Random pixels in exported images

#1

Checking some exports I noticed that formware creates various random pixels that repeat themselves for many layers.

Have a look at this image, I outlined the errors.

711
711.png1440x2560 33.1 KB

the pixels are even way outside the models

#2

That’s really bizarre…

If you regenerate that set of slices, do you see the spots happen in exactly the same place and layer numbers?

If you move your objects in the build space and reslice, do you see the spots in different location or the same place?

#3

the pixels start few layers from start and repeat themselves in every frame up to the end.

I just noticed them because I wanted to post-process the frames, I haven’t tried yet to move or change the objects, but I don’t feel comfortable using formware knowing that I could get random pixels.

I’ll keep you posted

#4

What a pity
I was just preparing my first test of formware with my brand new photon.
I need some custom supports with part on the build plate, and I will have to download chitu box now and look into it
I did not test it with MOAI due to lack of distortion correction.
SO for now formware is still just an icon on the desktop for me
Thank you for the info

#5

What i can think of that happens is that there are small horizontal in the mesh below it.
When you slice using the GPU; (as with the formware program) you will get these kinds of artifacts when your mesh contains horizontal holes.
Can this be the case?

I’m working on integrating our numerical slicer (svg/slc/sla) also as an alternative into the main slicing proces.
This will give you the option to prevent this kind of artifacts. Has other downsides though… (slower with many objects)

For more info how slicing works:
https://www.formware.co/article/SlicingSpeed

#6

the work above included 5 times the same object, still the pixels location is very random. shouldn’t the issue be the same for all 5 identical meshes?

I don’t think the 3d model has horizontal holes btw, but the STL could have some issue

#7

mm interesting… but are there parts below the random pixels? nothing?

Basically the GPU stencil buffer just sums up the negative and positive z compontents of normals.
So:
+1 for every downfacing normal (nz < 0)
-1 for every upfacing normal (nz > 0)
If the sum is bigger > 0, then you are inside
It’s hard for me to imagine anything going wrong there.

Afterwards the bytes are converted to a png image.
This might be something to look at.
Does it still happen when you switch to a 32bit png image format?
It might be in the PNG encoder from windows maybe? Also hard to believe; but you can’t know for sure

#8

regarding the 5 meshes; you are partially right.
The GPU has to rasterize it to pixels. (from triangle -> filled pixels)
Dragging of the objects in the 3dview is not on a grid. So you might have different pixel layouts per object…

Still it would be interesting to see the mesh below it.

#9

the PNG is already at 32bit.

below the random pixel there’s actually something. the random pixel just touch by a whisker the surface of the model and remains stuck on the image.

the weird thing is that it happens in two different random spots and not for all the 5 meshes.

also I sliced this on a laptop so maybe the slicing happened on CPU (can you slice on the integrated card?)

#10

Yes you can slice on an integrated card… should be no problem. Of course your memory is shared on your RAM. Find it hard to believe this would change anything…

If you could send me the model i could give it a try.

#11

I don’t have the model anymore, I used it for a test (that I will post in a moment in another thread) and then trashed BUT

I tried to slice another model right now and I got the same issue!

53
53.png1440x2560 31.6 KB

I’ll let you have the 3d model

#12

This is the model above

https://drive.google.com/open?id=1PY0FxuanUk_PAIOwEiifNEUKfTZcUa7q

Please note it’s just a test, I created the supports and exported, I don’t plan on printing it. It was just to check the “random pixel” issue.

I kinda believe every export has this issue…

#13

I had a look at the file, thanks.

So it seems i found 2 pixels in my export; both exactly on the inner border of the bottom base plate.
When you walk through the png’s in the folder with windows previewer you can see where they arise.

When i supported the model again without base plate and sliced again; i couldn’t detect any pixels on the eye.

So i think the base plate could be the bad guy in this case.

Could this be also in other cases where you noticed this?

#14

I tend to always use a base, so it could be indeed a problem with that.

I’ll check that and let you know

#15

I made some more tests. The problem is indeed with bases, any shape. Weird enough, by just moving the object by few mm sometimes solves the issue. this means that before printing you have to check is there are spurious pixels, if so move the object a little and reslice. viable but very annoying

#16

mmm ok thanks for checking. I will put it on the list to check.

It’s probably due to some overlaying triangles in the base generation.
Then the GPU doesn’t know what to draw.

So can be there is a mistake somewhere in the mesh generation.
Or the triangulation algoritm for the base plane… :frowning:

I think you could prevent it as well by using one of the different shapes.

#17

I’ve made various updates and improvements to the base plate.
Also found a couple of bugs in the generation of the shapes.

Double checked al generated meshes with Rhino for validity; all valid.

New release is uploaded.