Weird bands of disappearing texture in game?

[oktyabr 12]

I built a giant map with L3DT, assembled with Terrain Builder, compiled with makepbo. I had to scale up the png files in terrain builder so they would fit my terrain HF (I NEED a tutorial on how to work with really huge png images!) The mask, sat, and normal maps do not have any artifacts. But when I'm testing the terrain in game I get these long, straight bands of texture that cross the map, and crisscross, but disappear as I near the terrain they appear on, like a mirage! These also appear in the in game map. I also noticed blocks of textures were flashing around in buldozer but I decided to compile the map and see what happened anyway. The uniformity of them suggests some sort of overlapping problem but I have no idea where to look or what to look for.

Any help would be greatly appreciated! Thanks in advance!

---------- Post added at 09:35 PM ---------- Previous post was at 08:24 PM ----------

Disregard. Fixed the problem! Bad sampler settings. For posterity, in case anyone else finds this thread with the same problem, here is where I found the solution: http://forums.bistudio.com/showthread.php?177931-Satellite-Image-Overlap-Issue&highlight=make+forest

I'm still looking for a good tutorial on how to work with sat and mask images when you are working on a 40km.sq. terrain. That and the best way to plant forests without them looking like a gardener did it. :)

EDIT: I thought I had it fixed but no joy. 8192x8192 terrain, x5m = 40960. Sat, mask and normal are all 8192x png as well.

Edited April 30, 2015 by Oktyabr

[badluckburt 78]

The Tartan bug strikes again. There's no fix for this atm, except using a sat / mask image of different dimensions. Bohemia is aware of the problem but there's no ETA on when it will be fixed.

[Richie 330]

Welcome to the tartan club :) not much you can do until BI fix it, you can keep tabs on This Ticket for progress.

[oktyabr 12]

Thanks guys. :(

"There's no fix for this atm, except using a sat / mask image of different dimensions."

What do you recommend? The thread I linked earlier, it seems that at least one got rid of it by adjusting Samplers. I've tried just about every combination that Terrain Builder will accept but will pursue this further if it might prove fruitful? Or do you mean actually making sat/mask physically a different size than the xyz?

I see in the linked ticket, besides that it is supposedly "fixed in version", that this seems to only apply to "high resolution imagery (1/2/5m/pixel)"? I haven't tried switching to larger than 5m but I guess I would if it might help.

[badluckburt 78]

Thanks guys. :(

"There's no fix for this atm, except using a sat / mask image of different dimensions."

What do you recommend? The thread I linked earlier, it seems that at least one got rid of it by adjusting Samplers. I've tried just about every combination that Terrain Builder will accept but will pursue this further if it might prove fruitful? Or do you mean actually making sat/mask physically a different size than the xyz?

I see in the linked ticket, besides that it is supposedly "fixed in version", that this seems to only apply to "high resolution imagery (1/2/5m/pixel)"? I haven't tried switching to larger than 5m but I guess I would if it might help.

I'm afraid I have no recommendations, I've spent a night with Schulzit a few months ago trying to figure out 'working' settings for 20480x20480 map but no luck. Also, that's not the only ticket about this bug but since the bug-tracker doesn't seem to let me search my own comments, I can't find it atm but it's a lot older than Richie's. If I recall correctly, there are a bunch of Arma 2 user-maps that also exhibit this bug so it's probably a bug in the code that determines which tile goes where on the map when rendering.

I do hope they find some time this year to get it fixed because it's really stopping people from creating larger high-quality maps.

[Richie 330]

Just carry on as normal and hope it's fixed by the time you're ready to release :)

At least the problem has been narrowed down and the ticket is assigned.

[Uro 220]

I NEED a tutorial on how to work with really huge png images!

Save them as plain TIFF's or TIFF's using ZIP compression - this will dwindle your file sizes to something more manageable - The save time from Photoshop will be longer for the zip compressed version - depending on how you have PS configured (Mine take around 5 mins for a single 20480 surface mask image, ~8 mins for the sat mask of the same dimensions.).

TB Import time for a single 20480px tiff is ~8 minutes for me and I have 4x20480 sat & surface mask images.

When you compare that to waiting an era for PS to work with large PNG/BMP's and then crashing after a few hours, the 5+8 minute time spent working converting the filetypes is worth it. :)

Unlucky with the tartan dude, there are quite a few of us suffering from the issue, I created a feedback tracker issue about it - you might wanna hit it with an upvote to help push it through the system a little more - http://feedback.arma3.com/view.php?id=23685 :)

[oktyabr 12]

Just carry on as normal and hope it's fixed by the time you're ready to release :)

At least the problem has been narrowed down and the ticket is assigned.

LOL! And that's most likely how I will end up working it. I know I have at least months of work ahead of me, if I stay on this project, just carving roads, planting trees, and placing buildings, fences, and the like. Maybe it will be fixed before my hard work is ever released? ;)

But really, I get the gist of the problem but I don't really understand the important factor(s). Is it a problem with the terrain size itself? If so, how big is too big? Cell size? Reading the tickets it seems "no one" has had any issues at x10m (or larger). Or is there a way to get a smaller sat and mask to play nicely upsampled to fit a higher res height map? I don't really mind sacrificing sat quality IF that means I can keep a hires (5m?) terrain.

[Uro 220]

After Neo & Miro's investigation into the layers/rvmats generated from a few maps with these issues (which showed nothing wrong with them), he suspects it to be an issue with the engine itself. Since some A2 maps - that work correctly in A2 also have this issue when loaded in A3, we can only await the outcome of BIS' investigation into this issue, You can find his full explanation in the feedback tracker issue above.

It's not only satellite quality you will be sacrificing by going to a 10m/pixel sat/surface mask - it's surface detail.

There are a lot more pixels for a detailed surface mask in a 20480 mask image than in a 2048 mask image, and 1m/pixel looks way better in game than 10m/pixel, since 1 pixel = 1 surface texture.

I've spent literally weeks trying to figure out if this was an issue with my map and its sources et al, resizing/reproducing hundreds of GB of source files at different resolutions just to rule out anything in the processes I use for my source data.

It was only really when I posted up my issues in the map makers skype chat and explained everything I'd tried. Then the more experienced terrain guys started to look into my problem - along with many others having the same problem that we started to make some headway with the issue. The FB tracker issue has been assigned, so hopefully we will see some resolution to this issue soon.

In the meantime, keep on building! :)