Sign in to follow this  
Guest sgreenwood

FS max terrain resolution/LOD?

Recommended Posts

Does anyone know what the maximum mesh resolution FS2004 and FS2002 can display is?I know the max resolution was reduced in FS2004 but what is it at now, and what was the max allowed in FS2002?This discussion is very interesting:http://www.visualflight.co.uk/forums/topic.asp?TOPIC_ID=1742It suggests that the max resolution is FS2002 was 19m, not 9.6m, and the max res supported in FS2004 is 38m. I've read that FS2004 supports 19m, but at other places, I've read that it only supports 38m :-hmmmAny scenery designers why know more about this?

Share this post


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

Hi Jimmi,There are three issues here: * whether the mesh at a given resolution is used* how much of the data in that mesh is actually rendered in the sim* whether the difference in detail at different LOD values is meaningfulBoth versions will use/display mesh created with LOD values up to at least 13 (I haven't tested higher values).FS2004 shows no additional detail when the LOD is increased beyond 11. You can sometimes see differences between LOD 10 and LOD 11 mesh created from 10m source data, but probably only in screenshots. So LOD 10 is the maximum practical value.FS2002 shows no additional detail when the LOD is increased beyond 12, but does reveal additional detail when going from LOD11 to LOD12. But not much.The two versions handle mesh VERY differently, so it is difficult to compare the two directly. For example, FS2002 may show more detail with LOD12 mesh than with LOD11 mesh, but that detail may still be less than the detail revealed in FS2004 using LOD11 mesh! It will certainly be different.So, while we can talk about the resolution of the mesh we are providing, I don't think we can yet talk intelligently about the actual resolution we see in the sim. We can only describe the apparent effects of different source data. Certainly such a discussion would have to address CLOD issues, and none of the posts in the discussion you provided a link for mentions it. If the sim actually revealed all the detail in LOD11 mesh, it would only be at the very bottom of the screen. CLOD algorithms start collapsing polygons fairly quickly as you move up to the horizon.Since the rendered resolution of the terrain varies with the vertical position on the screen, screen resolution, aircraft elevation, distance from the subject, degree of zoom, ..., any attempt to document the effect of mesh on scenery so others can confirm it would have to include considerably more information than just the coordinates of the location!And how does theTMVL parameter work? Does it relax some constraints on the CLOD algorithm, directly or through a lookup table that may alter multiple parameters, based on a single setting?) Hmmm.In addition to the TMVL setting there are those other two pesky parameters introduced in FS2004, TERRAIN_MIN_DEM_AREA and TERRAIN_MAX_DEM_AREA. I saw no mention of those either in the referenced discussion. What impact do they have on the detail we see, and how do they interact with TMVL, CLOD effects?Here's a link to the web page I put up over two years ago when I first discovered the effect of the TMVL parameter. It's pretty basic, but does suggest one type of testing one can do to learn more about mesh.http://fs-traveler.com/tmv-example.shtmlIf you would really like to do some experimenting, download my mesh test data set for Maui. This is artificial mesh at LOD 9-13, each at a different elevation so you can see them all at once, or turn them off and on individually. Compare the appearance of the same mesh in FS2002 and FS2004. The dramatic difference must be considered in any discussion of how mesh is rendered. You won't learn much about detail rendered within an LOD, since there is none, but it does provide useful insights as to how mesh works. And it suggests many additional tests one might conduct after further editing of this basic mesh data. Tip: By enabling each LOD in turn and approaching the coast with the GPS terrain map visible, you can see when the artificial mesh reaches the coastline. You can then estimate the approximate effective radius of the mesh using the map and known distances between the islands.Here's the link (<100kb, since flat terrain compresses well):www.fs-traveler.com/cgi-bin/maui-lod-test.zipIf a good picture is worth 1000 words, I wonder how many words a good experiment is worth.It seems to me that it is almost hopeless to try to learn about mesh using complex real-world data. By creating test mesh, one has complete control over the source the sim is working with, and can then see exactly how the sim handles known data.And without a good understanding of how mesh works by itself, any effort to describe how it interacts with scenery is, well, ... ;-)StevePS: Thanks for bringing this discussion about mesh to THE mesh forum on the web!

Share this post


Link to post
Share on other sites

[table width=700][tr][td width=300]I decided to do a bit more testing myself!I constructed LOD10, LOD11, and LOD12 mesh from 10m NED data for central New Jersey (USA).Rather than trying to demonstrate the differences using traditional screenshots, which would be useless in this region, I ran my flatten utility, which reports ground elevations and coordinates, among other things.The image on the right includes readings with the default, LOD10, and LOD11 mesh (LOD12 was the same as LOD11, as expected).These results are repeatable. I swapped the mesh several times, at two locations, with comparable results.So, while there is virtually no visual difference in the rendering of LOD 11 mesh in the sim (vs LOD10), it is being used and differences can sometimes be seen, at least where the variation in terrain elevations is modest. The small differences between LOD10 and LOD11 mesh probably become irrelevant both as we move away from the aircraft and in more irregular terrain, where algorithms for rendering the screen image start aggregating the data.Steve[/td][td valign=top" align="center]http://forums.avsim.com/user_files/75936.gif[/td][/tr][/table]

Share this post


Link to post
Share on other sites

Hi Steve,thanks for the (continued) testing and neat summary; this is one of those threads that will be very useful for future reference and similar questions.Given the common belief that bigger is better it's good to have a "reality check" once in awhile; there are indeed boundaries in FS with regards to accuracy and visible detail.Cheers, Holger

Share this post


Link to post
Share on other sites

Come on guys! It would have been nice to have been told that an entire discussion was being so comprehensively rubbished on a different forum to the one on which it started (right of reply and all that?).I can understand that Steve feels that his approach is the only valid one, but I don't see how such offhand dismissal of other people's contributions is going to benefit the flight sim community in the end. We should be working together to build up and share knowledge, not knocking one another's efforts. I don't mind fair discussion, but I don't think this is it, so I will refrain from challenging the various factual inaccuracies and false assumptions as I see them in Steve's post - I don't think it will add anything to the current tone of the discussion.Steve, if you want to take another look at your reply, maybe we can give it another shot? In the meantime, if anyone wants to judge the content of the Visual Flight discussion for themselves, I would suggest that they read it here:http://www.visualflight.co.uk/forums/topic.asp?TOPIC_ID=1742NB. The discussion currently runs to two pages, and is a little heavy going in places, so stick with it.JohnVisual FlightCreator of [a href=http://www.visualflight.co.uk/vfrterrain]VFR Terrain[/a] and [a href=http://www.visualflight.co.uk/photoscenery]VFR Photographic Scenery[/a]

Share this post


Link to post
Share on other sites

Hi John,Please note measured delay in responding, lest this post appear "offhand" as well. ;-)Let me begin by complimenting you on the graph you presented in your forum. It does suggest avenues for further testing.There was no intent to offend you by posting here, this is where the question was asked and the forum I follow for current information about mesh. (I did respond in your forum as well, providing you with a new conceptual tool for your analytical toolbox.) Upon re-reading my post, however, I would like to re-classify it as a generic rant, certainly not directed at anyone in particular.I'm always happy to engage in discussions about how things work, but these are rarely fruitful unless there is a careful definition of the terms being used. I usually try to provide sufficient information, tools, and sometimes even data, so others can attempt to reproduce my findings, if interested. I've done a bit more testing. :-) I created two artificial mesh files (5000m, LOD11) with one column and two adjacent columns of 6000m elevations, respectively (available on request!). I can support your claim that increasing the TMVL setting in FS2004 to values above 20 had no effect (at least for my limited test conditions). What's going on with the rendering of the mesh/polygons/textures is far less clear. My test mesh does not involve any other scenery (roads, water, ...), so interactions are not considered at this point.Below is an image of the DC3 sitting on top of a similar ridge created using a flatten.bgl. If the wingspan is 29m as reported in the aircraft preview screen, then the width of the flatten is about 45m, not 38.2, so it seems the situation is even worse than you suggest. (How did you come up with your value?) This is the narrowest surface I can get FS2004 to display.. If I reduce the width any further, the top shrinks to a narrow ridge, probably a single elevation point wide.http://forums.avsim.com/user_files/77828.jpg"So, while we can talk about the resolution of the mesh we are providing, I don't think we can yet talk intelligently about the actual resolution we see in the sim. We can only describe the apparent effects of different source data."I'll now amend that statement, based on my tests, and note that is looks like the minimum width polygon FS2004 will display, directly under the aircraft sitting on the ground, is about 45m, (or 38, but certainly less than FS2002).Here's a link to a quickly constructed page with a few more related images:www.fs-traveler.com/ss-04-tmvl-lod.shtmlIntrigued by your graph, I've created a small utility that may interest you. http://forums.avsim.com/user_files/77829.jpgThe button is used to adjust for the different "heights" of different aircraft, allowing us to standardize the readings. By setting the aircraft on, for example, a lake surface (where mesh and polygons do coincide) and clicking this button, the utility calculates a Calibration Value, which is used to set the aircraft Altitude to that of the ground. From then on, all Difference values read with the plane on the ground will reflect the actual distance between the polygons and the mesh. Or something like that.It's available at www.fs-traveler.com/cgi-bin/fst-elevation.zipNow, about your evidently infrequent visits to this Forum for "working together to build up and share knowledge, ..." :-)Looking forward to more revealing graphs, Steve

Share this post


Link to post
Share on other sites

SteveI could not ask for a better reply. Worth the wait - thanks. :-) Now is probably a good time to post a link to a set of pages I wrote a few days before replying to your post, both to clarify my thoughts, but also anticipating the question you've now posed, namely "How did you come up with your value?". The answer can be found from page 6 onwards, but you may want to read the pre-amble as well:* [a href=http://www.visualflight.co.uk/vfrterrain/terrain_resolution/index.htm]FS2002 vs FS2004 Terrain Resolution[/a]There's a few things I wasn't happy with when I re-read the above just now, but given that it was written between midnight and four in the morning I hope you'll excuse any deficiencies (at least with web pages I don't get avsim telling me it's too late to edit!). I hope you find something of interest in it anyway.I'm right in the middle of a few other things right now, but I'll study your post and new web page in more detail as soon as I get a chance, and maybe discuss a couple of things from your previous post which I had put off for the time being."Now, about your evidently infrequent visits to this Forum for "working together to build up and share knowledge, ..." :-)"Not all that infrequent, but the Visual Flight forums do keep me fairly busy. :-)John

Share this post


Link to post
Share on other sites

Hi all,I read John's document on terrain resolution with great interest. Well, it didn't really come as a complete surprise, but now it's forever demonstrated that in Fs9:1. There is no hope of seeing any mesh detail smaller than 38m;2. TMVL can be happily set at 20 and left there forever, if one's machine power allows.A third conclusion appears to me as a direct consequence of the above, but I would appreciate your opinion:3. Any re-meshing should be done at LOD10 (or coarser).I still have two questions, just to understand if finer meshes can still be of some value:4. When Fs9 finds a LOD11 mesh file (at runtime), what does it do:4a. Ignores the file completely?4b. Considers only the subset of points coincident with a LOD10 grid?4c. Averages the elevation of each LOD10 point in some (weighted) way with that of the surrounding, not-LOD10 ones?5. When the resamplers (Fs2k2 and Fs9, separately) are fed with a LOD11 data file and asked to compile as LOD10, what do they do (same as above)?Cheers,AdrianoP.S.: About Steve's experiments with a DC-3, I found very convenient using the quasi-aircraft proposed by Lou Volland (chb_m_lv.zip) which is just a precise 50m-sided flat square box to be slewed around. I removed any and all contact points, so that it squats on terrain at no offset elevation - be very quick to press Y, or you'll crash. (I also extended the concept in order to be able to work on contour lines, but that's another story).

Share this post


Link to post
Share on other sites

Hi Adriano,I may be able to help a bit.>3. Any re-meshing should be done at LOD10 (or coarser).Generally true. The limited radius of higher LOD mesh is reason enough for some of us.>4. When Fs9 finds a LOD11 mesh file (at runtime), what does it do:4b seems likely, since the terrain appears to be identical. >5. When the resamplers (Fs2k2 and Fs9, separately) are fed with a LOD11 data file and asked to compile as LOD10, what do they do (same as above)?They do the same thing with all source data; use the NW corner coordinates and the cell spacing information to create a new terrain model, then calculate the elevation values for the positions which coincide with the grid of the LOD requested. More or less. Your quasi-aircraft approach is interesting. I went back and remeasured the distance using John's approach and agree with his measurement of 38m (thanks John, and nice work on the web page). I've just added another page to my website reporting the results of recent tests relating custom source data, how accurately it is handled by resample, and what it looks like in the sim. It looks like elevation data is rendered using its own set of polygons, relatively independent of the mechanism used to position textures (a second set of polygons?). http://www.fs-traveler.com/rendering.shtmlSteve

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