that the crosshatching is less visible if I roll back all the way to the original firmware, so the crosshatch problem may just be amplified by the new updates.

Also, when rendering I have a 3840x2160 clip in resolve, and my project settings are set to 3840x2160, and I render out to 3840x2160, the crosshatching remains in the rendered file. However, if I change the project settings to 1920x1080 and then render out as 3840x2160, the crosshatching is gone. This is dumb. Blackmagic wants to RMA my camera again, but I'm pretty sure they're just gonna grab one off the shelf that has the same problem, not look at it, and send it tome. That's what they did last time.
Offline
User avatar

Valentin Remy

  • Posts: 282
  • Joined: Thu Mar 24, 2016 3:52 pm
  • Location: Belgium

Re: Cross Hatching Wont Go Away

PostTue Oct 11, 2016 8:03 pm

They allowed me to RMA the camera, will keep you guys updated.

But, I'm afraid they will just downgrade the firmware to when the issue was less present.

I'll try to roll back myself and see if the issue is there, but the old firmware is a pain to use compared to the new one :'(

EDIT: well, no need, I also checked older 4k footage. It was there, but way less obvious.
At the pixel level (like 200% zoom), I couldn't see it, but when scaled it was there (which means the pattern also is, even if I didn't manage to clearly see it zoomed in compared to now).
http://www.instagram.com/valentinremy.be/
Offline

David Hessel

  • Posts: 274
  • Joined: Mon Mar 10, 2014 7:53 pm

Re: Cross Hatching Wont Go Away

PostTue Oct 11, 2016 9:56 pm

Valentin Remy wrote:They allowed me to RMA the camera, will keep you guys updated.

But, I'm afraid they will just downgrade the firmware to when the issue was less present.

I'll try to roll back myself and see if the issue is there, but the old firmware is a pain to use compared to the new one :'(

EDIT: well, no need, I also checked older 4k footage. It was there, but way less obvious.
At the pixel level (like 200% zoom), I couldn't see it, but when scaled it was there (which means the pattern also is, even if I didn't manage to clearly see it zoomed in compared to now).


I have just had the opportunity to look into some DNG's from a camera running 4.0 beta 2 and the same camera running firmware 3.3. Both had the same alternating rows of pixels in the green channel but the 3.3 was far more subtle. It is consistent with what you both are seeing with the crosshatching be less severe before version 4 firmware.
David Hessel
Offline

Phillip Bergman

  • Posts: 159
  • Joined: Mon Nov 02, 2015 6:36 pm
  • Location: Fresno, CA

Re: Cross Hatching Wont Go Away

PostWed Oct 12, 2016 6:34 am

Phillip Bergman wrote:
Simon Schubert wrote:So, let me join the happy complaining...

obviously, I'm not the only one with this issue... We got our UM46k in march. And man, I can't deny it's a real pleasure shooting with this cam! BUT lately we've been expieriencing the mentioned issue extremely. We don't produce much content above HD, but in our case, it gets really annoying when shooting HD >60p on a windowed sensor (no matter the codec or raw). The grid (or crosshatch) pattern - especially at ISO800 and above - is simply awful. I tested it on resolutions above HD, and it's definitely there for almost all resolutions in windowed mode. I'm actually not getting it, shooting on full sensor. (Maybe the position of the cropped area is badly calculated?)

(screenshots on request, but I think, we've seen the issue more than enough)

I Tried fiddling with all the settings in Resolve, Scaling, debayer quality, bla bla,... no real improvement.
Rendering in native timeline, leaves the issue in the final file...
The only thing that got rid of the grid (while sacrificing quality), was the "0.5 pan/tilt hack" (which implies a scaling problem - again, maybe wrong calculation of crop area? - but still leaves the question, why the heck it should be so difficult, to capture images, that don't provoke scaling issues.... c'mon ... every smartphone can do it!)

I reviewed footage, I recorded before updating the firmware to V4b2 and it's the same thing...
Then I went back to footage recorded before updating to V3.3 and voilà... nothing... Looks like 3.3 introduced the issue (maybe someone's able to confirm that). So I don't really believe in a hardware problem.
Checking back the release notes, V3.3 was intended to "Improve sensor response to lens shading". So maybe they screwed up something, or had to decide between improved sensor response and getting rid of the (let's call it) "Bayer issue"...

Sadly, BM doesn't seem to be very willig to solve this properly, so I'm stuck, unable to decide pro or contra RMA... But I guess the more related RMAs they have to deal with, the sooner they might provide official statements...

Greets


I think the crosshatching is there regardless of the firmware (at least in my cameras case), but I do notice that the crosshatching is less visible if I roll back all the way to the original firmware, so the crosshatch problem may just be amplified by the new updates.

Also, when rendering I have a 3840x2160 clip in resolve, and my project settings are set to 3840x2160, and I render out to 3840x2160, the crosshatching remains in the rendered file. However, if I change the project settings to 1920x1080 and then render out as 3840x2160, the crosshatching is gone. This is dumb. Blackmagic wants to RMA my camera again, but I'm pretty sure they're just gonna grab one off the shelf that has the same problem, not look at it, and send it tome. That's what they did last time.



After a few more test renders I've discovered that when I have my UHD footage on a 1080 Timeline, and then render to UHD, there is a loss in detail and sharpness (I'm guessing this is why the crosshatch pattern goes away). And when I render from a UHD timeline to UHD, the detail and sharpness remains in tact, but so does the crosshatch pattern. So the only way to get rid of the crosshatch pattern is by losing quality....this is not ok.

Does anyone have a camera that DOESN'T exhibit crosshatch patterns?
Offline

Simon Schubert

  • Posts: 10
  • Joined: Tue Oct 11, 2016 8:58 am

Re: Cross Hatching Wont Go Away

PostWed Oct 12, 2016 8:12 am

I think the crosshatching is there regardless of the firmware (at least in my cameras case), but I do notice that the crosshatching is less visible if I roll back all the way to the original firmware, so the crosshatch problem may just be amplified by the new updates.


I have just had the opportunity to look into some DNG's from a camera running 4.0 beta 2 and the same camera running firmware 3.3. Both had the same alternating rows of pixels in the green channel but the 3.3 was far more subtle. It is consistent with what you both are seeing with the crosshatching be less severe before version 4 firmware.



On 3.3 I also had the pattern and in my case it was just as noticeable as in the 4b2 (I deliberately skipped 4b1). I can't remember what was the original Version... 3.2 I guess... this one had (in my case) absolutely no pattern, in neither zoom level. Maybe my Sensor is overall somewhat less "griddy" so it wasn't visible back then...

The question "to RMA or not to RMA" remains... :|
Offline

David Hessel

  • Posts: 274
  • Joined: Mon Mar 10, 2014 7:53 pm

Re: Cross Hatching Wont Go Away

PostWed Oct 12, 2016 4:09 pm

Phillip Bergman wrote:
Phillip Bergman wrote:
Simon Schubert wrote:So, let me join the happy complaining...

obviously, I'm not the only one with this issue... We got our UM46k in march. And man, I can't deny it's a real pleasure shooting with this cam! BUT lately we've been expieriencing the mentioned issue extremely. We don't produce much content above HD, but in our case, it gets really annoying when shooting HD >60p on a windowed sensor (no matter the codec or raw). The grid (or crosshatch) pattern - especially at ISO800 and above - is simply awful. I tested it on resolutions above HD, and it's definitely there for almost all resolutions in windowed mode. I'm actually not getting it, shooting on full sensor. (Maybe the position of the cropped area is badly calculated?)

(screenshots on request, but I think, we've seen the issue more than enough)

I Tried fiddling with all the settings in Resolve, Scaling, debayer quality, bla bla,... no real improvement.
Rendering in native timeline, leaves the issue in the final file...
The only thing that got rid of the grid (while sacrificing quality), was the "0.5 pan/tilt hack" (which implies a scaling problem - again, maybe wrong calculation of crop area? - but still leaves the question, why the heck it should be so difficult, to capture images, that don't provoke scaling issues.... c'mon ... every smartphone can do it!)

I reviewed footage, I recorded before updating the firmware to V4b2 and it's the same thing...
Then I went back to footage recorded before updating to V3.3 and voilà... nothing... Looks like 3.3 introduced the issue (maybe someone's able to confirm that). So I don't really believe in a hardware problem.
Checking back the release notes, V3.3 was intended to "Improve sensor response to lens shading". So maybe they screwed up something, or had to decide between improved sensor response and getting rid of the (let's call it) "Bayer issue"...

Sadly, BM doesn't seem to be very willig to solve this properly, so I'm stuck, unable to decide pro or contra RMA... But I guess the more related RMAs they have to deal with, the sooner they might provide official statements...

Greets


I think the crosshatching is there regardless of the firmware (at least in my cameras case), but I do notice that the crosshatching is less visible if I roll back all the way to the original firmware, so the crosshatch problem may just be amplified by the new updates.

Also, when rendering I have a 3840x2160 clip in resolve, and my project settings are set to 3840x2160, and I render out to 3840x2160, the crosshatching remains in the rendered file. However, if I change the project settings to 1920x1080 and then render out as 3840x2160, the crosshatching is gone. This is dumb. Blackmagic wants to RMA my camera again, but I'm pretty sure they're just gonna grab one off the shelf that has the same problem, not look at it, and send it tome. That's what they did last time.



After a few more test renders I've discovered that when I have my UHD footage on a 1080 Timeline, and then render to UHD, there is a loss in detail and sharpness (I'm guessing this is why the crosshatch pattern goes away). And when I render from a UHD timeline to UHD, the detail and sharpness remains in tact, but so does the crosshatch pattern. So the only way to get rid of the crosshatch pattern is by losing quality....this is not ok.

Does anyone have a camera that DOESN'T exhibit crosshatch patterns?


While I have not looked into all of the files in this post the ones I have looked at didn't have any sign of crosshatching in the DNG, so this camera might not.

viewtopic.php?f=2&t=51936&start=50#p302096
David Hessel
Offline

David Hessel

  • Posts: 274
  • Joined: Mon Mar 10, 2014 7:53 pm

Re: Cross Hatching Wont Go Away

PostWed Oct 12, 2016 4:14 pm

Simon Schubert wrote:
I think the crosshatching is there regardless of the firmware (at least in my cameras case), but I do notice that the crosshatching is less visible if I roll back all the way to the original firmware, so the crosshatch problem may just be amplified by the new updates.


I have just had the opportunity to look into some DNG's from a camera running 4.0 beta 2 and the same camera running firmware 3.3. Both had the same alternating rows of pixels in the green channel but the 3.3 was far more subtle. It is consistent with what you both are seeing with the crosshatching be less severe before version 4 firmware.



On 3.3 I also had the pattern and in my case it was just as noticeable as in the 4b2 (I deliberately skipped 4b1). I can't remember what was the original Version... 3.2 I guess... this one had (in my case) absolutely no pattern, in neither zoom level. Maybe my Sensor is overall somewhat less "griddy" so it wasn't visible back then...

The question "to RMA or not to RMA" remains... :|


Hard to say. My feeling is that there is an issue that cannot be fully addressed with firmware which is why BM is allowing for RMA's however it appears the issue can be made better or worse with firmware . Unfortunately no one here knows what the underlying cause is.
David Hessel
Offline
User avatar

rick.lang

  • Posts: 17175
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Re: Cross Hatching Wont Go Away

PostWed Oct 12, 2016 4:45 pm

Phillip Bergman wrote:After a few more test renders I've discovered that when I have my UHD footage on a 1080 Timeline, and then render to UHD, there is a loss in detail and sharpness (I'm guessing this is why the crosshatch pattern goes away). And when I render from a UHD timeline to UHD, the detail and sharpness remains in tact, but so does the crosshatch pattern. So the only way to get rid of the crosshatch pattern is by losing quality...


Phillip, just a reminder if you are using Resolve for your renders in Deliver, the render will use your Timeline setting of HD and upscale it to UHD. you can have the final render using UHD by changing the timeline resolution to UHD and reverting to your original clips before your deliverables are generated.



Sent from my iPad using Tapatalk
Rick Lang
Offline

Phillip Bergman

  • Posts: 159
  • Joined: Mon Nov 02, 2015 6:36 pm
  • Location: Fresno, CA

Re: Cross Hatching Wont Go Away

PostWed Oct 12, 2016 5:32 pm

rick.lang wrote:
Phillip Bergman wrote:After a few more test renders I've discovered that when I have my UHD footage on a 1080 Timeline, and then render to UHD, there is a loss in detail and sharpness (I'm guessing this is why the crosshatch pattern goes away). And when I render from a UHD timeline to UHD, the detail and sharpness remains in tact, but so does the crosshatch pattern. So the only way to get rid of the crosshatch pattern is by losing quality...


Phillip, just a reminder if you are using Resolve for your renders in Deliver, the render will use your Timeline setting of HD and upscale it to UHD. you can have the final render using UHD by changing the timeline resolution to UHD and reverting to your original clips before your deliverables are generated.



Sent from my iPad using Tapatalk



Hey Rick,

That's the problem I'm running into though. If I shoot in UHD at a higher ISO I can see a crosshatch pattern in the footage. If I put it into Resolve on a 1080 timeline, and then Render it out as UHD, it upscales from 1080 to UHD and the Crosshatch pattern is no long present in the footage. However, because of that upscale, there is a loss of detail/quality in the footage.

Now, like you said, if I change the timeline settings to UHD and then render to UHD, it keeps the good detail/quality in the rendered file, but it also keeps the crosshatch pattern. So basically I have to sacrifice quality to get rid of the crosshatch pattern.
Offline

olan_collardy

  • Posts: 175
  • Joined: Sat Apr 05, 2014 10:59 am
  • Location: London

Re: Cross Hatching Wont Go Away

PostWed Oct 12, 2016 8:07 pm

Simon Schubert wrote:I reviewed footage, I recorded before updating the firmware to V4b2 and it's the same thing...
Then I went back to footage recorded before updating to V3.3 and voilà... nothing... Looks like 3.3 introduced the issue (maybe someone's able to confirm that). So I don't really believe in a hardware problem.
Checking back the release notes, V3.3 was intended to "Improve sensor response to lens shading". So maybe they screwed up something, or had to decide between improved sensor response and getting rid of the (let's call it) "Bayer issue"...


Correct - I did the same test. Version v3.1 doesn't have the cross hatch, v3.3 does (in small doses) and quite prominent in v4.0.
Offline
User avatar

rick.lang

  • Posts: 17175
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Re: Cross Hatching Wont Go Away

PostWed Oct 12, 2016 9:59 pm

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
Rick Lang
PreviousNext

Return to Cinematography

Who is online

Users browsing this forum: Johannes Hoffmann, Shawn Miller and 70 guests