Black "spikes" showing in some render frames #29921

New Issue

David Brennan · 2012-01-18T11:36:40+01:00

David Brennan commented

2012-01-18 11:36:40 +01:00

%%%I have a setup with a computer monitor, and then a simple plane slightly in front of it as its "screen". The screen is placed .001 units in front of the monitor.

Then, to simulate light glowing from the screen, I have a little area lamp, which is then .001 units in front of the plane (the screen).

So, in total, I have:

A computer monitor model
A plane as its screen, .001 units ahead
A light to simulate glowing, .001 units ahead of plane (.002 units ahead of monitor)

In some frames, a random black "spike" mark shows up on the screen. It is consistent. In other words, if the spike mark just shows up in frame 54, it will show up that way in other render attempts.

At the suggestion of others, I spread the distance between the three objects (the monitor, plane, and light) to varying degrees (as far as .01 blender units apart from each other). And, although the spike marks changed, there were usually still spike marks. (Maybe they would be on frame 70 instead of 54, but they would still be there.)

I've attached the .BLEND of a set-up in which the spike appears on frame 41. I've also attached a .PNG so that you can see it clearly.

Of course, this was not an insurmountable obstacle. There were various work-arounds to avoid this (such as using Gimp to touch up the frames with the spike mark). But I do believe it's a bug and it is somewhat frustrating.

Tremendous thanks as always.%%%

%%%I have a setup with a computer monitor, and then a simple plane slightly in front of it as its "screen". The screen is placed .001 units in front of the monitor. Then, to simulate light glowing from the screen, I have a little area lamp, which is then .001 units in front of the plane (the screen). So, in total, I have: - A computer monitor model - A plane as its screen, .001 units ahead - A light to simulate glowing, .001 units ahead of plane (.002 units ahead of monitor) In some frames, a random black "spike" mark shows up on the screen. It is consistent. In other words, if the spike mark just shows up in frame 54, it will show up that way in other render attempts. At the suggestion of others, I spread the distance between the three objects (the monitor, plane, and light) to varying degrees (as far as .01 blender units apart from each other). And, although the spike marks changed, there were usually still spike marks. (Maybe they would be on frame 70 instead of 54, but they would still be there.) I've attached the .BLEND of a set-up in which the spike appears on frame 41. I've also attached a .PNG so that you can see it clearly. Of course, this was not an insurmountable obstacle. There were various work-arounds to avoid this (such as using Gimp to touch up the frames with the spike mark). But I do believe it's a bug and it is somewhat frustrating. Tremendous thanks as always.%%%

David Brennan commented

2012-01-18 11:36:40 +01:00

Changed status to: 'Open'

Brecht Van Lommel commented

2012-01-18 13:58:25 +01:00

%%%There seem to be no file attached, the attachment system is a bit unreliable, and maybe the file was too big. Could you try attaching it again, or put it on pasteall.org? Thanks.%%%

David Brennan commented

2012-01-18 14:06:58 +01:00

%%%Yeah, the Upload kept on showing, "ERROR: Missing Parameters". Every technique I tried around this failed.

Currently, PasteAll.org is down in defiance against the government's SOPA evil, so I uploaded the two files via MediaFire:

BLEND File - http://www.mediafire.com/?myez5412dfz086k
Image - http://www.mediafire.com/?o6a538118g518z8

%%%

%%%Yeah, the Upload kept on showing, "ERROR: Missing Parameters". Every technique I tried around this failed. Currently, PasteAll.org is down in defiance against the government's SOPA evil, so I uploaded the two files via MediaFire: BLEND File - http://www.mediafire.com/?myez5412dfz086k Image - http://www.mediafire.com/?o6a538118g518z8 %%%

Tobias Oelgarte commented

2012-01-18 19:23:43 +01:00

%%%Had the same issue yesterday with 8x antialiasing (5x AA didn't show it). Very dark spots (3x3 pixels), which are -nan for all values. The only thing i know is, that it is at a point where a faces normal is in right angle to the camera. Guess it is something like a terminator problem for faces that aren't visible anymore for every sample. 7 samples hit the triangle, 8th sample misses it and returns -nan as value. Mitchel filter for AA makes 3x3 pixels out of it. This problem exist for a longer time. Had in a video that i made on 17.10.2011.

Both scenes used a raytraced soft lamp. But more importantly. It was always at the "edge" of objects with a baked normal map and the object has to be partially shadowed. If i remove other objects and it is not inside the shadow, it wont occur. I appended a screenshot "Bildschirmfoto-62.png" which shows the problem directly above the mouse.%%%

%%%Had the same issue yesterday with 8x antialiasing (5x AA didn't show it). Very dark spots (3x3 pixels), which are -nan for all values. The only thing i know is, that it is at a point where a faces normal is in right angle to the camera. Guess it is something like a terminator problem for faces that aren't visible anymore for every sample. 7 samples hit the triangle, 8th sample misses it and returns -nan as value. Mitchel filter for AA makes 3x3 pixels out of it. This problem exist for a longer time. Had in a video that i made on 17.10.2011. Both scenes used a raytraced soft lamp. But more importantly. It was always at the "edge" of objects with a baked normal map and the object has to be partially shadowed. If i remove other objects and it is not inside the shadow, it wont occur. I appended a screenshot "Bildschirmfoto-62.png" which shows the problem directly above the mouse.%%%

Tobias Oelgarte commented

2012-01-18 19:27:27 +01:00

%%%Oh, now as i looked at the screenshot, it are two different problems. Can we move my comment to a new bug report? Don't know if both issues are related.%%%

David Brennan commented

2012-01-18 19:31:48 +01:00

%%%Tobias,

Even though our two issues are not identical, I believe that they are still connected. I think this because you wrote, "The only thing i know is, that it is at a point where a faces normal is in right angle to the camera. Guess it is something like a terminator problem for faces that aren't visible anymore for every sample." This seems to match up with my issue - although the resulting artifact was different (a spike versus a block).

(Also, very great looking artwork, in my opinion!)%%%

%%%Tobias, Even though our two issues are not identical, I believe that they are still connected. I think this because you wrote, "The only thing i know is, that it is at a point where a faces normal is in right angle to the camera. Guess it is something like a terminator problem for faces that aren't visible anymore for every sample." This seems to match up with my issue - although the resulting artifact was different (a spike versus a block). (Also, very great looking artwork, in my opinion!)%%%

Tobias Oelgarte commented

2012-01-18 19:41:26 +01:00

%%%David,

I will keep my files as they are. If your issue is solved and mine is still there, then it are two different reasons. If im lucky, then my problem will also goes away.
PS: That was a strip down of the entire scene, trying to remove as much as possible while still having the error. Unluckily the objects are a part of different libraries, which makes it hard to upload for investigation. Thats why i will be waiting. ;-)%%%

%%%David, I will keep my files as they are. If your issue is solved and mine is still there, then it are two different reasons. If im lucky, then my problem will also goes away. PS: That was a strip down of the entire scene, trying to remove as much as possible while still having the error. Unluckily the objects are a part of different libraries, which makes it hard to upload for investigation. Thats why i will be waiting. ;-)%%%

David Brennan commented

2012-01-18 19:44:36 +01:00

%%%Tobias,

Since the bug/glitch was still visible, I believe that, even the stripped-down file you offered should be enough for the good people of Blender to investigate. Hopefully, yeah, they're connected and they debuggers can kill two birds with one stone. (Or many more birds, considering that there are probable other users with this problem who didn't bother to share it.)%%%

%%%Tobias, Since the bug/glitch was still visible, I believe that, even the stripped-down file you offered should be enough for the good people of Blender to investigate. Hopefully, yeah, they're connected and they debuggers can kill two birds with one stone. (Or many more birds, considering that there are probable other users with this problem who didn't bother to share it.)%%%

Brecht Van Lommel commented

2012-01-18 21:35:26 +01:00

%%%David, I can't redo the problem with your .blend file. The image in the render is not included, and replacing it with another image doesn't give me any artifacts. It would be useful if you could include the image, and/or strip down the .blend file removing other meshes, materials, images, lamps etc to pin down exactly which feature is causing the problem. Also you didn't mention your operating system and the exact blender version used.

Tobias, I think the problem you're seeing is a different one. That's clearly a NaN value problem which is caused by things like division by zero, the artifacts in this report are not because of NaN values, they wouldn't look "antialiased" then. Probably it's best to open a separate bug report.%%%

%%%David, I can't redo the problem with your .blend file. The image in the render is not included, and replacing it with another image doesn't give me any artifacts. It would be useful if you could include the image, and/or strip down the .blend file removing other meshes, materials, images, lamps etc to pin down exactly which feature is causing the problem. Also you didn't mention your operating system and the exact blender version used. Tobias, I think the problem you're seeing is a different one. That's clearly a NaN value problem which is caused by things like division by zero, the artifacts in this report are not because of NaN values, they wouldn't look "antialiased" then. Probably it's best to open a separate bug report.%%%

David Brennan commented

2012-01-18 21:53:21 +01:00

%%%Brecht,

Here is a link to the .PNG that was the texture of the plane: http://www.mediafire.com/?l3yjxmlu91ua8hb

However, I just rendered it without the texture and the black "spike" showed up just on the texture-less plane, too! (So I believe that we can rule out the texture as the issue.)

With the particular arrangement in the .Blend I uploaded, frame #41 was the only frame where the spike artifact appeared - at least for me.

OS - Windows, 32-bit
Blender 2.61
as well as r.43117

(NOTE: The reason that I use 43117 is because of the constant rendering issue - with the frames getting pale - that was reported and then fixed here: http://projects.blender.org/tracker/?func=detail&atid=498&aid=29692&group_id=9 )%%%

%%%Brecht, Here is a link to the .PNG that was the texture of the plane: http://www.mediafire.com/?l3yjxmlu91ua8hb However, I just rendered it without the texture and the black "spike" showed up just on the texture-less plane, too! (So I believe that we can rule out the texture as the issue.) With the particular arrangement in the .Blend I uploaded, frame #41 was the only frame where the spike artifact appeared - at least for me. OS - Windows, 32-bit Blender 2.61 as well as r.43117 (NOTE: The reason that I use 43117 is because of the constant rendering issue - with the frames getting pale - that was reported and then fixed here: http://projects.blender.org/tracker/?func=detail&atid=498&aid=29692&group_id=9 )%%%

David Brennan commented

2012-01-25 05:59:13 +01:00

%%%I would just like to note that, for other reasons, I re-did much of this Scene and even changed the texture. The "spike" problem in rendering is STILL there.%%%

Miika Hamalainen commented

2012-01-28 09:12:09 +01:00

%%%I can redo the problem on a 32-bit build, a 64-bit build it works fine. So it's clearly a precision issue.%%%

Campbell Barton commented

2012-02-01 05:22:13 +01:00

%%%attaching files linked above.%%%

Brecht Van Lommel commented

2012-02-02 22:46:31 +01:00

%%%Fix in svn, thanks for the report.%%%

Brecht Van Lommel commented

2012-02-02 22:46:31 +01:00

Changed status from 'Open' to: 'Resolved'

Brecht Van Lommel closed this issue

2012-02-02 22:46:31 +01:00

David Brennan commented

2012-02-05 22:09:13 +01:00

%%%Thanks very much to Brecht Van Lommel and anybody and everybody else who figured this issue out. I haven't tested out the SVN for several reasons, but I have no doubt that you guys solved it. Congratulations and thanks again!%%%

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Black "spikes" showing in some render frames #29921