Posts by Ian C

    I tried something recently that I'd never tried before - and it failed. I wonder if anybody has any thoughts as to why .....

    I was geoconverting a level 9 area where mostly the Bing images were the better of the two. However there was a patch where the Google images were superior. So what I did was to download both sets of images using AeroScenery. I then deleted the poor quality Bing images in the "18-stitched" folder and replaced them with the better quality Google images. I also changed the "g_" to "b_" for the Google images and did the same for the .aero files. When I geoconverted everything seemed to go OK, but when I flew over the area it was obvious that most of the tiles were missing in that area. (I was just seeing the low res tiles beneath.)

    Is it possible to do what I was trying to do? Did it fail because "g_18_stitch_5_4.png" is not necessarily exactly the same size as "b_18_stitch_5_4.png" (for example)? Was it even necessary to rename the png's? Was it necessary to bother about the .aero's at all?

    Interesting. Is that happening with the default terrain mesh? I had assumed that my problem was related to my custom mesh. Also I find that it never happens at the point where I start the simulator - it's only after I have flown a few km from my starting point.

    The size parameter is indeed a load point of the TSC. This parameter is measured in meters, but should only be used as a guide as there are many other variables used for when the TSC is loaded, some of these are based on what your graphical settings are set at. As a guide, lets say that your scenery area for any particular TSC is 5 nautical miles x 5 nautical miles, you first want to convert the miles to meters, so 9260 meters, now take half of that (since its your center point that's important here, so 4630 meters. Now you can use this number if you want the scenery to load as soon you are at its edge or you can add some meters to it so that the scenery loads from a ways out, lets say you want your TSC to load 1 nautical mile from the edge your number will be 6482.

    so size=6482 in the referring TSC will load into memory 1 nautical mile from any direction.

    This was my understanding of the "size" parameter - though I usually also multiply by root 2 (1.414) so that the scenery loads at the corners of my cultivation square. I see the point about performance and TOC size, but I never had the floating building problem before I added high res TTH files - so I suspect they must be the culprit. Also the buildings tend to float or sink where the high res mesh makes most difference - i.e. on steep slopes etc. The problem can be cured by restarting the simulator, but it then creeps in again after I've flown a few miles. Even then, 90% of the buildings are OK - and it never affects my trees for some reason.

    That's also the way I see it.

    As far as I understood, the size parameter in the TSC file is a markup in meters above the default load radius, which seems to be some 30km...



    There's something I'm confused about here. I've always assumed that the "size" parameter in the TSC file is measured from the longitude and latitude in the TSC file - i.e. from a fixed point on the earth's surface. I thought that it determined how far from that point the cultivation could be seen. In that case the size boundary is a fixed circle on the earth's surface so there would be no way of guaranteeing that you hit it after the load distance of the TTH files, since it's not centred on the aircraft. Maybe I've got the wrong idea?!

    Yes - there are some things that are a bit of a mystery. In a situation like this the only way is probably to experiment with various parameters and see what effect it has.

    I don't know about Aerofly but what you describe can be seen in the FSX based Simulators as well, reason being exactly what you guessed - buildings being positioned on a low res mesh before the higher res mesh loads in. I guess the only workaround would be to change the distance at which the high red meshes is loaded, which would probably decrease performance significantly. I think there's a reason why most simulators ship with an average resolution mesh by default, you have to make a compromise between visual quality, performance and susceptibility to artifacts and visual errors.

    I suspect you may be right. The ORBX Netherlands scenery has a 5 m terrain mesh (I think) but the Netherlands is a very flat country so you probably wouldn't notice anything unless your house was on top of a dyke!

    well, the size of the tocs must be smaller than the size off the max resolution elevation tile.

    then you have the size in the tsc file of the toc. This should not be bigger than the distance, the max-resolution-elevation tile gets loaded. Possibly you have not corrected the tsc file after splitting your tocs?

    At least for my comores islands it was the solution. If you don't reduce the size, the tocs possibly gets loaded when only the low res elevation is in memory.

    That's an interesting point about the relative sizes of the TOCs and high level tiles. I assume that means the total size of the TTH tiles within the circular area covered by the TSC "size" parameter? But I'm confused, because I would have thought the TOC would need to be the larger of the two, so that the buildings would load later? In any case I will experiment with it.

    I did correct my TSC "size" parameter to "5000" after splitting the large TOCS. Previously it was "80000".

    How big are your cultivation regions? Earth curvature is modeled in aerofly, which means big enough models will tend to hover above the terrain.

    Does your situation improve it you split the cultivation into smaller chunks?

    Splitting it into smaller chunks doesn't seem to help. I've gone down from 100 x 100 km to 20 x 20 km and it makes no difference. Also I never had the problem before I introduced higher res terrain mesh, even with 100 x 100 km areas - so I guess it's somehow related to the new terrain mesh. I tried putting the TTH files on the (SSD) C-drive, thinking that that would make them load faster - but that doesn't help either.

    I'm beginning to wonder whether this isn't inevitable if you have an improved terrain mesh. With the default mesh it doesn't make any difference if the higher level TTH tiles load later than the lower level ones (starting at level 7). It doesn't alter the shape of the terrain when they load because the heightmap data is low res (roughly level 7) anyway. However with higher res data it does make a difference to the shape of the terrain when the higher level TTH files load - so if a building "sees" a lower level TTH tile before it "sees" a higher level one it's going to "sit" on the lower level tile so to speak - and therefore be at the wrong height. Perhaps it would help if there was a way of making the buildings "wait" intil the mesh has fully loaded?

    This is pure speculation on my part BTW - and may be completely wrong!

    It's also a mystery why it doesn't affect my trees. Why should buildings behave differently from trees?


    I know that this has been talked about before but I've read a few different explanations (and possible cures), so I was just wondering what the latest thinking is. I also want to clarify my own undestanding of the issue.

    It's only become an issue for me since I started improving the terrain heightmap in certain areas. It's never a problem where I start from. It only appears after I've flown a few miles from my starting point. If I then restart at that point the problem goes away. Cultivation may be either floating or buried, and the effect seems to be greatest where the discrepancy between the default and high res terrain meshes is greatest (ie steep slopes, valleys etc). Oddly enough I've never noticed it happening with trees - only buildings.

    The first thing I thought to blame was my Ordnance Survey cultivation, which is based on 100 x 100 km areas and often involves TOC files well over 1 GB in size. So I decided to split the cultivation into 0.2 x 0.2 degree areas (roughly 20 x 20 km - a bit less than a level 10 square). Unfortunately, however, this has done nothing to improve the situation.

    My understanding is that it occurs because the buildings load before the higher res mesh, so they are placed on the lower res mesh. Is that correct? If so, it means that either the buildings need to load more slowly, or the mesh more quickly? Any thoughts would be greatly appreciated.

    OK thanks Rodeo - that's clarified it!

    That sounds like progress! Nice job! Regarding the mask files for terrain, I just had a closer look at my stuff here and didn't realize before that the terrain height masks don't have a corresponding non-mask file. I was confused by your question about that before because I thought they came in pairs like the orthophoto tiles. So... I have no idea how or if the elevation masks actually do work :D Based on the limited testing I did before when I converted a whole bunch of elevation, I remember concluding somehow that it did look like they were allowing partial terrain tiles to be built, but I could have been wrong about that. One of these days when I have some spare time I'll dig into it some more, unless you beat me to it :)

    With my new method I'm producing loads of masked TTH tiles. They don't seem to be producing any problems at the moment - but I need to fly around a bit more just to check. Also most of the cropped Geotiffs are overlapping, which geoconvert seems to take in its stride. I didn't realise you could do that before!

    If only there was some way to automatically generate the TFW files (like the INF to TFW converter) - that's the weak point in the process because it's so time-consuming if you have lots of cropped bits of Geotiff. AeroScenery just uses the AID files as far as I'm aware, so how does it create them? That's something I need to experiment with. If you could merge the cropped bits without creating large no-data areas that would also help - but I don't think (?) you can do that. In any case I'm not sure how the TFW file would work with an odd-shaped (non-rectangular) Geotiff.

    The other thing I've noticed is that it's possible to make the black (no data) areas transparent in the QGIS canvas using layer > properties > transparency - but unfortunately it doesn't seem to be possible to export the Geotiff and retain the transparency. As you say, that would be the best way of solving the problem if it were possible.

    UPDATE: I've managed to get round the problem of creating multiple TFW files for all the cropped bits of geotiff by using "raster > conversion > translate" in QGIS. Check "batch mode" and select the input and output directories. Check "creation options", click "+" and type "tfw" under "name" and "yes" under "value". I'll update my tutorial with the full details.

    OK - thanks qwerty - that's helped a lot.

    I think the problem was that I decided early on to banish masks from my photoscenery so I stopped trying to understand them after that! The other thing that confused me was that I imagined that maybe something strange was happening to them during the geoconversion process so that they no longer behaved like normal images.

    As for the TTH files I assume that the mask and image are inseparable for some reason?

    UPDATE: I've started using raster > extraction > clipper in QGIS to crop rectangles from my Geotiff, thereby avoiding the holes in the data. Each cropped Geotiff needs its own separate TFW file so it's quite time-consuming - but it does work. It also enables me to geoconvert all levels 12-14. Previously I was only geoconverting level 14, which was giving poor results in some areas - even though I had a base layer of level 11 derived from 50 metre data.

    The masks only create issues when you try to create adjoining scenery at the same level.

    Interesting what you say about masks - and I think I've partially (!) got my head around it. However some bits are still puzzling me:

    "The masks only create issues when you try to create adjoining scenery at the same level, because the mask blocks out the area it covers until it's removed, even if you replace the partial-coverage scenery tiles with full coverage."

    Suppose you geoconverted 2 adjacent areas and the boundary bisected level 15 tiles .....

    You say: "You only get black regions if you have incomplete tiles and *no* mask for the incomplete area."

    What confuses me is that I seem to remember getting black areas in my early attempts at scenery creation - when I had loads of masks for the incomplete tiles at the boundaries of my area. Or maybe I'm mis-remembering that. As you say, if you can see through masked areas then masked tiles shouldn't create any black areas (for TTC) or sea-level areas (for TTH). The worst you should get is blurry scenery at that point - or a low res mesh in the case of masked TTH files.

    The fact that the TTH files don't have separate masks confuses me too.

    Hey Ian, great work!!! I never tested this myself, but I wonder if you can use those empty voids to create masks for the terrain mesh. I know that Geoconvert generates terrain masks, and assuming they work like masks for orthophotos, you might be able to do this in a way that just shows the lower-level mesh in places where your higher-res stuff is missing.

    The trick is, I don't know what geoconvert looks for to create masks with the terrain data. It's not full black, or full white -- does geotiff support transparency? Maybe if you leave those empty spots transparent instead of black, it'll mask them for you...?

    This of course creates the issue of partial terrain tiles and masks, but that might be ok if your original dataset isn't going to ever fill in those voids.

    Hi qwerty - and thanks!

    I've been puzzling about this for ages and asking myself the same questions. As far as I can see, the problem with the "holes" in the 1 metre data is that there is actually data in the "holes" - but it's the wrong data. i.e. The holes are basically flat areas at sea level. It might be useful if the holes could be converted into missing data in the true sense of the term. I don't know whether there is a way of doing that. (I may be mistaken but I think there is some way in QGIS of setting the black areas to a certain height - but that doesn't really solve our problem.) Geoconvert doesn't currently produce any masked tiles where the holes are. I assume that this is because it doesn't see them as missing data.

    But then there is the question of what effect it would have if there were masked TTH files in those places. Masked TTC files seem to have the effect of hiding whatever tiles are below them in level - i.e. masked level 15 tiles hide level 14 tiles and below. (To be more precise they hide what's below where the mask is - i.e. where there is no ortho image.) This produces black patches in the scenery. By analogy you would think that masked TTH files would have the same effect - i.e. they would produce patches where you couldn't see the lower level TTH tiles. This would presumably produce flat areas at sea level as well - so there wouldn't be any advantage.

    Another thing that puzzles me is the fact that the masked TTC files come in pairs - i.e. a tile + a separate mask - whereas there is only one masked TTH file. This makes me wonder whether the TTC and TTH masks really behave in the same way.

    I find masks very confusing - and I'm not sure that I fully understand them!

    Just a heads up that Orbx is doing the entire UK for Aerofly as we speak........…rue-earth-for-aerofly-fs/

    Interesting ..... and thanks

    I know that I would still probably want to use my own ortho images because I always geoconvert to level 15 - and I doubt that ORBX will do that because at level 15 we're talking about more than 1 TB of data for GB.

    However, I would definitely still buy the product in order to use the cultivation - so maybe I won't invest too much effort in that for the time being except for using improved textures on the default buildings, which is fairly easy.

    The terrain mesh is then the big question. Will they use the Ordnance Survey 5 m mesh? The price for me is prohibitive, but I imagine it wouldn't be a problem for ORBX. That's the real dilemma for me, because I only have access to 50 metre and 1 metre data - and the latter is very time-consuming to use.

    Conclusion: (1) I'll continue to make high res level 15 scenery. (2) I'll go easy on 3D models for the time being, though I'll continue with the one(s) I have on the drawing board. (3) I'll probably also continue with my work on the 1 metre mesh because I'm impatient to see what the scenery looks like with a better mesh.

    This is just a bit of extra detail to go with the above tutorial on heightmaps for Great Britain.

    Currently I'm working on the SW peninsula of England (roughly Cornwall) and I've attached a couple of Geotiffs to give you an idea of what to expect if you want to have a go yourself. The first one comes from the Ordnance Survey 50 metre data (OS Terrain 50) and is complete with no missing data, as you can see. I use this for level 11. The second was compiled from DEFRA 1 metre data. In the latter there are an annoying number of black patches - i.e. areas with no data. If geoconverted, these would produce a flat landscape at sea level - so they need to be avoided. Fortunately the coast is complete. I use this data, where available, for level 14 tiles.

    To get the 50 metre heightmap a total of 33 .asc files were merged. The size of the Geotiff is about 10 MB.

    The 1 metre heightmap is the result of merging roughly 3000 .asc files, and it had to be done in two stages in batches of about 100. Its size is a stonking 18.5 GB. Although it is interesting here for illustrative purposes, it is actually useless from the point of view of geoconversion. Geoconvert doesn't seem to be able to cope with a single TIF of this size. It hangs for ever on 0 tiles - "for ever" being about 10 mins before I gave up! About a quarter of this area is do-able I would say. Then you have the "interesting" task of picking out and deleting all the tiles that coincide with black areas. This can be done by getting the lon-lat co-ordinates of the black patches from QGIS and mapping them onto a grid showing the Aerofly tiles - but it's quite time-consuming. (If anybody can think of a better way I'd be very interested*.) Level 14 tiles are only about 1.5 km square so there's a good chance of them fitting in between the black areas. Very few level 12 & 13 tiles are usable, so I've pretty much given up on them.

    *Actually, it's just occurred to me that I could somehow or other superimpose a semi-transparent image of the geotiff on an image of the Aeroscenery map with all the level 14 tiles selected. I'll try that tomorrow using GIMP.

    [Frustratingly, the Ordnance Survey also has complete 5 metre data, but they won't let you try it out for free unless you intend eventually to use it commercially (and pay a hefty fee). My application to use it for a hobby was rejected.]

    OK thanks Rodeo - that's been a big help.

    I've attached some samples of what I've done so far. They are for the "Celtic Fringe" - i.e. northern and western parts of the British Isles. They're probably not perfect yet, but they do make the scenery look massively more realistic - even in quite large towns.

    Thanks Rodeo - I'll have a look for those template BMP files. I think another problem I have at the moment is that my "gable" doesn't match my "wall". The wall is dark stone - the gable is light stone. How are the TTX files randomised? Does "res_gable_01" always go with "res_wall_01" for example - or could you get "res_gable_01" with "res_wall_02" say? If the latter, then I guess there will inevitably be some mismatches?

    OK - just had a look at your BMPs. You have a lot of tiles on your roof images. I may have to copy my roof image and join together the copies to get that many tiles. Same with the walls - you have a lot of windows in your wall images. I guess I need a lot of windows to be sure of getting at least a couple of windows in a small house?


    Many thanks for the info!

    Many thanks for the info!

    Hi again rodeo

    The images apparently need to be 512x512, 1024x1024 etc. Are they then stretched or cropped to fit the house? I'm trying to figure out how to size the image so as to make the roof tiles look a normal size. The slates/tiles look the right size on large buildings, but on small buildings the roof slates/tiles look far too large. I guess I need an image with smaller roof tiles on it, so I'll keep experimenting.

    Other than that the improvement is enormous. I'm cultivating Cornwall (UK) and, as can be seen from the screenshot below, the overall colour of the buildings (grey, stone and white) is way better now that I've got rid of the red roofs etc. If I could make the slates smaller the roofs would be perfect. (I have the same problem with the stonework texture on the side walls, but the walls are not so visible from the air so it doesn't matter so much.)


    OK thanks Rodeo - I'll have a play with it.

    OK thanks Rodeo - I'll have a play with it.

    Success - I've managed to get a TTX file from a trial image, I assume that the different TTX files in the default "building_textures" folder ..... 01 02 03 04 etc ..... are just alternatives that are chosen randomly? Can I have fewer options? (I'm only interesting in the residential ones - I'll keep the default commercial TTX files as they are.)

    A couple more questions:

    1. Are "color" and "light" simply names for different textures - or do they have different functions?

    2. Do I need an image of a roof from directly above for the "res_roof" textures (ie with the roof ridge across the middle)?