Sign in to follow this  
Justin Toposim

10m mesh - a limitation?

Recommended Posts

Or is it just me?It seems as if the sim will only display 10m mesh for a radius of about 6 miles (same as the hi-res textures?). I noted this while testing my 10m mesh for Hawaii (Diamond Head, Oahu) where the Diamond Head crater provides a good example because of the distinct shape and obvious shift in location. I have replicated the effect in the San Francisco area.The series of images below show the changes seen as I approach the crater.While the 10m mesh covers the entire island, only the default mesh version of the crater is rendered when at a distance of about 7 miles.http://ftp.avsim.com/dcforum/User_files/3df0d8c446afaa34.jpgMoving only a bit closer, we see the 10m mesh coverage now extends just past the edge of the correct location of the crater, indicated by the yellow arrow. The green vegetation appears on the small portion of the wall of the crater included in this quadrant. (Although not apparent in this screenshot, this quadrant is at the edge of the transition to the next lower texture resolution.)http://ftp.avsim.com/dcforum/User_files/3df0d8e047096407.jpgFinally, just a bit closer again, and the 10m mesh includes the entire crater, now completely replacing the default mesh.http://ftp.avsim.com/dcforum/User_files/3df0d8f9474dc58a.jpgThis behavior is very different from 30m and 100m mesh, which are rendered to a distance of about 40km. While this in not a very important limitation for those of us who enjoy exploring the terrain up close, this very small area of coverage has significant implications for the use of such high resolution mesh when flying at high altitudes. Can anyone else confirm/explain this phenomenon?Test conditions:Terrain Mesh Complexity=100TERRAIN_MAX_VERTEX_LEVEL=21Regards,Stevewww.fs-traveler.com

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

HI Steve,Yes, I can confirm it. Although I think its radius is a bit further than 6 miles. Maybe a bit more. There may be an FS2002.cfg hack, but I've not found it yet.This is also why I've added a 19.2m base to all my 9.6m terrain meshes. Without it, you'll see a "wall" off in the distance beyond the radius of the 9.6m between the 40 km radius of the default terrain. The only way I figured out to get around it is to lay down some 19.2m under the 9.6m. You may get away with it if 38.2m is under it, though. But installing 9.6m on top of the default 76.2m/153m will reveal the wall. Seems to me there has to be a more gradual incremental buildup of LOD.The 9.6m is really only good for low and slow flying in a piper cub or other light or ultra-light aircraft. Which is not to say that kind of flying is without merit. But there is absolutely no benefit above about 8000 feet, yet even those high flyers can benefit on final approach. And it does make a nice difference on the ground at the airport.Best regardsJustinhttp://www.fsgenesis.net

Share this post


Link to post
Share on other sites

Oh, I forgot to add that this would explain the floating trees, too. My theory is that the autogen loads at the same distance as the lower-res mesh and stays there. Then the 9.6m terrain loads and voila! you have the floating trees. Only workaround I've found is to pull down the World menu, click Scenery Library, then OK to reload the terrain with the autogen. The autogen will then be at the proper elevation--at least until you fly a few more kilometers, then it's back to floating trees again.There must be an fs2002.cfg setting that will force the 9.6m to load further away and fix this.Justin

Share this post


Link to post
Share on other sites

Thanks Justin,>this would explain the floating trees, too. I agree; this was the original problem that prompted me to look further. >Seems to me there has to be a more gradual incremental buildup of LOD.I would think that the DLOD alogorithm would probably be able to manage the transition and eliminate the cliffs if the 10m mesh extended out as far as the others. It appears as though the MS developers anticipated our enthusiasm for detail and built in a check for mesh that would tax our current systems too much. This data is clearly handled differently.>There must be an fs2002.cfg setting that will force the 9.6m to load >further away and fix this.I hope so, but have not seen anything promising yet. I tried a different approach to see if I could force it to extend the range, removing the default Hawaii mesh. The sim then used a much lower resolution mesh (dem4km.bgl?), making things even worse. I will try moving that file next, and will also test the mesh with a 30m base and see what happens. Steve

Share this post


Link to post
Share on other sites

>I would think that the DLOD alogorithm would probably be >able to manage the transition and eliminate the cliffs if >the 10m mesh extended out as far as the others. It appears >as though the MS developers anticipated our enthusiasm for >detail and built in a check for mesh that would tax our >current systems too much. This data is clearly handled >differently. This is also one of my theories, and it seems MS has been quite conservative for highres mesh.My other theory is that MS stuffed up and this is a bug. The 'DLOD threshold' variable may be calculated wrongly in FS2002. Why does adding adding lowres terrain fix the issue (I also observed the same thing)? Something really isn't right there and I doubt it's as simple as just fixing the FS2002 config...Cheers, Christian

Share this post


Link to post
Share on other sites

Hallo Justin,am I wrong (yes I am pretty sure) but the highest LOD for fs2k2 is at level 13 corresponding to a 19.8 m grid step? How do you get a "LOD14" (9.6 m grid step)?Cheers, Stefano.

Share this post


Link to post
Share on other sites

Hi Stefano,Theoretically, the highest LOD is 15 (1.2m). You get LOD=12 (9.6m) by setting the LOD value at the top of you INF, and using 10m source files.In order to see it in FS, you have to set the TERRAIN_MAX_VERTEX to at least 21 in fs2002.cfg.Hope this helpsJustin/FSG

Share this post


Link to post
Share on other sites

Hi Justin.Altering the TERRAIN_MAX_VERTEX to 21 does allow the viewing of more refined mesh, but that does present problems with coastal areas that need to use LWM flattens to "re-mesh".Once an area is flattened with an LWM, the mesh is forever gone. That presents a huge problem for people using photoreal or newer LWM methods to alter land/water shapes. The land may now extend into the old ocean, but the mesh has been destroyed by the LWM flattening of the old coastline.The only cure is a tedious "remeshing" of the affected area with tiny LWM polygons ( actually equivalent to an LOD13 mesh ). When the TERRAIN_MAX_VERTEX is set to 18 or 19, there is no visual problem, but when it is raised beyond that, the "remesh" becomes a very ugly oversample-looking job.Unfortunately, there is no other way to reclaim old LWM flattened areas. So that is why I suggest not to set the TERRAIN_MAX_VERTEX to over 19, as in the future, remeshing may become a popular means of reclaiming elevations.-----------------One thing I haven't tried is to reclaim LWM flattens with an LOD13 or greater mesh. I wonder if you have seen that effect near coasts.. with highly dense mesh ( > LOD12 ) destroying lake or ocean flattening? That might be a better "cure" for the remeshing problem, if it is possible.Dick

Share this post


Link to post
Share on other sites

HI Dick,I'm not sure I understand the issue. But I think I encountered something like what you describe when I was doing experiments in FS2000 with 1:100000 data and making rivers out of coast polygons. The default flatten polygons are not affected by exclusion areas and so they had to be disabled, renamed, or otherwise removed from an active state--and replaced by the new ones.Not sure I understand, then, the need for "remeshing." Given that if any part of the quadrant is touched by a flatten poly, the whole quadrant is flattened. An LOD=8 or LOD=7 quadrant could make a significant distortion. Is this what you mean? In which case, higher-res terrain mesh makes for smaller quadrants and the effect is minimized.Or maybe I'm just not grasping what you describe. That's possible since I've not yet delved very deeply into that area yet. (Been too busy with the simple stuff like terrain mesh and landclass :-)).JustinFSG

Share this post


Link to post
Share on other sites

Well, gentlemen, I have done some additional testing, with surprising results.First, some clarification of my original post:* the 10m mesh was constructed using an LOD of 12* while I referred to the 'DLOD alogorithm', because I think of it as dynamic, I should have used the proper jargon for this process: CLOD (continuous LOD)In my next round of testing I used 10m source data for Hawaii, with LOD levels of 10, 11, 12, and 13, and TMV (terrain_max_vertex) levels of 19 and 21. In this more comprehensive set of tests, I looked for the impact of LOD and TMV setting on differences in radius rendered with the high resolution mesh and in the amount of detail rendered in the foreground.Results:* The effect noted originally does not seem to be limited to 10m LOD12 data. The radius rendered with high resolution mesh is inversely proportional to the LOD of the data; the smaller the LOD, the further out it is rendered. These changes are quite distinct, and can be seen with either TMV value (but see below).* The detail in the foreground with the TMV value of 21 is essentially identical at all LOD levels. While switching between screenshots may reveal a few very minor differences, I do not believe anyone could identify which image was made with which mesh.* The tests for foreground detail with the default TMV value of 19 were particularly surprising. The improvement of LOD10 mesh over the default is significant. LOD11 mesh detail is identical to LOD10. LOD12 and LOD13 mesh were not rendered at all; the sim used the default mesh instead! (I confirmed this effect with mesh for Mt Washington NH as well). My personal conclusion (at least in the context of these results):* LOD10 is the optimal LOD for 10m source data: it is displayed over the largest radius, requires the least amount of processing overhead, will be used even with the default TMV value of 19, produces much smaller files (~6MB vs ~19MB for LOD11 and ~60MB for LOD12 - for all the major Hawaiian Islands), and no redundant mesh is required as a "base". All this with no significant loss of visual detail while flying. And of course the TMV value of 21 provides a substantial improvement in detail.So much for conventional wisdom and common sense. :)Steve

Share this post


Link to post
Share on other sites

HI Steve,So, you're saying that regardless of the LOD setting, the output will conform the resolution of the source data?IOW, if you tell the resampler to output LOD=10 (38.2m) with 10m source files, the resampler will put out 9.6m terrain? So LOD=10 output is just as detailed as LOD=12 output if you use 10m source to begin with? (referring to "with no significant loss of visual detail" above).JustinFSG

Share this post


Link to post
Share on other sites

Hi Justin,Not exactly. I assume the resample process is operating as we imagine - the output file sizes are certainly different. I am only looking at screenshots taken while flying in the sim, and assume the output (final display) is conforming to (constrained by) limitations in the sim, not the source data (bgl). That said, it is very curious that, with TMV=21, we see virually the same image created from 800k and 18MB bgl files. This could be an artifact of the resample process but it seems more likely to be some limitation in the amount of resolution the sim can handle. I expected us to run up against some limit eventually, but that all tested LODs are resolved to the same result is disconcerting. (And I do not believe increasing the TMV value further will help.)I did not test lower LOD values here, but I seem to recall seeing less detail with 30m data at LOD09 than at LOD10. This would support the idea that the limitation is in the display resolution and not in the resampler. The sim is now able to discriminate between the different resample output bgl files and render them accordingly.Maybe I will do some more testing, if I can determine some way to distinguish where the loss takes place.From a practical perspective, it does not really matter, in the short term. LOD10 seems to be ideal for now, but if the display resolution increases much in future versions, we may have to consider resampling at a higher level. Meanwhile, I have hundreds of 10m LOD12 files to reconsider!Steve

Share this post


Link to post
Share on other sites

Hi Justin.Actually, all LWM flattening ( default coasts, lakes, and airfields in FS2002 ) are forced at LOD13. And you can actually adjust a single point ( about 4.8 meters squared ), which is 1/65536th of the LOD13 Area. The LOD13 Area is split into a point grid 256x256. FS2002 LWM land, water and flattening have nothing in common with FS2000.As I stated, once a coastal flatten is applied for the water, it forever destroys the underlying mesh. You can delete the LWM BGL causing the problem, but then be prepared to spend months making a replacement LWM BGL for the coast, as the span covered by the default LWM BGLs is usually quite large, and our tools are rather crude.Photoreal designers are particularily hard-hit, as often their coasts do not match the default.So "remeshing"... faking a mesh with tiny LWM elevation polys is the only sane recourse... and tools are going to be developed to simplify that process.Then you don't need to replace any default BGLs. But with TERRAIN_MAX_VERTEX set to 21, instead of not greater than 19, this remeshing is terrible looking, instead of acceptable. But as I said, it's our only recourse. So I am advising everyone to not raise the TERRAIN_MAX_VERTEX to over 19.As more designers start adjusting coastlines, and remeshing problem spots with LWMs to raise reclaimed coastal areas, this will become an issue. Just a heads-up.Dick

Share this post


Link to post
Share on other sites

HI Steve,Yeah, me too. I may do another localized area or two, but any plans I may have had for wide-area 10m stuff is back on the shelf for now.I think we'll let that genre to Raimondo.JustinFSG

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this